Your SlideShare is downloading. ×
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Mobile healthcare solutions for biomedical applications
Upcoming SlideShare
Loading in...5
×

Thanks for flagging this SlideShare!

Oops! An error has occurred.

×
Saving this for later? Get the SlideShare app to save on your phone or tablet. Read anywhere, anytime – even offline.
Text the download link to your phone
Standard text messaging rates apply

Mobile healthcare solutions for biomedical applications

8,716

Published on

mobile solution for healthcare improvement

mobile solution for healthcare improvement

Published in: Technology, Business
0 Comments
1 Like
Statistics
Notes
  • Be the first to comment

No Downloads
Views
Total Views
8,716
On Slideshare
0
From Embeds
0
Number of Embeds
0
Actions
Shares
0
Downloads
96
Comments
0
Likes
1
Embeds 0
No embeds

Report content
Flagged as inappropriate Flag as inappropriate
Flag as inappropriate

Select your reason for flagging this presentation as inappropriate.

Cancel
No notes for slide

Transcript

  • 1. Mobile Health Solutionsfor BiomedicalApplicationsPhillip OllaMadonna University, USAJoseph TanWayne State University, USA Medical Information science reference Hershey • New York
  • 2. Director of Editorial Content: Kristin KlingerSenior Managing Editor: Jamie SnavelyManaging Editor: Jeff AshAssistant Managing Editor: Carole CoulsonTypesetter: Larissa VinciCover Design: Lisa TosheffPrinted at: Yurchak Printing Inc.Published in the United States of America by Information Science Reference (an imprint of IGI Global) 701 E. Chocolate Avenue, Suite 200 Hershey PA 17033 Tel: 717-533-8845 Fax: 717-533-8661 E-mail: cust@igi-global.com Web site: http://www.igi-global.com/referenceand in the United Kingdom by Information Science Reference (an imprint of IGI Global) 3 Henrietta Street Covent Garden London WC2E 8LU Tel: 44 20 7240 0856 Fax: 44 20 7379 0609 Web site: http://www.eurospanbookstore.comCopyright © 2009 by IGI Global. All rights reserved. No part of this publication may be reproduced, stored or distributed in any form or byany means, electronic or mechanical, including photocopying, without written permission from the publisher. Product or company names used in this set are for identi.cation purposes only. Inclusion of the names of the products or companies doesnot indicate a claim of ownership by IGI Global of the trademark or registered trademark.Library of Congress Cataloging-in-Publication DataMobile health solutions for biomedical applications / Phillip Olla and Joseph Tan, editors. p. ; cm. Includes bibliographical references and index. Summary: “This book gives detailed analysis of the technology, applications and uses of mobile technologies in the healthcare sector byusing case studies to highlight the successes and concerns of mobile health projects”--Provided by publisher. ISBN 978-1-60566-332-6 (hardcover : alk. paper) 1. Telecommunication in medicine. 2. Mobile communication systems. 3. Wireless communication systems. 4. Cellular telephones. 5.Medical technology. I. Olla, Phillip. II. Tan, Joseph K. H. [DNLM: 1. Telemedicine. 2. Ambulatory Monitoring. 3. Cellular Phone. 4. Computers, Handheld. 5. Medical Records Systems, Com-puterized. 6. User-Computer Interface. W 83.1 M6865 2009] R119.9.M58 2009 610.28--dc22 2008040451British Cataloguing in Publication DataA Cataloguing in Publication record for this book is available from the British Library.All work contributed to this book is new, previously-unpublished material. The views expressed in this book are those of the authors, but notnecessarily of the publisher.
  • 3. NeedAdvanced Ad
  • 4. Editorial Advisory BoardGeorge Demiris, University of Missouri, USANayna Patel, Brunel University, UKThomas M. Deserno, RWTH Aachen University, GermanyJyoti Choudrie, University of Hertfordshire, UKPaul Hu, University of Utah, USAPatrice Monthrope, University of West Indies, JamaicaRichard Hull, University of Newcastle upon Tyne, United KingdomElena Qureshi, Madonna University, USAFrancis Lau, University of Victoria, CanadaVenus Olla, Nottingham University, UKDave Parry, Auckland University of Technology, New ZealandMathew Guah, Erasmus University, The NetherlandsJim Warren, University of Auckland, New ZealandH. Joseph Wen, Southeast Missouri State University, USAYvette Miller, University of Toronto, CanadaYufei Yuan, McMaster University, CanadaDaniel Zeng, University of Arizona, USAKai Zheng, The University of Michigan, USAJacqueline Brodie, Napier University, ScotlandCarla Wiggins, Idaho State University, USABendik Bygstad, Norwegian School of IT, Norway
  • 5. Table of ContentsPreface . ...............................................................................................................................................xiii Section I Mobile Health Applications and TechnologiesChapter IEvaluation of Two Mobile Nutrition Tracking Applications for Chronically Ill Populationswith Low Literacy Skills ........................................................................................................................ 1 Katie A. Siek, University of Colorado at Boulder, USA Kay H. Connelly, Indiana University, USA Beenish Chaudry, Indiana University, USA Desiree Lambert, Trilogy Health Services, USA Janet L. Welch, Indiana University School of Nursing, USAChapter IIAccessing an Existing Virtual Electronic Patient Record with a Secure Wireless Architecture .......... 24 Ana Ferreira, Center for Informatics, Faculty of Medicine in Porto, Portugal Luís Barreto, Instituto Politécnico de Viana do Castelo, Portugal Pedro Brandão, LIACC at Faculty of Science in Porto, R. Campo Alegre, Portugal Ricardo Correia, Center for Informatics, Faculty of Medicine in Porto, Portugal Susana Sargento, Universidade de Aveiro, Portugal Luís Antunes, LIACC at Faculty of Science in Porto, R. Campo Alegre, PortugalChapter IIIPersonal Health Records Systems Go Mobile: Defining Evaluation Components............................... 45 . Phillip Olla, Madonna University, USA Joseph Tan, Wayne State University, USAChapter IVMedical Information Representation Framework for Mobile Healthcare ............................................ 71 Ing Widya,University of Twente, The Netherlands HaiLiang Mei,University of Twente, The Netherlands Bert-Jan van Beijnum,University of Twente, The Netherlands Jacqueline Wijsman,University of Twente, The Netherlands Hermie J. Hermens,University of Twente, The Netherlands
  • 6. Chapter VA Distributed Approach of a Clinical Decision Support System Based on Cooperation...................... 92 Daniel Ruiz-Fernández, University of Alicante, Spain Antonio Soriano-Payá, University of Alicante, SpainChapter VIManaging Mobile Healthcare Knowledge: Physicians’ Perceptions on KnowledgeCreation and Reuse.............................................................................................................................. 111 Teppo Räisänen, University of Oulu, Finland Harri Oinas-Kukkonen, University of Oulu, Finland Katja Leiviskä, University of Oulu, Finland Matti Seppänen, The Finnish Medical Society Duodecim, Finland Markku Kallio, The Finnish Medical Society Duodecim, Finland Section II Patient Monitoring and Wearable DevicesChapter VIIPatient Monitoring in Diverse Environments ..................................................................................... 129 Yousef Jasemian, Engineering College of Aarhus, DenmarkChapter VIIIMonitoring Hospital Patients Using Ambient Displays....................................................................... 143 Monica Tentori, CICESE, Mexico Daniela Segura, CICESE, Mexico Jesus Favela, CICESE, MexicoChapter IXTowards Easy-to-Use, Safe, and Secure Wireless Medical Body Sensor Networks........................... 159 Javier Espina, Philips Research Europe, The Netherlands Heribert Baldus, Philips Research Europe, The Netherlands Thomas Falck, Philips Research Europe, The Netherlands Oscar Garcia, Philips Research Europe, The Netherlands Karin Klabunde, Philips Research Europe, The NetherlandsChapter XSensing of Vital Signs and Transmission Using Wireless Networks................................................... 180 Yousef Jasemian, Engineering College of Aarhus, Denmark
  • 7. Chapter XITowards Wearable Physiological Monitoring on a Mobile Phone...................................................... 208 . Nuria Oliver, Telefonica Research, Spain Fernando Flores-Mangas, University of Toronto, Canada Rodrigo de Oliveira, State University of Campinas, Brazil Section III Context Aware SystemsChapter XIIA Framework for Capturing Patient Consent in Pervasive Healthcare Applications.......................... 245 Giovanni Russello, Imperial College London, UK Changyu Dong, Imperial College London, UK Naranker Dualy, Imperial College London, UKChapter XIIITechnology Enablers for Context-Aware Healthcare Applications..................................................... 260 Filipe Meneses, Universidade do Minho, Portugal Adriano Moreira, Universidade do Minho, PortugalChapter XIVModeling Spatiotemporal Developments in Spatial Health Systems.................................................. 270 Bjorn Gottfried, University of Bremen, GermanyChapter XVContext-Aware Task Distribution for Enhanced M-health Application Performance......................... 285 Hailiang Mei, University of Twente, The Netherlands Bert-Jan van Beijnum, University of Twente, The Netherlands Ing Widya, University of Twente, The Netherlands Val Jones, University of Twente, The Netherlands Hermie Hermens, , University of Twente, The NetherlandsCompilation of References................................................................................................................ 308About the Contributors..................................................................................................................... 332Index.................................................................................................................................................... 341
  • 8. Detailed Table of ContentsPreface . ...............................................................................................................................................xiii Section I Mobile Health Applications and TechnologiesChapter IEvaluation of Two Mobile Nutrition Tracking Applications for Chronically Ill Populationswith Low Literacy Skills ........................................................................................................................ 1 Katie A. Siek, University of Colorado at Boulder, USA Kay H. Connelly, Indiana University, USA Beenish Chaudry, Indiana University, USA Desiree Lambert, Trilogy Health Services, USA Janet L. Welch, Indiana University School of Nursing, USAIn this chapter, the authors discuss two case studies that compare and contrast the use of barcode scanning,voice recording, and patient self reporting as a means to monitor the nutritional intake of a chronicallyill population.Chapter IIAccessing an Existing Virtual Electronic Patient Record with a Secure Wireless Architecture .......... 24 Ana Ferreira, Center for Informatics, Faculty of Medicine in Porto, Portugal Luís Barreto, Instituto Politécnico de Viana do Castelo, Portugal Pedro Brandão, LIACC at Faculty of Science in Porto, R. Campo Alegre, Portugal Ricardo Correia, Center for Informatics, Faculty of Medicine in Porto, Portugal Susana Sargento, Universidade de Aveiro, Portugal Luís Antunes, LIACC at Faculty of Science in Porto, R. Campo Alegre, PortugalThe main objective of this chapter is to model, develop and evaluate (e.g. in terms of efficiency, com-plexity, impact and against network attacks) a proposal for a secure wireless architecture in order toaccess a VEPR. This VEPR is being used within a university hospital by more than 1,000 doctors, on adaily basis. Its users would greatly benefit if this service would be extended to a wider part of the hos-pital and not only to their workstation, achieving this way faster and greater mobility in the treatmentof their patients.
  • 9. Chapter IIIPersonal Health Records Systems Go Mobile: Defining Evaluation Components............................... 45 . Phillip Olla, Madonna University, USA Joseph Tan, Wayne State University, USAThis chapter provides an overview of Mobile Personal Health Record (MPHR) systems. A Mobilepersonal health record is an eclectic application through which patients can access, manage, and sharetheir health information from a mobile device in a private, confidential, and secure environment. Specifi-cally, the chapter reviews the extant literature on critical evaluative components to be considered whenassessing MPHR systems.Chapter IVMedical Information Representation Framework for Mobile Healthcare ............................................ 71 Ing Widya,University of Twente, The Netherlands HaiLiang Mei,University of Twente, The Netherlands Bert-Jan van Beijnum,University of Twente, The Netherlands Jacqueline Wijsman,University of Twente, The Netherlands Hermie J. Hermens,University of Twente, The NetherlandsThis chapter describes a framework which enables medical information, in particular clinical vital signsand professional annotations, be processed, exchanged, stored and managed modularly and flexibly in amobile, distributed and heterogeneous environment despite the diversity of the formats used to representthe information.Chapter VA Distributed Approach of a Clinical Decision Support System Based on Cooperation...................... 92 Daniel Ruiz-Fernández, University of Alicante, Spain Antonio Soriano-Payá, University of Alicante, SpainThis chapter presents an architecture for diagnosis support based on the collaboration among differentdiagnosis-support artificial entities and the physicians themselves; the authors try to imitate the clinicalmeetings in hospitals in which the members of a medical team share their opinions in order to analyzecomplicated diagnoses.Chapter VIManaging Mobile Healthcare Knowledge: Physicians’ Perceptions on KnowledgeCreation and Reuse.............................................................................................................................. 111 Teppo Räisänen, University of Oulu, Finland Harri Oinas-Kukkonen, University of Oulu, Finland Katja Leiviskä, University of Oulu, Finland Matti Seppänen, The Finnish Medical Society Duodecim, Finland Markku Kallio, The Finnish Medical Society Duodecim, Finland
  • 10. This chapter aims to demonstrate that mobile healthcare information system may also help physiciansto communicate and collaborate as well as learn and share their experiences within their work commu-nity. Physicians’ usage of a mobile system is analyzed through a knowledge management frameworkknown as the 7C model. The data was collected through the Internet among all of the 352 users of themobile system. The results indicate that frequent use of the system seemed to improve individual physi-cians’ knowledge work as well as the collective intelligence of a work community. Overall, knowledgemanagement seems to be a prominent approach for studying healthcare information systems and theirimpact on the work of physicians. Section II Patient Monitoring and Wearable DevicesChapter VIIPatient Monitoring in Diverse Environments ..................................................................................... 129 Yousef Jasemian, Engineering College of Aarhus, DenmarkThis chapter intends to explore the issues and limitations concerning application of mobile health systemin diverse environments, trying to emphasize the advantages and drawbacks, data security and integritysuggesting approaches for enhancements. These issues will be explored in successive subsections byintroducing two studies which were undertaken by the author.Chapter VIIIMonitoring Hospital Patients Using Ambient Displays....................................................................... 143 Monica Tentori, CICESE, Mexico Daniela Segura, CICESE, Mexico Jesus Favela, CICESE, MexicoIn this chapter the authors explore the use of ambient displays to adequately monitor patient’s healthstatus and promptly and opportunistically notify hospital workers of those changes. To show the feasibil-ity and applicability of ambient displays in hospitals they designed and developed two ambient displaysthat can be used to provide awareness patients’ health status to hospital workers.Chapter IXTowards Easy-to-Use, Safe, and Secure Wireless Medical Body Sensor Networks........................... 159 Javier Espina, Philips Research Europe, The Netherlands Heribert Baldus, Philips Research Europe, The Netherlands Thomas Falck, Philips Research Europe, The Netherlands Oscar Garcia, Philips Research Europe, The Netherlands Karin Klabunde, Philips Research Europe, The NetherlandsWireless Body Sensor Networks (BSNs) are an indispensable building stone for any pervasive healthcaresystem. Although suitable wireless technologies are available and standardization dedicated to BSNcommunication has been initiated, the authors identify key challenges in the areas of easy-of-use, safety,
  • 11. and security that hinder a quick adoption of BSNs. To address the identified issues we propose usingBody-Coupled Communication (BCC) for the automatic formation of BSNs and for user identification.They also present a lightweight mechanism that enables a transparent security setup for BSNs used inpervasive healthcare systems.Chapter XSensing of Vital Signs and Transmission Using Wireless Networks................................................... 180 Yousef Jasemian, Engineering College of Aarhus, DenmarkThis chapter deals with a comprehensive investigation of feasibility of wireless and cellular telecom-munication technologies and services in a real-time M-Health system. The chapter bases its investiga-tion, results, discussion and argumentation on an already developed remote patient monitoring systemby the author.Chapter XITowards Wearable Physiological Monitoring on a Mobile Phone...................................................... 208 . Nuria Oliver, Telefonica Research, Spain Fernando Flores-Mangas, University of Toronto, Canada Rodrigo de Oliveira, State University of Campinas, BrazilIn this chapter, we present our experience in using mobile phones as a platform for real-time physiologicalmonitoring and analysis. In particular, we describe in detail the TripleBeat system, a research prototypethat assists runners in achieving predefined exercise goals via musical feedback, a glanceable interfacefor increased personal awareness and a virtual competition. We believe that systems like TripleBeat willplay an important role in assisting users towards healthier and more active lifestyles. Section III Context Aware SystemsChapter XIIA Framework for Capturing Patient Consent in Pervasive Healthcare Applications.......................... 245 Giovanni Russello, Imperial College London, UK Changyu Dong, Imperial College London, UK Naranker Dualy, Imperial College London, UKIn this chapter, the authors describe a new framework for pervasive healthcare applications where thepatient’s consent has a pivotal role. In their framework, patients are able to control the disclosure of theirmedical data. The patient’s consent is implicitly captured by the context in which his or her medical datais being accessed. Context is expressed in terms of workflows. The execution of a task in a workflowcarries information that the system uses for providing access rights accordingly to the patient’s consent.Ultimately, the patient is in charge of withdrawing consent if necessary. Moreover, the use of workflowenables the enforcement of the need-to-kwon principle. This means that a subject is authorised to accesssensitive data only when required by the actual situation.
  • 12. Chapter XIIITechnology Enablers for Context-Aware Healthcare Applications..................................................... 260 Filipe Meneses, Universidade do Minho, Portugal Adriano Moreira, Universidade do Minho, PortugalThis chapter focuses on how context and location can be used in innovative applications and how touse a set of solutions and technologies that enable the development of innovative context and location-aware solutions for healthcare area. It shows how a mobile phone can be used to compute the level offamiliarity of the user with the surrounding environment and how the familiarity level can be used in anumber of situations.Chapter XIVModeling Spatiotemporal Developments in Spatial Health Systems.................................................. 270 Bjorn Gottfried, University of Bremen, GermanyThis chapter introduces spatial health systems, identifies fun¬damental properties of these systems, anddetails for specific applications the methods to be applied in order to show how problems are solved inthis field. On the one hand, this chapter gives an overview of this area, on the other hand, it is writtenfor those who are interested in designing spatial health systems. The result is that different spatial scalesand pur¬poses require different representations for describing the spatiotemporal change of objects,that is their spatiotemporal development, showing how fundamental aspects of spatial health systemsare dealt with.Chapter XVContext-Aware Task Distribution for Enhanced M-health Application Performance......................... 285 Hailiang Mei, University of Twente, The Netherlands Bert-Jan van Beijnum, University of Twente, The Netherlands Ing Widya, University of Twente, The Netherlands Val Jones, University of Twente, The Netherlands Hermie Hermens, , University of Twente, The NetherlandsAs well as applying the traditional adaptation methods such as protocol adaptation and data prioritization,the authors investigate the possibility of adaptation based on dynamic task redistribution. In this chapter,the authors propose an adaptation middleware that consists of a task assignment decision mechanismand a task redistribution infrastructure. The decision mechanism represents task assignment as a graphmapping problem and searches for the optimal assignment given the latest context information. Oncea new assignment is identified, the member tasks are distributed accordingly by the distribution infra-structure. A prototype implementation based on the OSGi framework is reported to validate the taskredistribution infrastructure.Compilation of References................................................................................................................ 308About the Contributors..................................................................................................................... 332Index.................................................................................................................................................... 341
  • 13. xiiiPrefacePervasive healthcare environment, focusing on the integration of mobile and ubiquitous technology toreform working and living conditions for individuals and organizations in the healthcare sector, sets thestage for an innovative emerging research discipline. Healthcare systems are experiencing a variety ofchallenges including the prevalence of life-style related conditions, growing consumerism in healthcare,the need to empower patients with information for better decision making, requests for better tools forself-care and management of deteriorating health conditions, the need for seamless access for healthcareservices via the Internet and mobile devices, and the growing costs of providing healthcare. Mobile health (m-health) is an integral and significant part of the emerging pervasive healthcare field.M-Health contains three core components integrated into the healthcare environment. The first componentis the availability of a reliable wireless architecture; the second component is the integration of medicalsensor or wearable devices for monitoring; and the final component is a robust application and servicesinfrastructure. M-Health relates to applications and systems such as telemedicine, telehealth, e-health,and biomedical sensing system. The rapid advances in information communication technology (ICT),nanotechnology, bio-monitoring, mobile networks, pervasive computing, wearable systems, and drugdelivery approaches are transforming the healthcare sector and fueling the m-health phenomenon. M-Health aims to make healthcare accessible to anyone, anytime, and anywhere by elimination constraintssuch as time and location in addition to increasing both the coverage and quality of healthcare. Mobile and wireless concepts in healthcare are typically related to bio-monitoring and home moni-toring; however, more recently the trend to incorporate mobile technology has become more prevalentacross almost the entire healthcare data acquisition task domains. Bio monitoring using mobile networksincludes physiological monitoring of parameters such as heart rate, electrocardiogram (ECG), electro-encephalogram (EEG) monitoring, blood pressure, blood oximetry, and other physiological signals.Alternative uses include physical activity monitoring of parameters such as movement, gastrointestinaltelemetry fall detection, and location tracking. Using mobile technology, patient records can be accessedby healthcare professionals from any given location by connecting the institution’s internal network.Physicians now have ubiquitous access to patient history, laboratory results, pharmaceutical data, in-surance information, and medical resources. These mobile healthcare applications improve the qualityof patient care. Handheld devices can also be used in home healthcare, for example, to fight diabetesthrough effective monitoring. A comprehensive overview of some of these mobile health applicationsand research has been presented in this book. This book provides an international perspective on the benefits of mobile health technology to illus-trate different examples and applications implemented in the global healthcare sector. The work features32 contributing authors representing six countries including the United States, United Kingdom, Spain,Portugal, Italy, and Denmark. Even though the healthcare policies and governance of healthcare systems
  • 14. xivin these countries differ, the benefits to be realized from a future of implementations of mobile healthtechnology are not inconsistent among the countries. The book may be divided into three major sections:1. Mobile Health Applications and Technologies2. Patient Monitoring and Wearable Devices3. Context Aware Systems in Healthcare The first section “Mobile Health Applications and Technologies” provides an analysis of the technol-ogy. Case studies highlighting the successes and challenges of mobile health projects offer real-worldillustrations of applications and uses of mobile technologies in the healthcare sector. M-Health is abroad area transcending multiple disciplines and utilizing a broad range of technologies. “Evaluation ofTwo Mobile Nutrition Tracking Applications for Chronically Ill Populations with Low Literacy Skills,”authored by Katie A. Siek, Kay H. Connelly, Beenish Chaudry, Desiree Lambert, and Janet L. Welch,discusses two case studies that compare and contrast the use of barcode scanning, voice recording, andpatient self reporting as a means to monitor the nutritional intake of a chronically ill population. Chapter II “Accessing an Existing Virtual Electronic Patient Record with a Secure Wireless Archi-tecture” by Ana Ferreira, Luís Barreto, Pedro Brandão, Ricardo Correia, Susana Sargento, and LuísAntunes presents the concept of a virtual electronic patient records system that enables the integrationand sharing of healthcare information within heterogeneous organizations. The VEPR system aims toalleviate the constraints in terms of physical location as well as technology in order to access vital patientrecords. The use of wireless technology attempts to allow access to patient data and processing of clinicalrecords closer to the point of care. The ubiquitous access to information can minimize physical as wellas time constraints for healthcare, enhancing users’ mobility within the institution. The next chapter inthis section entitled “Personal Health Records Systems Go Mobile: Defining Evaluation Components”is authored by Phillip Olla and Joseph Tan. It provides an overview of Mobile Personal Health Record(MPHR) systems. A Mobile personal health record is an eclectic application through which patients canaccess, manage and share their health information from a mobile device in a private, confidential, andsecure environment. Chapter IV focusing on “Medical Information Representation Framework for Mobile Healthcare”was written by Ing Widya, HaiLiang Mei, Bert-Jan van Beijnum, Jacqueline Wijsman, and HermieHermens. This chapter describes a framework which enables medical information such as clinical, vitalsigns and professional annotations to be manipulated in a mobile, distributed and heterogeneous envi-ronment despite the diversity of the formats used to represent the information. It further proposes theuse of techniques and constructs similar to the internet to deal with medical information represented inmultiple formats. Chapter V is “A Distributed Approach of a Clinical Decision Support System Basedon Cooperation,” authored by Daniel Ruiz-Fernández and Antonio Soriano-Payá. This chapter discussesan architecture that supports diagnosis based on the collaboration among different diagnosis-supportartificial entities or agents and the physicians themselves. The proposed systems architecture, whichwas tested in a melanoma and urological dysfunctions diagnosis, combines availability, cooperation andharmonization of all contributions in a diagnosis process. Chapter VI, the final chapter in this section,“Managing Mobile Healthcare Knowledge: Physicians’ Perceptions on Knowledge Creation and Reuse”was authored by Teppo Räisänen, Harri Oinas-Kukkonen, Katja Leiviskä, Matti Seppänen, and MarkkuKallio. This chapter focuses on mobile access to medical literature and electronic pharmacopoeias, aim-ing to demonstrate that using these recourses effectively may help physicians to communicate and col-
  • 15. xvlaborate as well as learn and share their experiences within their user community. The chapter presentsa case study of the users of Duodecim mobile healthcare information system. The second section presents research on Patient Monitoring and Wearable Devices. Chapter VII, thefirst chapter in this section, is titled “Patient Monitoring in Diverse Environments” and is authored byYousef Jasemian. This chapter discusses the benefits of recording of physiological vital signs in patients’real-life environment by a mobile health system. This approach is useful in the management of chronicdisorders such as hypertension, diabetes, anorexia nervosa, chronic pain, or severe obesity. The authorexplored the issues and limitations concerning the application of mobile health system in diverse envi-ronments, emphasizing the advantages and drawbacks, data security and integrity while also suggestingapproaches for enhancements. The following chapter, Chapter VIII, is titled “Monitoring Hospital Patientsusing Ambient Displays” authored by Monica Tentori, Daniela Segura, and Jesus Favela. This chapterexplores the use of ambient displays to promptly notify hospital workers of relevant events related totheir patients. To highlight the feasibility and applicability of ambient displays in hospitals, this chapterpresents two ambient displays aimed at creating a wearable connection between patients and healthcareproviders. The authors also discuss issues and opportunities for the deployment of ambient displaysfor patient monitoring. Chapter IX is titled “Towards Easy-to-uUse, Safe, and Secure Wireless MedicalBody Sensor Networks” and is authored by Javier Espina, Heribert Baldus, Thomas Falck, Oscar Garcia,and Karin Klabunde. This chapter discusses the use of wireless body sensor networks (BSNs), whichare an integral part of any pervasive healthcare system. It discusses suitable wireless technologies andstandardization dedicated to BSN communication and highlights key challenges in the areas of easy-of-use, safety, and security that hinder a quick adoption of BSNs. To address the identified challenges,the authors proposed the use of body-coupled communication (BCC) for the automatic formation ofBSNs and for user identification and presented a lightweight mechanism that would enable a transparentsecurity setup for BSNs used in pervasive healthcare systems. Chapter X is titled “Sensing of Vital Signs and Transmission Using Wireless Networks” and is authoredby Yousef Jasemian. This chapter investigated the feasibility using wireless and cellular telecommu-nication technologies and services in a real-time m-health system. He based his investigation, results,discussion and argumentation on an existing remote patient monitoring system. His results indicatedthat the system functioned with a clinically acceptable performance, and transferred medical data witha reasonable quality, even though the system was tested under totally uncontrolled circumstances duringthe patients’ daily activities. Both the patients and the healthcare personnel who participated expressedtheir confidence in using the technology. The author also suggested enhancing features for more reliable,more secure, more user-friendly and higher performing M-Health system in future implementations. Chapter XI, “Towards Wearable Physiological Monitoring on a Mobile Phone” by Nuria Oliver,Fernando Flores-Mangas, and Rodrigo de Oliveira discusses the experience gained from using mobilephones as a platform for real-time physiological monitoring and analysis. The authors presented twomobile phone-based prototypes that explore the impact of real-time physiological monitoring in the dailylife of users. The first prototype is called HealthGear; this is a system to monitor users while they aresleeping and automatically detect sleep apnea events; the second is TripleBeat, a prototype that assistsrunners in achieving predefined exercise goals via musical feedback and two persuasive techniques: aglanceable interface for increased personal awareness and a virtual competition. The third and last section focuses on research and on the theme of Context Aware Systems in thehealthcare arena. Chapter XII, the first chapter in this section, is titled “A Framework for CapturingPatient Consent in Pervasive Healthcare Applications.” It is authored by Giovanni Russello, ChangyuDong, and Naranker Dualy and describes a new framework for pervasive healthcare applications wherethe patient’s consent plays a pivotal role. In the framework presented, patients are able to control the
  • 16. xvidisclosure of their medical data. The patient’s consent is implicitly captured by the context in whichhis or her medical data is being accessed. Context is expressed in terms of workflows. The executionof a task in a workflow carries information that the system uses for providing access rights accord-ingly to the patient’s consent. Ultimately, the patient is in charge of withdrawing consent if necessary.Chapter XIII is titled “Technology Enablers for Context-Aware Healthcare Applications” authored byFilipe Meneses and Adriano Moreira. This chapter discusses how context and location can be used ininnovative applications and how to use a set of solutions and technologies that enable the developmentof innovative context and location-aware solutions for healthcare area. The chapter highlights how amobile phone can be used to compute the level of familiarity of the user with the surrounding environ-ment and how the familiarity level can be used in a number of situations. The increasing availabilityof mobile devices and wireless networks, and the tendency for them to become ubiquitous in our dallylives, creates a favourable technological environment for the emergence of new, simple, and added-valueapplications for healthcare. Chapter XIV is titled “Modeling Spatiotemporal Developments in SpatialHealth Systems” is authored by Bjorn Gottfried and discusses Spatial health systems and the supportthese systems can provide to disabled people and the elderly in dealing with everyday life problems.The author also addresses every kinds of health related issues that can develop in space and time. Thework focuses on how spatial health systems monitor the physical activity of people in order to determinehow to support the monitored individuals. Chapter XV, the final chapter in this section, titled, “Context-Aware Task Distribution for Enhanced M-Health Application Performance” authored by Hailiang Mei,Bert-Jan van Beijnum, Ing Widya, Val Jones, Hermie Hermens. This chapter describes the importanceof context-aware mobile healthcare systems. Due to the emergence of new medical sensor technologies,the fast adoption of advanced mobile systems to improve the quality of care required by today’s patientscontext aware healthcare systems is clearly needed . The authors propose an adaptation middleware thatconsists of a task assignment decision mechanism and a task (re-) distribution infrastructure. The deci-sion mechanism represents task assignment as a graph mapping problem and searches for the optimalassignment given the latest context information. The research presented in this book is important due to the emergence of pervasive computing andhealth care systems that provide quality patient care services. By reviewing the diverse chapters pre-sented a healthcare provider or practitioner will learn about the potential applications that will becomethe norm in the future.
  • 17. Section IMobile Health Applications and Technologies
  • 18. Chapter I Evaluation of Two Mobile Nutrition Tracking Applications for Chronically Ill Populations with Low Literacy Skills Katie A. Siek University of Colorado at Boulder, USA Kay H. Connelly Indiana University, USA Beenish Chaudry Indiana University, USA Desiree Lambert Trilogy Health Services, USA Janet L. Welch Indiana University School of Nursing, USA ABSTRACT In this chapter, the authors discuss two case studies that compare and contrast the use of barcode scanning, voice recording, and patient self reporting as a means to monitor the nutritional intake of a chronically ill population. In the first study, they found that participants preferred unstructured voice recordings rather than barcode scanning. Since unstructured voice recordings require costly transcrip- tion and analysis, they conducted a second case study where participants used barcode scanning or an integrated voice response system to record nutritional intake. The authors found that although the latter input method provided participants with a faster method to input food items, participants had difficulty using the system despite training. Copyright © 2009, IGI Global, distributing in print or electronic forms without written permission of IGI Global is prohibited.a
  • 19. Evaluation of Two Mobile Nutrition Tracking Applications INTRODUCTION select a picture. Health professionals could eas- ily administer the intervention and evaluate data Chronic diseases, such as chronic kidney disease without intermediate steps of electronic transcrip- (CKD) and heart disease, are among the leading tion. The low literacy chronically ill participants causes of death and disability in the world. At least benefit from using the application because they half of the chronic disease related deaths could can use the application anytime they consumed a be prevented by adopting a healthy lifestyle, such food item, receive immediate visual feedback on as good nutrition, increased physical activity, and their nutritional intake, and make decisions on a cessation of tobacco use. Researchers believe that prospective basis. In addition, the interface and the world must put a higher priority on interven- content could be customized for populations with tions to help prevent and successfully manage varying literacy and computing skills. chronic illness (Preventing Chronic Diseases: A In this chapter, as part of a larger study, we Vital Investment, 2005). will compare and contrast the use of barcode scan- Current interventions to help chronically ill ning, integrated voice response system (IVRS), populations improve their nutritional health and and patient self reporting as a means to monitor self-manage therapeutic diets include paper- their nutritional intake relative to their dietary based food diaries, 24 hour recalls, and food prescription of CKD patients. In the first case study frequency questionnaires (Dwyer, Picciano, we found that participants preferred unstructured Raiten, 2003; Resnicow et al., 2000). Patients who voice recordings rather than barcode scanning. use these interventions must have high literacy Unstructured voice recordings are difficult to and memory recall skills. Unfortunately, over a automatically parse and require transcription. We quarter of the United States population do not had to find out if patients would use a menu-based have the necessary literacy or numeracy skills structured voice input system, such as IVRSs for needed to successfully self-monitor themselves automated recognition. In the second case study, (Kirsch et al., 1993). If people cannot self-moni- we explored participant use of an IVRS and found tor themselves, they cannot manage their chronic although the system provided participants with a conditions (HRSA Literacy) and may lead them to quicker way to input food items, participants had worse health outcomes (Schillinger et al., 2002). difficulty using the system and some could not In addition, to administer current interventions use the system despite training. We will discuss medical professionals must spend a significant the methodology and findings from these two amount of time evaluating the data from paper- case studies. We will conclude the chapter with based forms. lessons learned during the user study and provide We are currently developing a mobile handheld considerations for future areas of research. application to assist CKD patients on hemodialy- sis monitor and maintain their nutritional intake. Initially, we thought a personal digital assistant RELATED WORK (PDA) would be the best solution for health pro- fessionals and patients (Connelly, Faber, Rogers, PDAs with scanner input and mobile phones Siek, Toscos, 2006). Participants could scan used for IVRS input gather information in many barcodes on food items for their primary input or domains. PDAs and scanners have been used select items from an interface as a secondary input. to show clinicians videos about specific unit These input mechanisms are ideal for low literacy appliances (Brandt, Björgvinsson, Hillgren, populations because there is no reading required Bergqvist, Emilson, 2002), save and search – participants only have to identify a barcode or for information about food products, music, and TRCTRT
  • 20. Evaluation of Two Mobile Nutrition Tracking Applicationsbooks (Bernheim, Combs, Smith, Gupta, 2005), dinner. The nutritional analysis is given on aand obtain information about an environment separate screen. Researchers at Indiana Universityfrom embedded barcodes (Fitzmaurice, Khan, studied how three people with CKD used Diet-Buxton, Kurtenback, Balakrishnan, 2003). MatePro to monitor nutritional consumption overMobile phones used for IVRSs have been used a three-month period. They found participants hadfor patient counseling to enhance time spent difficulty navigating standard PDA menu naviga-with health professionals (Glasgow, Bull, Piette, tion and preferred using a large PDA screen with Steiner, 2004) and assess patient status with touch sensitive icons (Dowell Welch, 2006).chronic illnesses such as depression, cancer, Sevick and colleagues evaluated how five CKDheart failure, and diabetes (Piette, 2000). In this participants used BalanceLog over a four-monthsection, we discuss specifically how PDAs and period. They found that participants improvedmobile phones have been used for interventions their dietary intake using the electronic nutritionand nutritional monitoring. monitoring system (Sevick et al., 2005). Both applications evaluated in these studies requiredPDA Nutrition Monitoring significant literacy and cognitive skills.Interventions Stephen Intille et al. created a proof-of-concept PDA application that provides users with a way toCurrently, there are many PDA applications that scan food items and obtain nutritional informationcan assist with the self-monitoring of nutritional to assist users in making healthy choices (Intille,intake. The United States Department of Agri- Kukla, Farzanfar, Bakr, 2003). The applicationculture (USDA) has a PDA nutrient database that did not have an extensive UPC/nutrition databaseprovides people with a mechanism for looking up because none are freely available. Although thethe nutritional information of foods. Users must application does not allow users to save intakecorrectly type the first few letters of a food item information, the application shows that integrationthey are looking for into a search box and then click of scanners and nutrition information is possiblethrough a series of menus to find the appropriate given enough resources.food item based on portion size and preparation Researchers at Microsoft created a generic(“USDA Palm OS Search,” 2008). barcode look-up system that gave participants DietMatePro ( http://www.dietmatepro.com) the opportunity to look up product informationand BalanceLog (http://www.healthetech.com/) available online about specific food items. Duringuse the USDA database along with other fast food their five-week study with twenty participantsnutritional information to create a PDA program familiar with PDA technology, they found par-that provides users with a way to save consump- ticipants had mixed reactions to the system intion information for a set of specific nutrients. terms of enjoyment and usefulness. Similar to aCalorieKing (http://www.calorieking.com/) uses recent mobile phone study at Georgia Tech (Patel,its own nutritional database and provides users Kientz, Hayes, Bhat, Abowd, 2006), participantsthe ability to save consumption information. In in the Microsoft study did not always bring theaddition, it has a nutritional tracking application PDA with them despite being enthusiastic PDAspecific to diabetic populations. The applications owners (Bernheim et al., 2005).are similar to the USDA database in that users In addition to PDA monitoring of nutrition,must be able to spell the first few letters of food there have been great strides in mobile phoneitems. Unlike the USDA database, users must nutrition monitoring applications. Those whotype in portion size. Food items are also broken use the commercial application myFoodPhoneup into three subsections - breakfast, lunch, and take pictures of foods they are consuming with
  • 21. Evaluation of Two Mobile Nutrition Tracking Applications their mobile phone and post the pictures to an needs among 207 homeless adults, finding some online food journal to receive feedback from evidence of greater disclosure of risky behaviors a nutritionist (http://www.myfoodphone.com/). with IVRS. However, users must have access to a computer Long-term IVRS usage has had mixed report- and be able to properly upload the information. ing rates and health-related quality of life benefits. Tsai and colleagues developed a mobile phone A 91 day coital study by Schroder et al. (2007) application where participants input food items found a significant decrease in self-reports over via the keypad and immediately receive feedback time, while a two-year study with daily reports of on caloric balance on the phone screen. During alcohol consumption by Helzer et al. (2006) had the month-long feasibility study with 15 college- a 91.7% reporting rate, but compensated partici- educated participants, they found participants pants per call. Daily alcoholism reports among preferred the mobile phone input system to tra- HIV patients found a decrease in drinking over ditional paper and pen journaling methods (Tsai time (Aharonovich et al., 2006). In contrast, an et al., 2006). These applications use mobile phone IVRS intervention with diabetes patients found input via pictures or key presses, but a more natu- no measurable effects on anxiety or health-related ral input interaction would be voice recognition quality of life (Piette et al., 2000). software. In the next subsection, we discuss the Disease management IVRSs that act as diaries use of IVRSs in health interventions. have improved participant satisfaction over paper diaries (Hays et al., 2001). Two recent studies have Integrated Voice Response Systems challenged this result (Weiler, Christ, Woodworth, in Interventions Weiler, Weiler, 2004; Stuart, Laraia, Ornstein, Nietert, 2003). Weiler et al. (2004) conducted a IVRSs in healthcare have been used for reminders, 3-week, 3-way, cross-over trial including 87 adults surveys, screening and assessments, and disease with allergic rhinitis recording daily through management (Lavigne, 1998). A review of IVRS an IVRS or paper diary. A majority (85%) of feasibility studies in populations with chronic ill- the participants preferred the paper instrument, nesses such as depression, cancer, heart failure, whereas only 4% preferred the IVRS. Stuart et and diabetes led Piette to conclude that IVRSs are al. (2003) conducted a year-long study with 642 feasible for chronically ill populations, including patients to enhance antidepressant medication populations that have mental health problems compliance. One of three different treatment or low-income (Piette, Weinberger, McPhee, strategies included a 12-week IVRS component, 2000). According to Mundt et al. (2002), IVRSs yet no significant differences in patient compli- benefit healthcare because they ensure procedural ance were found and 50% of the 232 patients standardization, automatic data scoring, direct assigned to the IVRS component either never electronic storage, and remote accessibility from used the system or stopped before the 12 weeks multiple locations. were completed. Long-term alcoholism and coital studies have IVRSs in healthcare typically limit response supported the feasibility of interventions (Aharo- input to yes/no or numeric responses (Levin novich et al., 2006; Helzer, Badger, Searles, Rose, Levin, 2006). Recent work exploring how Mongeon, 2006; Mundt et al., 2002; Hays, IVRS vocabulary is expanded in a two week Irsula, McMullen, Feldblum, 2001; Schroder pain monitoring study by Levin et al. found that et al., 2007), though the populations are well edu- number of sessions per subject ranged from 1 to cated and technically savvy. Notably Aiemagno 20, accumulating 171 complete sessions and 2,437 et al. (1996) assessed substance abuse treatment dialogue turns. Only 2% of responses recorded RS
  • 22. Evaluation of Two Mobile Nutrition Tracking Applications Table 1. Overview of case study 1 Study Length of Motivating Research Question(s) Phase # Phase Phase 1 1 week 1. Can participants find, identify, and successfully scan barcodes on food items? Break 3 weeks Phase 2 2 weeks 1. Will participants remember how to use this application after a 3 week break? 2. Will participants actively participate without meeting with researchers every other day? were out-of-vocabulary. Though volunteers in participants input food items into an electronic the evaluation were not trained, the results sug- intake monitoring application. The study required gested that training sessions could have significant that participants complete PDA application train- value and that IVR-based data collection is not ing exercises, meet with researchers during di- a replacement for existing data collection, but alysis sessions three times per week, and use the simply another option for healthcare providers Barcode Ed application during two study phases and researchers. for a total of three weeks. Table 1 shows that there Whereas the research discussed in this section was a three week break between the two phases primarily focuses on how well educated, techni- that allowed researchers to evaluate the data and cally savvy users interact with various technology decide on future directions for the application. All interventions for monitoring in their everyday interactions with participants were done during lives, our work deals with how non-technical us- dialysis treatment in an urban, hospital-based, ers with varying literacy skills use two different outpatient dialysis unit. We documented how types of input mechanisms. The IVRS literature we conducted user studies in a dialysis ward in especially shows how compliance is studied with previous work (Siek Connelly, 2006). this technology, but it does not research if partici- pants could use the system and how the system Methodology can be improved. We are iteratively studying input mechanisms because our target population will In this section, we discuss why we selected the depend on the application for their personal health hardware and application used for this case and thus will have to find using the application study. efficient and enjoyable for long-term adoption. This chapter details two case studies that provided Hardware insight into finding the ideal input mechanism for nutrition monitoring. We chose an off-the-shelf Palm OS Tungsten T3 PDA for our study. The Tungsten T3 has an expandable screen, large buttons, voice recorder, C STUDY 1: BARCODE AND SDIO slot, 52 MB of memory, and Bluetooth. We UNSTRUCTURED VOICE chose an off-the-shelf PDA so the results could RING be useful to the consumer health informatics community for future studies. In this section we present our initial formative The Socket In-Hand SDIO card scanner study that examines what, when, and how CKD (Socket Scanner) was chosen as the barcode scan- SSTBRCSTRCTRC
  • 23. Evaluation of Two Mobile Nutrition Tracking Applicationsner because it was small, easy to use, and gave PDA beeps and shows appropriate feedback whenvisual and audio feedback to users. Participants participants have successfully scanned a barcode.must press the predefined scanning button, line Previous studies have shown that CKD patientsup the scanning light perpendicular to the bar- can use the Tungsten T3 and Socket Scannercode, and hold the PDA and object steady. The (Moor, Connelly, Rogers, 2004)Figure 1. Screen shots from Barcode Ed. (a) Home Screen; (b-c) Voice recording and playback screens;(d-e) Barcode Scanning feedback screens
  • 24. Evaluation of Two Mobile Nutrition Tracking ApplicationsApplication Design If the food item was not successfully scanned, a red “X” would appear on the Barcode scanningWe created a simple application, Barcode Ed, unsuccessful page and participants could decidebecause we wanted to isolate participants’ abil- whether to scan again or return to the home screenity to scan and yet have an alternative input and voice record the item instead.mechanism (e.g., voice input) to record all food The application recorded the time the par-items consumed. In initial interviews, half of the ticipant first pressed a Scan or Voice button,CKD patients said they did not eat any foods with the barcode number or voice recording, and thebarcodes. However, once they were prompted, time the recording was saved. We also recordedwe found they primarily ate frozen, canned, and how many times participants played back theirprepared foods. Thus, for participants to use voice recordings. We did not record how manyan easy input mechanism like scanning, they failed barcode scans were attempted because itwould have to learn how to identify barcodes was difficult to differentiate when a participantand use the scanner. We only used scanning and was scanning the same object or gave up andvoice recording in this study because we did attempted to scan a new object during the samenot want to overburden novice computer users period of time. Also, participants sometimes didwith a complex interface because they may have not use the scan button on the Barcode scanningdecreased cognitive function during treatment unsuccessful page - instead they went to the Home(Martin-Lester, 1997). screen and then pressed the scan button again. Barcode Ed consists of five screens as shown The times recorded assisted us in determiningin Figure 1. Since our user group had low literacy when participants recorded what they consumed.skills, we relied on icons 11mm large with some Recording the number of voice recording play-text for navigation. We found these CKD patients backs gave us insight into how participants usedcould view icons 10mm or larger (Moor et al., the application.2004). When participants turned on the PDA,they would view the Home screen. Participants Participantscould choose to voice record by pressing theVoice button or scan a barcode by pressing the Participants were asked to participate in the studyScan button. As soon as participants pressed during their dialysis session. They had to be (1)the Voice button, the application would begin over 21 years of age, (2) able to make their ownvoice recording and show participants how many food or have the ability to go out and purchaseminutes and seconds they recorded on the Voice food, (3) willing to meet with researchers duringrecording screen. When participants were finished each dialysis session during the week, and (4)recording, they could press the Stop button and willing to carry the PDA and scanner with themplay back their recording on the Voice recording and input food items consumed. Ten participantsplay back screen. When participants were satis- volunteered for the study. During the first phase,fied with their recording, they could return to one participant could not participate anymorethe Home screen. When participants pressed the because of a medical emergency and anotherScan button, participants could see a red laser participant dropped out because he did not wantline emitted by the scanner. Participants lined the to record what he was eating (n = 8). We lost twoscanner line perpendicularly across the barcode participants during phase two for similar reasonsthey were attempting to scan. If the food item was (n = 6).successfully scanned, a green check mark would The average age of participants was 52 yearsappear on the Barcode scanning success screen. old (s.d. = 16.28). Half of the participants were
  • 25. Evaluation of Two Mobile Nutrition Tracking Applicationsmale; all of the participants were black. One a food item that could have had a barcode. Par-participant completed an associate degree, four ticipants returned the PDAs at the end of eachparticipants graduated from high school, and one phase of the study, talked to researchers aboutparticipant completed 10th grade. Participants had their experience, and verbally completed a modi-been receiving dialysis treatments on average of fied Questionnaire for User Interface Satisfactionfive years (s.d. = 3.5 years). (QUIS) (Chin, Diehl, Norman, 1988) survey. Only four participants reported using a Participants received ten dollars (U.S.) for everycomputer. Usage frequency ranged from every time they met with researchers for a total of thirtycouple of months to once a week for a half hour. dollars during phase 1. For phase 2, participantsParticipants primarily played games and surfed received five dollars each time they met with thethe Internet. Only two of the participants owned researcher for a total of fifteen dollars.a mobile phone that they used for emergencies Competency skills tests were administered atonly. the end of the second and fourth meeting of the The participants were equally divided about first phase and during the first and last meeting ofhow many food items they consumed had bar- the second phase to test basic Barcode Ed skillscodes - some thought all and some did not think - turning the PDA on; inserting the scanner; scan-any food items had barcodes. Five patients said ning three to five objects with different physicalthey did not have to monitor any nutrients or qualities; voice recording with play back; and dofluid. However, by the end of the first phase, the a combined barcode scanning and voice record-researcher had established a trusting relationship ing sequence. The items participants had to scanwith the participants and found that all of them ranged from a cardboard soup mix box that is easyhad to monitor fluid and nutrients such as sodium, to scan because of the material; a can of chips thatpotassium, phosphorus, and protein. None of the is somewhat difficult to scan because of materialpatients recorded their fluid or nutrient consump- and barcode orientation; and a bag of candy thattion prior to the study. is difficult to scan because it is amorphous and made of shiny material. Researchers measuredDesign and Procedure how many times it took participants to success- fully complete each task. We measured the timeWe met with participants during dialysis sessions it took to complete each competency skill withfour times during each phase of the study for ap- the Barcode Ed application.proximately 30 minutes. During the first session, Participants were instructed to scan or voicewe collected background information and taught record food items when they consumed theparticipants how to turn the PDA on, insert the items. Participants were instructed to scan thescanner, and use the application. Participants barcodes on food items first and voice recordingpracticed scanning various food items and voice items only if they could not scan the barcode orrecording messages. Researchers met with par- if a food item did not have a barcode. When par-ticipants during the study sessions to discuss any ticipants mastered scanning and voice recording,problems participants may have had with the researchers encouraged participants to note viaPDA, retrain participants how to do certain tasks voice recording how much they were consuming(e.g., barcode scanning), and collect recordings and the portion size. Each participant was givenand barcodes from the PDAs via Bluetooth. The a phone number of a researcher to contact if theyresearchers played back the voice recordings to had any questions during the study. Participantsensure the correct information was transcribed were given a visual state diagram of the applica-and informed participants if they voice recorded tion to assist them with any questions regarding
  • 26. Evaluation of Two Mobile Nutrition Tracking Applicationsuse of the application that had images similar to Barcode Scanning and Voicethose shown in Figure 1. Recording FrequencyFindings One of the motivating factors for the first phase of the Barcode Education study was to teachThe key findings of our study were: participants how to identify and scan barcodes. In Figure 2, we see that there was a learning• Participants preferred voice recording once curve associated with identifying and scanning they mastered the application barcodes during the first study phase. Participants• Participants with low literacy skills needed voice recorded more individual food items during extra instruction on how to sufficiently the first few days of the study because they were describe food items for voice recordings either unsure of where the barcode was located on• Participants reported more individual food the food item or were unable to scan the barcode. items with the Barcode Ed application than Gradually during the week, we noticed an increase what they thought they consumed of barcode scans up until the last day of the first• Electronic monitoring provides researchers study phase when participants barcode scanned with ways to identify participant compli- more than they voice recorded. ance A goal of the second study phase was to see if this trend of increased barcode scans would In this section, we present the results in more persist and if participants would continue activelydetail. participating in the study without meeting withFigure 2. Graph of the number of voice recordings and barcode scans participants input over the twobarcode education study phases (dotted line denotes study break). Faces underneath each day denotewhen researchers met with participants
  • 27. Evaluation of Two Mobile Nutrition Tracking Applicationsresearchers every other day. The first two days of recordings. Since the participants were unable tothe second study phase were promising because read the name on the food item, they were not ableparticipants were scanning everything they con- to say what they were eating (e.g., Lucky Charmssumed and only voice recorded items without cereal). Instead, participants said, “I had cereal forbarcodes (e.g., fresh produce). However, after the breakfast.” When we met with participants andsecond day, participants realized everything had played the recordings for transcription, we werebarcodes and were overwhelmed with the amount able to suggest ways to be more descriptive (e.g.,of time it took to scan each individual food item. describe what is on the box) to help us identify theThus, during the third and fourth day of the study, food items. After two to three sessions, the lowparticipants began voice recording food items they literacy participants recorded more descriptivehad previously scanned to save time. input (e.g., I ate the cereal with the leprechaun and The lack of items input at the end of phase one rainbow on the box) and it was easier to identifyshown in Figure 2 can be attributed to not seeing a what they were eating. However, even with de-study researcher to encourage them to participate scriptive input, we were unable to identify threeat the end of the week. Indeed, three participants of the items mentioned in the 195 recordings.acknowledged that they had forgotten to inputfoods on more than one occasion because they had Barcode Ed vs. Self Reported Foodnot been visited by a researcher. Participants were Itemsmore likely to forget to input foods on weekends(days six, seven, thirteen, and fourteen). In pre-study interviews, participants told us they During the second week of the second study had good and bad days that affected how muchphase, participants rarely scanned barcodes and they consumed and discussed how many mealstypically voice recorded what they consumed. The they typically consumed on each of these days.voice recordings listed multiple food items in an The participants usually had a good and bad dayunstructured manner. For example, one partici- fairly recently and could easily describe to uspant recorded, “I ate a small apple, a lunch meat the exact number of items they consumed. Wesandwich, and a boost for lunch. I ate … eggs, asked participants if they had a good or bad dayand bacon for breakfast. Tonight for dinner I am each time we met during the first study phase.planning on eating…” We then compared how many items they elec- When we asked participants why they scanned tronically input to how many items they said theymore on the 13th day of the study, they told us would consume, including the type of day theythat they had remembered they would see a re- were having in the calculation. Participants atesearcher on the following day to finish the study. more than they estimated for an average of threeOf course, the researchers called the participants days (s.d. = 2.875) during the seven day period.to remind them to bring the PDAs to the last day When participants did consume more than theyof the study. estimated, they typically consumed on average 3.5 more items than estimated – nearly doublingVoice Recording Food Items their normally recorded intake of 4.4 items (s.d. = 3.27)1.We thought voice recording food items was aneasy alternative input method when participants Participant Compliancecould not scan. However, participants with lowliteracy skills were initially unable to give suf- For this study, we loosely defined compliance asficient identifying information in their voice inputting at least one food item a day. Similar10
  • 28. Evaluation of Two Mobile Nutrition Tracking Applications Figure 3. Example of voice recordings, barcode scans, and voice recordings that should have been bar- code scans (wrong record) a participant made during the first phase. The participant did back filling as shown by the green circle and increased input during the end of the study. The dotted lines denote the next day. Faces denote when researchers met with participants to traditional monitoring methods, participants and increases participation in hopes the researcher could back fill and modify their compliance re- will not notice. cord. However, unlike traditional methods, with We discussed earlier that once participants electronic nutrition monitoring, researchers can realized everything had a barcode on it, partici- identify this behavior more quickly. For example, pants began to voice record more. We see this a participant back filled entries in Figure 3 (green behavior in Figure 3– the participant starts to circle) by recording what he had consumed for scan items, but then starts to hoard consumption the last two days since he had not actively par- information in one voice recording a day. The ticipated. Another indicator of back filling is the participant told us in a post-study interview that number of times a participant recorded a food reporting everything he ate in one voice recording item that could be scanned during a short time was more time efficient. interval since participants cannot scan items that have been consumed and discarded. Participants were unaware that we were record- CASE STUDY 2: BARCODE AND ing the date and time of inputs and thus assumed IVR if they said, “Today, on February 11, I ate…” the researcher would not know that it was recorded In this section we present our follow-up study that on February 12. When we showed participants examines what, when, and how CKD participants similar graphs as shown here, participants at- input food items into an electronic intake moni- tempted to decrease backfilling or were more toring application and an IVRS with a borrowed truthful in disclosing lack of participation. In mobile phone. Similar to the first case study, addition to backfilling, we see in Figure 3 an participants complete PDA application and mobile example of End-Of-Study compliance where the phone training exercises, meet with researchers participant realizes the end of the study is near during dialysis sessions, and use either the PDA 11CSSTBRCR
  • 29. Evaluation of Two Mobile Nutrition Tracking Applicationsbarcode monitoring application or the mobile We provided participants with a Nokia 6682phone IVRS over a two week period. Participants mobile phone to provide participants the abilitywere recruited and trained at the same dialysis to record food at any time. The phone has a high-unit from the first case study. resolution color screen and large buttons. As with the PDAs, we provided soft leather cases withMethodology belt clips to the participants. We programmed the phone so that pressing any button would dial theIn this section, we discuss the hardware selected number for recording their food items.for the study and design of the applications usedfor capturing participant input. Application DesignHardware The scanning application was similar to the Bar- code Ed application used in the first case study.We designed an application to run on a PDA with The only difference in the application was thatan attached barcode scanner to test participants’ participants did not have the ability to recordability to scan barcodes of food items. For the unstructured voice recordings. If the food itemPDA, we chose an off-the-shelf Pocket PC from did not have a barcode, the participant could notHewlett Packard: the iPAQ hx2495b. We decided record the food item.to use an iPAQ for the second case study because We implemented an IVRS that could be ac-the Windows CE operating system provides a cessed with any phone to test participants’ abilitybetter rapid prototyping environment with Visual to use structured voice input. As Figure 4 shows,Studio .NET CF. The iPAQ hardware includes a we implemented the IVRS by transferring a calllarge, color, touch screen, stylus and large buttons. through a Session Initiation Protocol (SIP) gate-We used the same SDIO In-Hand Scan Cards way to Voxeo, an IVRS platform provider. The(SDSC Series 3E). caller identifier was then submitted to our webFigure 4. Integrated voice response system overview12
  • 30. Evaluation of Two Mobile Nutrition Tracking Applicationsserver where a CGI script selected participant before completion. Two people dropped out aftergrammar files (Nuance GSL Grammar Format), the second day due to lack of interest and onereturning a VoiceXML form to collect items. person was forced to drop out at the end of the The initial grammar included 152 food items first week because she had to undergo emergencyand 2 command operators, ‘done’ and ‘wrong.’ surgery and remained in the hospital during theThe same grammar was available at every prompt. second week of the study. This high dropout rate‘Done’ submitted the results and terminated the is consistent with our previous studies and is acall. ‘Wrong’ incremented a counter, such that if result of working with this type of chronically illsaid twice without an intervening positive rec- population. Here, we report on the six participantsognition, the participant was prompted to voice who completed the study (n=6).record the item for addition to the grammar. With The participants’ average age was 55 years,food items, 45 were single words (e.g., bagel), with a standard deviation of 10.9 years. The12 were compound words (e.g., fish sticks), 27 youngest participant was 36 and the oldest wasused optional phrase operators where a portion 65. Four of the participants were female. Fiveneed not be uttered (e.g., French fries; French is participants identified themselves as Black orconditional) and 50 optional phrase operators African American, and one as White. Oneinitially existed. There were 4 subset uses of the participant had a ninth grade education, twodisjunction operator [] (e.g., ([green baked] beans) had completed high school and three had someis valid for ‘green beans’ or ‘baked beans’). community college. We updated the grammar throughout the study One participant had undergone dialysis for 23based on participant interviews and the items voice years. The remaining participants ranged fromrecorded through IVRS interaction. The Voxeo 2-5 years of dialysis treatment. Two participantsplatform also provided detailed logs of each call, said they did not try to keep track of their nutrientidentifying the caller and the interaction sequence or fluid consumption. Two participants did notbetween the participant and VoiceXML prompts. keep track of nutrients, but attempted to limitThe interaction sequence logs included timeouts, their fluid intake by either not drinking liquidsgrammar recognition errors labeled No Match, over the weekend or “staying conscious” of howprompts, and recognitions. much they drank. Two participants claimed to With a completed call, two lists of items and keep track of both nutrients and fluid. One usedcounter variables were submitted to a MySQL a journal and was conscious of portion sizes; theDatabase—a list for food items misinterpreted other could not describe their method of moni-by the IVRS when identified as wrong by the toring but said they carefully monitored sodiumparticipant and a list of identified food items. and potassium intake. We have found in previousWhen a participant recorded an item for addition studies that participants in this population oftento their grammar, the WAV file was submitted to tell researchers what they think they want to hearour web server, written to disk, and a VoiceXML in regards to their nutrient and fluid consumption,file returned to continue prompting for additional regardless of the reality.food items. Two participants were very familiar with com- puters. One took surveys on the Internet, whileParticipants the other used his laptop daily, including bringing it to the dialysis sessions. One participant hadWe used the same criteria for selecting participants some familiarity with computers. This partici-as we described in case study one. Nine people pant had a computer at home, but did not use itvolunteered for the study, but three dropped out very often. The final three participants said they 13
  • 31. Evaluation of Two Mobile Nutrition Tracking Applicationswere not familiar with computers, although one Participants were paid ten dollars (US) at thehad three years of typing experience and said she end of each week of the study, for a total of twentycould use a keyboard. Three participants owned dollars. Payment did not depend on the numbermobile phones. of times they recorded food itemsDesign and Procedure FindingsFor most participants, the study lasted a total of The key findings of our second case studytwo weeks. However some participants had extra were:time with one of the applications because badweather caused them to miss the dialysis session • Participants spent less time recording inputin which they were supposed to change technol- with the IVRSogy. For these participants, we extended the total • Participants performed better with the scan-length of the study to ensure they had a minimum ner application on non-dialysis days andof one week with each technology. better with the IVRS on dialysis days We primarily used the same methods described • Participants can record more items consumedin the first study. In this section, we describe ad- with the IVRS, but the scanner applicationditions we made to the methods. For the phone is more usable for a larger audienceapplication, we taught participants how to turn the • Input mechanism preference is not alwaysphone on and off, how to dial the number to record linked with the participants’ performancetheir meals and how to record food items with with the technologythe voice recognition application, making sure tospeak one food item at a time very clearly. Barcode Scanning and IVRS During each session, the researcher asked Frequency of Useparticipants about any problems they were hav-ing with the application, if there were any food Despite participants using each technology for atitems they did not record, why they did not least seven days, we found that in reality partici-record a food item, when and how they used the pants used the PDA to scan items on average onlyapplication and their general opinions about its five days (s.d. = 1.4 days) and the mobile phoneusefulness. In addition, we asked participants to to input items with the IVRS on average of 4.5list the foods they had eaten in the last 24 hours days (s.d. = 2.95 days). We found that participantsso that we could compare their recall with what who used the technologies on most of the studythey recorded with the applications. days did so because they enjoyed using the ap- Similar to the first study, competency tests plication systems and wanted to tinker with thewere given to participants during all but the final technology to identify breaking points. In addition,day of the study. For the mobile phone, partici- participants mentioned a desire to help advancepants were asked to record their last meal, which medical research to help themselves and theirrequired them to turn the mobile phone on, dial peers. Participants also mentioned the compensa-the number, and follow the prompts to record the tion rewards, although the compensation was notmeal. We recorded the number of times partici- dependent on frequency of use. Participants whopants attempted to complete each task and noted did not use the technologies regularly in the studyany difficulties they were having. If necessary, sometimes forgot the PDA in their homes andwe retrained and retested the participant. expressed a reluctance to integrate technologies14
  • 32. Evaluation of Two Mobile Nutrition Tracking ApplicationsTable 2. Number and length of time (minutes:seconds) of sessions for each device. Averages are cal-culated per week PDA CP #sessions (avg.) length (avg.) #session (avg.) length (avg.) 1 18 (2.57) 72:23 (4:01) 10 (1.43) 24:10 (2:25) PDA 2 16 (2.29) 29:07 (1:49) 25 (3.57) 28:19 (1:08) 3 4 (0.57) 5:27 (1:22) 4 (0.57) 0:04 (0:01) 4 19 (2.71) 48:48 (2:34) 22 (3.14) 15:26 (0:42) CP 5 6 (0.86) 9:17 (1:33) 13 (1.86) 17:41 (1:28) 6 7 (1.00) 16:14 (2:19) 8 (1.14) 0:52 (0:07)into their daily routines. We found no correlations be to use these systems in their everyday lives. Ifbetween personal computer and mobile phone us- technology is going to take too much time, thenage outside of the study and their willingness to individuals will not be willing to use it. We see inincorporate the technology into their lives. Table 2 that participants spent less time on input We examined usage patterns more closely by sessions when using the IVRS in comparison tolooking at participant input sessions. We defined the PDA scanning application. Scanning tookan input session for the PDA scanner application more time because (1) occasionally the scanneras events that occurred within 10 minutes of each popped out of the SDIO card holder and had toother because we found participants took longer be replaced multiple times and (2) participantsto scan items in realistic situations (e.g., cooking were multitasking during scanning sessions andmeals). We defined an input session for the IVRS input food items as they were doing an activityas any time a participant called into the system. (e.g., cooking a meal) instead of input all at once When we analyzed usage of each technology later on (e.g., right after eating). Participants’ whoon a per input session basis, we found participants multi-tasked with the PDA application showed thatoverall had more input sessions with the IVRS they are willing to integrate the technology intothan with the PDA (13.67 input sessions versus their lives. However, it also shows that raw input11.67 input sessions), but they had similar amount times may not be the best measure of efficientof input sessions when averaged over the week usage of the PDA application.(1.95 input sessions versus 1.67 input sessions).In Table 2, we show the total and average num- Performanceber of sessions each participant had with eachdevice, and the total and average time spent in Besides the actual usage of the technologies ineach session. Participants 1-3 had the PDA the this study, we wanted to study the participantfirst week of the study, while participants 4-6 had performance with each input mechanism. Forthe mobile phone. this study, we defined performance as the ratio of Looking at the time participants spent on unsuccessful to successful attempts at recordinginput gives us insight into how realistic it would food items. We observed that performance was 15
  • 33. Evaluation of Two Mobile Nutrition Tracking Applicationsnot consistent on all days. The ratio of unsuccess- Electronic Input vs. Self Reported Foodful to successful barcode scans on dialysis days Itemswas two times higher than on non-dialysis days(2.43 to 1.11). Conversely, we found participants We asked participants to recall all of the foodperformed better with voice recording on dialysis they ate in the last 24 hours each time we metdays – they had better performance on three out of with them. We then compared their 24 hourthe four non-dialysis days. Thus, on non-dialysis recall to the foods they electronically input intodays participants performed better with the scan- either the scanning program or IVRS with Vennner application and on dialysis days, participants diagrams shown in Figures 5 and 6 . The relativeperformed better using the IVRS. ratios between these three numbers provide us We also studied how participants interacted insight into how participants used the electronicwith the IVRS. Unlike the first study, participants application.would have to say items one at a time and use The Venn diagrams for voice and scanningcommand operators to record food items. We show that participants did not record everythingfound on average that 53% of the time participants they ate. Indeed, participants were somewhatdid not use command operators correctly during limited with their ability to electronically recordIVRS sessions. Participants did not say, “Wrong,” because the scanning application required allwhen items were not recognized by the IVRS for recorded items to have barcodes and the IVRS27% of the total calls. Participants did not say, required the items be in the database to be rec-“Done,” when they finished their calls 26% of ognized. We found that sometimes participantsthe total calls. These errors effect how the IVRS electronically recorded items they did not eat.interprets the input and thus could affect giving One participant in particular recorded non-foodparticipants feedback on their food consumption items. Overall, it appears that participants canin future implementations. capture more items they consume with the IVRS.Figure 5. Venn diagram of food items in 24 hour recall and items scanned16
  • 34. Evaluation of Two Mobile Nutrition Tracking Applications Figure 6. Venn diagram of food items in 24 hour recall and items reported to IVRS However, more participants with varying abilities application because they were comfortable with can capture items they consume with the scanning using phones and with practice, could improve applications as shown by only one participant using the IVRS. not using the scanning application as opposed to two participants not using the IVRS success- fully. We also see that providing alternative input DISCON mechanisms, scanning or IVRS, did not motivate participants 4, 5, and 6 to input a majority of the Even though barcode scanning is a quick method food items they consumed during the study. for inputting individual food items, our results show that it may not be usable over an extended Electronic Input Preference period of time when participants do not receive immediate feedback about their nutritional intake. At the end of the study, we asked participants Participants were overwhelmed with the amount which device they preferred. Overall, two par- of work associated with scanning every food ticipants preferred scanning and four preferred item they consumed. However, participants did voice. Once we identified their preferred device, think that this application would be helpful for we looked at their performance with each input CKD patients who have recently been diagnosed mechanism as described in previous sections and with the chronic illness to assist them in learning pictorially compared preference with performance about the restrictive diet. Participants thought as shown in Figure 7. We found that performance CKD patients in their first year of dialysis treat- influenced preference in only 3 participants. Par- ment would be more likely to spend extra time ticipants 4 and 6 chose the IVRS input despite scanning barcodes if it meant clinicians could not being able to successfully use the system. give them better feedback about their diet and They told us that they still preferred the mobile health. Another possibility for an electronic self phone despite moderate success with the scanning monitoring application would be to have people 17SS
  • 35. Evaluation of Two Mobile Nutrition Tracking ApplicationsFigure 7. Participant preference of electronic input mechanism and overall performance with eachinput mechanismuse it periodically (e.g., quarterly when dietitians used in the previous studies. Indeed, the standardare conducting nutritional assessments with pa- deviation for days participants ate more than theytients) to raise awareness and help them stabilize estimated is large for our small sample. This istheir diet. significant because of the participants’ restrictive We did not anticipate the amount of training diet – overconsumption of the restricted nutrientsparticipants needed to create descriptive voice is dangerous and can result in death.recordings. In retrospect, it made sense that people Backfilling and hoarding are subject to ret-with low literacy skills would not be able to gather rospective biases and may not completely be ac-enough data from the food item to identify it. curate. In addition, researchers have shown thatTranscribing the data was time consuming, but memory recall is undependable – thus participantswas easier as the study continued because the may not be able to accurately describe what theyparticipants typically consumed the same food had consumed during the past days even if theyitems. Researchers need a better understanding are attempting to be accurate (Stone, Shiffman,of their user group so they can accurately identify Schwartz, Broderick, Hufford, 2003). The endfood items that may be culturally or economically of study compliance we discussed is similar toinfluenced. Since our user group has a restrictive Rand’s parking lot compliance where participantsdiet, not being able to identify food items is unac- attempt to be compliant by complying with theceptable since it can have such a drastic change study procedure in the parking lot of the researchin participants’ overall health. facility. Since it is difficult to scan food items once Participants’ underestimation of what they they are consumed (or disposed of), participantsthought they would consume in comparison with increased participation before the end of eachwhat they actually consumed has been document- study phase with voice recording or wrong records.ed by other nutrition researchers (Dwyer et al., It is difficult to determine if patients were increas-2003; Resnicow et al., 2000). However, electronic ing participation before dialysis sessions whereself monitoring gives more detailed information they met with researchers because participants(e.g., date, time, food item) than 24 hour recalls may have been having a bad day (e.g., not feelingand food frequency questionnaires as had been well due to dialysis session recovery).18
  • 36. Evaluation of Two Mobile Nutrition Tracking Applications We occasionally had difficulties with partici- speaker phone so they can hear participants’ ut- pants forgetting the devices, especially the PDAs, terings and the system response. at home when we met with them. Since the par- One weakness that all monitoring methods ticipants were not use to having these devices in have is that we are not sure if participants are their lives, it is not surprising that they forgot them truthfully recording what they consume. Without sometimes. In a recent study with “enthusiastic” subjecting participants to costly blood work or PDA owners, three out of the eight participants requiring participants to wear an invasive device forgot their PDA during a scheduled observation that could detect what a person is eating, we can time (Bernheim et al., 2005). In addition, we had a only assume participants are being truthful. As number of participants who had to stop the study we discuss above, electronic self monitoring can early because of medical concerns or a lack of mo- help researchers identify noncompliant, untruthful tivation to complete the study. Losing participants trends more quickly and discuss non-compliance is not localized to chronically ill populations - in with participants, but this is not a fault proof the enthusiastic PDA owner study, out of 20 total method. participants, only 5 participants finished all the We recognize that the case studies presented tasks in the study (Bernheim et al., 2005)! in this chapter are relatively small. Although re- In case study two, the IVRS we used had dif- searchers have shown that conducting usability ficulties recognizing inputs when the vocabulary studies with 4-6 participants will sufficiently became too large. Participants became increas- provide enough data to determine the effective- ingly frustrated when the system could not recog- ness and usability of a system (Nielsen, 2002; nize even the simplest word, such as “egg” during Virzi, 1992), we are currently conducting a larger the second week of the study. Participants voiced scale study with a fully functioning version of the their frustrations by attributing human traits to system. We recommend researchers who work the hardware (e.g., “It [IVRS] was dragging last with chronically ill populations conduct smaller night.”). IVRSs that use large vocabularies must studies to better understand their target user group be robust and able to handle slight variations better before conducting a larger study. between words. Despite updating the vocabulary each night and thereby increasing it by 30% by the end of FUTURE WORK the study, participants continued to voice record new food items not in the vocabulary showing that The research discussed in this chapter provides although participants typically eat the same foods, many avenues for future research projects. In- there is some variation that must be considered terface designers must find a way to visually when designing a nutrition monitoring system. display portion sizes that low literacy populations Another difficulty we encountered with the understand. Visualizing portion sizes is fairly IVRS was that participants did not use the com- complicated because the type of visualization mand operators correctly. It would be difficult to must be customized based on the type of food. create an IVRS without some command operators For example, water would have a different portion to provide the system information about cor- visualization than bread or meat. In addition, the rectness and when to store the information. We portion size visualization has to be informed by attempted to use a minimal amount of command current methods dietitians use to educate CKD operators, but participants did not use them half patients about portion sizes. of the time. We would encourage researchers to We must find a way to verify consumption conduct more thorough training sessions with a to ensure self monitoring assistive applications 19TR
  • 37. Evaluation of Two Mobile Nutrition Tracking Applications can provide participants with accurate measures intake relative to their dietary prescription of of their dietary intake. Consumption verification CKD patients. When we found that participants could provide participants with reminders to re- preferred unstructured voice recordings rather cord what they consume instead of estimate time than barcode scanning in the first case study, we reminders not based on actual context. Indeed, decided to study structured voice recording in there is already work being done in this area a follow-up study. We found in the second case (Amft Troster, 2006), but we need to continue study that although the system provided partici- development to make less obtrusive or invasive pants with a faster method to input food items, devices for everyday use. participants had difficulty using the system despite In the second case study, one team member training. We are continuing to study if patients spent a significant amount of time updating the will increase their usage of nutrition monitoring IVRS vocabulary each night. We could decrease systems if they receive immediate feedback. the update time and distribute the work load if we better utilize all research members’ time to help listen to and decipher unrecognized phrases ACKNOWLEDGMENT throughout the day. This idea builds on the hu- man solver attack for Completely Automated Katie A. Siek was supported in part by the Na- Public Turing test to tell Computers and Humans tional Physical Science Consortium and Sandia Apart (CAPTCHA). Websites, such as web mail National Laboratories/CA during case study 1. and blogs, make users who want an account or This work was supported by NSF grant EIA- post a comment identify the wavy characters in 0202048, a grant from the Lilly Endowment, and a picture – this challenge-response is known as Grant R21 EB007083 from the National Institute a CAPTCHA. In the CAPTCHA human solver of Biomedical Imaging and Bioengineering to attach, a computer script would automatically fill J.L. Welch. out an online form, identify a CAPTCHA, and then pass the CAPTCHA to a high traffic website and promise the Internet surfer something in return LIST OF MAIN ACRONYMS (e.g., free porn) for identifying the characters in the CAPTCHA. The computer script would then PDA – Personal Digital Assistant take this response, enter it in the form, and create a CKD – Chronic Kidney Disease malicious third party account (Doctorow, 2004). In IVRS – Integrated Voice Recognition System an IVRS update setting, team members would be CAPTCHA - Completely Automated Public Tur- prompted throughout the day to identify phrases. ing test to tell Computers and Humans Apart Depending on ethics board approval, this method could be distributed among a broader Internet community for faster turn-around time. REFERENCES Aharonovich, E., Hatzenbuehler, M., Johnston, CONCLUSION B., O’Leary, A., Morgenstern, J., Wainberg, M., Yao, P., Helzer, J., Hasin, D. (2006). A low-cost, In this chapter, we highlighted results from two sustainable intervention for drinking reduction case studies that compared and contrasted the in the HIV primary care setting. AIDS Care, use of barcode scanning, IVRS, and patient self 18(6), 561-568. reporting as a means to monitor the nutritional 20CSTSTCRR
  • 38. Evaluation of Two Mobile Nutrition Tracking ApplicationsAiemagno, S. A., Cochran, D., Feucht, T. E., Dwyer, J., Picciano, M., Raiten, D. (2003).Stephens, R. C., Butts, J. M., Wolfe, S. A. Estimation of Usual Intakes: What We Eat in(1996). Assessing substance abuse treatment America-NHANES. J. Nutr., 133(2), 609S-623.needs among the homeless: a telephone-based Fitzmaurice, G., Khan, A., Buxton, W., Kurte-interactive voice response system. Am J Public nback, G., Balakrishnan, R. (2003). SentientHealth, 86(11), 1626-1628. Data. Queue, 1(8), 52-62.Amft, O., Troster, G. (2006). Methods for detec- Glasgow, R., Bull, S., Piette, J., Steiner, J.tion and classification of normal swallowing from (2004). Interactive behavior change technology:muscle activation and sound. In Pervasive Health A partial solution to the competing demands ofConference and Workshops, 2006, (pp. 1-10). primary care. American Journal of PreventiveBrush, B. A. J., Turner, C. T., Smith, M. A., Medicine, 27(2, Supplement 1), 80-87.Gupta, N. (2005). Scanning objects in the wild: Hays, M., Irsula, B., McMullen, S., Feldblum,Assessing an object triggered information sys- P. (2001). A comparison of three daily coital diarytem. In UbiComp 2005: Ubiquitous Computing, designs and a phone-in regimen. Contraception,3660/2005, 305-322. Springer Berlin / Heidel- 63(3), 159-166.berg. Health Resources and Services Administra-Brandt, E., Björgvinsson, E., Hillgren, P.-A., tion. Health literacy. Available at www. hrsa.Bergqvist, V., Emilson, M. (2002). PDA’s, gov/healthliteracy/barcodes and video-films for continuous learningat an intensive care unit. Paper presented at the Helzer, J., Badger, G., Searles, J., Rose, G., NordiCHI ‘02: Proceedings of the second Nordic Mongeon, J. (2006). Stress and Alcohol Con-conference on Human-computer interaction. sumption in Heavily Drinking Men: 2 Years of Daily Data Using Interactive Voice Response.Chin, J., Diehl, V., Norman, K. (1988). De- Alcoholism Clinical and Experimental Research,velopment of an instrument measuring user 30(5), 802-811.satisfaction of the human-computer interface.Paper presented at the CHI ‘88: Proceedings of Intille, S., Kukla, C., Farzanfar, R., Bakr, W.the SIGCHI conference on Human factors in (2003). Just-in-time technology to encouragecomputing systems. incremental.Connelly, K. H., Faber, A. M., Rogers, Y., Siek, K. Kirsh, I. S., Jungeblut, A., Jenkins, L., et al. (1993).A., Toscos, T. (2006). Mobile applications that Adult literacy in America: A first look at the resultsempower people to monitor their personal health. of the National Adult Literacy Survey. Washing-e i Elektrotechnik und Informationstechnik, ton DC: National Center for Education Statistics,123(4), 124-128. United States Department of Education.Doctorow, C. (2004). Solving and Creating CAPT- Lavigne, M. (1998). Interactive Voice ResponseCHAs with free porn. Boing Boing. http://www. in Disease Management: Providing Patient Out-boingboing.net/2004/01/27/solving-and-creating. reach and Improving Outcomes (pp. 1-16).html Levin, E., Levin, A. (2006). Evaluation ofDowell, S., Welch, J. (2006). Piloting the Use spoken dialogue technology for real-time healthof Electronic Self Monitoring for Food and Fluid data collection. J Med Internet Res, 8(4).Intake. Nephrology Nursing Journal, 33(3), 271-277. 21
  • 39. Evaluation of Two Mobile Nutrition Tracking ApplicationsMartin-Lester, M. (1997). Cognitive function in to Sexual Behavior Self-reports: A Comparisondialysis patients. Case study of the anemic patient. of Three Methods. AIDS and Behavior, 11(2),ANNA J, 24(3). 313-323.Moor, K. A., Connelly, K. H., Rogers, Y. Schillinger, D., Grumbach, K., Piette, J. et al.(2004). A Comparative Study of Elderly, Younger, (2002). Association of health literacy with diabetesand Chronically Ill Novice PDA Users (No. outcomes, JAMA, 288, 475-482.TR595). Sevick, M. A., Piraino, B., Sereika, S., Starrett,Mundt, J., Bohn, M., King, M., Hartley, M. T., Bender, C., Bernardini, J., et al. (2005). A(2002). Automating Standard Alcohol Use As- preliminary study of PDA-based dietary self-sessment Instruments Via Interactive Voice monitoring in hemodialysis patients. J Ren Nutr,Response Technology. Alcoholism: Clinical and 15(3), 304-311.Experimental Research, 26(2), 207-211. Siek, K. A., Connelly, K. H. (2006). LessonsNielsen, J. (2002). Why you only need to test Learned Conducting User Studies in a Dialysiswith 5 users. From http://www.useit.com/alert- Ward. Paper presented at the Extended Abstractsbox/20000319.html of CHI 2006: Workshops - Reality Testing.Patel, S., Kientz, J., Hayes, G., Bhat, S., Abowd, Stone, A. A., Shiffman, S., Schwartz, J. E.,G. (2006). Farther Than You May Think: An Broderick, J. E., Hufford, M. R. (2003). PatientEmpirical Investigation of the Proximity of Us- compliance with paper and electronic diaries.ers to Their Mobile Phones. In UbiComp 2006: Control Clin Trials, 24(2), 182-199.Ubiquitous Computing (pp. 123-140). Stuart, G. W., Laraia, M. T., Ornstein, S. M., Piette, J. D. (2000). Interactive voice response sys- Nietert, P. J. (2003). An interactive voice responsetems in the diagnosis and management of chronic system to enhance antidepressant medication com-disease. Am J Manag Care, 6(7), 817-827. pliance. Top Health Inf Manage, 24(1), 15-20.Piette, J. D., Weinberger, M., McPhee, S. J. Tsai, C., Lee, G., Raab, F., Norman, G., Sohn, T.,(2000). The effect of automated calls with tele- Griswold, W., et al. (2006). Usability and Feasi-phone nurse follow-up on patient-centered out- bility of PmEB: A Mobile Phone Application forcomes of diabetes care: a randomized, controlled Monitoring Real Time Caloric Balance. Papertrial. Med Care, 38(2), 218-230. presented at the Pervasive Health Conference and Workshops, 2006.Preventing Chronic Diseases: A Vital Investment(2005). World Health Organization. USDA Palm OS Search: Health Tech (2008). ht t p://w w w.a r s.u sd a .gov/Se r v ice s /do cs.Resnicow, K., Odom, E., Wang, T., Dudley, W., htm?docid=5720Mitchell, D., Vaughan, R., et al. (2000). Valida-tion of Three Food Frequency Questionnaires and Virzi, R. A. (1992). Refining the test phase of us-24-Hour Recalls with Serum Carotenoid Levels ability evaluation: how many subjects is enough?in a Sample of African-American Adults. Am. J. Human Factors, 34(4), 457-468.Epidemiol., 152(11), 1072-1080. Weiler, K., Christ, A., Woodworth, G., Weiler,Schroder, K., Johnson, C., Wiebe, J. (2007). R., Weiler, J. M. (2004). Quality of patientInteractive Voice Response Technology Applied reported outcome data captured using paper and22
  • 40. Evaluation of Two Mobile Nutrition Tracking Applications interactive voice response diaries in an allergic Endno rhinitis study; is electronic data capture really bet- ter? Program and Abstracts of papers presented 1 The standard deviation is large because it during Scientific Sessions - AAAAI 60th Annual depends if participants were having a good Meeting, 113(2, Supplement 1), S78. or bad day in terms of consumption and physical health. 23e
  • 41. 24 Chapter II Accessing an Existing Virtual Electronic Patient Record with a Secure Wireless Architecture Ana Ferreira Ricardo Correia Center for Informatics, Faculty of Medicine in Center for Informatics, Faculty of Medicine in Porto, Portugal Porto, Portugal Luís Barreto Susana Sargento Instituto Politécnico de Viana do Castelo, Universidade de Aveiro, Portugal Portugal Luís Antune Pedro Brandão LIACC at Faculty of Science in Porto, R. LIACC at Faculty of Science in Porto, R. Campo Alegre, Portugal Campo Alegre, Portugal ABSTRACT Virtual electronic patient records (VEPR) enable the integration and sharing of healthcare information within large and heterogeneous organizations by aggregating known data elements about patients from different information systems in real-time. However, healthcare professionals need to access a terminal every time they treat a patient. This may not be trivial as computers are not available around every corner of big healthcare institutions. The use of wireless technology can improve and fasten healthcare treatment because it can bring information and decision to the point of care allowing also for health- care professionals’ mobility. However, as healthcare information is of a very sensitive nature, it has to comply with important security requirements. The wireless technology makes it more difficult for these requirements to be achieved as it is harder to control disruptions and attempts to access information can be more common and less simple to detect. The main objective of this chapter is to model, develop Copyright © 2009, IGI Global, distributing in print or electronic forms without written permission of IGI Global is prohibited.a
  • 42. Accessing an Existing Virtual Electronic Patient Record and evaluate (e.g. in terms of efficiency, complexity, impact and against network attacks) a proposal for a secure wireless architecture in order to access a VEPR. This VEPR is being used within a university hospital by more than 1,000 doctors, on a daily basis. Its users would greatly benefit if this service would be extended to a wider part of the hospital and not only to their workstation, achieving this way faster and greater mobility in the treatment of their patients. INTRUCTION ficult integration of VEPRs into medical processes, within large environments such as hospitals (Ben- Virtual electronic patient records (VEPR) enable son, 2002). The lack of security increases users’ the integration and sharing of healthcare informa- reluctance for VEPRs’ acceptance. Both patient tion within heterogeneous organizations (Blobel, and healthcare organization trust can be seriously 2004). Hospitals are an example of such healthcare damaged if no proper security is provided (Denley institutions with great turnover in terms of health- S. W. Smith, 1999) care professionals. However, there are usually Furthermore, wireless technology adds a some constraints in terms of physical location as higher level of security issues. Disruptions and well as technology in order to access it. Healthcare attempts to access information can be more com- professionals need to access a terminal in order to mon and easier to try, and less simple to detect get information about the patients they are treating. and control; so security needs to be studied and This may not be easy to attain within a big and analysed thoroughly before wireless networks are complex healthcare institution where computers implemented in a larger scale within a hospital are not available around every corner. (Dixie B. Baker, 2003). Security requirements The use of wireless technology tries to take need to be considered and applied from the be- this integration further. It allows access to patient ginning to the end of a system’s development and data and processing of clinical records closer implementation (Ana Ferreira, Ricardo Correia, to the point of care. The ubiquitous access to A. Costa-Pereira, 2004; Ana Ferreira et al., information can minimize physical as well as 2005; Ana Ferreira et al., 2004). time constraints for healthcare, enhancing us- This chapter proposes a wireless architecture ers’ mobility within the institution. There have in order to model access to an existing VEPR been some experiences with the use of wireless within a university hospital that can provide an technology in the healthcare environment. These extra security layer to the wired system. The next have shown that healthcare professionals were section describes the VEPR architecture along usually satisfied with the use of portable devices with the security requirements for the wireless to access patient information. They save time and version. The third section presents the wireless are bound to improve patient care (McAlearney, architecture that uses the latest wireless standards Schweikhart, Medow, 2004). The most com- and security protocols and takes into account the mon devices include mobile wireless patients’ security services that were implemented within health monitoring systems. These equipments the wired version of the system. Section four add more security concerns (Ramon Marti describes an evaluation of the proposed solutions Jaime Delgado, 2003) but are out of the purpose against network attacks and its efficiency in terms of this research. of complexity and impact on the network. The last Among other problems, the lack of security section discusses the results and shows some of processes is one of the main reasons for the dif- the challenges where to focus future research. 25CT
  • 43. Accessing an Existing Virtual Electronic Patient Record THE VIRTUAL ELECTRONIC MAID collects clinical reports from various PATIENT RECORD hospital Departmental Information Systems (DIS), and stores them on the central repository With the objective to face one of the major prob- (CRep) consisting of a database holding references lems within large and complex health organiza- to these clinical reports and a file system where tions - data retrieval and integration - a VEPR was reports are stored. After searching the database, built within a University Hospital with over 1350 VEPR users can access the integrated data of a beds, by the Biostatistics and Medical Informatics particular patient through the web-based interface Department, at the Faculty of Medicine in Porto. (VIZ). When selecting a specific report, its content This system provides a cost-effective solution for is downloaded from the central repository file most clinical information needs (Ricardo Cruz- system to the browser. MAID (the agents’ server) Correia et al., 2005). communicates with the DIS using XML. MAID Currently, more than 1000 doctors use the connects to the database server through JDBC1 system on a daily basis. Other healthcare pro- and operates the files using NFS protocol2. The fessionals (namely nurses) are expected to start application in the Web Server (VIZ) communi- using it soon. cates with the CRep database server using OCI/ PHP (Oracle Call Interface with PHP: Hypertext Architecture Preprocessor Language) functions and operates the files using NFS protocol. The Web browser This VEPR allows the collection, integration and client accesses the Web Server using HTTPS availability of clinical reports providing an up- protocol. The Web services connect to the CRep to-date overview of a patient medical history at database server, SONHO server and IEG server all points of care. The system uses a traditional using JDBC, and use SOAP messages to deliver three layered approach composed by presentation, information to the GUI Components. business and data layers. The VEPR has been working for 4 years, The presentation layer is composed by a web regularly scanning eleven DIS and collecting a application (VIZ) and a package of graphical user mean of 3000 new reports each day (currently interface components to be used by third party holds about 3 million documents). A viewing applications. The web-interface was designed module for the VEPR was made available in to include graphical components and layouts to October 2004. Integrated DISs have evolved to summarise past patient data (patient chronological send different documents to the VEPR without bars), and folders that reproduce the traditional the need of any type of adaptation. types of patient record organisations (source, chronological and problems views). Integration and Communication The application layer is composed by an inte- gration engine (Multi-Agent system for Integra- The integration of hospital data in VEPR is ac- tion of Data – MAID), and a set of web-services complished with the use of different agents as- that allow access to the data layer. The data signed to different tasks. Some collect reports’ layer includes all repositories, namely the CRep references and others the actual reports from the that comprises the VEPR database and clinical DISs. When a user requests a report whose file documents file system, the central patient system is not in CRep, there is an explicit report request (SONHO) and the hospital statistics system (IEG) made directly to MAID, by the VIZ module. This (Figure 1). request activates the express agent from the agent 26TRTCTRCTR
  • 44. Accessing an Existing Virtual Electronic Patient Record Figure 1. VEPR generic architecture platform in order to get the report requested by the with staff identification numbers, which were user, from the right department (Figure 2). reused after staff members left the hospital. Several integration models had to be used to achieve the necessary integration level. The Statistics selection of appropriate integration model was often conditioned by the maturity of the IS being VIZ was made available for testing in October integrated, and by the resources available at that 2004 but has only been known and routinely time. It should be noted that the development of used since December 2004. The number of ses- communication interfaces was simpler for the sions and report views has been growing steadily eight applications that had Web interfaces because since. The number of sessions increased 147% in they were already using standard communication 2006, and 70% in 2007. The number of distinct protocols such as HTTP. Web-services and shared users using the VEPR has also grown in the same graphical components were very useful in deliver- period, representing an annual growth of 29% ing an integrated view to other ISs. users in the 4th quarter of 2006 and 41% in the The process of integration of heterogeneous 4th quarter of 2007. Currently, 4th quarter of 2007, clinical information systems has shown the exis- 1.24 reports are viewed per session, 0.4 reports tence of organisational or technical problems and, are viewed per patient encounter and 82.4 reports indirectly, contributed to their solution. While are viewed per user. Also the use of the VEPR is some reports cannot be associated with identi- more widespread by the hospital computers (975 fied hospital patients (e.g. outpatients who are not computers in 4th quarter of 2007). administratively considered as Hospital patients), The number of report views per user per pa- some patients had multiple, rather than unique, tient encounter has stabilized around 3.8 views identification numbers, making their correct iden- per user per 10 000 encounters since the first tification difficult. A similar problem was found quarter of 2006. 27S
  • 45. Accessing an Existing Virtual Electronic Patient Record Figure 2. UML sequence diagram of the VEPR The number of direct access to the VEPR to return to their workstation. This allows over- from the computer desktop hyperlink has been coming most physical and logical obstacles that diminishing since the first quarter of 2006, whilst the hospital offers, therefore increasing VEPR the number of accesses that originate in the Hos- availability. pital Patient Record (SAM) as been growing. The number of report views from other referrals Security and Monitoring is small when compared with direct access and hospital patient record. The number of views per VEPR present many security challenges namely session for direct access is 1.81, for the DISs is the need to provide protection to patient’s sensi- 1.20, for the Hospital Patient Record is 1.18 and tive information. The implementation of security for the Emergency Department IS is 1.05. mechanisms was thought from the beginning of the The introduction of wireless technology will project’s development and implementation, allow- allow the access to this VEPR system to a wider ing for its better integration and acceptability (Ana number of people and locations. The healthcare Ferreira et al., 2004). This subject was tackled ac- professionals will be able to access patient in- cording to the three main security characteristics: formation whenever they need without having integrity, confidentiality and availability. One of 28S
  • 46. Accessing an Existing Virtual Electronic Patient Record the main security issues relies in the information services that are not working and even improper collected in the stored patient reports. Digital behaviour. As an example, to detect users that signatures are security mechanisms that provide share their logins and passwords the logs of ses- the integrity of a report by enabling the detec- sions from October 2004 until December of 2007 tion of unauthorized modifications. If the digital were analysed. The suspicious behaviour that was signature does not match the report contents then searched for was users working for more than 24 this report is marked as not valid (Ana Ferreira hours (in some cases doctors work for 24 hours et al., 2004). Confidentiality relates mainly to consecutively). All user sessions that started less the access to sensitive information by authorized than 10 hours from the last session were considered individuals. It is obtained by controlling access to to be referring to the same working day. information and by protecting it while in transit The number of suspicious cases found was 508; along network communications. Access control the calculated working days ranged from 24 to 63 policies were defined by the hospital administra- hours (average = 29 hours). These working days tion after a proposal from a specifically assigned referred to 139 of 1434 logins (rVPR=9.7%). The 10 committee defining roles and levels of access to logins that more frequently have suspicious behav- VIZ. These policies were implemented using role- iour referred to the following medical specialties: based access control (RBAC) (Ferraiolo, Sandhu, Anaesthesiology (4 logins), Emergency (2 logins), Gavrila, Kuhn, Chandramouli, 2001), an ac- Infectious Diseases (2 login), Cardiothoracic Sur- cess control model used for large organizations gery (1 login), Gastroenterology (1 login). (Ana Ferreira et al., 2005). In order to provide Although technical solutions exist to provide for an efficient way for user identification and secure access control, they demand a clear defi- authentication, development of access control nition of permissions for each group of actors. tools was based on ENV 12251 European pre- Healthcare organisations must comply with standard (CEN, 1999). As the network wiring current legislation, ethical rules and internal pro- and equipment is spread all over the hospital, it cesses which are very difficult to be objectively is necessary to protect the network infrastructure defined into access control rules. The number of from eavesdropping. This was accomplished us- shared logins found may probably just represent ing TLS authentication protocol (B. Aboba D. the tip of the iceberg. However, it is high enough Simon, 1999) which provides encryption of all to raise concern. information whilst in transit. Availability focuses on means to provide for the continuous access to Scurity Requirements information by authorized users. Equipment and power redundancy, backups and system monitor- All the security services implemented for the ing were all put in place to guarantee availability wired VEPR mentioned in the previous section are of the system at all times. The number of reports obviously valid for the wireless architecture. daily retrieved from each DIS is compared to The wireless technology stresses however the what is expected and the number of sessions of need for extra layers of security. In order for a different users is monitored. Any deviation from healthcare professional to access the VEPR with expected values triggers an alert message to the a wireless device, there are 3 main security is- system administrator. sues to address: Monitoring sensors have also been developed within the VEPR in order to detect problems in 1. Authentication and authorization from the any of the three security characteristics, as well wireless to the wired network; as for instance systems’ malfunctions, errors, 29R
  • 47. Accessing an Existing Virtual Electronic Patient Record 2. Secure communications of information in cess. They need quick and reliable access to carry transit; out their job, or the system will be circumvented 3. Integrity trust in the information that is (Lehoux, Sicotte, Denis, 1999). requested and visualized by the users. Another important concept is the requirement to access the VEPR infrastructure from outside For (1) there is the need to create an access the local network (eg. from the internet) (Yu control infrastructure that will prevent problems Jothiram, 2007). Also the security in pervasive of confidentiality such as masquerading and sensor networks for healthcare monitoring (Ng, password sniffing. Also, policy rules need to be Sim, Tan, 2006) is another relevant trend. set so that access from the wireless to the wired These subjects are however outside the scope of network is properly controlled. Still, the process this discussion. of access control must be transparent to the users This section describes some possible solutions and simple to use and manage. to support security in WLANs. These include a Point (2) requires that information in tran- general framework to communicate authentication sit must travel encrypted at all times to avoid details (EAP) to allow or deny network access eavesdropping. It should always be available in and exchange cryptographic material (802.1X). a certified and trusted manner. Building on these, WPA and 802.11i (WPA2) For (3) there is the need for the means to guar- are able to control the access to the network and antee that the information in transit within the provide encryption of the communications. IPsec wireless network is protected from unauthorized addresses authentication and encryption at the or accidental modifications. Healthcare profes- network (IP) layer whereas the previous tech- sionals must trust the information they use to treat nologies lie on the data link (medium) layer. The patients. The most accurate and correct it is the next sub-sections describe all these technologies better and adequate the treatment will be. in more detail. Extensible Authentication Protocol PROPOSED WIRELESS ARCHITECTURE The Extensible Authentication Protocol (EAP) (B. Aboba, L.Blunk, J. Vollbrecht, J. Carlson, As previously mentioned, users of healthcare H. Levkowetz, 2004) is a general authentica- environments would greatly benefit with the tion protocol defined by the IETF. It was origi- availability of information anywhere through a nally developed to be used with a point-to-point wireless local area network (WLAN). Usually, the protocol. EAP provides an interface to several healthcare institution where the WLAN is going to authentication mechanisms, as Kerberos, public be deployed has already a LAN in use. Setting a key ciphering or one time passwords. WLAN on top of this one is seldom trivial. Build- EAP cannot be used independently as an au- ing dimensions, user locations, connectivity and thentication protocol. It is just a set of rules of how the security requirements previously mentioned an authentication server and a client can exchange account for the stringent issues. The need for a messages and provides a pluggable architecture good location map and distribution is essential for for different security protocols. EAP uses the data tackling the first two issues. The last two will be link layer for message exchange, and so does not the focus of this section. Healthcare professionals require IP addresses for communication. must trust not only the technology they use (e.g. A network with EAP capabilities has three robust, usable) but also the information they ac- independent identities: the client (also known as 30RSSSRCTCT
  • 48. Accessing an Existing Virtual Electronic Patient Record supplicant), the authenticator and the authentica- each client and session. This means that keys have tion server. The client has to deliver the authenti- to be regularly changed, thus reducing repetition cation credentials (a certificate or a username and problems. a password). The authenticator is the equipment The 802.1X three main processes are the that implements security at the port level and does mutual authentication between the client and also network access control. According to the EAP the server, the cryptographic keys dynamically authentication protocol used, the authenticator generated after authentication and the centralized re-transmits the necessary messages, between the policy control. client and the authentication server, acting as an 802.1X is not a protocol; it is an authentica- intermediary and enforcer in the authentication tion and key management process. In a wireless request. The authentication server specifies the network it defines authentication and the dynamic authentication protocol to be used and validates generation of cryptographic keys. The ciphering the credentials delivered by the client. is accomplished using any of the wireless security EAP enables the support of multiple authen- protocols. tication protocols without the need to configure the authenticator with each specific authentication WPA – Security and Architecture mechanisms. EAP allows also the authentication server to control which authentication protocols WPA (“(Wi-Fi Protected Access)”) was devel- should be supported. These features increase oped with the aim of decreasing the problems flexibility to the process and allow greater in- associated to Wired Equivalent Protocol (WEP)3 teraction. (Walker, 2003). WPA is based on the principles of the IEEE802.11i standard (IEEE 802.11i, 2004) 802.1X with some simplifications to be compatible with the equipments at the time WPA was released. IEEE 802.1X (IEEE 802.1X, 2004) is a network WPA uses a robust cipher algorithm and intro- security specification initially developed for duces user authentication, one of the WEP missing wired networks, with its concepts and utilization characteristics. extended afterwards to wireless networks. 802.1X WPA is intended to be implemented in a defines a network access control based in ports. It home/office environment and is available in all was developed to deny or accept requests based Access Points (APs) and Network Interface Cards on user authentication information (credentials). (NICs) currently available4. 802.1X uses EAP for authentication. The ac- To improve data codification, WPA uses the cess control is performed at the Medium Access Temporal Key Integrity Protocol (TKIP) (IEEE Control (MAC) level and is independent from 802.11i, 2004) which, when compared to WEP, the physical layer. A port in 802.1X is any type improves data level ciphering by using temporal of controlled access element (i.e. switch, router, and per packet keys. WPA also has a key mixing AP). In this context, the association between one function for each packet, a Message Integrity client and one AP is called a virtual port and the Check (MIC), extended initialization vectors access to the network is seen as another virtual (IV) with sequential rules and a key renewal port. The client associates first if the port is avail- mechanism. able and uses this connection to authenticate. If WPA makes use of 802.1X for user authenti- the authentication is successful the AP gives ac- cation, making it possible to use one of the EAP cess to the network (thus granting access to the methods. For security matters in these environ- network virtual port). 802.1X provides keys for ments, the EAP- Transport Layer Security (TLS) 31S
  • 49. Accessing an Existing Virtual Electronic Patient Record (B. Aboba D. Simon, 1999) method is used. For connectivity between the different net- This method uses digital certificates for each user works a layer 2 or 3 switch is used. This type of authentication. A central authentication server switch adds a new layer of filter/protection to the is used to manage mutual authentication, which system with the use of Virtual LANs (VLANs) apart from authenticating the user, it eliminates and, if needed, allows to route data between the the danger of rogue APs. The authentication different networks. This solution provides an ac- server usually employed is the Remote Access cess level linked to the security standard used by Dial-In User Service (RADIUS) (C. Rigney, A. the clients. The proposed architecture uses two Rubens , W. Simpson , S. Willens, 1997). The security VLANs. These VLANs are configured RADIUS server authenticates the WLAN user and in such a way that only WPA and 802.11i clients determines the session key to be used. RADIUS are able to access all network services. The is only used to communicate between the AP and VLANs distinguish, transparently to the user, the the authentication server; in the WLAN, EAP security protocol used by the client and trigger is used between the user and the AP (Figure 3). all the necessary and specific procedures needed Notice that other Authentication, Authorization for authentication and authorization. and Accounting (AAA) protocols (Kim Afifi, The implementation of a WPA system requires 2003) could be used such as Diameter (Ventura, the development of an 802.1X infrastructure. 2002), COPS (Durham, Boyle, Cohen, Rajan, All the necessary elements for building a WPA Sastry, 2000) or TACACS (Finseth, 1993) server. network are shown in Figure 4. However, RADIUS is used for WPA. It is worth noting that there is a possibility of A Lightweight Directory Access Protocol using a password based user authentication (for (LDAP) (J. Hodges R. Morgan, 2002) server either WPA or 802.11i). However, this approach is can also be used for a centralized user authentica- not recommended in high security infrastructures tion. All RADIUS implementations can interact (Moskowitz, 2003). with an LDAP server, making it possible to use a central point of administration of all users, thus 802.11i Security and Architecture creating a strong security policy. Other centralized user authentication implementations that can use In June 2004, the Institute of Electrical and LDAP are Active Directory (Microsoft, 2004) and Electronics Engineers (IEEE) ratified the 802.11i Novell eDirectory (Novell, 2004). standard, also called Robust Security Network (RSN)5 (IEEE 802.11i, 2004). This security stan- Figure 3. Authentication architecture 32S
  • 50. Accessing an Existing Virtual Electronic Patient Record Figure 4. WPA and RSN/IEEE802.11i architecture dard includes the following functionalities: uses compliant to 802.11i and WPA2 (including some the Advanced Encryption Standard (AES) (NIST, PDAs)6. 2001) block cipher to encrypt the data packets, 802.11i actually defines three protocols for data 802.1X for user authentication and TKIP for the protection: the Counter Mode with Cipher Block management of the cipher keys. The standard Chaining Message Authentication Code Protocol also recommends a set of new improvements to (CCMP) (Whiting, Housley, Ferguson, 2003), WEP in 802.11b NICs. Some NICs, due to design the Wireless Robust Authenticated Protocol (IEEE limitations, cannot support AES but are able to 802.11i, 2004) and TKIP. CCMP will be the ‘de support TKIP with a small update. facto’ IEEE802.11i cipher protocol. It is based in 802.11i requires that all clients announce their AES counter mode. This protocol derives from cipher capabilities in their AP association requests. lessons learned with 802.10 (IEEE 802.10, 1998) The AP and the wireless client then establish the and IPsec protocols. It uses strong cipher primi- appropriate channel for data ciphering. This key tives, which makes it reliable against all (currently) agreement is based on their mutual cryptographic known attacks. capabilities and configured in one of the security As with WPA, for implementing an 802.11i policies (eg.: “allowing only associations with solution it is necessary to deploy an 802.1X in- AES clients”). Moreover, 802.1X authentication frastructure. assures key renewal during a session. Figure 4 shows the required elements to sup- AES is currently widely recommended for con- port an 802.11i architecture. fidentially. However, AES entails more demanding cryptographic functions. This means that older IPSec – Security and Architecture devices do not have processing capacity to handle AES and keep a normal network performance. To The two previous solutions are specially designed circumvent the problem 802.11i enables the use for wireless networks. However, it is also possible of TKIP as the cipher protocol. This method is to protect these networks with a network layer more feasible for less capable devices. Nonethe- protocol originally developed for wired networks, less, there is already a wide selection of products like IP Security (IPSec) (B. Aboba et al., 2004). 33S
  • 51. Accessing an Existing Virtual Electronic Patient RecordThis protocol, though intended to protect Internet The IPsec standard includes two security pro-communications and wired networks, has some tocols: the Authentication Header (AH) (Kent characteristics that make it suitable to protect Atkinson, 1998a) that provides data integrity andwireless communications. While the previously the Encapsulating Security Payload (ESP) (Kentmentioned solutions protect the information at the Atkinson, 1998b) that adds confidentiality. Alldata link layer, IPsec protects the information at IPsec parameters are negotiated using the Internetthe network layer. This functionality makes it a Key Exchange (IKE) (Harkins Carrel, 1998)versatile protocol, which can be used to protect protocol. IKE uses digital certificates for endany kind of IP network, and is independent of the points authentication. ESP makes use of cipherapplication and type of data flow. It comprises a techniques for data confidentiality, and digitalset of protocols for the development of Virtual signatures for source authentication, while AHPrivate Networks (VPNs). only uses digital signatures for source authentica- IPsec VPNs are a very common method for tion (AH does not cipher data). Thus ESP shouldprotecting data that traverses public networks be used when confidentially is an issue.(or non-protected networks). IPsec adds secu- Figure 5 shows an IPsec VPN adapted to arity through a set of tunnelling and ciphering wireless network and the elements required formechanisms: it implements network layer au- an IPsec protected wireless network.thentication and ciphering; keeping end-to-end The network has wireless terminals with VPNsecurity within the network architecture. Its main client software. This software should be able toadvantage is that it can protect any kind of data start protected tunnels between the terminalspacket routed through the network independently and the gateway. The firewall assures the rightof the source application7. Its main disadvantage establishment of a tunnel and also guarantees thatis its complexity. only specified devices can establish that tunnel. IPsec has two modes of operation: tunnel and Recent Windows OS have a native VPN client.transport. In tunnel mode IPsec protects a com- The wireless terminal connects to the AP thatpletely normal IP packet, thus its payload is an offers, between the wireless and the wired net-IP packet. This mode is used when the IP packet works, initial filters to the IP protocol. Betweenneeds to be sent unchanged to the destination. the AP and the wired networks there is a layer 2Transport mode IPsec is integrated with IP and switch responsible for the connectivity. Recentthus transports an UDP/TCP packet from the models of this kind of switch allow Virtual LANtransport layer. Access Control Lists (VACL), which adds a newFigure 5. Wireless network IPsec VPN34
  • 52. Accessing an Existing Virtual Electronic Patient Record filter/protection layer to the system (as discussed and two different security protocols. The APs previously). As in the previous architectures, are enabled with both 802.11i and WPA. This LDAP and RADIUS servers are used for central- configuration creates a secure logical network, ized user authentication. allowing healthcare professionals to have a secure and controlled access to the VEPR. A RADIUS Wireless Architecture Proposal server acts as the policy enforcement point (PEP), configured with different access control policies This section discusses a secure wireless ar- for each SSID9. These policies define the data chitecture for accessing the VEPR taking into protection protocol, the key management protocol account the specific characteristics of a health and the key length used with a specific SSID. care institution (importance of security) and the The RADIUS server is coupled with the actual characteristics of the available solutions. This VEPR solution in terms of user management. The architecture uses the WPA-TLS protocol and also previous sections discussed the use of LDAP for considers the use of the new 802.11i standard. All the VEPR. For this case, the RADIUS authenti- existing equipments can, with a small firmware cation should use the LDAP of the VEPR. This upgrade, support WPA-TLS and therefore, be is very important as it will enable the use of the reused reducing implementation costs. WPA-TLS current VEPR access control in the new wireless should only be considered a transition solution architecture. until all devices support 802.11i. As expected, all terminal/client equipments As such, the aim is to support WPA and 802.11i should support either WPA-TLS or 802.11i. into a single network. The way to accomplish this Figure 6 shows the proposed architecture, is by dividing the physical network into separate where the two logical secure access networks logical security networks. Most of the last genera- are presented. tion APs support WPA and 802.11i protocols, as well as the ability to create separate service set identifiers (SSIDs)8. EVALUATION AND INSIGHTS Therefore, in the proposed architecture, each AP is configured with two different SSIDs This section presents an evaluation of the security (SSID=802.11i-VEPR and SSID=WPA-VEPR) and performance capabilities of WPA EAP-TLS. Figure 6. VEPR secure wireless architecture 35TSTS
  • 53. Accessing an Existing Virtual Electronic Patient Record IPsec. 802.11i is not addressed due to the unavail- Security Experiments ability of 802.11i compliant devices at the time the experiments where undertaken. The discussion The network reaction to network attacks was comprises the evaluation of the proposed solutions observed in order to evaluate the security of the against network attacks and its efficiency in terms proposed solutions. These attacks comprise man- of performance and impact on the network. in-the-middle (MITM), impersonation, Denial of Service (DoS) and session hijacking. General Testbed In the MITM attack an intruder tries to see (“sniff”) the information exchanged between The testbed built to perform the experiments is the active hosts and insert itself in the middle. depicted in Figure 7. Unless otherwise mentioned, This allows the intruder to eavesdrop the com- all the experiments hereby described were built munications and even alter the data exchanged. upon open source operating systems and tools. The A basic approach for this attack, when no secu- FreeRADIUS (FreeRADIUS , 2008) implementa- rity is used, is a technique called arp spoofing tion was used as the RADIUS server. For the pub- (Whalen, 2001). lic key infra-structure the OpenSSL (OpenSSL, In the impersonation attack an intruder tries to 2007) suite was used. The IPSec infrastructure use the same IP address and the same hostname, was implemented on FreeSwan (FreeS/WAN as one of the valid clients of the network, to get Project, 2004). The software was installed in access to network resources. It differs from the computers running the Linux Operating System. MITM attack in that the attacker’s objective is In the IPsec tests, open source software was also only to access the network. So the intention is used to implement Access Points: HostAP (Ho- not to eavesdrop or alter the data exchanged by stAP team, 2007). This software allows building the valid host. a fully functional AP. In the WPA infrastructure, The session hijacking consists in an intruder the wpa_supplicant software (HostAP team, 2007) trying to obtain full control of a client successful was employed. session. It is an extension of the impersonation The ettercap tool (Ettercap Team, 2005) was attack, where the attacker needs to use the session used to perform the tests/security attacks. credentials/identifiers from the valid host to steal Figure 7. Wireless architecture used for testing procedures 36TS
  • 54. Accessing an Existing Virtual Electronic Patient Record its current session. It may use a MITM attack to it is possible with a WPA client to trigger this acquire such information. behaviour with fake network access messages. The Denial of Service (DoS) attack consists of This issue makes it possible to do a DoS attack disabling some (or all) of the network services (for against WPA, since it is just necessary to activate example denying authentication) by overwhelm- a WPA client and ask an AP for network access. ing the targeted service. The ultimate objective The AP verifies the message and, if it detects a is to deny network access. fake message, it blocks all network access, and stops all communications, including the access IPSec Results of valid clients. It is important to refer that, with the new 802.11i standard, this vulnerability has In the IPsec solution, the DoS attack was only not been solved (Wullems, Tham, J. Smith, successful before the establishment of the IPsec Looi, 2004). tunnel; after the establishment of the tunnel the attack did not succeed. For the MITM attack, the Comments arp spoofing option was used. We observed that, with the IPsec tunnel established, the MITM at- From the above experiments we can conclude tack did not succeed (it was not possible to see or that the IPsec and WPA EAP-TLS solutions are detect any kind of data flow). The impersonation very efficient against MITM, impersonation and attack also did not produce any result. For this session hijacking attacks. Both solutions are not attack an intruder used the same network address efficient against DoS attacks. It is possible to and hostname of a recognized client and then tried successfully perform DoS attacks using freely to establish an IPsec tunnel. As IPsec uses digital available tools. For systems where availability certificates for client authentication, the intruder is is essential, it is necessary to complement those not authenticated and the tunnel is not established solutions with mechanisms that reduce the risk of as was expected. Finally, the same negative results such attack. It is thus necessary to use tools like were achieved with session hijacking. Intrusion Detection Systems (IDS) and vulner- ability scanners. WPA/EAP-Tsults Complexity Experiments The same tests were performed to the WPA EAP- TLS implementation. One advantage of the WPA The system performance was measured in order solution is that it is a link layer security protocol. to evaluate the complexity introduced in the As ettercap is a tool that relies on the network layer, network elements. For this purpose the sysstat it was not possible to do MITM, impersonation (Systat Team, 2008) and vmstat (“vmstat Man and session hijack attacks. Other tools were also page”) tools were used. These tools allow evaluat- used to try to break the security of WPA such as ing CPU utilization, memory and interrupts. The Cain e Abel (Oxid IT Team, 2005) and Kismet results given by those tools are shown in Figure (Kismet Team, 2004). However, none of them 8 and Figure 9. The WPA experiment impacts on was able to achieve a successful result. On the the WPA client and RADIUS server; in the IPsec other hand, DoS attacks were performed with a experiment, the impact is on the VPN compo- high percentage of success. WPA disconnects nents10 (see Figure 4 and Figure 5 for the archi- the network for 1 minute if it detects an attack tectures). The scale is the percentage of resource against the MIC, this is done as part of a protection utilization except for the processes and interrupts against brute force attacks. Although difficult, that are absolute values. The pictures only show 37RSC
  • 55. Accessing an Existing Virtual Electronic Patient RecordFigure 8. System performance – sysstat resultsFigure 9. System performance – vmstat resultsthe RADIUS impact results for the WPA-TLS The results of WPA are similar to the ones ofexperiment, as they were negligible in the IPsec the plain system, introducing low impact in theexperiment. The presented results represent the network elements.average values obtained by 35 simulations, with From Figure 8 and Figure 9 we can observea stochastic confidence interval of 90%. that different absolute results are obtained by each An UDP flow of 54Mbps was used to represent tool. This is due to the specific requirements ofa fully loaded network. These results show that each tool and its design, i.e. the number of pro-the IPsec system requires more: CPU utilization, cesses, memory usage and number of interruptsmemory, interrupts and processes, therefore, its is influenced by the specific characteristics ofimpact on devices’ performance is not negligible. each tool.38
  • 56. Accessing an Existing Virtual Electronic Patient Record Note that the processing of the WPA packets throughput; it also adds more overhead, since itis done in the WPA client and the AP. Thus, conveys less information per bytes transferredthe encryption/decryption occurs at these two (total amount of data transferred for each TCPelements. In the IPsec case the secure tunnel is window) than the WPA solution. The throughputbetween the VPN client and the VPN Gateway, and transferred bytes of WPA is larger than IP-thus not impacting the AP (Figure 5). sec, but obviously lower than the plain network (without security).Impact on Data Flows These results are due to the larger complexity introduced by IPsec (ESP with tunnel mode wasTo evaluate the impact on data flow when the used, which adds a new header and new authen-security mechanisms are in place, we performed tication field). WPA does not make significantexperiments using TCP and UDP traffic, and changes to a packet, just ciphers it and adds an IVconsidering a network with and without security field. The same experiment was done for differentimplemented. TCP window sizes11, which also confirmed the For traffic generation, IPERF (Iperf Team, fact that IPsec is the solution with less throughput2005) and Crude (Crude team, 2002) tools were and bytes transferred.used. All traffic was generated after the negotia- To evaluate the jitter12 and the number of losttion of the specific security protocol (IPsec and packets, IPERF was used with UDP flows inWPA-TLS). networks with bandwidths of 10 Mbits/s and 54 Figure 9 shows the results of throughput and Mbits/s. These consisted of 5 flows with durationtransferred bytes of a TCP flow with a duration of 60 seconds, simulating a voice communication.of 120 seconds and a default window size of 85.3 The obtained results represent the average resultsKbytes, when no security, WPA and IPsec are in of 20 simulations with a stochastic confidenceplace. The presented results represent the aver- interval of 95%. Figure 11 and Figure 12 show theage values obtained by 48 simulations, with a results for a network bandwidth of 10 Mbits/s.stochastic confidence interval of 92%. As can be The results demonstrate that, due to its com-seen, IPsec is the mechanism that achieves lower plexity and packet processing, IPsec has worseFigure 10. Throughput and bytes transferred 39
  • 57. Accessing an Existing Virtual Electronic Patient RecordFigure 11. Jitter of UDP flows in a 10 Mbits/s networkFigure 12. Lost packets of UDP flows in a 10 Mbits/s networkjitter results. Regarding the number of lost packets, These data flow results led to the naturalIPsec is the security solution that has better results. conclusion that for TCP communications (e.g.This is due to the fact that the process of packet with file transfers), the WPA implementation hasprotection happens between the VPN gateway and more advantages. For UDP communications thethe client, while in the WPA solution this is done IPsec protocol achieves lower loss rates.between the AP and the client. As the gatewayhas more capacity for processing the packets, it Deployment Discussioncan keep its buffer available and the number oflost packets is reduced. The results obtained with The deployment of the infrastructure requires54 Mbits/s and with CRUDE confirm the ones of studies regarding the location of access pointsIPERF with 10 Mbits/s. for the intended coverage, as mentioned in the40
  • 58. Accessing an Existing Virtual Electronic Patient Record introduction. The costs associated with the hard- believed to be very similar to the WPA solution. ware (APs, wireless cards, Ethernet switches and The impact of the WPA security is negligible servers) would depend on the required coverage in terms of the performance of the system. The and the number of users enabled with this access. throughput achieved was slightly worse in WPA However, notice that current laptops and PDAs than in a plain system. However, the difference have already wireless capabilities supporting should not be noticeable to users. 802.11i. As discussed, the existing VEPR wired solu- The software associated with the framework tion was designed and implemented with all the is already available with the hardware except for security requirements; adding this extra layer of the servers (LDAP, RADIUS). Nevertheless, they security results in an easier process, as long as it are readily available in reliable free open source respects the security goals of the VEPR. packages (with support available separately) and With the proposed architecture, secure ac- also in commercial products. cess to the current system is increased due to the In terms of user impact, the access to the net- wireless connectivity advantages (e.g. mobility, work could be coupled with the existing credential everywhere access and access to wider range of system, thus easing the needed user interaction. devices). This access provides secure authentica- However, as a first approach these two authen- tion and authorization, secure communications tication points in the network and in the VEPR and also maintains the integrity of the retrieved should be done separately. The final purpose is to information, thus preserving the security goals build a single-sign-on system that would provide of the VEPR. This is very important and justifies only one authentication control. the need for similar studies when implementing wireless solutions. CUSiON Open Challenges Discussion As future work, a prototype will be implemented within the real scenario so that the wireless solu- The wireless architecture discussed above is tion can be evaluated. Several issues need to be able to provide wide as well as mobile and flex- tested and enhanced. These include performance, ible access to the VEPR implemented within a access control, availability issues (such as DoS), healthcare institution. The architecture is modular access point correct distribution and usability. and flexible in order to adapt itself to the existing Further issues are related to the presentation features so that it can be incorporated when a LAN of the VEPR within wireless devices. This needs is already in place. In particular, the proposed proper study as its usefulness and success may architecture takes into account the fact that the depend upon it. existing devices can be reused with WPA/EAP- TLS; it also integrates the recent 802.11i standard, making it versatile and upgradeable. ACKknowledgmen To account for the security and performance of the system, several studies and tests were This VEPR has already won 2 prizes for its in- made with the presented technologies. The only novation and results from Portuguese Government exception is the recent 802.11i because no compli- Institutions. As such, the first author would like ant devices were available at the time of testing. to thank all the parties that collaborated in its Nevertheless, its overall observed performance is implementation, specially the Security Commis- 41CS
  • 59. Accessing an Existing Virtual Electronic Patient Record sion of Hospital S. João, LIACC and CINTESIS Durham, D., Boyle, J., Cohen, R., Rajan, R., for their interest and support. Sastry, A. (2000, January). The cops protocol. Retrieved February 14, 2008, from http://www. rfc-editor.org/rfc/rfc2748.txt. Refeen Ettercap Team. (2005, May 29). Ettercap ng. Re- trieved February 12, 2008, from http://ettercap. Aboba, B., Simon, D. (1999). Rfc 2716 ppp eap sourceforge.net/. tls authentication protocol. IETF. Retrieved from http://tools.ietf.org/html/rfc2716. Ferraiolo, D. F., Sandhu, R., Gavrila, S., Kuhn, D. R., Chandramouli, R. (2001). Proposed nist Aboba, B., Blunk, L., Vollbrecht, J., Carlson, J., standard for role-based access control. ACM Trans. Levkowetz, H. (2004). Rfc 3748 extensible authen- Inf. Syst. Secur., 4(3), 224-274. tication protocol (eap). IETF. Retrieved February 8, 2008, from http://tools.ietf.org/html/rfc3748. Ferreira, A., Correia, R., Costa-Pereira, A. (2004). Securing a web based epr: an approach Baker, D. B. (2003). Wireless (in) security for to secure a centralized epr within hospital. In 6th healthcare. In Advocacy White Paper. Science International on Enterprise Information Systems, Applications International. 3(pp. 54-59). Benson, T. (2002). Why general practitioners Ferreira, A., Correia, R., Antunes, L., Palhares, use computers and hospital doctors do not---part E., Farinha, P., Costa-Pereira, A. (2005). How 2: scalability, BMJ, 325(7372), 1090-1093. doi: to start modelling access control in a healthcare 10.1136/bmj.325.7372.1090. organization. In 10th International Symposium for Blobel, B. (2004). Authorisation and access control Health Information Management Research. for electronic health record systems. International Ferreira, A., Cruz-Correia, R., Antunes, L., Journal of Medical Informatics, 73(3), 251-257. Palhares, E., Marques, P., Costa, P. et al. (2004). CEN. (1999). Health informatics - secure user Integrity for electronic patient record reports. identification for healthcare - management and In 17th IEEE Symposium on Computer-Based security of passwords. CEN. Medical Systems (pp. 4-9). Crude team. (2002, September 13). (c)rude - rude Finseth, C. (1993, July). Rfc 1492 - an access crude. Retrieved February 12, 2008, from control protocol, sometimes called tacacs. Re- http://rude.sourceforge.net/. trieved February 14, 2008, from http://www.faqs. org/rfcs/rfc1492.html. Cruz-Correia, R., Vieira-Marques, P., Costa, P., Ferreira, A., Palhares, E., Araújo, F. et al. (2005). FreeRADIUS . (2008, January 22). Freeradius Integration of hospital data using agent technolo- server. Retrieved February 12, 2008, from http:// gies - a case study. AICommunications special www.freeradius.org/. issue of ECAI, 18(3), 191-200. FreeS/WAN Project. (2004, April 22). Frees/wan. Denley, I., Smith, S. W. (1999). Privacy in clini- Retrieved February 12, 2008, from http://www. cal information systems in secondary care. BMJ: freeswan.org/. British Medical Journal, 318(7194). Retrieved Harkins, D., Carrel, D. (1998). Rfc 2409 the March 10, 2008, from http://www.pubmedcentral. internet key exchange (ike). IETF. Retrieved from nih.gov/articlerender.fcgi?artid=1115718. http://tools.ietf.org/html/rfc2409. 42e
  • 60. Accessing an Existing Virtual Electronic Patient RecordHodges, J., Morgan, R. (2002). Rfc 3377: light- Lehoux, P., Sicotte, C., Denis, J. (1999). Assess-weight directory access protocol (v3): technical ment of a computerized medical record system:specification. IETF. Retrieved from http://tools. disclosing scripts of use, Evaluation and Programietf.org/html/rfc3377. Planning, 22(4), 439-453. doi: 10.1016/S0149- 7189(99)00034-8.HostAP team. (2007, December 2). Host ap linuxdriver for intersil prism2/2.5/3 wireless lan cards Marti, R., Delgado, J. (2003). Security in aand wpa supplicant. Retrieved February 12, 2008, wireless mobile health care system. In Universitatfrom http://hostap.epitest.fi/. Pompeu Fabra.IEEE 802.10. (1998). Ieee standards for McAlearney, A. S., Schweikhart, S. B., Me-local and metropolitan area net works: dow, M. A. (2004). Doctors’ experience withstandard for interoperable lan/man security handheld computers in clinical practice: qualita-(sils). Retrieved from http://standards.ieee.org/ tive study. BMJ, 328(7449), 1162. doi: 10.1136/getieee802/download/802.10-1998.pdf. bmj.328.7449.1162.IEEE 802.11i. (2004). Part11: wireless lan medium Microsoft. (2004). Windows server 2003 activeaccess control (mac) and physical layer (phy) directory. Retrieved February 15, 2008, fromspecifications amendment 6: medium access con- http://www.microsoft.com/windowsserver2003/trol (mac) security enhancements . Retrieved from technologies/directory/activedirectory/default.http://standards.ieee.org/getieee802/download/ mspx.802.11i-2004.pdf. Moskowitz, R. (2003, November 4). WeaknessIEEE 802.1X. (2004). Ieee standards for local and in passphrase choice in wpa interface. Wi-Fimetropolitan area networks—port-based network Net News. Retrieved from http://wifinetnews.access control. Retrieved from http://standards. com/archives/002452.html.ieee.org/getieee802/download/802.1X-2004.pdf. Ng, S., Tan. (2006). Security issues of wirelessIperf Team. (2005, May 3). Nlanr/dast : iperf sensor networks in healthcare applications. BT- the tcp/udp bandwidth measurement tool. Re- Technology Journal, 24(2), 138-144. doi: 10.1007/trieved February 12, 2008, from http://dast.nlanr. s10550-006-0051-8.net/Projects/Iperf/. NIST. (2001). Fips-197: advanced encryption stan-Kent, S., Atkinson, R. (1998a). Rfc 2402 ip dard. Natioanl Institute of Standards (NIST).authentication header. IETF. Retrieved from Novell. (2004). Novell edirectory vs. micro-http://tools.ietf.org/html/rfc2402. soft active directory. Retrieved February 15,Kent, S., Atkinson, R. (1998b). Rfc 2406 ip 2008, from http://www.novell.com/collater-encapsulating security payload (esp). IETF. Re- al/4621396/4621396.pdf.trieved from http://tools.ietf.org/html/rfc2406. OpenSSL. (2007, October 19). Openssl: the openKim, H., Afifi, H. (2003). Improving mobile source toolkit for ssl/tls. Retrieved February 12,authentication with new aaa protocols. Retrieved 2008, from http://openssl.org/.February 15, 2008, from http://citeseer.ist.psu. Oxid IT Team. (2005). Cain abel. Retrievededu/article/kim03improving.html. February 15, 2008, from http://www.oxid.it/cain.Kismet Team. (2004). Kismet. Retrieved February html.18, 2008, from http://www.kismetwireless.net/. 43
  • 61. Accessing an Existing Virtual Electronic Patient Record Rigney, C., Rubens, A., Simpson, W. Willens, EndNoTES S. (1997). Rfc 2138 remote authentication dial in user service (radius). IETF. Retrieved from 1 Java version of the Open DataBase Con- http://tools.ietf.org/html/rfc2138. nectivity (ODBC) designed by Microsoft to provide a common API for accessing Systat Team. (2008, January 6). Sysstat. Retrieved databases. February 12, 2008, from http://pagesperso-orange. 2 Network File System is an IETF protocol fr/sebastien.godard/. to allow client systems to access remote Ventura, H. (2002). Diameter next generation’s storage as if it were locally available. aaa protocol. 3 WEP is part of the original 802.11 stan- dard. Vmstat man page. Retrieved December 17, 2007, 4 Some older products that do not support from http://linuxcommand.org/man_pages/vm- directly WPA can (most likely) be software stat8.html. upgradable. Walker, J. (2003). 802.11 security séries part ii: 5 The Wi-Fi Alliance certifies products com- the temporal key integrity protocol. Intel Corpora- pliant to 802.11i as WPA2. tion. Retrieved from http://softwarecommunity. 6 See http://certifications.wi-fi.org/wbcs_cer- intel.com/articles/eng/1905.htm. tified_products.php?advanced=1. 7 Note that WPA and 802.11i also are inde- Whalen, S. (2001, April). An introduction to arp pendent of the source application. spoofing. Retrieved February 15, 2008, from 8 SSIDs identify the network that a device is http://www.node99.org/projects/arpspoof/. connecting to. Whiting, D., Housley, R., Ferguson, N. (2003, 9 For technical reasons the AP needs to map September). Rfc 3610 - counter with cbc-mac SSIDs with VLANs. The AP marks all IP (ccm). Retrieved March 8, 2008, from http://www. packets with the VLAN associated with the faqs.org/rfcs/rfc3610.html. corresponding SSID. For interconnecting the AP and the RADIUS server, a layer 2 Wpa . Retrieved February 8, 2008, from http:// or 3 switch is used. www.wi-fi.org/knowledge_center/wpa/. 10 It was not technically possible to evaluate Wullems, C., Tham, K., Smith, J., Looi, M. the impact on the Access Point. (2004). A trivial denial of service attack on ieee 11 The TCP window size controls the number 802.11 direct sequence spread spectrum wireless of packets that can be sent without being lans. Wireless Telecommunications Symposium, acknowledged. Increasing its size will mean 129-136. that a higher number of packets can be sent but if the receiver’s buffer can not cope with Yu, W. D., Jothiram, V. (2007). Security in wire- the amount it will mean that the sender will less mobile technology for healthcare systems In have to re-send more packets. , e-Health Networking, Application and Services, 12 Jitter pertains to the variation of packet de- 2007 9th International Conference on (pp. 308- lay; the delay is composed by sender delay, 311). doi: 10.1109/HEALTH.2007.381659. travelling in the network delay and receiver delay. The variability of this total delay is measured by jitter. 44e
  • 62. 45 Chapter III Personal Health Records Systems Go Mobile: De ning Evaluation Com onents Phillip Olla Madonna University, USA Joseph Tan Wayne State University, USA ABSTRACT This chapter provides an overview of mobile personal health record (MPHR) systems. A Mobile personal health record is an eclectic application through which patients can access, manage, and share their health information from a mobile device in a private, confidential, and secure environment. Personal health records have evolved over the past three decades from a small card or booklet with immuniza- tions recorded into fully functional mobile accessible portals, and it is the PHR evolution outside of the secure healthcare environment that is causing some concerns regarding privacy. Specifically, the chapter reviews the extant literature on critical evaluative components to be considered when assessing MPHR systems. Inoduion the healthcare industry and the recent attention and increased activity in the adoption of Personal Information technology (IT) is dramatically Health Record (PHR) systems. By distinction, transforming the delivery of healthcare services. PHR systems have not established a similar height This can be seen through the increased activity of interest as the EHR (Tang, 2006), but this is in Mobile Health (M-Health) and promotion of changing as more government bodies such as the the Electronic Health Record (EHR) systems in U.S. Secretary of Health and Human Services, Copyright © 2009, IGI Global, distributing in print or electronic forms without written permission of IGI Global is prohibited.a
  • 63. Personal Health Records Systems Go Mobile the National Coordinator for Health Information The PHR migration to the mobile platform Technology, and the Administrator of the Centres offers immense benefits such as portability and for Medicare and Medicaid Services (CMS) have convenience in the accessing and transmitting all identified PHRs as a top priority. In addition of personal health records from a single loca- to the government organizations involvement, tion, the empowerment of the health consumers standards organization such as Health Level Seven to control, verify, and manage their own health (HL7) have began the standard definition process information, and the potential enhancement of to formalize a system model for PHRs. patient-caregiver relations. Unfortunately these PHR aims to allow individual health consum- benefits can be overshadowed by the concerns ers the ability to monitor and manage their personal regarding security, privacy, mobile technology health information from multiple sources in a choice, and validity of information. This chapter single repository. Research shows that maintain- will highlight the important evaluation compo- ing a PHR encourages personal participation in nents that need to be considered when the PHR healthcare and cultivates an increased emphasis is modified to support mobility. on communication between the individual and the The discussion is structured as follows. Fol- healthcare provider (Kupchunas, 2007). The use lowing the introduction, the next section describes of a PHR provides the opportunity for healthcare the history and background of PHRs. Against providers to monitor and educate patients on this backdrop, an overview of the literature and health matters and lifestyle changes, and it also progress being made on PHR research will be acts as a tool for enhancing health literacy. The highlighted. This will be followed by a discussion PHR will eventually improve the decision making on the four categories of PHR systems, namely capabilities as the patients become more proficient “individually maintained”, “tethered” to a health at recording and monitoring vital health informa- plan or employer, “comprehensive” or “Health tion (Lee, Delaney, Moorhead, 2007). The goal 2.0”. Once the advantages and disadvantages of utilizing personal health records would be to of the various types of PHR systems have been enhance and optimize the healthcare practices presented, the discussion will converge on a while allowing patients to manage their own framework for Mobile PHR systems evaluation, health care decision-making. For the caregivers, which is then followed by the review of three PHR technology can improve efficiency, cost- commercial Mobile PHR systems using compo- effectiveness, timeliness, safety, and efficacy of nents from the framework. The chapter will then the care processes, whereas for the individual conclude with a summary of thoughts on future consumers, it can help improve their quality of growth and development in this area. life. Large organizations and government bod- ies have recently gained an interest in the PHR phenomenon; for example, Intel, Wal-Mart and Ba and HISTORY of BP have formed a consortium, called Dossia, to PHR supply PHRs for their employees; Medicare and Medicaid Services are trialing PHR with Medi- The Personal Heath Record (PHR) is not a com- care claims; and Google and Microsoft have also pletely new phenomenon; accordingly, one of entered this market with new products such as the earliest references to a PHR can be found in Microsoft Health Vault and Google Health. In ad- an article by Okawa (1973) entitled, “A personal dition, Verizon Communications in combination health record for young female students.” Several with WebMD now offers a password-protected references to personal health records surfaced site for PHR (Reese, 2007). prominently as “paper records” up until the mid 46kgoundioyR
  • 64. Personal Health Records Systems Go Mobile1990’s, when the computerized version of the that it is not possible or even desirable to attemptPHR appeared. With the diffusion and general a unitary definition of a PHR. The NCVHS diduse of computers, individuals became interested however state that it is possible and useful toin maintaining an electric copy of their personal characterize PHRs by their attributes. Attributeshealth information. Sittig (2002) conducted a associated with PHRs are elements such as thesearch of available Internet-based PHR’s and nature of the PHR’s content, the source or sourcescame up with 27 identified programs or usable of information, and the functions that they offertools; by April of 2003, only 7 of these tools (Sprague, 2006). In retrospect, the consumer andwere relevant and still accessible (Sittig, 2002). health care provider will be able to maintain thisThe demise of these Internet-based companies health care information in a way that is assessableoccurred with the meltdown of the dotcoms, to both parties.poor business plans, inefficiency, and decreased More recently, President Bush and Secretaryvalue to the individual. (Personal Health Working Leavitt brokered a plan that would allow patientsGroup, 2005). Even though the early PHR’s were and their health care providers the ability to accessultimately unsuccessful, they provided valuable their personal health records through the use ofinsight into the next generation of PHR’s by sup- technology (Gellman, 2008). Having this techno-porting the fact that by “adopting EMRs, providers logical capability ensures that patient and healthand health care delivery systems can facilitate the care provider can access healthcare information atdevelopment and implementation of PHR tools any time when seeking medical care. As a result,and PHR’s should offer clear benefits to users and in 2001, the National Committee on Vital andcritical stakeholders.” (Personal Health Working Health Statistics identified areas in deliverance ofGroup). healthcare services that required improvements In summary, PHRs have evolved over the past in systems in order to promote quality patientthree decades from a small card or booklet with care services, continuity of care and beneficialimmunizations recorded on it to a computerized treatment modalities. Technological systemsapplication that stores an individual’s personal were further explored in order to change the wayhealth information. A report by the Markle Foun- personal health information was maintained bydation, Connecting for Health (2003), defines the patient and the health care provider.PHR’s specifically as The Hurricane Katrina saga in the USA has caused the healthcare industry to recognize the“An electric application through which individu- importance of seeking ways to secure personalals can access, manage and share their health health records. For years, people had collected,information, and that of others for whom they are maintained and stored their health informationauthorized, in a private secure, and confidential on paper, in memory, and other manual means.environment.” Page 4 During such disaster, all of the information kept by patients and health care providers became Imperative to the usefulness and functioning inaccessible. But with the advances in technol-of the PHR is the establishment of a common data ogy, consumers are provided with tools andset. At the very least, it should contain information mechanisms to maintain their own health recordsregarding items such as: allergies and adverse drug through the capability of computer-based ap-reactions, illness and hospitalizations, surgeries plications. The consumer then has the ultimateand procedures, vaccinations, lab test results, and responsibility of keeping track and updating theirfamily history. The National Committee on Vital personal information so that their medical careand Health Statistics (NCVHS) reported in 2005 is effective and efficient. Additional benefits of a 47
  • 65. Personal Health Records Systems Go Mobile PHR include access to health information, data on system functionality of PHRs along with the collection, improved health, disease management unique elements associated with a PHR. and tracking, improved communication with health care providers, and although not well docu- mented, it is also believed to lower costs related PHR Funionali and to chronic disease management and wellness Componen programs (Tang, Ash, Bates, Overhage, Sands, 2006) According to Kupchunas, “maintaining a There have been considerable research and in- PHR encourages increased personal participation vestment into PHRs over the past decade. This in healthcare and fosters a greater emphasis on is evident in the growing number of publications communication between the individual and the and patents registered in this area. Based on a caregivers” (Kupchunas, 2007: pg 185). review of the extant literature, the main areas of functionality are illustrated in figure 1. PHRLiterature Review As shown in figure 2, the functionality can be broken down into various aspects of con- PHRs aim to enhance and optimize the healthcare sumer functionality (Denton, 2001). Two major practices while allowing patients to manage their aspects, which are considered to be among the own health care decision-making. On the side most promising uses for PHR system, are the use of healthcare practices, this would improve ef- of PHR for health education and for managing ficiency, effectiveness, timeliness, safety; on the specific patients conditions(Arbogast Dodrill, side of the patients, it would help them to improve 1984; Bent, Bolsin, Creati, Patrick, Colson, their quality of life. A comprehensive review of 2002; Bhuyan, 2004). Another aspect is linking the extant literature published on issues related the PHR to decision support. This is also a key to PHRs using scropus tools (www.scropus.com) research component with obvious benefits to the is now presented. patients choosing and understanding a suitable Scopus is the largest abstract and citation treatment plan (Abidi Goh, 2000; Ackerman, database of research literature and Web sources. 2007). Other important aspects include consumer Scropus contains over 33 million abstract and information (managing clinical and personal data) provides access to publications in 15,000 peer- and the growing use of PHR systems to support reviewed journals and 200 book series from and improve administrative support functions. more than 4,000 publishers. The system also has As issues relating to the area of privacy and access to over 1200 Open Access journals, 500 security (Agrawal C. Johnson, 2007; Alban, conference proceedings and 600 trade publica- Feldmar, Gabbay, Lefor, 2005; Albright, 2007; tions. The literature search undertaken not only Alhaqbani Fidge, 2008; Al-Salqan, Jaganna- investigated the bulk of the academic literature than, Davis, Zhang, Reddy, 1995; Anderson, but also reviewed data from 386 million scientific 1996; Armitage et al., 2008; Blobel, Pharow, Web pages and 22 million patent records from 5 Spiegel, Engel, Engelbrecht, 2001) have now patent offices. Even though our search criteria become more important than ever given the nature covered a period of 20 years, we have chosen to of the new generations of PHR systems, we will present data that are only from the last decade. examine them separately and more closely in a The search excluded medical health records, later section of this chapter. . electronic health records, and electronic medical Although there seems to be a peak in academic records. This essentially directs the review to focus research around 2006 (see Figure 1), this has not 48Ry
  • 66. Personal Health Records Systems Go Mobile Figure 1. Personal health records publications slowed down the commercial field. There is an Business Models increase in Web based activities and patent ap- plications. This shows that as the research matures A study by Adler, highlighted the fact that 74.6% more commercial offerings are being launched. of patients surveyed were willing to pay a small The industry-based publications are focusing on annual fee for one or more of the following online issues such as standards, interoperability, security services: viewing parts of their medical record, and integration (Ball Gold, 2006). messaging with their physician, medication refills, Personal health records are maturing rap- appointment requests, and billing inquiries (Adler, idly and several diverse actors’ have full-scale 2006). A variety of sources, such as healthcare implementation of PHR systems - these actors providers, insurers, employer or commercial include employer groups such as DOSSIA, and suppliers offer a wide range of products that are commercial vendors such as Google and Micro- available to help you create your own PHR. Some soft. The emergence of large organizations such PHRs include products that have free tools, others as Google who already hold significant volumes offer products for purchase, Figure 3 summarizes of data about individuals is creating serious the various existing business models. concerns. The concept map below provides a An array of complex business models exists snapshot of the current work being undertaken in in the PHR domain. However, from a general the field, the research covers a broad spectrum, marketing perspective, these can be grouped into which highlights the multidisciplinary approach three commercial types – consumer purchase, to research in this field. consumer subscription, and a combination of the two (Gellman, R. 2008). For the purpose of this 49B
  • 67. Personal Health Records Systems Go MobileFigure 2. Concept map of personal health records research from literaturereview, the business models have been classified 3. Free model: With this model the PHR isinto five categories: free to consumers because the service is supported by advertising.1. Consumer subscription model: The con- 4. Employee support model: With this model, sumer is responsible for an ongoing service an employer or health plan will contribute fee to access the data and related systems. part or all the fees to run the service. This2. Consumer purchase model: With this is seen as preventative medicine as there is approach the consumer pays a fixed fee to strong evidence that they could save money purchase the software that provides the core on health care costs in the long run. functionality of the PHR outright. This type 5. Combination: There is also the possibility is typically a desktop application. The sub- that a combination of these various models scriber may also have the option to purchase could be in effect. For example a PHR service a support contact. paid for by an employer or health plan may allow advertising.50
  • 68. Personal Health Records Systems Go MobileFigure 3. Business model for personal health records Although these models may be the obvious certification. Most sites stated they protect thesource of revenues, there may be other elements privacy of personal health information (PHI) andin play such as affiliation models with informa- will not share your PHI. A PHR privacy policytional Web sites, niche search engines, articles, study in 2007 conducted by the Department ofsurveys, software downloads, and a variety of Health and Human Services (DHHS) stated thatproducts that are not directly associated with the “only 3 percent, or one in 30, of PHR privacyPHR system. policies stated that explicit consumer consent Prices and formats vary widely, with diverse was necessary prior to the vendor sharing any oflevels of technical abilities and functionality the data in the PHR” (Gellman, 2008, p. 7). Evenamong the different product lines. From the though numerous PHRs are free to consumers,perspective of Health Insurance Portability and a PHR vendor is operating a business, in whichAccountability Act (HIPPA), many of the PHR revenues are generated primarily through adver-sites state in their guidelines that they ‘comply tising and marketing. Therefore, it is imperativewith’, or ‘we voluntarily operate within the guide- that consumers educate themselves when sharinglines’, but most mentioned nothing about HIPPA their information with PHR vendors. 51
  • 69. Personal Health Records Systems Go Mobile Categorization of PHR Systems of collaboration, information exchange, and Based on Ownership knowledge transfer; and focus on delivering value added services that empower health participants There are many ways to categorize PHR systems, (patients, physicians, providers, and payers) with but for the purpose of this chapter, the concept freedom, choice, and accountability for health of ownership is used. This approach was taken outcomes. to highlight the potential issues that arise from There is currently a lot of debate regarding transferring ownership. Most of the existing this new concept. Some believe that the these applications fall into three main categories: indi- companies are providing a wider movement to vidually maintained, “tethered” to a health plan or reform the entire US health care system, while employer, or comprehensive (Sprague, 2006). others believe that these are merely tools and A new category, which has also emerged re- technologies to support the current system. There cently and has been added for the purpose of this are considerable concerns when companies that review, is called health 2.0. It is this new category are not within the health care industry take active of PHR applications that is providing serious con- roles in storing and maintaining personal health cerns from a security and ethics perspective. data. Most of the companies that operate in this In order to assist readers attempting to compare domain are strong advocates for mobility and are and contrast among the different categories, we keen to integrate mobile technology into their will also highlight the respective advantages and Web portal solutions. disadvantages associated with each category. Google is implementing a pilot at Cleveland Clinic hospital in Cleveland as the pilot site for a new personal health records initiative. Between MoBIile HEALTH 2.0 PHRS 1,500 and 10,000 patients will partake in the project. Patients will have their current MyChart This category refers to the new generation of Web electronic health records migrated to their Google 2.0 healthcare applications that support mobility. accounts. Once the PHRs, are shared with Google, Health 2.0 is just as difficult to define as the Web patients will have the capability to access them 2.0 concepts. The Health 2.0 conference defini- outside of the Cleveland Clinic. Google is not tion focuses on user-generated aspects of Web2.0 the only technology giant looking to change the within health care but not directly interacting with healthcare industry. AOL founder, Steve Case, the mainstream health care system. The problem has recently launched a new organization called with this definition is that it is very difficult to Revolution Health (http://www.revolutionhealth. separate the user generated and mainstream com/); InterActiveCorp has also invested in healthcare systems without generating duplica- several health-related start-ups (http://www. tion and redundancy. Given that there are several healthcentral.com/); and Microsoft has been definitions, we have decided to share the approach very active with a medical record service called taken by Scott Shreeve from CrossOver Healthcare HealthVault. due to the pragmatic nature of his definition of a One of the ways that companies are providing Health 2.0 Company: customers with additional benefits is by providing “Next generation health companies that secure access to the health records stored on the leverage the principles of openness, standards, portals via a mobile device. At this point, we will and transparency; utilize the technology tools highlight the respective advantages and disadvan- tages associated with the Health 2.0 category. 52CRSealh
  • 70. Personal Health Records Systems Go Mobile Advantages Individually Main PHR • The data is available from anywhere The simplest form of a PHR is one that is main- • Interoperability and the use of open in- tained by the individual. This sort of PHR is terfaces mean that data can be imported created, updated, and controlled strictly by the directly from the healthcare provider in individual (Sprague, 2006). Such a PHR allows some cases. the individual to organize and retrieve their own health information and it captures the patient’s Disadvantages concerns, symptoms, emergency contacts, and other relevant information (Endsley et al., 2006). • This approach has already raised serious This type of PHR can be software driven and may privacy concerns due to the migration of reside on a person’s computer or be Web-based. private data into the commercial domain, The Wed-based format is maintained by a third- and critics of such projects have already party. Other devices such as “smart cards,” USB begun to make themselves heard. drives, and CDs can also be utilized for this type • These third-party PHR systems are not of PHR (Endsley et al., 2006). covered by the HIPPA, which has been in effect since 1996 and requires individuals to Advantages be notified when a party other than a patients doctor wants to access confidential medical The individually maintained PHR has a limited information. number of advantages, aside from it being con- • There will be some costs associated with trolled by the individual; the security may be the business model. the biggest advantage of this type of PHR. The • Access to the Internet is required to access individual PHR provides more control over ac- data via the mobile device. cess to the data contained within the PHR (Tang et al., 2006). Figure 4. Personal health records categories 53ainedR
  • 71. Personal Health Records Systems Go Mobile Disadvantages who changes jobs or insurance companies may lose access to the personal health information The question regarding the individual PHR is (Sprague et al., 2006). Other disadvantages include how often will the individual update their PHR? security and privacy issues and the question of The individual PHR, as maintained by the in- standardized language (Tang et al., 2006). dividual, may not be updated as often as they should. Another disadvantage is that it may not be considered a trusted conduit for transmission Compehen PHR of medical information among clinical offices or health care institutions. Another aspect is the fact A more sophisticated PHR is made available that the individual PHR may not have enough through the electronic health record (EHR). A back-up systems in place in case of any technical care provider or organization, such as a hospital, malfunction (Tang et al., 2006). As we discussed physician, or an integrated delivery system, usu- the back-up system, we need to also take into con- ally maintains the EHR. The EHR is designed sideration the literacy of the individual. How well to be a repository of clinical information on a versed is the individual with regards to medical patient and to accept information from a variety and technical information? Does the individual of sources. The sources of information may in- understand and comprehend the information and clude physicians, laboratories, and consumers. can they relay the information technically (Tang The capability of this type of PHR allows the et al., 2006). consumer access to some portion of his or her clinical data, under rules set by the provider; it may also allow secure e-mail messaging, access “Teed a HEALTH Plan to condition-specific information, appointment omploye scheduling, and many other functions (Sprague, 2006). Different organizations will maintain a The “tethered” PHR, populated with claims data range of somewhat differing policies and proce- and typically available to the consumer through dures with respect to availability, accessibility, a secure Web portal, is created by a health plan portability, release and use of personal health or an employer Web. information captured in the PHR. Advantages Advantages The main advantage of the “tethered” PHR is the The biggest advantage is the access that patients fact that this type of integrated PHR can provide will have to a wide array of credible health infor- the patient with much more relevant data. It may mation, data, and knowledge (Tang et al., 2006). also provide the patient with a better back-up A secondary but still very important advantage system, due to the fact that the integrated PHRs is the potential to lower communication barriers have a larger back-up system (Sprague, 2006). between the patient and health care provider (Tang et al., 2006). Integrated or comprehensive PHRs Disadvantages provide an ongoing connection between patient and physician (Tang et al., 2006). A major disadvantage to the “tethered” PHR is the lack of portability; for example, the individual 54heoealhRive
  • 72. Personal Health Records Systems Go Mobile Disadvantages portable (flash drive, CDROM, DVD or smart card) or on mobile devices. Many of the disadvantages are the same for the comprehensive as for the “tethered” PHR. The issues of security and privacy and the use PHRon Peonal of standardized language are being raised with Compu regards to the comprehensive PHR. The World Privacy Forum recently issues a report on why PHR products that involved health information many PHRs can actually threaten security. The residing on one’s own computer is one of the most next section will discuss some of the important common types of PHR service. . The user typically issues surrounding privacy and security of PHR uses a CD or downloads the PHR template from systems the Internet to the personal computer. Informa- tion about one’s health would then be entered and Categorization Based on Storage maintained by the user on his or her computer. Medium There are now organizations that collected, PHR on UNIVERSAal SERIial BUS organized, summarized, and then make avail- (USB) DRIVES able electronically copies of all of the medical information. In some cases, the information is A USB, (Universal Serial Bus), also known as provided to user in a wallet or regular CD, or it thumb drive or jump drive, is an inexpensive, por- could be accessed through the Internet. Typically table, electronic device used to store information. an electronic PHR can be maintained in various The USB is a NAND flash device, with memory formats including paper-based, PC, Internet or stored in chips. This architecture allows for low Figure 5. Categorization of PHRs based on storage medium 55CBSheeRniveSeBuSBive
  • 73. Personal Health Records Systems Go Mobilepower consumption, fast speed of use, and high identity, and ability to perform on-line back updensity, allowing for storage of large amounts of support. Security technology available includesinformation (Axelson, 2006). Device memory is encryption, password protection, requirements“nonvolatile”, easily allowing the user to write and for complex passwords, password reset capability,erase information. Moreover, these devices are biometric identification verification, and “lost anddesigned to spread the write-erase cycles evenly found” features.across the components of the device in order toprolong the device life. DisadvantagesAdvantages USB device issues that can be problematic for PHR utilization include potential for device dam-Utilization of a USB device for storage of the PHR age and data corruption. In addition, the costs ofpromotes an important goal, which is, encouraging USB devices increase as the technological andindividuals to become actively involved in their security features become more advanced andhealthcare. With a USB device, the individual complex. Healthcare information is secure onmanages health information data input and storage the USB device when in possession of the owner,(Ball, Smith, Bakalar, 2007). Once health in- however, if the device is lost and data on theformation is stored on a portable USB device, this device is not encrypted or password protected,information can be quickly accessed anywhere, a privacy can become jeopardized (Tang et al.,factor especially important during emergencies 2006). While the goal of a PHR is to increase an(Shetty, 2007). USB access is a standard feature individual’s engagement in their own healthcare,on all computer systems, and the USB devices those involved in the evaluation and promotion ofallow easy insertion and disconnection without PHRs question whether consumers are capable andinterruption of the system (www.intel.com). willing to assume the tasks and costs involved in Individuals who advocate for development maintaining a stand-alone/USB PHR (Nationaland utilization of PHRs stress that portability of the Committee on Vital and Health Statistics, 2006).PHR is important as individuals move through the The ongoing work involved in keeping the USBhealthcare system. A USB would allow individuals PHR up to date will most likely be seen as a burdento input and integrate data from many different to many consumers. Many providers would notsources, such as electronic health records, labo- likely see a PHR that contains information storedratories, radiology departments, and pharmacies. and maintained solely by individual consumersWith increased consumer awareness of the impor- as a trusted or valid source. Similarly, providerstance of privacy, confidentiality and information may deem the information stored by consumerssecurity, a USB device that is individually carried as “clinically irrelevant”, and if the informationand controlled can be an ideal solution (Ball et al., were excessive, it would be overwhelming for2007). One design characteristic of USB devices providers to review (Tang et al., 2006). If providersthat makes them desirable for information stor- decide that information contained in the PHR isage is durability, as the devices have no moving unimportant, the value of the PHR is adverselyparts with the casing protecting the components affected (Ball et al., 2007). Finally, providers(Axelson, 2006). Research of various USB ven- may be concerned with the legal issues involveddors provided information on features available, in the utilization of PHR data, that is, when theincluding continuously increasing storage capac- treatment decisions they made are based on in-ity, overwrite/modification prevention, display accurate or invalid patient-entered informationof remaining storage capacity, display of owner (Tang et al., 2006).56
  • 74. Personal Health Records Systems Go Mobile While utilization of USB devices for PHRs A number of U.S. companies are working together promotes individual’s engagement in their to develop “Dossia”, a Web-based framework to healthcare, the ability to interface and integrate assist employees and retirees to create and main- healthcare information over time is a priority that tain lifelong personal health records, of themselves an USB/stand-alone record cannot provide as the and their dependents. sole PHR source. Advantages PHRon WORLld WIDE WEB There are many advantages to having PHR on the Web for patients, physicians, employers, Personal health records on the Internet are a and pharmaceutical companies. “Technology growing phenomenon. These systems typically can allow the use of personal health records that consist of a patient’s personal health information patients themselves can maintain, can allow and on the Internet, entered by them and/or possibly promote telehealth systems, and can actually by their caregivers. Patients can record their enhance consumer choice” (Colorafi, 2006) Page personal information, demographic information, 3. Several of these sites “allow patients and phy- emergency contacts, insurance, medications, al- sicians to share patient-charted information and lergies, immunizations, tests, hospitalizations, diagnostic test results. The benefit to the physician surgeries, advance directive, spiritual affilia- and office staff is that it enhances the physician’s tion, and even their care plan (Colorafi, 2006). teaching efficiency and reduces communication There are several Web sites on the Internet pro- bottlenecks when the telephone is the sole com- moting personal health records. In 2001, if you munication tool” (Smithline Christenson, 2001). performed a general search on the Internet for Another push towards the Internet for physicians personal health records revealed over 19 sites, is the ease of use with their handheld devices. some of the examples found included Dr. I-Net, They are able to research medications, diseases, HealthCompass, MedicalEdge, MedicalRecord. treatments, as well as patient labs, test results, com, MedicData, Medscape, AboutMyHealth, and and even billing, coding, and dictation abilities. many more (Kim, Johnson, 2001), this number Electronic prescribing systems are also on the rise has grown significantly. The Web services begin and a great advantage to the physician and patient. with a registration process that involves the user These systems increase patient safety and physi- choosing a username and password. Through a cian efficiency (Smithline Christenson, 2001). Web interface, users then complete information There are many benefits to the use of the Internet about their (or a family member’s) health that is and personal health record. “The most important stored in a secure server maintained by the PHR benefit the Internet will bring to health care will company. Users can then access that information be the integration of information” (Smithline (and/or authorize access to others such as emer- Christenson, 2001). gency contacts, physicians, or ER departments) by logging-in and providing their password. Disadvantages At this point not too many patients are aware of personal health records. First, many of them 1. With all the positive aspects of PHR and are not computer savvy; or since some of the Web the Internet, some major disadvantages in- sites are subscription based. However, there are cludecost and training. In addition, with the some companies that are willing to help out their increased use of computers and the Internet, employees, and are encouraging the use of a PHR. 57heoidee
  • 75. Personal Health Records Systems Go Mobile there is an increase of technical issues and Advantages system downtime. 2. Training staff can be tricky, especially since 1. Patients can create and maintain compre- a large number of health care providers did hensive online PHR accounts via the mobile not grow up with the internet or e-mail. devices. 3. There will be a large cost for training staff 2. The mobile devices provide the capability in addition to hardware costs, software costs, to easily update and manage that PHR at implementation, maintenance, and produc- any time, and from any location. tivity improvements (Colorafi, 2006). 3. Easy and convenient to use for wellness and health monitoring. PHRS on Mo devi Disadvantages There is a new trend that is enabling patients to 1. If the device is lost and the health data is use their mobile devices to access details from stored on the mobile device, there could their PHR. Patients can use the mobile device for be serious security implications for the a variety of functions including: maintaining a patient. real-time health diary, and tracking vital health 2. The screen for the mobile device may not measures such as blood glucose levels, blood display all the information clearly due to pressure, carbohydrate intake, height, weight. This the size and those users who are not used approach can also be used to record and monitor to mobile displays may find it difficult to physical activity such as diet, calorie intake or navigate and access information. exercise. As people are becoming more reliant 3. Only limited information can be viewed, on mobile technology to organize and manage while images and notes may be difficult day to day routines, accessing and maintaining to understand. The mobile version is not a personal health information on wireless devices replacement for the Web portal, and the role is a natural progression. is to provide an interface to the portal. One approach taken by system developers is to provide functionality that will allow emer- Creating a Framework for Mobile gency details along with important data such as PHR Systems Evaluation immunization records, insurance details, and allergy information from a PHR portal to be The Mobile PHRS framework presented here downloaded to a secure module within a mobile has been inspired by the Personal Health Record devices operating system. Another approach is System Functional Model (PHR-S FM), a model to provide access to the PHR via Smartphones / proposed by the Health Level Seven (HL7), along mobile device using the Internet and the mobile with the Evaluation model proposed by (Kim network to view and update the records held in the K. Johnson, 2002). The approach by Kim (2002) PHR portal. The benefit of the latter approach is provides a comprehensive view of the PHR func- the ability to gain access to more detailed infor- tions, and identified five prospective functions mation such as clinical records, medical history of PHRs. The model outlined requirements for and interventions. accurate entry of information and verification of reported test and study results. The criteria also outlined requirements for the provision of different 58RileeCS
  • 76. Personal Health Records Systems Go Mobileroutes of access, links to consumer health care Portability of Records: The HL7 EHR Techni-information, functions to process and interpret cal Committee was created in 2005 by the PHRinformation, and functions to provide secure Working Group - the group has members fromcommunication between patients and providers. heathcare providers, consumers, vendors, andThis evaluation was constrained by Web based payers. The group recently announced that itcriteria and does not take mobility into consider- had entered into a memorandum of understand-ation. Another important element omitted from ing (MOU) with America’s Health Insurancethe Kim model is the data storage medium. Plans (AHIP) and the Blue Cross and Blue Shield HL7 is a premier health care information Association (BCBSA) to create a collaborativetechnology standards development organization process for the maintenance of portability stan-boasting an extensive national and international dards for PHRs. AHIP and the BCBSA haverepresentation. The main purpose of the PHR-S already developed an implementation guideFM is to define the set of functions that may be (Implementation Guide for the Personal Healthpresent in PHR systems. The PHR-S also presents Record Data Transfer Between Health Plans)a set of guidelines that “facilitate health infor- containing technical standards, a data dictionary,mation exchange among different PHR systems and operating rules for the transfer of PHR dataand between PHR and EHR (electronic health elements between health insurance plans. Underrecords) systems,” The HL7 group advocates that the MOU, AHIP and the BCBSA have agreed“The PHR-S FM can be applied to specific PHR to turn over the maintenance of the technicalmodels (stand-alone, Internet-based, provider- standards components of the Implementationbased, payer-based, or employer-based models). Guide to HL7 and ASC X12 to ensure long-termAt the same time, the Functional Model is flex- maintenance of the standards.ible enough to encourage product innovation.” With considerable research and investment intoThe mobile model presented here also takes into personal health records over the past decade, andconsideration the mobility aspects. The model is the growing number of publications and patentscurrently not an American National Standards registered in this areas, our review unveiled thatInstitute (ANSI)-accredited standard. The ANSI the main areas of PHR functionality or concernaccreditation process will take 2 years. This means can be grouped into four functional areas as il-that the PHR-S FM will become a U.S. standard lustrated in Figure 6 and Figure 7.for PHRs at around 2010. Once the PHR-S FM Personal health records are maturing rapidlyis finalized by HL7, it will ensure that standards and several diverse actors’ have fullscale imple-are available to the health care industry and the mentation of PHRs, - these groups include sub-general public for use in PHR development. scribers, employer groups such as DOSSIA, andThere is currently a profusion of PHR systems commercial vendors such as Google. There are fewin existence but there is a lack of a functional Web based systems that are fully integrated intostandard to which these systems should conform. ambulatory or hospital-based EMR systems.HL7’s PHR-S FM will be the first major indus- There are considerable challenges to imple-try standard to specify functionality for PHR ment the ideal PHR system, and there are im-systems. HL7 proclaims that the development portant lessons that can be learned from theof standardized, interoperable PHRs is a major early adopters. A study by Halmaka et al (2008)component in the U.S1. DHHS plans, which is to identified a set of unique implementation issuesmake health information available electronically and concerns from three case studies MyChart atthrough a National Health Information Network Palo Alto Medical Foundation, PatientSite at Beth(NHIN) by the year 2014. Israel Deaconess Medical Center, and Indivo at 59
  • 77. Personal Health Records Systems Go MobileFigure 6. Mobile personal health record functional overviewChildren’s Hospital Boston. They identified the • PHRs are institution-based and patients willfollowing implementation challenges from 1999 want a single PHR that works with all theirto 2007, postulating that further challenges are sites of care – how can this be achieved?likely to evolve over the next five years. • Should PHRs support electronic data input Current challenging questions facing imple- from outside institutions?mented PHR systems include: • How do you allow patients to integrate knowledge sources on the Internet with their• Should the entire problem list be shared? PHRs?• Should the entire medication list and allergy • How do you connect patients using social list be shared? networking tools? Patients with specific dis-• Should all laboratory and diagnostic test eases may want to connect to communities results be shared with the patient? of others with similar diagnoses• Should clinical notes be shared with the • Patients may wish to participate in clinical patient? trials, post market pharmaceutical vigilance,• How should patients be authenticated to or public health surveillance via their PHR access the PHR? – how is this possible without compromising• Should minors be able to have their own security? private PHR and should patients be able to • How do you securely incorporate the concept share access to their PHR via proxies? of mobility in a PHR system?• Should PHR include secure clinician/patient messaging? The next section will discuss the wireless and information management element of the evalua- Future challenging questions that may arise tion framework, which are important evaluationin 2008 and Beyond include: components of a mobile PHR.60
  • 78. Personal Health Records Systems Go Mobile Figure 7. Mobile personal health record functionality Mobile PHR Information 1. Interoperability: Interconnectivity among Infrastructure systems is important and managing relation- ships with various healthcare providers in There are no clear guarantees that the use of any a seamless and efficient manner along with IT applications in healthcare is going to be ef- providing user-friendly processes and inter- fective due to the technical complexity of Health faces to perform administrative functions IT systems. In the past healthcare software and are key features that must be considered in hardware markets were considered to be less the design of a PHR. mature than other Industries and for medical 2. Information management: Question about technologies (Chiasson Davidson, 2004). This how the data is to be stored, how often will notion is changing due to the development of new it be backed up, and what storage medium innovative software applications and availability is in place are key to successful information of hardware specifically targeted to the growing management. healthcare market. The key to the growth in this 3. Record security: A variety of options may area has been the launch of software that improves be available such as password protection, effectiveness by providing functionality to man- biometrics, and encryption, but the challenge age the administrative and support functions of is to fit the best mechanism to the purpose healthcare. and design of the PHRs. It is important that any PHRS system should 4. Audit capabilities: With growing security take into consideration Information Infrastructure and privacy concerns, measures must be in from the following Dimensions. place to provide detailed audit of access to the records. 61R
  • 79. Personal Health Records Systems Go Mobile It is vital that any MPHRS is evaluated on the transferring data need to provide secure intercon-audit capabilities. Clear and comprehensive audit nection capabilities between the host systems andpolicies must be defined that describe the use of the PHR database. The data must be protected inpatient medical records within the system. The terms of data integrity and patient privacy.policies should not only aim to protect the con- Security policies for personal health infor-fidentiality and integrity of data but also protect mation must be carefully designed in order tothe patient. One of the important features of the limit the number of people, clinical physicians,new generation of PHR systems is their ability to insurance companies, nurses, and others, thatinterconnect electronically using predefined inter- can access the patient record, and to control thefaces or XML based interfaces. All sub systems operations that may be applied to the record itself (Anderson 2006).Figure 8. Information infrastructure components62
  • 80. Personal Health Records Systems Go Mobile Jelena (2007) defines clear policies that are It is important to understand how the device will appropriate for wirless clinical information sys- connect to the Internet to access private health tems, these policies have been adapted to a PHR. information as some networks are more secure These policies are discussed below as a number than others. The ease of access that wireless net- of security procedures: works offer is matched by the security and privacy challenges presented by the networks. • Each record must have an associated ac- One of the key concerns surrounding the cess control list - a list that restrict access implementation of Mobile PHRS is the issue of to the records other than those individuals security. Moving a PHR into the mobile realm and groups identified on the access control compounds these security fears. There are four list. types of security breaches that can occur. • There must be an individual on the access control list that must have administrator 1. Data duplication: PHRs raise the possibil- privileges and/or rights, i.e., the power to ity of storing health data in multiple storage add other users to the access control list. locations. For example in the EMR, Hospitals It is critical that the administrator notifies databases, and the PHR. In a mobile scenario the patient of any changes of names on the this issue is compounded because the data access control list to any part of the patient may also exist on the mobile device. records. 2. Data transmission issues: Using open • An audit log of usage activity must be pre- unsecured networks such as wireless local sented to the user. Each time the record is area networks (wifi) to transmit personal accessed the following information must be health information will leave users open to presented - the name of the user performing security vulnerabilities. The system must the access, the date and time of access, and detect when using an unsecured network the manner of access (including records read, and prevent the release of information in updated, stored, and/or deleted) - and has that scenario. to be recorded. 3. Lost devices: If the device is lost, the data • When the patient is incapacitated, the own- server must prevent that device from access- ership of the records should belong with the ing medical information from that device. legal guardian or another person with the If the functionality is embedded within appropriate power of attorney, not the person the SIM card the device must prevent the with the patients’ mobile device. embedded functionality as well. 4. Virus and malware: There is a growing trend of Viruses being targeted at mobile WieleETWORK devices due to the proliferation of mobile Componen devices. The adoption of mobile technologies in healthcare is on the increase and technologies such as Wire- MoBIile PHR SYSTEemReview less Local Area Networks (WLAN) that use dif- ferent protocols from the standard digital mobile There are a variety of configurations that can be technologies such as 2G, 2.5 and 3G technologies. employed for a mobile PHR system. Each of the A summary of these technologies are presented approaches provides a variety of benefits along below along with the speeds and range covered. with potential security vulnerabilities. The section 63ewokRSy
  • 81. Personal Health Records Systems Go Mobile Figure 9. Wireless system components will use the evaluation framework defined in the The user has the capabilities to modify what previous section to review three commercial PHR pieces of their information will be accessible in systems that support mobile phone access. an emergency situation by using a Web applica- tion. The full medical record resides within the System 1: In Case of Emergency HealthVault eco-system and is transferred into the (IC) PHR Mobile icePHR. The icePHR emergency data are viewed by three methods. This PHR product is a combination of CapMed products, icePHR and Microsoft HealthVault. The 1. displayed through a personalized icePHR aim here is to make appropriate medical records Web site, freely available “In Case of Emergency (ICE)”. 2. wallet-size emergency card, 64SCR
  • 82. Personal Health Records Systems Go MobileTable 1. Wireless networks Networks Speed Range and Coverage Main Issues for M-Health 2nd Generation GSM 9.6 kilobits per second (KBPS) World wide coverage, Bandwidth limitation, Interference. dependent on network operators roaming agreements. High Speed Circuit Between 28.8 KBPS and 57.6 KBPS. Not global, only supported by Not widely available, scarcity of Switched Data (HSCSD) service providers network. devices. General Packet Radio 171.2 KBPS Not global, only supported by Not widely available. Service (GPRS) service providers network. EDGE 384 KBPS Not global, only supported by Not widely available, scarcity of service providers network. devices UMTS 144 KBPS - 2 MBPS depending on When fully implemented Device battery life, operational mobility should offer interoperability costs. between networks, global coverage. Wireless Local Area 54 MBPS 30–50 m indoors and 100–500 Privacy, security. m outdoors. Must be in the vicinity of hot spot. Personal Area Networks 400 KBPS symmetrically 10 – 100m Privacy, security, low bandwidth. – Bluetooth 150 -700 KBPS asymmetrically Personal Area Networks 20 kb/s – 250 KBPS depending on 30m Security, privacy, low bandwidth. – Zigbee band WiMAX Up to 70MBPS Approx. 40m from base Currently no devices and networks station. cards. RFID 100 1m Security, privacy. KBPS Non line-of-sight and contact less transfer of data between a tag and Reader. Satellite Networks 400 to 512 KBPS new satellites have Global coverage. Data costs, shortage of devices with potential of 155MBPS. roaming capabilities. Bandwidth limitations.3. Mobile client side wallet. application must be hosted on the server side of the provider of the Vault Server software. The user The Wallet is the client-side portion of the has the capability to view, submit or send, via fax,application and resides on the mobile phone or email or SMS, the user’s selected data.personal computer. The personal health informa- icePHR Mobile is a mobile device accessibletion located externally or privilege information software application that provides the capabilityis accessed right to the wallet. The mobile device to store and manage emergency medical infor-would need to receive a software installation which mation and contact numbers in a mobile phone.can be delivered over-the-air. Vault Server: the This product requires that the user subscribes toWallet communicates with the MobiSecure Vault icePHR subscription. Mobile devices must meetServer for data synchronization and management. the requirements of installing and running theThe role of Server-based software is to securely custom icePHR Mobile application to allow thehost, manage and retrieve personal user data from data to be stored on the phone negating the needexternal data sources. The Personal Health record for a data link connection. The application will 65
  • 83. Personal Health Records Systems Go Mobile allow subscribers the editing capability to insert, application where medical data are stored on the update and delete medical information directly device, using the WAP access method no infor- on the icePHR server when Internet access is mation is stored on the device nor can the user available. edit the records over WAP.      The icePHR application is only supported on specified networks and on a number of certified System 2: No More Clipboard handsets, which includes smartphone, blackber- M-PHR ries and standard mobile phone handsets that run java Midlet. NoMoreClipboard.com is an online, patient-con- The phones that do not incorporate the tech- trolled personal health record management system nology to install and run the icePHR Mobile™ designed to consolidate medical information application can use the WAP to view medical in one convenient and secure location for easy information stored on the server. To use WAP, the retrieval and updates. NoMoreClipboard.com handset must have a built in Web browser such provides Web-based solutions to maintain an as Media Net. Unlike the Java icePHR Mobile online personal health record (PHR). The system Table 2. Evaluting three commercial MPHR Systems Criteria System 1 System 2 System 3 System Name ICE PHR No More Clipboard allOne Mobile System Description software application that Mobile Web based patient- AllOne Mobile cell phone-based provides the capability to store controlled personal health record application. to manage personal and manage emergency medical management system designed to health information. Patient information and contact numbers consolidate medical information information the a mobile phone. Network 2G or 3G Cellular requires data Access via mobile Internet Access via mobile Internet plan and SMS Plan browser using WiFi browser using WiFi Device - OS RIM 3.6, J2ME 1.1, Windows Any device with Internet access Blackberry RIM Java, Windows mobile for Pocket PC and Smart mobile for Pocket PC and Smart phone 2003+ phone 2003+ Devise - Type Mobile Phone (singular, sprint Any mobile device with Internet Any mobile device with Internet devices), Smartphone or PC capabilities capabilities Information Application is downloaded and Application is run from the Application is downloaded and management stored on the device server. The user accesses the stored on the device information via navigating to the home page on the mobile device Data Storage Stored on Device. (There is also All the data is stored on the Data is stored on both the server an option to store data on server server and device for non supported handsets PHR Capabilities Provided by ICEPHR and Provided by Nomoreclipboard. Integrates with online PHRs Microsoft Health Vault com. Can integrate with via community of Care (CCR) Microsoft health vault standards Business Model Add on to the ICE PHR Subscription only Subscription only subscription Access technology Uses WAP for non supported Via mobile browser. Future Via mobile browser. devices enhancement will allow cellphone technology to access sever 66SCR
  • 84. Personal Health Records Systems Go Mobile also provides cell phone integration capabilities. accesses the patients PHR stored securely on the Patients with Internet-enabled Smartphones such Internet. The PHR system allows updates from as the Palm Treo or Apple iPhone can interact the mobile device. The application supports all with a PHR if they have an account with No- smartphones and the majority of non-business MoreClipboard. This approach requires the user mobile consumer devices. The AllOne Mobile to be connected to the Internet as no data is down- application supports the following features. loaded to the device. To gain access to the Web portal from a mobile device the users points their • Storage of confidential personal informa- Smartphone browser to the PHR portal address tion, including provider and insurance and a version of the PHR application optimized information, allergies, immunizations, and for mobile devices is presented. medications Future functionality will include the capability • Synchronize their mobile device with Web to send reminders, receive appointment alerts, based PHR. medications reminders, or follow care plan direc- • Fax PHR information from a Mobile de- tives. The Patients will also have the capability to vice capture pain levels or illness symptoms as they • Control access to receipt of relevant and occur and insert this information directly into the timely communications on health care–re- PHR. Patients will also be able to capture and store lated topics images from their phone into their PHR. From a security perspective the site uses Secure Socket AllOne Mobile integrates with existing health Layer (SSL) account security. Although the infor- care information systems and applications, includ- mation is secured behind a firewall, accessing the ing existing online PHRs. information over non-secured network can lead to security vulnerabilities. An application under development that uses Binary Runtime Environ- Conlu ment for Wireless (BREW)will allow patients with standard Code division multiple access (CDMA) The goal of the MPHR system is to provide based wireless devices to access their PHRs. secure and controlled access to personal health A Short Messaging Service (SMS) interface is informationat anytime via a mobile device to also being developed to enable communications improve health outcomes. Mobile PHRs can between patients and the PHR system. provide patients with a variety of functions in- cluding: maintaining a real-time health diary, System 3: AllOne Mobile MPHR and tracking of vital health measures such as blood glucose levels, blood pressure, carbohydrate AllOne Mobile cell phone-based application was intake, height, weight. created by the AllOne Health Group Inc. to manage This chapter presented a framework adapted personal health information. Patient information from the HL7 PHR functional model and PHR is stored behind a password-protected, encrypted literature.The purpose of the framework is to channel using Diversinet wireless security appli- define the set of functions that may be present in cation. AllOne Mobile uses mobile technology to Mobile PHR systems and to highlight important facilitate the exchange of critical health informa- components that must be taken into consideration tion between individuals, providers, and health when evaluating Mobile PHR systems. plans. This application downloads wirelessly a The growing number of large technology small application to the mobile device, which organizations entering the PHR landscape will 67SRion
  • 85. Personal Health Records Systems Go Mobile eventually help the industry by driving standards Informatics, 76(5-6), 471-479. Retrieved from forward, developing open interfaces and generat- http://www.scopus.com/scopus/inward/record. ing awareness of the products and implementing url?eid=2-s2.0-33947621113partnerID=40re solutions that incorporate mobility on the one l=R7.0.0. hand, and satisfying standards requirements and Alban, R., Feldmar, D., Gabbay, J., Lefor, A. regulatory policies such as HIPAA privacy and (2005). Internet security and privacy protection security rulings on the other hand. for the health care professional. Current Surgery, The key challenge that is likely to evolve 62(1), 106-110. Retrieved from http://www.scopus. from the influx of non healthcare organization com/scopus/inward/record.url?eid=2-s2.0-13844 delivering PHR systems runs parallel with Elec- 266517partnerID=40rel=R7.0.0. tronic Medical Records (EMR) and EHRs prior to HIPPA rules and regulations, enacted by the USA Albright, B. (2007). Prepping for PHRs. The congress in 1996. These problems included the growing trend of consumer empowerment in- denial of health care coverage to individuals with cludes the speedy rise of personal health records. chronic and genetic predispositions to diseases Healthcare informatics: the business magazine for and the release of personal health information. information and communication systems, 24(2). Ultimately consumers must exercise extreme Retrieved from http://www.scopus.com/scopus/ caution when utilizing and implementing a mobile inward/record.url?eid=2-s2.0-34247121143par or Web based PHR. tnerID=40rel=R7.0.0. Alhaqbani, B., Fidge, C. (2008). Access control requirements for processing electronic health re- Refeen cords, 4928, 371-382. Retrieved from http://www. scopus.com/scopus/inward/record.url?eid=2- Abidi, S., Goh, A. (2000). A personalised s2.0-40549129015partnerID=40rel=R7.0.0. Healthcare Information Delivery System: push- ing customised healthcare information over the Al-Salqan, Y. Y., Jagannathan, V., Davis, T., WWW. Studies in health technology and infor- Zhang, N., Reddy, Y. (1995). Security and con- matics, 77, 663-667. Retrieved from http://www. fidentiality in health care informatics. In Proceed- scopus.com/scopus/inward/record.url?eid=2- ings of the ACM Workshop on Role-Based Access s2.0-0034574134partnerID=40rel=R7.0.0. Control (pp. 47-51). Retrieved from http://www. scopus.com/scopus/inward/record.url?eid=2- Ackerman, M. (2007). The personal health record. s2.0-0029427930partnerID=40rel=R7.0.0. The Journal of medical practice management: MPM, 23(2), 84-85. Retrieved from http://www. Anderson, R. J. (1996). Security policy model scopus.com/scopus/inward/record.url?eid=2- for clinical information systems. In Proceedings s2.0-38449122521partnerID=40rel=R7.0.0. of the IEEE Computer Society Symposium on Research in Security and Privacy (pp. 30-43). Adler, K. (2006). Web portals in primary care: An Retrieved from http://www.scopus.com/scopus/ evaluation of patient readiness and willingness inward/record.url?eid=2-s2.0-0029697680part to pay for online services. Journal of Medical nerID=40rel=R7.0.0. Internet Research, 8(4). Arbogast, J., Dodrill, W. (1984). Health main- Agrawal, R., Johnson, C. (2007). Securing elec- tenance and the personal computer. Journal of tronic health records without impeding the flow Family Practice, 18(6), 947-950. Retrieved from of information. International Journal of Medical http://www.scopus.com/scopus/inward/record. 68e
  • 86. Personal Health Records Systems Go Mobileurl?eid=2-s2.0-0021234056partnerID=40re Denton, I. (2001). Will patients use electronicl=R7.0.0. personal health records? Responses from a real- life experience. Journal of healthcare informationArmitage, J., Souhami, R., Friedman, L., Hil- management: JHIM, 15(3), 251-259. Retrievedbrich, L., Holland, J., Muhlbaier, L., et al. (2008). from http://www.scopus.com/scopus/inward/The impact of privacy and confidentiality laws record.url?eid=2-s2.0-0035464326partnerID=on the conduct of clinical trials. Clinical Trials, 40rel=R7.0.0.5(1), 70-74. Retrieved from http://www.scopus.com/scopus/inward/record.url?eid=2-s2.0-4094 Gellman, R. (2008). Personal Health Records9122714partnerID=40rel=R7.0.0. and Personal Health Record Systems. The World Privacy Forum. Retrieved March 1, 2008,Ball, M., Gold, J. (2006). Banking on health: from http://www.worldprivacyforum.Personal records and information exchange. org/pdf/WPF_PHR_02_20_2008fs.pdfJournal of healthcare information management:JHIM, 20(2), 71-83. Retrieved from http://www. Halamka, J., Mandl K., Tang, C. (2007).scopus.com/scopus/inward/record.url?eid=2- Early Experiences with Personal Health Recordss2.0-33744500251partnerID=40rel=R7.0.0. Journal of the American Medical Informatics Association, 15(1), 1-7.Bent, P., Bolsin, S., Creati, B., Patrick, A., Colson, M. (2002). Professional monitoring and Jelena, M., Vojislav, B. M. (2007). Implementa-critical incident reporting using personal digital tion of security policy for clinical informationassistants. Medical Journal of Australia, 177(9), systems over wireless sensor networks Ad Hoc496-499. Retrieved from http://www.scopus. Networks, 5, 134–144.com/scopus/inward/record.url?eid=2-s2.0-0037 Kim, M., Johnson, K. (2002). Personal health020972partnerID=40rel=R7.0.0. records: Evaluation of functionality and utility.Bhuyan, K. (2004). Health promotion through self- Journal of the American Medical Informat-care and community participation: Elements of a ics Association, 9(2), 171-180. Retrieved fromproposed programme in the developing countries. http://www.scopus.com/scopus/inward/record.BMC Public Health, 4, 1-31. url?eid=2-s2.0-0036491265partnerID=40re l=R7.0.0.Blobel, B., Pharow, P., Spiegel, V., Engel, K., Engelbrecht, R. (2001). Securing interoperability Kupchunas, W. (2007). Personal health record:between chip card based medical information New opportunity for patient education. Ortho-systems and health networks. International paedic Nursing, 26(3), 185-191. Retrieved fromJournal of Medical Informatics, 64(2-3), 401-415. http://www.scopus.com/scopus/inward/record.Retrieved from http://www.scopus.com/scopus/ url?eid=2-s2.0-34249788684partnerID=40rinward/record.url?eid=2-s2.0-0035188372part el=R7.0.0.nerID=40rel=R7.0.0. Lee, M., Delaney, C., Moorhead, S. (2007).Chiasson, M. W., Davidson, E. (2004). Push- Building a personal health record from a nurs-ing the contextual envelope: developing and dif- ing perspective. International Journal of Medi-fusing IS theory for health information systems cal Informatics, 76(SUPPL. 2). Retrieved fromresearch. Information and Organization, 14(3), http://www.scopus.com/scopus/inward/record.155-188. Retrieved from http://www.sciencedi- url?eid=2-s2.0-34548216363partnerID=40rrect.com/science/article/B6W7M-4C40PF5-1/1/ el=R7.0.0.bf7efac96eb75891f0f22febc1672ffd. 69
  • 87. Personal Health Records Systems Go Mobile NoMoreClipboard.com (2008). PHR Launches scopus.com/scopus/inward/record.url?eid=2- Cell Phone Integration. Business Wire. Oct 29, s2.0-39049183450partnerID=40rel=R7.0.0. 2007. FindArticles.com. 25 Jun. 2008. http:// Tang, P., Ash, J., Bates, D., Overhage, J., Sands, findarticles.com/p/articles/mi_m0EIN/is_2007_ D. (2006). Personal health records: Definitions, Oct_29/ai_n21068069 benefits, and strategies for overcoming barriers Okawa, T. (1973) A personal health record for to adoption. Journal of the American Medical young female students. Japanese Journal for Informatics Association, 13(2), 121-126. Retrieved Midwife, 27(11), 36-40. from http://www.scopus.com/scopus/inward/re- cord.url?eid=2-s2.0-33644682163partnerID=4 Sittig, D. (2002). Personal health records on the 0rel=R7.0.0. internet: A snapshot of the pioneers at the end of the 20th Century. International Journal of Medi- Thielst, C. B. (2007). The New Frontier of Elec- cal Informatics, 65(1), 1-6. tronic, Personal, and Virtual Health Records. Jour- nal of Healthcare Management, 52(2), 75-78. Sprague, L. (2006). Personal health records: the people’s choice? NHPF issue brief / National Health Policy Forum, George Washington Uni- Endno versity, (820), 1-13. Retrieved from http://www. 1 www.hl7.org/ehr. 70e
  • 88. 71 Chapter IV Medical Information Representation Framework for Mobile Healthcare Ing Widya Jacqueline Wijsman University of Twente, The Netherlands University of Twente, The Netherlands HaiLiang Mei Hermie J. Hermens University of Twente, The Netherlands University of Twente, The Netherlands Bert-Jan van Beijnum University of Twente, The Netherlands ABSTRACT In mobile healthcare, medical information are often expressed in different formats due to the local poli- cies and regulations and the heterogeneity of the applications, systems, and the adopted Information and communication technology. This chapter describes a framework which enables medical information, in particular clinical vital signs and professional annotations, be processed, exchanged, stored and managed modularly and flexibly in a mobile, distributed and heterogeneous environment despite the diversity of the formats used to represent the information. To deal with medical information represented in multiple formats the authors adopt techniques and constructs similar to the ones used on the Internet, in particular, the authors are inspired by the constructs used in multi-media e-mail and audio-visual data streaming standards. They additionally make a distinction of the syntax for data transfer and store from the syntax for expressing medical domain concepts. In this way, they separate the concerns of what to process, exchange and store from how the information can be encoded or transcoded for transfer over the internet. The authors use an object oriented information model to express the domain concepts and their relations while briefly illustrate how framework tools can be used to encode vital sign data for exchange and store in a distributed and heterogeneous environment. Copyright © 2009, IGI Global, distributing in print or electronic forms without written permission of IGI Global is prohibited.a
  • 89. Medical Information Representation Framework for Mobile HealthcareInoduion with healthcare data which are represented in multiple format standards due to the differentMobile healthcare applications receive more policy or regulations and the heterogeneity ofand more attention due to the ability to reshape applications, systems and ICT technology.healthcare delivery, for example, enabling self- This chapter describes a framework whichmanagement of patients whilst they pursue their enables healthcare data, in particular (digitized)daily activity. Information and communication continuous-time patient’s vital signs and profes-(ICT) technology and infrastructures which pro- sional annotations, be processed, exchanged,vide the necessary ubiquitous connectivity enable stored and managed modularly and flexibly in athese applications. Competitive value-add ICT mobile, distributed and heterogeneous environ-providers moreover facilitate these applications ment. A framework is often described as a basicwith alternatives to computation and communica- conceptual structure to compose somethingtion services. Today’s environment for networked from fitting parts. In the context of this chapter,applications is therefore rich in ICT services a framework is an integrative (standardized) con-which are accessible anywhere and anytime, for ceptual structure which brings together a set ofexample by prepaid or subscription contracts components which themselves may be standardsbetween users and ICT service providers or by such as vital signs format encoding standardscollaboration contracts between these providers. (Blair Stefani, 1998). It therefore addressesSuch environment enables applications to select questions like:(wireless) connections of required quality andtechnology which are considered best for their • How to deal with healthcare data expressedpurpose. A mobile application may for instance in accordance with several data formatseamlessly switch over between GSM, UMTS or standards and how to encode the data to fitWiFi 802.11 (Schiller, 2003) connections that are to the characteristics of the provided con-offered by competing providers. These develop- nections to enable effective and efficientments enable mobile healthcare applications in data transfers;choosing the appropriate situations with adequate • How to deal with professional (textual,ICT support that permit healthcare to be delivered graphical or multimodal) annotations andwhere previously it was difficult or impossible to derived (i.e. trend) signs in sync with thedo so (Wootton, 2006). analyzed vital sign segments; Due to these ICT and business advancements, • How to manage vital sign data sets of aa travelling patient with a chronic disorder can patient that originate from the same measure-be monitored continuously everywhere in the ment session in a (distributed) study, whichcountry of residence as well as abroad. If his typically process data in several steps usinghealth condition requires, he may be examined processing tools with specific parameter set-at a care centre abroad that uses equipment dif- tings. Similarly, how to manage vital signferent than at his country of residence. This may data sets (of the same patient and the samefurther imply that the format of the processed measurement session) in different formats,healthcare data differs from the format used at e.g. if the returning traveling patient, whohis residential care centre. Local care centre’s has been monitored and diagnosed in a carepolicy or local governmental health regulations centre abroad, consults his general practi-may also impose the use of a different healthcare tioner, who then inspects the annotationsdata format standard. In (near) future mobile and the vital signs measured and processedhealthcare therefore, we typically need to deal using a locally certified system to confirm72
  • 90. Medical Information Representation Framework for Mobile Healthcare the annotations, the diagnosis and treatment the transfers of sequences of bits or characters, of his colleague abroad. respectively). For example, an echocardiogram needs to be formatted as a sequence of pictures, The proposed framework should furthermore serialized, and encoded further to suit the pipe, fit to the practices used in ICT to manage the use transferred via the pipe, and at the receiving end of multiple format and encoding standards, as reconstructed (i.e. decoded). This chain of data discussed in the next sections. formatting and encoding steps requires a suitable In the next section, we discuss some of the end-to-end quality to preserve the clinical inter- issues of information exchange using computer pretations of the echocardiogram. Mechanisms networks and illustrate the need for a framework and techniques for data formatting and encoding which flexibly supports exchange of healthcare have been widely investigated and developed in data, in particular digitized continuous-time vital the area of computer networking. In this chapter, signs and professional annotations, in a distributed we present the data representation model of the and heterogeneous environment. Thereafter, we Open Systems Interconnections (OSI) of Inter- analyze the functional requirements of mobile national Organization for Standardization (ISO) healthcare stakeholders on the framework. We (MacKinnon, 1990). This model provides clarity address only those stakeholders that influence the to and better understanding in the structures of functional aspects of a framework for multiple many format and encoding standards like MPEG, formatted vital signs for use in a heterogeneous JPEG, H.261, or DICOM (Le Gall, 1991; NEMA, distributed environment. Stakeholders addressing 2007a). This is due to the distinction between financial aspects like insurance companies are abstract syntax representation, which is suitable therefore beyond the scope of this chapter. In the for the entities that exchange information, and section thereafter, we address the representational transfer syntax representation, which is suitable model, which distinguishes between the syntax for the pipe that transfers the serialized and en- for data transfer and store from the syntax for coded data. This distinction therefore separates the expressing medical domain concepts. Then, we concerns of exchanging concepts of the domain discuss the information model of the framework ontology from the concerns of serializing and and some ECG standards. Thereafter, we address transferring the encoded concepts in a meaning some other syntax notations and briefly discuss preserving way. tool based translations of conceptual or abstract As in multimedia, several formats and encod- syntax to transfer syntax. The last section presents ings have been proposed or developed for vital our conclusions. signs, in particular for electrocardiograms (ECGs). We may identify de-jure standards developed by standardization bodies, such as the CEN/SCP- Ba ECG (CEN/TC251 prEN 1064, 2002), which is developed by the European Committee for Stan- One of the issues of transferring information in dardization CEN (CEN/TC251, 2007) and defined an ICT environment is to preserve meaning de- specifically for ECGs, or HL7 (Hinchley, 2005), spites the dynamic property of the data transfer which is developed by an organization cooperat- characteristics of the connections and the different ing with standardization bodies and accredited ways of representing information at the computer by the accrediting organization for US national systems at the connection endpoints. A connec- standards, but which has a larger scope than only tion in this environment can be modeled by a bit addressing monitored healthcare data like ECGs. or a character pipe (i.e. a model which supports Another example of a de-jure standard that can be 73kgound
  • 91. Medical Information Representation Framework for Mobile Healthcareused to represent ECGs, or vital signs in general, also used by other standards, for example HL7,is VITAL (Weigand, 2005). We may also iden- to capture the association semantics betweentify de-facto standards, i.e. standards that were the healthcare domain concepts. In this chapter,developed by industrial or research consortia, or however, we specify the vital signs informationproprietary standards used by vendors of medical model from the perspective of the different formatequipment or proposed by a research institute, e.g. and encoding standards. This approach fits to ourecgML (Wang, 2003). For our convenience, we objective to develop a framework for processing,denote ECG data format and encoding proposals transferring and storing vital signs in a multiplefound in the literature as (proprietary) standards. formats environment. Our information modelECG data representation standards vary in their se- therefore includes multiple structures for (repli-mantic expression levels. Some of these standards cated) ECG data that are specified by the differentfocus only on the waveform representation, some standards and it includes structures to express theirothers additionally provide heart physiological or relations, for example the applied conversion tools,bioelectrical domain concepts like the notion of the settings of the tool and the actor in charge ofP or ST waves. These differences may imply loss the conversion, or the processing algorithm andof interpretation power when ECG data has to be settings that derive a trend sign.converted from one onto another standard (lossy To deal with the exchange of medical informa-conversion). In this chapter, we show how the OSI tion represented in multiple formats we adopt simi-data representation model (MacKinnon, 1990), in lar techniques and constructs as are used on theparticular the abstract syntax, can be used to iden- internet. In particular, we are inspired by MIMEtify these differences in semantic expression level (Multipurpose Internet Mail Extensions) (Freed,and how to associate equivalent ECG segments 1996) which enables users to exchange text, pic-formatted in different standards. We also discuss tures, video clips, excel sheets, etc. independentlyhow the abstract syntax can be used to specify of the computing devices, software packages orprofessional annotations or derived (/trend) signs the operating systems involved. For example,like heart-beats such that rendering tools are able the MIME construct “multipart/alternative” canto visualize these annotations or trend signs in be used to express the relation between two orsync with the associated data segment. more ECG segments of the same measurements We apply the Unified Modeling Language of a patient but formatted differently, e.g. one in(UML) (Booch, 1999) as a (graphical) abstract the CEN/SCP-ECG format and the other in thesyntax language to express the concepts of vital DICOM waveform format (NEMA, 2007b). Thesigns, in particular ECGs; this results in an in- latter can be a conversion of the first to match theformation model of the framework. Some of the format of the data to the software or the equip-ECG standards format ECGs as sampled time- ment of the professional, for example in the earlierdomain bio-signals (e.g. the format described in illustrated case of the travelling patients. As this(Browns, 2002)), others include bio-electrical or construct specifies that the multiple parts areheart physiological concepts like the notion of alternatives of one another, an ECG viewer toolP-waves and QRS complexes (e.g. the standard can select the part that is encoded in a preferreddescribed in (CEN/TC251 prEN 1064, 2002)). The format as indicated by a profile of preferences.specification of ECGs using UML, addressed in Moreover, a policy that regulates tools to ignore(Concalves, 2007), has elaborated several ECG parts that are encoded in a format unknown toontological models from different perspectives the tool provides flexibility when introducinglike the heart physiological, bio-electrical, includ- new formats without influencing existing systemsing the recording session perspectives. UML is (upwards compatibility and open-endedness with respect to new features or new functionality).74
  • 92. Medical Information Representation Framework for Mobile Healthcare Furthermore, to enable synchronization of care Extensible Markup Language - XML (Bray, 2004) professional’s (textual or graphical) annotations constructs for vital signs representations. These with segments of analyzed vital signs, we adopt constructs were discussed in (Mei, 2006) and a construct similar to MIME “multipart/parallel” several simplified scenarios were used in (Mei, to inform a rendering tool that the annotations 2006) to illustrate the benefits of the framework could better be visualized together with the cor- which accommodates these constructs. responding vital sign segments. We distinguish three mobile healthcare stake- We additionally adopt a similar technique as holders who typically influence the vital sign is applied in MPEG (Le Gall, 1991) for joining representations and their use (Figure 1): and splitting types of media, e.g. synchronized under-titles with video, to merge professional • End-users; stakeholders who use the services annotations, trend signs or other auxiliary data provided by the mobile healthcare systems. on the fly. As in MPEG, the framework includes End-users include both patients and the identifiers to distinct between the data types at healthcare professionals, for example the abstract as well as at transfer syntax level. medical specialists, nurses, physiothera- Besides the discussed facilities for healthcare pists; data processing, transfer and store, this chapter • Mobile healthcare system providers; stake- also addresses the facilities to manage the dynam- holders who are involved in the provisioning ics experienced by mobile healthcare applications of mobile healthcare systems for clinical due to changes in patient’s health conditions or remote monitoring and treatment. In the fluctuations of the ICT infrastructural resources context of this chapter, these providers are due to environment data traffic or roaming pa- assumed to be aware of the applied informa- tients. tion and communication technologies; A framework for multiple formatted vital signs • Care centers, such as the primary care therefore needs to adopt the discussed techniques centers, healthcare call centers (also called or constructs. In the next section, we justify these healthcare portals), and the secondary care needs by analyzing the requirements of healthcare centre’s (e.g. corporate hospitals with their stakeholders that are relevant for the framework’s departments of different specialties). For our functionality. convenience, regulatory bodies as well as medical ethical committees are categorized as this stakeholder. That is, the care centers SakeholdeConain are assumed aware of the healthcare regu- and RReemen lations that influence the way of handling patient’s vital signs. We analyze the needs of three mobile healthcare stakeholders to identify the functional needs that Requirements from End-Users have to be accommodated by the framework. For this analysis we use our experiences collected From the healthcare professional’s point of view, during several mobile healthcare projects (Mo- the vital sign representation should be suitable biHealth, 2002; HealthService24, 2005; Myotel, for effective clinical interpretation as required 2008) and our study of several healthcare systems by the health condition of the patient and in ac- reported in the literature. Some of the identified cordance with the working practices of these needs were examined during the development of professionals: 75uiR
  • 93. Medical Information Representation Framework for Mobile Healthcare Figure 1. Stakeholders of vital sign representation framework • Healthcare professionals typically access In some mobile healthcare applications, pa- units of interpretable segments of vital signs tients typically generate vital signs by attaching in a quality appropriate for the purpose of sensors on their body and initializing the sensing the clinical task, e.g. patient’s ECG filtered devices. These patients, especially mobile patients, from noise and movement artifacts and vi- may need to check and calibrate the sensors’ read- sualized in a resolution necessary to inspect ings from time to time to ensure accurate (local) ventricular contraction; monitoring and treatment feedback. For example, • Healthcare professionals may need to patients may need to re-attach sensors in case of correlate signs that belong to a group of bad skin contacts. For this, vital signs visualization coherent vital signs, e.g. patient’s oxygen or other feedback modality has to have a resolution saturation, heart beat, blood pressure, and suitable for patient’s interpretation. respiration that together form an indicator Moreover, medical and sensor technologies are of the oxygenation of the patient’s brain in evolving and new vital signs or sensors may be trauma care; developed for measuring patient’s health condi- • Healthcare professionals may have priorities tion in mobile environments. Therefore, vital sign regarding the importance of vital signs, e.g. representations should be extensible to enable the doctors may prefer to see trend signs and introduction of new vital signs or the integration only in case of abnormalities, they need the with new data like professional annotations. underlying vital signs; • Healthcare professionals may need to an- Requirements from System notate vital sign segments; Providers • Healthcare professionals may need to know how vital sign data was measured and pro- Mobile healthcare system providers have the cessed for evidence based treatment. mission to facilitate the computation and com- munication needs of the patient’s care process. 76RS
  • 94. Medical Information Representation Framework for Mobile Healthcare In a remote monitoring and supervised treatment InfoRMmaRepenion session, the healthcare system regularly matches Model the computation and communication needs of the supported care process with the resource capabil- A model suitable for information transfer in a het- ity and capacity of the ICT infrastructure. These erogeneous environment, which accommodates systems often apply a hunting strategy to collect different (wireless) communication technologies the available ICT resources of the contracted ICT and qualities and different computer systems, is providers. They often apply an adaptation strat- the OSI Presentation Layer model (MacKinnon, egy to control the vital sign data processing and 1990) (Figure 2). This model uses three kinds of transmission. For example, by down-sampling, syntaxes to represent information. The earlier prioritizing or discarding some of the vital sign described abstract syntax represents the domain packets, a system may improve the utility of trans- ontological structure of the information in respect ferring vital signs in a meaning preserving and of the entities exchanging the information. It is adequate way. Therefore, a vital sign framework therefore the vocabulary and the structuring rules should enable prioritized transfers of important used to represent the information. This syntax signs and deferred transfers of remaining signs, is considered useful in a meaningful meaning which may traverse other delivery routes and at preserving transfer, in which the sending and the cheap data communication hours. Consequently, receiving entities share a common universe of the framework should further support aggrega- discourse. An abstract syntax enables these enti- tion and resynchronization to reconstruct the set ties to interpret the exchanged information in the of vital signs. same way. The earlier described transfer syntax is the syntax used to represent data in transfer. Rquirements from Care Centers Information expressed in a transfer syntax is therefore represented as sequential groups of bits Care centers, especially corporate hospitals, or characters sequences. Groups, in turn, associate often accommodate a diversity of specialized to terms of the abstract syntax vocabulary. The systems, each of which may apply specific vital third kind is the local syntax which is the syntax sign formats. If furthermore, these centers also used to represent stored data at the involved com- treat travelling patients, interoperability between puter systems. In a heterogeneous environment, these remote systems needs to be supported. In the local syntaxes used by the communicating such cross-platform environments, vital sign computer systems can be different, e.g. one uses representations require an open environment to a Java based local syntax and the other a C based facilitate multiple vital signs formats. syntax in a Unix system. Healthcare data is considered private and has An abstract syntax is therefore not a concrete to be subjected to privacy rules. Monitoring and syntax as are transfer and the local syntaxes. treatment protocols described in the trial designs ECGs specified from a specific perspective in which were proposed to the Medical Ethical an abstract syntax result in a conceptual model Committees in the earlier mentioned healthcare of the ECGs. This model can be used to reason projects address healthcare data privacy, such about the elements of the ECG, for instance the as password protected and role based access to bio electrical properties of the heart or the heart recorded and processed data, vital signs are also condition if the model is defined from those per- made anonymous. The framework should enable spectives. In the perspective of interoperability transferring, processing and storing of vital signs in an environment that uses multiple standards, subjected to privacy rules. the ECG model at abstract syntax level should en- 77Cionea
  • 95. Medical Information Representation Framework for Mobile Healthcare Figure 2. Information representational model able the identification of the same ECG segments efficient. In an e-mail application, a plain text which are formatted using different standards message can be encoded amongst others as an and should further enable conversion from one ASCII characters sequence or a base64 character format to another. sequence (Freed, 1996). Base64 encodes 6 bits of An abstract syntax moreover enables the de- the abstract syntax representation to one base64 velopment of information exchange techniques transfer syntax character. Three (8 bits) characters and mechanisms for a heterogeneous environ- of the plain text message will therefore be encoded ment. Information conceptually represented in to 4 base64 characters. However, binary data can an abstract syntax can be encoded to different be encoded using the base64 encoding to fit to transfer syntaxes. Information encoding from a character pipe as used by internet e-mail. The abstract syntax to a transfer syntax is virtual, benefit of binary data encoded in base64 is the because in reality the information is represented availability of many internet protocols to convey at a computer system in a local syntax. Informa- the data using computer networks. On the other tion encoding in reality is therefore the conver- hand, conversions of digitized ECGs to base64 sion from a local syntax to a transfer syntax. The and back to a digital form at the receiving end rules needed for the conversion can be derived point consume processing capacity and a lot of from the encoding rules from abstract syntax to time, a bit oriented transfer syntax is much more transfer syntax. efficient in such cases. As mentioned earlier, information represented in an abstract syntax can be encoded in several transfer syntaxes, each of them binary or charac- Vi-SIGNS INFORMmaion ter sequence oriented. Moreover, some transfer Model syntaxes are more suitable for efficient process- ing rather then generating compact codes; others In this section, we discuss the information model generate compact codes but are not processing of the conceptual structure that binds together the 78alSignnfo
  • 96. Medical Information Representation Framework for Mobile Healthcareabstract syntax level structures of vital signs, in In general, a many to many, many to one or a oneparticular ECGs, as defined in the various vital to many association may exists, for example in thesign standards. In particular, the model specifies case of multiple vital signs types or in the case thatthe different kinds of relations between the vital the abstract syntax of the source vital sign segmentsign structures as identified in the stakeholder’s standard is much richer than the abstract syntaxanalysis section. As discussed in that section, of each of the destination standards, but togetherseveral kinds of relations need to be addressed: these destination abstract syntaxes span the source abstract syntax. This is for example useful in a• Similarity relation: This relation expresses case in which an annotated ECG segment which that the related segments of vital signs are includes both a time based signal representation similar to one another in respect of a de- and the physiological phenomena like P-wave and fined context, such as the context of their QRS complexes is converted to standards that use which reflects the purpose of the vital support time based signal representations, but signs. Similarity is used here to associate only one of them is able to represent physiologi- vital signs that reflect the same (physi- cal phenomena but, on the other hand, does not ological) phenomena but are represented and support annotations. In this example, annotations structured in different ways in the different but not physiological phenomena are supported standards. For example, an ECG P-wave may by the other destination standards. be represented as sampled amplitude values Vital sign segments which are similar are also of the wave and parameterized by a sample equivalent. That is, segments which are similar distant variable specified in another part of also have the reflexive, symmetry and transitiv- the ECG standard. This wave may similarly ity property of an equivalency relation because be represented in terms of the wave onset, the related vital signs are supposed to reflect the duration and peak value. A converted ECG same (physiological) phenomena. As discussed segment, which is formatted in a standard earlier, these properties are defined in the sense other than the original one but considered of the applied conversion tools and settings. That having the same interpretation and quality is, similar ECG segments reflect the same heart in the perspective of the addressed context, condition in respect of the resolution of the ap- is defined here as being similar to the origi- plied tool. For example, an ECG formatted in nal source segment. This similarity relation the CEN/SCP-ECG standard, which is converted therefore needs to contain the context of to the DICOM wave form standard and the lat- the similarity; it for example includes the ter converted again in ecgML (Wang, 2003), is identity of the conversion tool or algorithm, considered equivalent and even similar to the the parameter settings, the actor in charge ECG representation in ecgML in the context of of this conversion and the actor’s comments the applied tools and parameter settings. Remark for example to further detail the context of that one of the applied tools needs only to convert similarity. This similarity relation originates the wave form of the ECG and can be unaware from the need of the care center stakeholder of the physiological phenomena expressed by the to enable a multi standards environment and data. Therefore, the resulting ECG formatted in the policies of the regulatory bodies at the ecgML is not necessarily completely identical different points of care. to the CEN/SCP-ECP formatted ECG, but in the context of use, which is reflected by the applied In many cases, this relation associates one tools and settings, they are considered equallysource segment to one other converted segment. useful for the clinical purpose because at the 79
  • 97. Medical Information Representation Framework for Mobile Healthcareresolution of the cascaded tools they both reflect transfer of vital signs which are consideredthe same heart condition. This cascade of conver- important for the diagnosis or treatmentsions is usually called lossy if a CEN/SCP piece tasks. In case of severe bandwidth degrada-of data representing an abstract syntax concept tion, vital signs which are considered lessis not represented in ecgML. important may for example be stalled; • Aggregation and splitting relations:• Enhancing relation: This relation expresses these relations express that the related seg- that the enhanced segments of vital signs are ments of vital signs are aggregated or split, better conditioned in respect of the context respectively, from the others. As discussed of use. For example, ECG segments filtered in the previous cases, especially the simi- from undesirable noise or EMG movement larity relation, the aggregated segments are artifacts are enhanced if compared against equivalent to the source segments in the the source ECG segments. Although the sense of the aggregation tool resolution. enhanced segments are more appealing This equivalence is therefore specified by for use, the originating source segments the aggregation tool and parameter settings. essentially have the same effectiveness in We may apply the same justifications for the respect of the context of use. This enhanc- splitting relations. However, in the latter ing relation is meant to express vital sign relation, we additionally may deal with the segments which contain the same bio-elec- downscaling of vital signs, for example to trical or physiological phenomena relevant fit the data onto an available transmission for the medical purpose, but the enhanced channel of a specific quality that otherwise is segments are considered better conditioned not able to transfer the vital signs. Although for the medical purpose, for example more the quality may be reduced, the resulting efficient for use. As in the case of the simi- vital signs are considered useful for the larity relation, this relation needs to contain professional; otherwise the downscaling the specification of the context of use, for was meaningless, thus not executed. In this example it needs to include the identity of context of use, the related vital signs are the vital sign enhancing tool or algorithm, also considered equivalent. As in the other the parameter settings, the actor in charge of cases, it is therefore necessary to specify the this enhancement and the actor’s comments, aggregation or splitting tool, tool settings, for example, to further detail the intended the actor in charge of the aggregation and context. In this perspective, the enhanced the splitting or the splitting strategy and the and the source segments are equivalent in the actor’s comments. The tool that splits vital specified context, that is one may replace the signs may use the priority of the vital sign other without influencing the interpretation discussed earlier to determine the splitting. of the clinical data in the addressed monitor- Aggregation of vital signs, on the other hand, ing and treatment context. This enhancing may also be used to concatenate (digitized relation originates from the need of the continuous-time) vital signs which other- professional stakeholder to provide adequate wise are located remotely; this improves vital sign units of interpretations. availability or efficiency of the processing• Priority relation: This relation expresses or the vital sign analysis by a professional. the inter vital signs degree of importance. It These aggregation and splitting relations is a means for the mobile healthcare applica- come mainly from the system provider’s tions to ensure continuity of processing or requirements analysis and partly from the professional needs.80
  • 98. Medical Information Representation Framework for Mobile Healthcare Concepts of vital signs at abstract syntax level Clinical Data of Patients can be expressed by languages like ASN.1 (ASN.1, 2008; MacKinnon, 1990), XML schemas (Malik, Figure 3 shows that patient’s clinical data (repre- 2008) or UML (Booch, 1999). Figure 3 describes sented by the UML class PatientClinicalData) is a the information model of the framework expressed collection of vital signs data (represented by the in the Unified Modeling Language (UML) class abstract class VitalSignData explained in the next diagram. A class is symbolized by a rectangular section). In Figure 3, the set of vital sign data is with a class name at the top, attributes in the cell represented as a (genuine) part of patient’s clinical in the middle, and operations at the bottom. In data by the black diamond composition symbol. this chapter, we do not detail the operations of a The clinical data is anonymous, because it is iden- class and only provide those attributes that are tified by some patient identification number. Via relevant to explain the framework. The associa- the patient’s electronic health record PatientEHR, tions between the classes are represented by the however, a patient’s clinical data can be associ- lines between the related classes. ated to the patient, but the other way around is not specified (in UML, this unidirectional association Figure 3. Vital sign information model 81C
  • 99. Medical Information Representation Framework for Mobile Healthcare is symbolized by the arrow, which arrow-head is a specialization of the class Equivalency. As denotes the navigation direction between the discussed earlier, the equivalence between vital involved classes). The 1 to 1 multiplicity of this sign data is defined in a specific contextual set- association indicates further that patient’s clinical ting. In the model, this context for equivalence is data represents the whole collection of measured, specified by the attributes actor_id, which identifies processed and stored vital signs of this anonymous the responsible actor for the relation between the patient. Alternatively, one may replace the left vital sign data, actor_comment, which denotes the value “1” with “1 .. *”, which indicates a range comments of the actor, and also the time and date of one or more collections of vital signs of the information. As discussed earlier, the context is patient identified by patient_id. In the context of also defined by the applied tool and its settings, this chapter, we assume the availability of one set both are represented by the class Tool. An addi- of clinical data per patient. tional design choice is that we define the similarity relation only for vital signs that are encoded in Vital Sign Data different standardized formats. This constraint is not shown in the figure, however, it can be As mentioned earlier, the vital sign data is repre- expressed by a UML note or specified in Object sented by the abstract class VitalSignData in Fig- Constraint Language (OCL) (OMG, 2003). ure 3. The class is a UML abstract class, because Analogous to the similarity relation, the en- the class is only conceptually defined, other (non- hancing relation, which is expressed by the class abstract) classes will refine (i.e. specialize) this Enhancing (Figure 3), is a specialization of the class. In UML, an abstract class can be identified class Equivalency. In this model, we define an by the class name written in italics. For example, enhancing relation only for vital signs that are the abstract class ECG_Data is a specialization formatted in the same standard. of the abstract class VitalSignData (specializa- Aggregation and split relations are also spe- tion is an “is-a” relation and is symbolized by the cialization of the class Equivalency. Aggregation open triangular symbol in UML). This abstract is a many to one relation between vital signs for- class ECG-Data may be specialized further for matted in conformance with the same standard example by the classes DICOMStudy and SCP- and the other way around, the split relation is a ECG_Record. In this chapter, the class DICOMStudy one to many relation. represents ECG data formatted in accordance with the DICOM waveform standard and the class SCP- Derived Vital Signs ECG_Record represents ECG segments formatted in accordance with the CEN/SCP ECG standard. Trend signs or, in general, derived vital signs are The information model can be further extended frequently used in care programs as first indica- with other ECG data formatted in other (de jure, tors of the condition of the patients. Instead of a de facto or proprietary) standards. plethysmogram, care programs like emergency services or COPD programs use the derived Vital Sign Relations oxygen saturation O2sat (or SpO2) parameter as a measure for the oxygenation of blood. Heart Rate Via the abstract class Equivalency, Figure 3 also and Heart Rate Variability are other examples shows that some source vital sign data can be of trend signs, typically derived from one of the related to some other destination vital sign data. ECG leads. In contrast to the similarity and the In the figure, the similarity relation discussed enhancing relations, we specify derived signs as earlier is represented by the class Similarity, which specializations of the abstract class Extracted- 82SR
  • 100. Medical Information Representation Framework for Mobile Healthcare Feature, which in turn is specified as a component tion at the level of the types of vital signs rather of the class VitalSignData (in UML symbolized than at a more detailed level, for example at the by a black diamond (cf. Figure 3)). As the case for level of digitized vital sign samples. This choice equivalence relations, the applied tool, tool setting, has the additional benefit that vital sign sets for- actor in charge of the trend data processing and matted in a specific standard can be treated as a the actor’s comments refines ExtractedFeature black box; an approach which intends to preserve even further. We model trend or derived signs the structures defined in standards as atomic units. as a component of the original vital signs, rather This could be necessary in case of handling vital than modeling them via the equivalence rela- signs formatted in proprietary standards whose tion, because it better fits to the way vital sign internal structures are unknown to the applica- standards deal with derived signs and because tion developers of the mobile healthcare system of the complexity of the required constraints due provider stakeholder. In this kind of cases, third to the transitivity of equivalency relations. For party tools that have knowledge of these structures example, Heart Rate and Heart Rate Variability are needed to enable processing, rendering or are derived signs but they represent different conversion of the vital sign sets. This black-box concepts; therefore they are not equivalent. Other approach is for example supported by MIME via features which can be extracted from ECG leads the “-x” constructs. are for instance the high and the low frequency In case of multi-valued or multi-channeled components, including their ratio. vital signs (e.g. the leads of ECGs) or in case of (multiple) trend signs, the earlier mentioned prior- Care Program Dependent Priority of ity attribute can be refined further to priority of Vital Signs these values, channels or trends (e.g. represented by the attribute t_priority in the class TrendSign). As discussed earlier, in mobile healthcare, data Consequently, these intra vital sign priorities transfer bandwidth especially from wireless com- depend on the attribute priority of the class Vital- munication channels like GPRS may fluctuate. SignData. This dependency is represented by If available bandwidth drops below the required the dashed arrow in Figure 3. level, less important vital signs can be stored locally in favor of the transmission of the more important ones. The management modules of ECanda the mobile healthcare applications may (semi) automatically decide which type of vital signs Several de-jure, proprietary and de-facto format to stall and which to transfer or process further and encoding standards are suitable for ECGs, if these vital signs are prioritized. Sophisticated amongst others CEN/SCP-ECG, DICOM wave- prioritizing structures which are care program form, ecgML, FDA-ECG, HL7 and VITAL. We or clinical task dependent can be developed, but express some of them in UML class diagrams to in this chapter we use a simple priority attribute. illustrate the use of the information model of the If necessary, this attribute can be extended with framework. It is not in the scope of this chapter to a reference to the professional actor in charge of provide a complete list of ECG standards neither prioritizing vital signs for the care program. to provide detailed UML class diagrams of all We specify the attribute priority in the abstract these standards. class VitalSignData to enable priority based selec- 83CSd
  • 101. Medical Information Representation Framework for Mobile Healthcare CEN SCP-ECG STAanda (e.g. Section 11). Sections have a common header structure, in the figure represented by the general- The Standard Communication Protocol for com- ized class CEN/SCP Section. A high level SCP-ECG puter-assisted Electrocardiography (SCP-ECG) is structure, expressed in a UML class diagram, is a standard developed by CEN/Technical Com- given in Figure 4. mittee (TC) 251 (CEN/TC251, 2007). Besides Section_6 contains a black-box of ECG data. ECG data, SCP-ECG additionally defines ECG To render the individual ECG leads from Sec- related data to enable the specification of patient’s tion_6, attributes of Section_3, which represent demographic data, the measurement settings, the metadata specifying the number of leads and the performed signal processing on the ECG the leads description, have to be accessed first. data, the compression and manufacturer specific This dependency of Section_6 from Section_3 is information. represented in UML by the dashed arrow between In SCP-ECG, the entire ECG data set is called these two classes. a record. A record is further decomposed into “section” parts (indicated with section numbers from 0 to 11), each of which carrying a specific DICOM ECG Wavefo aspect like patient (demographic) information, Supplemen compression tables, the ECG lead definitions, the ECG lead data, the reference ECG beat(s) of the DICOM (Digital Imaging and Communications leads, including the physiological complexes like in Medicine) standards (NEMA, 2007a) are de- QRS, and also interpretive annotations. veloped by a joint committee of the American Eleven types of sections are defined in SCP- College of Radiology (ACR) and the National ECG. Table 1 presents the eleven sections and a Electrical Manufacturers Association (NEMA), brief description. Some sections are mandatory often in liaison with other organisations like CEN (e.g. Section 0 or Section 1), others are optional TC251, JIRA in Japan, IEEE and the American Figure 4. CEN/SCP-ECG model 84CSCCSdm
  • 102. Medical Information Representation Framework for Mobile Healthcare Table 1. CEN/SCP-ECG sections Section No. Title Description 0 Pointer the sections and their locations in the data set record 1 Header Information patient and acquisition related information 2 Huffman tables the Huffman compression tables 3 ECG lead definition the leads, the sample numbers and their relativity to a reference beat (cf. Section 4) 4 QRS location and Reference beat the location of the QRS complexes and the position of the reference beat 5 Reference beat encodings parameters like encoding flag, sample distance, gain. 6 Rhythm data the ECG data 7 Global measurements info pacemaker spikes and QRS complexes like the P-, QRS-, T- on-/offsets, QT intervals 8 Interpretive statements text based (diagnostic) annotations 9 Manufacture specific statements manufacturer specific diagnostic annotations 10 Lead measurement leads information and fields reserved for manufacturer data 11 Universal ECG interpretive statements universal statement codes (cf. SCP-ECG standard) and most recent annotations which have to be consistent with annotations in other sections National Standards Institute (ANSI). Although The figure reflects the clinical procedure by using the DICOM organisation originally addresses terms like studies (class DICOMStudy) and series imaging standards, it also developed a standard of clinical data (class Series). Although not shown to exchange waveforms. This latter is therefore in the figure, these terms include the specifica- suitable for ECGs. tion of the responsible professional, the clinical DICOM uses an object based model, therefore protocols, the waveform identifications (incl. the not only specifying the structure of the medical acquisition time), the annotations, the waveform data content as information objects, but also data (which may be multiplexed bio signals, there- the operations on the data (i.e. services). The fore also includes the multiplexing parameters, functional units in DICOM define the classes the sampling rate, etc.) and also the corresponding of the information objects and the correspond- equipment used to generate the data. That is, the ing services, the so-called Service-Object Pair class Waveform may contain several multiplexed classes (SOP classes). One of the SOP classes is vital sign channels (represented by the classes for instance meant for a waveform store. MultiplexGroup and Channel in Figure 5) In DICOM, a waveform information object is decomposed into information entities, each of which stored in data modules. Examples of in- FDA EC SPECIifiion formation entities are patient, (clinical or patient) study, clinical data series within a study, equip- As observed in the previous section, the DICOM ment which creates the series, and waveforms as waveform standard is to some extent based on part of the series. clinical procedures and accordingly the data is Figure 5 presents a simplified UML model of represented in terms of studies, the FDA format DICOM’s waveform related information entities. for waveforms (Browns, 2002) is based on the 2D 85Spea
  • 103. Medical Information Representation Framework for Mobile Healthcare Figure 5. DICOM waveform model property of sampled waveforms; it emphasizes the FRAamewo IMPLEMENTATIion viewing representation of waveforms. ASPpe Figure 6 shows the FDA waveform information model. As in the earlier discussed standards, the In this section, we discuss some of the imple- FDA model also provides manufacturer informa- mentation aspects that illustrate the use and tion, patient (/subject) identification, and annota- the benefit of the framework. First we discuss tions. In the figure, the class PlotGroup models refinements of the information models presented an ECG data set and aggregates data of the class earlier. These refinements enable the translation XYPlot, each of which representing a piece of ECG and serialization of the abstract syntax to the data of a particular lead. transfer syntax. Refinement of the Specialization Constructs In the earlier discussed information models, we ap- ply the object oriented specialization construct, for 86kmplemenaC
  • 104. Medical Information Representation Framework for Mobile HealthcareFigure 6. FDA Waveform modelexample, to distinguish between vital signs that are code enables (de-)multiplexing of serialized piecesformatted in accordance with different standards. of vital signs, for example necessary for a 24/7This specialization can be refined by additional continuous monitoring of patients. Together withdiscriminating attributes, for example, an attribute the attribute priority (Figure 7), these discriminatingidentifying the vital sign type (i.e. vitalsign_type in attributes can be used to split a vital sign set, forFigure 7b) and an attribute identifying the format example necessary in case of severe bandwidthand encoding standard (i.e. standard_id in Figure degradation along the healthcare delivery path.7b). The advantage of this refinement is that the This (de-)multiplexing technique is proven use-encoded attributes in the transfer syntax can be ful in multimedia communication using MPEG,used as header fields in the transfer syntax, for which analogously applies process identifiers toexample to indicate that the subsequent payload join or remove language channels and to mergeblock of data contains vital sign data formatted in and split television program channels on the fly.the transfer syntax of the identified standard. This 87
  • 105. Medical Information Representation Framework for Mobile Healthcare Figure 7. Refinement with discriminating attributes Refinement of Many to Many Abstract and Transfer Syntax Relations Notations The information model can cope with one to many Other languages are available to express concepts or many to many associations between sets of vital at abstract syntax level, for example XML Schema sign data, for example the similarity between a and ASN.1. Using XML tools a character-based source set of ECG data formatted in CEN/SCP- XML document can be derived that contains the ECG and destination data sets formatted in FDA- vital sign data specified in accordance with the ECG and in DICOM waveform standard. In this XML Schema. This document can then be seri- case, a cardiologist may want to visualize both alized by reading it from left to right and from FDA and DICOM sets simultaneously in case top to bottom, yielding a sequence of characters that the CEN/SCP-ECG data set is not available (transfer syntax) suitable for transfer using inter- on the premises or a rendering tool for the latter net protocols. Tools are also available to visual- format is not available either. ize XML Schemas as XML Schema diagrams. As discussed earlier, the MIME type and sub- However, tool based development kits are also type construct informs applications which tools available to develop XML Schemas from UML to use and how tools should render the data. For (Malik, 2008). example, in e-mail applications, the MIME value Abstract Syntax Notation One (ASN.1) is “multipart/parallel” indicates that the aggregated defined by the International Organization for data sets have to be visualized simultaneously. Standardization ISO (ISO 8824, 1994). It is a Similar to MIME, we can refine the equiva- notation for specifying data at abstract syntax lency class with additional attributes that specify level. Associated with ASN.1 are encoding rules which vital sign sets need to be rendered simul- for generating binary transfer syntaxes from the taneously. abstract syntax (ISO 8825, 1994). ASN.1 and its encoding rules provide compact transfer syntax code. 88RTS
  • 106. Medical Information Representation Framework for Mobile Healthcare A joint committee of ISO and International ties of connections. This approach is considered Telecommunication Union (ITU-T) has produced useful for healthcare delivery in which patients several standards on the mapping of XML Schema are mobile and self managing. It is expected to ASN.1 and vice versa. A web accessible on-line useful for new clinical pathways for mobile and tool which translates XML Schemas to ASN.1 is distributed healthcare delivery that involves col- for example available (ASN.1, 2008). A framework laborating actors of different medical specialty, that accommodates the collection of the previ- possibly acting in new roles and each of them ously described tools provides a development needing medical information that are represented environment that enables the translation of vital in accordance with the (new) working practices signs specified via UML class diagrams to concise of their specialty. binary transfer syntax code. The proposed framework, which amongst others is an integrative conceptual structure that binds methods, techniques and mechanisms for Conlu interoperability of different format and encodings of medical information, needs to be supplemented We propose a framework for flexible and modular further with other ECG and vital sign standards. processing, storing and transferring (segments That is, the framework needs to be populated by of) medical information in a mobile, distributed relevant format and encoding standards. Conse- and heterogeneous environment. The framework quently, the information model described in this adopts an ICT information representation model, chapter needs to be refined further, for example, to which separates the concerns of information provide tool developers the necessary hooks (e.g. transfer and store from the concerns of expressing, class attributes, object methods and dependency converting, splitting, synchronizing and joining relations) to design medical information conver- information. The abstract syntax level methods, sion, splitting and joining tools. Such refinements techniques and mechanisms, which address the not only require details of the format and encoding latter mentioned concerns, provide the necessary standards which populate the framework but may support for processing medical information in also need abstract syntax level knowledge of the an environment that contains multiple standards ontology of the corresponding bio physiologi- for data format and encoding. On the other hand, cal or bio electrical phenomena, for example, to the transfer and local syntax level methods, tech- specify in details the transitivity constraints of niques and tools, associated to the first mentioned the similarity relation. concerns, enable transfer and store of medical Another topic for future work is for example information in an efficient and dependable way. the specification of guidelines or rules to up- or The framework, which also contains the vital down-scale digitized continuous-time vital signs sign information model discussed in this chapter, in transfer automatically to match to the fluctua- therefore supports the exchange of medical infor- tions of the properties of the end to end connections mation in a meaning preserving way despite the within the tolerance specified by care programs use of different format and encoding standards or professionals. These guidelines supplement and the fluctuations of the property of the end to the framework further and improve its use for end data transfer connections. mobile healthcare delivery in a heterogeneous This chapter discusses the framework at con- environment. ceptual level. It provides a generic approach to deal with multiple formats and fluctuating proper- 89ion
  • 107. Medical Information Representation Framework for Mobile Healthcare ACKknowledgmen Brazilian Symposium on Software Engineering (SBES). João Pessoa, Brazil. The authors thank Bayu Erfianto, who did his Freed, N., Borenstein, N. (1996). Multipurpose MSc project on XML and ASN.1 based repre- Internet Mail Extensions (MIME) Part One: For- sentations of ECGs. His project and results have mat of Internet Message Bodies. IETF RFC 2045. been a starting point for our continued research From http://www.rfc-editor.org/rfc/ rfc2045.txt. in this area. This work is part of the Freeband AWARE- HealthService24 project eTEN-C517352 (2005). NESS project (http://awareness.freeband.nl). EC eTEN Programme. Retrieved Feb. 24, 2006, Freeband is sponsored by the Dutch government from http://www.healthservice24.com. under contract BSIK 03025. Hinchley, A. (2005). Understanding Version 3, A primer on the HL7 Version 3 Communication Stan- dard. Munich: Alexander Mönch Publishing. Refeen ISO 8824 (1994). Information Processing System ASN.1 Information site. (2008). Retrieved March – Open Systems Interconnection – Abstract Syntax 28, 2008, from http://asn1.elibel.tm.fr. Notation 1 Specification. ISO/IEC JTC1/SC21. Blair, G., Stefani, J-B. (1998). Open Distributed ISO 8825 (1994). Information Processing System Processing and Multimedia. Addison-Wesley. – Open Systems Interconnection – Basic Encod- ing Rules for Abstract Syntax Notation 1 (ASN.1). Booch, G., Rumbaugh, J., Jacobson, I. (1999). ISO/IEC JTC1/SC21. The Unified Modeling Language: User Guide. Addison-Wesley. Le Gall, D. (1991). MPEG: A Video Compression Standard for Multimedia Applications. Commu- Bray, T. et al. (2004). Extensible Markup Language nications of the ACM, 4(34). (XML) 1.0. 3rd edition. Retrieved 2005, from http:// www.w3.org/TR/2004/REC-xml-20040204. MacKinnon, D., McCrum, W., Sheppard, D. (1990). An Introduction to Open Systems Intercon- Browns, B., Kohls, M. Stockbridge, N. (2002), nection. New York: Computer Science Press. FDA XML data format design specification. Draft of the US Food and Drug Administration. Malik, A. (2008). Design XML schemas using UML. Retrieved March 28, 2008, from http:// CEN/TC251 prEN 1064 (2002). Health Informat- www.ibm.com/developerworks/xml/library/x- ics – Standard Communication Protocol – Com- umlschem/. puter-assisted Electrocardiography. CEN/TC251 prEN 1064. Mei, H., Widya, I., Halteren, A. van, Erfianto, B. (2006). A Flexible Vital Sign Representation CEN/TC251 (2007). CEN website. Retrieved June Framework for Mobile Healthcare. 1st Inter- 8, 2007, from http:// www.centc251.org. national Conference on Pervasive Computing Concalves, B., Guizzardi, G., Pereira Filho, J. Technologies for Healthcare 2006. Nov. 29th – Dec G. (2007). An Electrocardiogram (ECG) Domain 1st, 2006. Innsbruck, Austria. Ontology. In Proceedings of the Second Brazilian MobiHealth project IST-2001-36006 (2002). EC Workshop on Ontologies and Metamodels for Soft- programme IST. Retrieved Feb. 24, 2006, from ware and Data Engineering (WOMSDE’07). 22nd http://www.mobihealth.org. Brazilian Symposium on Databases (SBBD)/21st 90e
  • 108. Medical Information Representation Framework for Mobile HealthcareMyotel (2008). Myofeedback based Teletreat- Schiller, J. (2003). Mobile Communications. Ad-ment Service Project. EU programme eTEN dison-Wesley.– 046230. Retrieved Feb. 2008, from http://www. Wang, H., Jung, B., Anuaje, F., Black, N. (2003).myotel.eu. ecgML: Tools and Technologies for multimediaNEMA (2007a). Digital Imaging and Communica- ECG Presentations. Proceedings of XML Europetions in Medicine (DICOM), Part 1: Introduction Conference. London.and Overview. Virginia: NEMA. Weigand, C. (2005). VITAL: Use and Implemen-NEMA (2007b). Digital Imaging and Communi- tation of a Medical Communication Standardcations in Medicine (DICOM), Part 3: Information in Practice. Computers in Cardiology, 32, 319-Object Definitions. Virginia: NEMA. 322.OMG ptc/03-10-14 (2003). UML 2.0 OCL Speci- Wootton, R., Craig, J., Patterson, V. (Ed.).fication. OMG Adopted Specification (ptc/03- (2006). Introduction to Telemedicine. 2nd edition.10-14). Retrieved in 2007, from http://www. London: Royal Society of Medicine Press Ltd.omg.org. 91
  • 109. 92 Chapter V A Distributed Approach of a Clinical Decision Support System Based on Cooperation Daniel Ruiz-Fernández University of Alicante, Spain Antonio Soriano-Payá University of Alicante, Spain ABSTRACT The incorporation of computer engineering into medicine has meant significant improvements in the diagnosis-related tasks. This chapter presents an architecture for diagnosis support based on the col- laboration among different diagnosis-support artificial entities and the physicians themselves; the au- thors try to imitate the clinical meetings in hospitals in which the members of a medical team share their opinions in order to analyze complicated diagnoses. A system that combines availability, cooperation and harmonization of all contributions in a diagnosis process will bring more confidence in healthcare for the physicians. They have tested the architecture proposed in two different diagnosis, melanoma, and urological dysfunctions. INTRODUCTION from elements that provide proofs of diagnosis such as medical image acquisition systems, for Medicine has been one of the most important example, radiographies, echographies, CAT, disciplines in society since mingled with magic PET images, etc. (Rangayyan, 2004); till techni- and religion in the Egyptian era. The importance cal support applied to treatments, for example, that medicine represents in society makes it one of electro-stimulation in rehabilitation or prosthesis the major destinations of technological advances: (Vitenzon, Mironov, Petrushanskaya, 2005) or Copyright © 2009, IGI Global, distributing in print or electronic forms without written permission of IGI Global is prohibited.aTRCT
  • 110. A Distributed Approach of a Clinical Decision Support System Based on Cooperationtelecommunications applied to medicine (Moore, we find the corresponding symptoms: physicians1999; Wootton, Craig, Patterson, 2006). must not only know the name and treatment for Although technology is used to apply certain diseases, they must also be to identify their di-treatments or to make diagnosis tests, it is still not agnostic signs and distinguish them from othersconsidered as a real aid to the main medical task: corresponding to similar diseases.the diagnosis decision process. On the other hand, The evolution of medicine has also led to themedical diagnosis is defined as “the discovery and gradual change in diagnosis techniques (Adler,identification of diseases from the examination 2004; Porter, 2006). In the early days of medicine,of symptoms” (Collins, 2003). This definition diagnosis was based exclusively on clinical data,involves two steps in any act of medical diagnosis. that is to say, on the symptoms and the physicalFirstly, the “research” task in which the specialist examination of the patient. With medical advancestries to determine the symptoms of a patient by and the application of technology, new diagnosisusing his medical record and diagnostic tests. tests and laboratory analysis were incorporated.Secondly, a task of analysis of these symptoms The discovery of new diseases and their group-and the decision, based on the medical knowledge, ing into families and specialities has facilitatedof which illness is associated to the symptoms the development of differential diagnosis, whichwith the greatest probability. An important detail consists of determining the different illnesses thatis noting that medical diagnosis is essentially a could affect a patient, after a comparative studydecision-making process based on the lesser or of the symptoms and injuries suffered.greater probability of a patient’s symptoms of The large number of diseases and organicbeing related to specific information. dysfunctions coupled with the growing number Medicine has evolved since the days of Escu- of diagnostic signs (that increase thanks to newlapio, when the physician was a wise expert on all diagnostic tests) are paradoxically hinderingthe medical knowledge, problems and treatments; the process of diagnosis. Computer engineeringresearch and discoveries have broadened the field has techniques for the treatment of knowledgeof medical knowledge, making necessary the that may be useful for the processes of medicalcreation of specialities: neurology, traumatology, diagnosis (Burstein Holsapple, 2008; Greenes,rheumatology, urology or gerontology (one of the 2007). Most of these techniques are based onlast specialities incorporated). Moreover, most artificial intelligence and have been drawn fromof these specialities are divided into two groups: biology to be applied to computer science asadult and paediatric specialities (Weisz, 2005). neural networks or genetic algorithms (Haas Have you ever wondered how many known dis- Burnham, 2008; Morbiducci, Tura, Grigioni,eases are presently now? We might have a slight 2005; Rakus-Anderson, 2007). These techniquesidea of the number of known diseases by check- can classify patients into groups according toing the International Classification of Diseases whether or not they have certain diagnostic signs.proposed by the World Health Organization in its There are many examples of researching applica-last revision (ICD-10) (WHO, 2005): the group of tions of these techniques to diagnosis support:infectious and parasitic diseases is divided into 21 in (Roberts, 2000) a system based on Bayesiansubgroups (and each subgroup includes dozens of networks is proposed to assist the diagnosis ofdisease families), the group of tumours is divided breast cancer; (Georgopoulos Malandraki,into 19 subgroups, the group of nervous system 2005) shows a soft computing system to help in thediseases has 11 subgroups, the group of circulatory differential diagnosis of dysarthrias and apraxiadysfunctions is subdivided in 10 subgroups, etc. of speech which is able to distinguish among sixAlong with this enormous amount of diseases, types of disarthria and apraxia; systems of clinical 93
  • 111. A Distributed Approach of a Clinical Decision Support System Based on Cooperationdecision support are also applied with diagnostic Research UK at London in collaboration withtests based on images, for example in diuresis Infermed Ltd.renography (Taylor, Manatunga, Garcia, 2007) Isabel (Ramnarayan et al., 2006; Ramnarayanor in radiographs for diagnosing lung diseases et al., 2003) is a web system created by physicians(Katsuragawa Doi, 2007). to offer diagnosis decision support at the point A good example of the importance of the clini- of care. Isabel covers all ages and specialitiescal decision support systems is the OpenClinical as Internal Medicine, Surgery, Gynaecology,project. OpenClinical is an international non- Paediatrics, Geriatrics, Oncology, etc. Isabelprofit organization created and maintained as a gives the physician an instant list of diagnosespublic service with support from Cancer Research for a given set of clinical features (symptoms,UK. One of the objectives of OpenClinical is to signs, results of tests, etc.). Isabel consists of apromote decision support and other knowledge proprietary database of medical content and amanagement technologies in patient care and tutored taxonomy of over 11000 diagnoses andclinical research. This organization, through its 4000 drugs and heuristics.web site, presents an interesting group of resources Medicine is a context where the massiverelated with support systems. quantity of knowledge makes it perfect to use There are several clinical decision support distributed knowledge; in fact, as it has been pre-systems that have gone beyond the field of re- viously explained, clinical knowledge is dividedsearch and are being continually applied in health into medical specialities: oncology, cardiology,centres (Coiera, 2003; Greenes, 2007). Examples urology, etc. In computer science, distributedof these systems are GIDEON®, ERA and Isabel. knowledge is implemented in different ways andGIDEON® (Berger, 2001) is a system designed to levels. On a basic level, distributed databases arediagnose infectious diseases, based on symptoms, used (Luo, Jiang, Zhuang, 2008; Waraporn,signs, laboratory testing and dermatological pro- 2007); information can be stored in different data-files. It is made up of four basic modules: diagnosis, bases according to a location criterion (store datawhich enables the user to introduce all the signs near the location where it is going to be used) or aand symptoms and provides a ranked list of dif- homogeneity criterion (store information about theferential diagnosis; the epidemiological module same topic in the same location). One importantlets the user retrieve epidemiological parameters advantage of distributed databases is that eachor access a list of the world wide distribution of instance (of the database) can be managed by aany disease; the therapy module provides detailed different group of experts, so there is an implicitinformation about choices in drug therapy; the sharing of knowledge (apart from sharing data).microbiology module provides full laboratory Distributed data mining techniques can be usedcharacteristics for almost 900 organisms. to extract knowledge implicit in a distributed The Early Referrals Application (ERA) database (Cannataro, Congiusta, Pugliese, Talia,(Coiera, 2003) is a system to support physicians Trunfio, 2004; Han Kamber, 2005).in identifying those patients with suspected cancer In a higher level of distributed of knowledgethat should be referred to a specialist in a short we have the agents paradigm (Lin, 2007; Pop,time period. ERA is intended to be used within the Negru, Sandru, 2006). In this paradigm, theconsultation so the workflow has been designed knowledge is shared in an explicit way betweento be as simple as possible: during a typical user different entities called agents. An agent analyzessession just four web pages, clear and concise, external information and makes actions (accord-will be encountered. ERA was developed by the ing to its internal logic) to achieve an objective.Advanced Computation Laboratory of Cancer There are different types of agents but all of them94
  • 112. A Distributed Approach of a Clinical Decision Support System Based on Cooperation are based in three main components: perception, decisions thanks to its capacity to extract and decision and execution. With the perception, the manage information. It is important to note that, agent obtains information from the environment; many times, the information provided by a DSS in the process of decision, the agent uses its becomes another variable of the group of variables “knowledge” (represented internally for example involved in the problem; therefore, the experts with an artificial intelligence method: neural should determine the value of the decision auto- networks, rules, fuzzy logic, etc.) to decide what matically provided by the DSS. Decision support action to take in order to achieve an objective; systems are used in a wide range of fields such as finally, with the execution, the agent tries to in- economics (Modarres Beheshtian-Ardekani, fluence the environment to achieve the objective. 2005), industry (Delen Pratt, 2005), medicine An important characteristic of the agents is their (Vihinen Samarghitean, 2008), etc. ability to interact with other agents. Thanks to DSS can be classified according to different this interaction, agents can exchange informa- criteria such as the mechanism used to represent tion or share knowledge. This characteristic of knowledge; this is the case of rule-based DSS, the agents paradigm is the basis of the idea used DSS based on decision trees, etc. (Burstein Hol- in the cooperative approach presented in this sapple, 2008). Another criterion used to classify chapter. We can find an example of cooperation DSS is their operation autonomy: solicited advice, between agents in business processes in which when the system helps the user when requested; software agents must negotiate in order to reach unsolicited advice, when the system provides an objective, for example, the best price. In our diagnosis information without any request from scheme, several entities with different knowledge the user. In the latter group, there are, for example, about the same topic, collaborate to get a better intelligent alarm systems that analyse constants diagnosis. and warn about a possible diagnosis that requires In the following sections, we are going to an emergency treatment. introduce the concept of clinical decision sup- When we use a DSS to support the medical port systems, detailing their features, advantages diagnosis task, we have what is known as Clini- and disadvantages. Then, we will introduce the cal Decision Support System (CDSS), that could natural evolution of these systems into coopera- be defined as “active knowledge systems which tive systems, propose their design and explain its use two or more items of patient data to generate performance in a scenario. Next, we will show case-specific advice” (van Bemmel Musen, some examples and results of the application of 1997). In this definition, we have the three main cooperative systems to diagnosis support. Finally, elements of a DSS: the knowledge of medical we will present different future lines regarding diagnosis, the patient data or the information to this subject. be analysed, and the results provided by the DSS as a recommendation. A result coming from a DSS applied to diagnosis will usually indicate the CINICAL DECISION SUPPORT probability that the symptoms will correspond to SYSTEMS a particular illness. The operation of a traditional CDSS is as follows: a user inputs data associated A decision support system (DSS) can be defined to any illness; the system analyses the information as a multi-model, interactive system used by a and provides one or more results along with their decision maker to perform an exploration (Berner, corresponding probabilities of success. 2007; Pomerol, 1997). The main objective of a Difficulties found in traditional medical di- DSS is to provide an aid to make unstructured agnosis can be divided into two main groups: 95CSRTST
  • 113. A Distributed Approach of a Clinical Decision Support System Based on Cooperationsources of error in the examination of the symp- • To support a clinical decision. Sometimes,toms and problems in forming hypotheses. The if the symptoms are not well defined, thefirst group covers the problems stemming from physician may have doubts about severalconfused information provided by the patient, diagnoses. The CDSS may help increaselow reliability of this information or deliberate the physician’s confidence in a particularconcealment of the true cause of consultation (due diagnosis.to cultural reasons or personal embarrassment). • To propose alternatives. The large numberThe second group includes those situations as- of existing illnesses can cause a physiciansociated with direct problems of the diagnosis overlooks diagnostic tests for less commontask: patients that present incomplete signs or diseases. A CDSS is able to propose ancommon illnesses with unusual manifestations, alternative diagnostic that could be easilylow frequency of diseases (rare diseases), mim- refuted or accepted by means of anothericking diseases, and, in general, the difficulty the diagnostic test or symptom the physicianspecialist has in associating the symptoms to a had not initially thought about.specific disease. • To question a clinical decision. A CDSS Although a CDSS can be useful for both can provide a more objective diagnosisgroups, in the first one, its support capacity is support than a medical specialist. This canmore limited, as problems are found in the input cause that sometimes the CDSS issues, as adata. Additionally, the same difficulties found by first option, a diagnosis result relegated bythe specialist are going to be found by the CDSS. the physician due to subjective reasons (forIn the second group, the problems are related to example, an emotional feeling).the enormous amount of symptoms and diseases(impossible to remember for a specialist) coupled Advantages and Disadvantageswith human diversity (the same illness can bemanifested in different ways in different patients). The use of computer technology in the medicalThese problems are directly associated with the field, particularly in diagnosis, involves bothability to analyze and compute, a question on advantages and disadvantages. Next we willwhich a computer system can be particularly describe some of the advantages related to thehelpful. use of a CDSS: Two parameters are used to validate a diagnosissystem in medicine, sensitivity and specificity • Permanence. A computer system does not(van Bemmel Musen, 1997). Sensitivity is the age or lose power over the course of time.probability of classifying a diseased individual It may need maintenance and updating ascorrectly (as diseased); that is to say, a high sen- well as specialists must constantly updatesitivity of a CDSS implies a low number of false their knowledge, but once it stores a spe-negatives. On the other hand, specificity is the cific knowledge, this information lasts theprobability of classifying a healthy individual as duration of its working life (without lost ofhealthy and is directly related to false positives. reliability).In general, it is a priority that a CDSS has a high • Duplication. Initially, the cost of develop-sensitivity and, therefore, a low number of false ing a CDSS may be high, but once it isnegatives (diseased individuals classified incor- implemented, duplication is simple andrectly as healthy). inexpensive. A CDSS may be helpful to a medical specialist • Reliability. The reliability of a computerin several aspects: system is independent of external condi-96
  • 114. A Distributed Approach of a Clinical Decision Support System Based on Cooperation tions such as fatigue, personal affinity or aware of his limitations; this flexibility is pressure. not possible with a computer system. • Ubiquity. A computer system can be ac- • Need for structured knowledge. Any knowl- cessed via communication networks from edge that is incorporated in a computer anywhere; computers can work in environ- system requires a task of structuring. ments that are hostile or dangerous for a • Lack of feelings. Its operation is governed human being. by strict rules that have nothing to do with • Availability. Access can be permanently human sensitivity, often necessary in the available, achieving an availability of 24 relationship between patient and doctor. hours a day, 7 days a week. • Ethical problems. This limitation especially arises when computer systems are applied in The limitations of an expert system that pro- health sciences. As computer systems cannot vides diagnosis support are directly associated to accept responsibility for their own decisions, the limitations of any computer system when it their operations should be supervised by a is responsible for tasks that may require human medical specialist at all times. skills. Some of these limitations are: Classical Design • Lack of common sense or limitation of knowledge. Without the proper knowledge, a The typical architecture of a decision support male could be diagnosed 80 weeks pregnant system consists of three main elements: the user by a computer system. interface, the reasoning module and the knowledge • Inability to maintain an informal conversa- database on which the reasoning is based. This tion in natural language. There are subtle architecture is schematically shown in figure 1. differences when a patient is expressing his The user interface has evolved over the course symptoms that could be essential for the of time, from the input of data and questions via diagnosis. console until the most modern graphical systems • Flexibility. When issuing a diagnosis, a hu- that give access ubiquity properties thanks to man specialist can be flexible because he is interconnection networks. Additionally, inter- Figure 1. DSS Architecture 97C
  • 115. A Distributed Approach of a Clinical Decision Support System Based on Cooperation Figure 2. Interface of a DSS for urologists faces have been adapted to the main users, the the possibility of failure; moreover, maintenance physicians, who are not technical experts. Figure tasks become more difficult. Cooperation between 2 shows a screenshot of a user interface for clini- simple CDSS can manage both problems, the lack cal decision support system in urology that has of a complete knowledge of the domain and the been adapted to the clinical methodology used way to improve the reliability easily. by the urologists. The knowledge database is the part of the system that contains the basic information for the COOPERATIVE CLINICAL DECISION diagnosis issuing. This information may include SPPORT SYSTEM both structured knowledge and data used to extract information or to train the reasoning module. In the previous section, we have explained what a Finally, the reasoning module implements an clinical diagnosis support system consists of and algorithm that analyses the input data and provides how it can be a tool to help medical specialists. a result that basically consists of a classification: The evolution of the CDSS causes them to meet in the case of medical diagnosis, certain symptoms the same limitations that the physicians have: the are classified as belonging to a certain disease large amount of medical knowledge makes a CDSS with a degree of certainty. specialise in just one kind of disease. The main problem of the classical design for a The approach presented in this point is the CDSS is the limitation to tackle problems inside design of a cooperative clinical decision support the domain of the CDSS but not well represented system (CCDSS). The cornerstone of this architec- in the knowledge database. Another problem is ture is the improvement of the system reliability that the only way to improve the functioning of the thanks to the participation of several entities, CDSS is to increase the knowledge database and to with diagnosis functions, which work together to improve the reasoning module. This increases the provide a single diagnosis. Such cooperation can complexity of the CDSS and, therefore, increases be oriented towards various directions; different 98CRTSRTST
  • 116. A Distributed Approach of a Clinical Decision Support System Based on Cooperationentities can be specialised in different illnesses The user interface will be the means throughor organ systems affected by the same illness which healthcare professionals will be able toand their collaboration facilitates the diagnosis. request the CCDSS opinion. It will also be inAdditionally, the use of several diagnosis entities charge of presenting the consensus informationoperating in parallel increases the global system that must be evaluated by the specialist that re-availability. quested it. Cooperation in diagnostic tasks, thanks to the The diagnosis entities may be physicians, andparticipation of several physicians, is a common they will be human entities, or decision-makingworking manner in many medical teams. When support software, and they will be then softwarefaced with cases of difficult diagnosis, several entities. The software entities will receive thephysicians work together to issue different diagno- symptoms and diagnostic tests and, next, they willsis ideas. They all together discuss ideas, propose issue a set of possible diagnoses along with theirdiagnostic tests and, finally, reach a consensus. corresponding probabilities of certainty.The architecture of a CCDSS wants to automate The system core will be responsible for receiv-this working manner. ing diagnosis requests and distributing them to The architecture of a CCDSS is primarily the appropriate diagnosis entities. It must keepmeant to satisfy the reliability requirements. the system security, preventing intrusions of falseMaintaining a high availability of the system is entities that will destabilize the system. In addi-also very important. To meet these criteria, we tion, it will also be responsible for collecting thedefine a distributed architecture based on the results from the different entities and providingparadigm of intelligent agents. The components a consensus.of the architecture are: the user interface, the The communications subsystem consistsdiagnosis entities, the system core and the com- of both the communications network and themunications subsystem (figure 3). protocol to be used to transmit information. TheFigure 3. Distribution of the components in the CCDSS 99
  • 117. A Distributed Approach of a Clinical Decision Support System Based on Cooperationstructure of the communications network must ner; in addition, the size of the screens is oftenpreserve the security of the communications and smaller than that in the PC environment. Finally,the system integrity. the PDA environment involves working with a device with graphical limitations, touch screensUser Interface and a very small size. Once the device environment has been select-This is a key module of the system, as it will ed, the system will need to format the informationinfluence the degree of user acceptance. The according to the role and user permissions as a lastinterface must be friendly and adapted not only step. The system, as a part of the security mecha-to the user but also to the device from which is nisms, will distinguish three levels of accessibil-being operated. ity according to a set of roles and individualized Authentication will be the first contact with permissions. There will be six roles:the system, as it is necessary to maintain theappropriate security levels and to provide each • Healthcare reviewer. This will be an ac-user with the most suitable interface. After the credited professional in charge of reviewingauthentication process (figure 4), the system core the ethics of the system, both the operationwill be responsible for adapting the data accord- of the artificial entities and the activities ofing to the device from which the system is being consulting physicians and medical officers.accessed; this involves the design of different It will be responsible for supervising theenvironments to facilitate the access from the consensus operations.most common devices used to access a data net- • Medical officer. This is the role for thosework: PC, Tablet PC and PDA. These three kinds professionals who are working together inof devices include the main differences between a diagnosis, giving their opinion on certainworking environments. For the PC environment, symptoms.there are no graphical limitations and the size • Consulting physician. Users that use theof objects can be small, as the commonly used system to get a possible diagnosis will bescreens allow it. For the Table PC environment, authenticated under this role.the touch screen function implies the adaptation • Practice. If users use the system under theof the graphical interface to this working man- practice role, only artificial entities willFigure 4. System access process scheme100
  • 118. A Distributed Approach of a Clinical Decision Support System Based on Cooperation participate and the results of clinical cases Diagnostic Entities will not be used to update the system.• Student. This role allows the system to serve The second component of a CCDSS is a set of as a teacher. The operation will be the oppo- diagnostic entities, each one understood as an site: the system will provide some symptoms element that, after receiving information about the and the user will issue the diagnosis. Thus, symptoms of a patient, issues a possible diagnosis the CCDS becomes a system that could help with a certain degree of certainty. These entities with the task of teaching future physicians, may be either human or artificial. Diagnostic hu- thanks to the enormous amount of clinical man entities are consulting physicians who want cases stored in the database. to participate voluntarily in diagnosis activities• Administrator. This will be the engineer within the system. Artificial entities are software responsible for the proper operation of the that can propose a classification and, therefore, a technical side of the system. The activities possible diagnosis, thanks to statistical or artificial involved will be coordinated with those intelligence techniques. corresponding to the healthcare reviewer Diagnostic entities are distributed and get the in charge of the medical supervision. symptoms and report their results to the system core through the communications subsystem. If a consulting physician accesses the system, There is a quality value associated to each diag-the data input will depend on the speciality related nostic entity. This quality value will be used byto the issue he wants to consult. This does not the system core to make a decision on the finalimply that the information will also be analysed diagnosis. We can distinguish between human andby entities with different specialities if deemed artificial entities to assign these quality values.necessary. Moreover, the consulting physician is Since it is difficult to automatically evaluate themore familiarised with a certain kind of informa- quality of a diagnosis from a physician participat-tion related to his speciality and this is why this ing in the system, the healthcare reviewer will befeature will be taken into account. the person who will assign the quality value of a After introducing the symptoms and diagnostic particular doctor when a physician is registeredtests, a diagnosis will be required from the system, in the system. This value may be altered at anyindicating a time limit for issuing this diagnosis. time by the healthcare reviewer.This time will allow the adjustment of the ac- In the case of diagnostic artificial entities, thecuracy level of some artificial entities as well as quality value can be automatically calculated. Inindicate the possibility of participation or not of order to do this, artificial entities must pass anhuman diagnosis entities (when available during initial testing phase through which the probabil-that time limit). Once time has expired, the system ity of diagnosis success is calculated as well aswill provide a list of possible diagnoses along with sensitivity and specificity. Taking into accounttheir corresponding probabilities of confidence. these three variables, a value of certainty for theFurthermore, the listing will show which entities entities can be calculated, for example, calculatinghave supported each diagnosis, indicating whether the average. Moreover, this value can be updatedthey are artificial entities or physicians and their with the performance of the artificial entity soconfidence level within the system. that the diagnosis successes improve its value of certainty and the failures decrease this value. This 101
  • 119. A Distributed Approach of a Clinical Decision Support System Based on Cooperation process of constant updating gives the system an of perception or interaction with the environ- added value, because as new artificial entities ment. The diagnostic module has the algorithm are joined to the system (with improved artificial to classify the information received (diagnostic classification techniques), old entities go obsolete data) and decides whether this information cor- automatically and their diagnoses lose importance responds to a healthy individual or to a diseased with regard to the final consensus. person. The local storage will be in charge of For the proper operation of a CCDSS, it is storing the information relevant to the diagnosis very important that all artificial entities follow the process; furthermore, it will store the data related same method to calculate their value of certainty, to the entity itself such us the entity designer, because if this value is obtained by different means the last update date, the achieved percentages of the consensus process will lose reliability. Differ- classification, sensitivity and specificity, etc. The ences between human and artificial entities should module of interaction with the environment is re- not affect the consensus negatively thanks to the sponsible for receiving the data, adapting them to control by the healthcare reviewer; additionally, the analysis to be done by the diagnostic module the final consensus will include the influences of and, afterwards, formatting the results in a form both kinds of entity. However, different CCDSS suitable for transmission; besides, this module will can use different methods to evaluate the value also update the diagnosis algorithm and provide of certainty because, if we wanted to combine information about the entity itself when required the results of several CCDSS, we would take the by the system core of the CCDSS. final consensus results and not the particular ones This structure is similar to that corresponding corresponding to the entities working together in to software agents (Ferber, 1999; Lin, 2007), which the consensus. are able to perceive changes in the environment, As explained, medical officers are identified to consider what actions to take and to try to influ- with the diagnostic entities. Every medical officer, ence the environment to achieve an objective. In when accessing the system by the authentication, the case of a medical agent, the perceptions cor- will find a set of diagnosis possibilities related to respond to the symptoms; the actions would be the his speciality and will be able to choose which diagnoses (that will give rise to treatments), and ones he wants to use to issue a diagnosis. It is the main objective would be to give an accurate important to say that, with regard to the consult- diagnosis in order to cure the patient. ing physician (who is requesting a diagnosis for certain symptoms), the detail of diagnosis will System Core not include any identification of the medical of- ficers who have participated in the diagnosis. If The system core is a key element in the CCDSS he needs any explanation about the diagnosis, management. This module is responsible for dis- he should ask the healthcare reviewer. Despite tributing the diagnosis requests to the entities, for this, due to legal and ethical reasons, detailed reaching a consensus of the different diagnoses information of every diagnosis from a medical and for controlling accesses and information officer will be stored and only accessible to the transmitted through the system. The structure healthcare reviewer. Depending on the CCDSS, of the system core has three different parts: the the medical officer may act in an altruistic way or consensus module, the security control and the receive remunerations for his contributions. request manager. Figure 5 shows the design of the diagnostic The consensus module is, after receiving the artificial entities. Each artificial entity consists of diagnoses from the different entities and taking a diagnostic module, a local storage and a module its values of certainty as a reference, elaborates 102SC
  • 120. A Distributed Approach of a Clinical Decision Support System Based on CooperationFigure 5. Diagnostic artificial entitya consensus listing possible diagnoses and their will have more value that those coming from thecorresponding probabilities. This is one of the remaining entities.most complex tasks of the CCDSS, as there are Security control is mainly performed in twonumerous factors involved and always surrounded ways: user authentication and entities control.by degrees of uncertainty. The consensus algo- User authentication involves not only the accessrithm must be able to solve extreme situations; control and the selection of the appropriate role butfor example, a case in which several artificial also a complete monitoring of their actions withinentities with a high degree of certainty agree on a the system. It is important to prevent a maliciousdiagnosis and the opinion from a medical officer user that has taken a medical officer role fromdiffers completely. Moreover, depending on the modifying a diagnosis, causing a malfunction oforganization that manages the CCDSS, there can the CCDSS. On the other hand, the system corebe other factors that may influence the decision also manages the artificial entities, which must besuch us the economic cost of the treatment or registered in a database containing informationwhether a diagnosis has a favourable prognosis; about the entity, the engineer responsible for thethus, faced with two diagnoses, being the second development, the speciality, the initial degree ofmore likely to has a better prognosis, the decision certainty, etc. All this information will be used byalgorithm could consider this factor and change the system to control to which artificial entities itthe diagnoses order, keeping an optimistic ap- is possible to request a diagnosis. The healthcareproach. There are a lot of decision algorithms reviewer is not only in charge of distributingthat can be used and a lot of studies done about the profiles among the physicians with accessthe decision problems, especially in economics permissions, he is also in charge, along with the(White, 2006). administrator, of allowing an artificial entity to Another aspect that must be controlled by join the system (and of deciding the deleting).the consensus module is whether the request is The request manager is responsible for distrib-related to several specialities. In this case, maybe uting the diagnosis requests and for collecting thethe decision from all entities has the same value results before transmitting them to the decisionor maybe there is a major speciality and second- module. In order to make the distribution, theary specialities; in this case, the diagnosis from database with the register data of the entities (boththe entities belonging to the major speciality artificial and human) is taken into account, and 103
  • 121. A Distributed Approach of a Clinical Decision Support System Based on Cooperation the symptoms input by the consulting physician RIM specifications, data are modelled as obser- are sent to those entities that correspond to the vations, both the data coming from the medical speciality or specialities related to the request examination (joined to the diagnostic tests) and (which have been input by the consulting physi- the diagnosis. cian as well). Along with the transmission of information necessary for the diagnosis, it should be included SCNARIOS the time limit for the entities to issue the diagnosis. Once the time has expired, the manager will submit The first proposed scenario is located in the sur- all diagnoses received with their corresponding gery of a family doctor from a city; we assume that probabilities, stating the entity from which they there are no limitations with regard to the resources come in order to consider the relative importance for making diagnostic tests. The physician, after of each diagnosis in the consensus task. checking the medical record of the patient and examining him, concludes that the patient suffers Communications Subsystem and from an intestinal problem. In order to determine Protocols the diagnosis, the physician will order urine and blood tests and ask the patient to come back with The communications subsystem is the basis for the results after three days. the operation of a CCDSS. The information flows During this period, the physician may use the between different components through secure system to consult experts on the digestive sys- channels and standardised protocols. Artificial tem, sending the patient symptoms and waiting entities work like web services and they use Hyper for a result within three days, before the patient Text Transfer Protocol (HTTP) and Simple Object comes back. The results provided by the system Access Protocol (SOAP) to inter-component com- can confirm the physician’s diagnosis, but also munication. In order to ensure a secure channel, can help him look for alternative diagnoses when HTTP is secured by using the Secure Socket Layer checking the laboratory results that the patient (SSL) protocol, which makes communication will bring on his next visit. In this case, the use secure on the transmission level. Moreover, bet- of the system has avoided the referral of the pa- ter security levels can be achieved by encrypting tient to a specialist on the digestive system, if the the information on the implementation level. The laboratory tests did not confirm the hypothesis of artificial entities secure the messages they send the family doctor, and he had not thought about and receive with the XML-Encryption specifica- any alternatives. In this scenario, the system has tion of the World Wide Web Consortium (W3C). helped reduce waiting lists for specialists. They also sign and validate the messages with the Another scenario could consist of a patient XML-signature specification, also of the W3C. living in a rural environment with a widely dis- The structure of the messages transmitted seminated population, away from the city and from within the system should be based on a standard. the most sophisticated healthcare resources. Faced For this project, we intend to use the HL7 stan- with the case of a patient with an acute backache, dard, so that the XML messages are consistent the physician should decide whether to order more with the Reference Information Model (RIM) of tests and, therefore, refer the patient to a medical the HL7 version 3 specification (Hinchley, 2005). centre (perhaps, many kilometres away), where The data considered to be managed are only the resources necessary for the tests are available. diagnostic data, as the aim of the posed system One of the ways of interacting with the system in is the diagnosis process. In accordance with the this case could be to determine which diagnostic 104CSR
  • 122. A Distributed Approach of a Clinical Decision Support System Based on Cooperation tests to do, thus reducing the number of trips the Although the architecture proposed is focused patient should make to the hospital. The physi- on the diagnosis of human diseases, it could be cian would check the patient’s symptoms with the used in any decision-making support in which a system while he prescribes an analgesic treatment consensus among different opinions is useful. If to see if the pain goes away. When the patient we stay in the healthcare field, we have the same comes back to the medical centre, the physician situation in veterinary medicine. A CCDSS could will already have a listing of possible diagnoses be also used to propose economical questions provided by the entities connected to the system (maybe related to decisions on stock market values) and will be able to order diagnostic tests at the or auto mechanics. hospital, which confirm any diagnosis. In this case, the system has helped the physician select the diagnostic tests needed to obtain a differential EXPERMENTATION diagnosis; at the same time, inconveniences for the patient due to the several trips to the hospital, In the research group of bioinspired engineering far away from his home, has been avoided. and healthcare computing, we have developed a Finally, we propose a different scenario that prototype of this architecture. The main objec- does not involve problems related neither to spe- tives for this first development were, on the one cialised knowledge (as in the first case) nor to hand, to confirm that more accurate diagnoses resources (as in the second case). A patient goes can be obtained thanks to a consensus among to the hospital with strange symptoms that do the diagnoses provided by different CDSS. On not correspond to any illness. In this case, a team the other hand, we wanted to study the overall of physicians takes responsibility for the patient performance of the system with the incorpora- and, usually, begins to study the case in order to tion of medical opinions and to know what the get a differential diagnosis. In this scenario, the physicians’ acceptance level was. system could be regarded as one more expert In order to study whether the consensus among opinion (with the value that the team leader wants several diagnostic artificial entities improved the to give it), taking into account that this opinion is individual diagnosis provided by each entity, a consensus of many diagnoses; therefore, it is as we selected the diagnosis of melanoma as an if the team of physicians at the hospital had at its experimental test. Two classification algorithms disposal another team of external physicians, also were implemented: a Bayesian classifier and a taking part in the same differential diagnosis. multilayer perceptron (Greenes, 2007). To in- With these three scenarios, we want to show crease the number of diagnostic artificial entities, the reader several possible uses of a CCDSS, in multiple instances of the multilayer perceptron situations with some kind of restriction as well were created by modifying the number of hidden as in cases in which the systems becomes one neurons and the training set. Finally, in addition more opinion to be considered depending on the to the entity based on the Bayesian classifier, we evaluation of a medical director. Although we also obtained four entities based on the multilayer have provided only three scenarios, there are perceptron (with 3, 4, 5, 6, 7 and 8 hidden neu- many more cases in which the system could be rons), which had acceptable classification rates. useful; for example, in a prison or war the location Table 1 shows the performance of the multilayer characteristics make it difficult for a complete perceptron with regard to the number of neurons medical team to work together, so a CCDSS would of the hidden layer. be a viable option. Instead of working with a list of symptoms, the prototype was adapted to work with vector 105T
  • 123. A Distributed Approach of a Clinical Decision Support System Based on Cooperationfeatures taken from a pre-processing of the mela- to solve this problem, we have selected, from allnoma images. Regarding the consensus module, possible artificial entities, only those with clas-we used a simple voting system based on the sification rates over 75%.classification rate of the entities: the importance Figure 6 shows a comparison of the classifica-of one entity was equivalent to its classification tion rate between the Bayesian algorithm, the bestrate. Thus, if an entity with a classification rate of multilayer perceptron entity (7 neurons in the hid-85% diagnosed a skin injury as healthy and two den layer) and the CCDSS (obtaining a consensusmore entities with rates of 80% and 82% diag- from all the artificial entities). The classificationnosed the injury as melanoma, the final consensus rate of the CCDSS (91.32%) is very similar to thewould be melanoma. This consensus mechanism one obtained with the multilayer perceptron entity.can involve the following problem: two entities Regarding the specificity and the sensibility, bothwith very low classification rates (45% and 47%) measures are higher in the CCDSS: 93.58% andcan induce a wrong diagnosis within the system, 78.04%, respectively. The Bayesian entity hasover the diagnosis provided by an entity with a a classification rate of 81.25%, a sensibility ofclassification rate of 90% (45+4790). In order 93.15% and a specificity of 76.70%.Table 1. Performance of the multilayer perceptron entities Hidden neurons Classification Rate (%) Specificity (%) Sensibility (%) 3 70.59 22.54 46.08 4 74.22 55.02 52.94 5 81.00 70.89 55.00 6 84.37 92.00 70.58 7 90.62 93.20 77.35 8 87.50 91.15 76.47Figure 6. Comparison between CCDSS and other DSS diagnosing melanoma106
  • 124. A Distributed Approach of a Clinical Decision Support System Based on Cooperation Figure 7. Classification rates for different artificial entities and CCDSS in urological dysfunctions In the other experiment, we developed artificial CONCLUSION entities to diagnose different urological disorders based on urodynamic tests (Ruiz Fernández, Gar- Throughout this chapter, we have tried to introduce cia Chamizo, Maciá Pérez, Soriano Payá, 2005). the reader to the relevance that the decision-mak- In this experiment, we used a urology expert and ing support systems may have in the healthcare several family doctors interested in participating field, particularly in medical diagnosis. Moreover, as consulting physicians. The experimentation given the characteristics of the medical environ- outcome is shown in figure 7. In particular, we ment, we have proposed a new kind of system used two different multilayer perceptron (MLP1 that involves the cooperation between different and MLP2, in which we modify the hidden layer entities as the core of the diagnosis support pro- and the training data set) and a Kohonen Self- cess. Thanks to cooperation among diagnostic Organising Map (SOM) to classify three types entities, physicians and artificial entities, we want of urological dysfunctions: obstruction, hyper- to achieve higher precision in the final decision, reflexia and effort incontinence. As it is possible as well as provide wider system availability. to observe in the graph, the classification rate The proposed architecture is based on the obtained with the consensus is higher than the paradigm of intelligent agents and distributed individual rates for each entity for all the dysfunc- computing. The diagnostic entities are distributed tions studied. In the consensus we used the same to the system and act as web services when asked voting system as in the other experiment, so the by a control element (system core); consequently, weight of one entity in the final consensus was a lot of diagnosis processes can be performed in equivalent to its classification rate (obtained in a parallel. These processes are carried out by physi- previous test stage). cians and software programs. Finally, the system The most important conclusion of this second performs a consensus process among the overall experiment was the physicians’ opinions with results in order to give the user a set of possible regard to the system; they found the system very diagnoses with their corresponding probabilities useful in supporting those diagnoses involving of occurrence. very specific fields in which they do not have deep knowledge. 107CS
  • 125. A Distributed Approach of a Clinical Decision Support System Based on Cooperation The distribution of knowledge is a powerful journals or electronic health records in hospitals tool in a clinical decision support system. By allows for an automatic and constant update of adapting the system to a distributed architecture, the knowledge rules of the clinical decision sup- any number of future sources of knowledge could port systems. be integrated into the network, generating an An important future goal is the interaction expanding knowledge-base. In addition, differ- between different decision support systems. ent knowledge sources specialized in the same These systems should evolve in the same way problem can provide different points of view, a as medicine. Each CDSS should be an artificial key factor in decision support. expert in one speciality or even in just one group The experimentation carried out by our re- of diseases. This implies not only incorporation search group has confirmed that cooperation be- of information about the diagnosis and therapy tween different diagnostic entities can determine of diseases but also including information about a diagnosis from a set of symptoms, improving others organic systems that can be affected by a the individual performance of each entity. We disease (in order to interact with the CDSS expert have presented two examples, one related to the in these organic systems). The consensus between melanoma diagnosis and another based in the di- the different artificial entities will be more difficult agnosis of urological dysfunctions. In both cases, because each entity will be an expert in just one diagnosis results obtained with the cooperative topic so each entity will have a relative impor- approach are better than individual results. It is tance in the final diagnosis. It will be necessary important to say that at no time did we propose to improve the consensus algorithms. that a CCDSS replace a qualified professional As a summary we can suggest three specific at the diagnosis activity; the CCDSS operation objectives for the future. First, it will be neces- cannot be understood without the supervision of sary to involve a higher number of healthcare physicians that check the proper operation of the professionals in order to validate the system artificial entities and evaluate the quality of the when collaboration of entities expert on different consensus diagnoses provided by the system. specialities is required. Another objective will be to improve the diagnostic capability of the artifi- cial entities, particularly during the self-learning FUTURE TRENDS process, based on their performance within the CCDSS and the feedback which is available after One problem with implementing a CDSS is the the physician approves the consensus. Finally, it acquisition of the knowledge related with the will be find necessary to do an in depth study of diagnosis. It is necessary to contact a group of the consensus algorithms used, in order to ensure experts and organize their knowledge, before more precise results and to ensure that they can including the diagnosis rules or heuristics in the take into account additional variables like the system; all these actions take a lot of work and morbidity of the diagnosed illness. time. In the future, the knowledge should be ex- tracted by more automatic means which take less time. Journals and conference proceedings are an REFERENCES easy alternative way to collect useful information about diagnosis; on the other hand, in hospital Adler, R. E. (2004). Medical Firsts: From information systems there are huge databases Hippocrates to the Human Genome (1st ed.): with data related to diseases and their diagnosis Wiley. evaluation. Furthermore, the use of data from 108TRTRSC
  • 126. A Distributed Approach of a Clinical Decision Support System Based on CooperationBerger, S. A. (2001). GIDEON: A Computer Han, J., Kamber, M. (2005). Data Mining:Program for Diagnosis, Simulation, and Infor- Concepts and Techniques (2nd ed.): Morganmatics in the Fields of Geographic Medicine and Kaufmann.Emerging Diseases. Paper presented at the 2000 Hinchley, A. (Ed.). (2005). Understanding VersionEmerging Infectious Diseases Conference. 3 - A Primer on the HL7 Version 3 Communica-Berner, E. S. (Ed.). (2007). Clinical Decision tion Standard (3rd ed.).Support Systems. Theory and Practice (2nd ed.): Katsuragawa, S., Doi, K. (2007). Computer-Springer. aided diagnosis in chest radiography. Comput-Burstein, F., Holsapple, C. W. (Eds.). (2008). erized Medical Imaging and Graphics, 31(4-5),Handbook on Decision Support Systems 1: Basic 212-223.Themes. Springer. Lin, H. (Ed.). (2007). Architectural Design ofCannataro, M., Congiusta, A., Pugliese, A., Talia, Multi-Agent Systems: Technologies and Tech-D., Trunfio, P. (2004). Distributed data mining niques. IGI Global.on grids: services, tools, and applications. IEEE Luo, Y., Jiang, L., Zhuang, T. (2008). A Grid-Transactions on Systems, Man, and Cybernetics, Based Model for Integration of Distributed Medi-Part B, 34(6), 2451-2465. cal Databases. Journal of Digital Imaging.Coiera, E. (2003). Guide to Health Informatics Modarres, M., Beheshtian-Ardekani, M. (2005).(2nd ed.). London: Hodder Arnold. Enterprise support system architecture: integrat-Collins (Ed.) (2003) Collins English Dictionary. ing DSS, EIS, and simulation technologies. In-Collins. ternational Journal of Technology Management, 31(1/2), 116-128.Delen, D., Pratt, D. B. (2005). An integrated andintelligent DSS for manufacturing systems. Expert Moore, M. (1999). The evolution of telemedicine.Systems with Applications, 30(2), 325-336. Future Generation Computer Systems, 15(2), 245-254.Ferber, J. (1999). Multi-Agent Systems. An In-troduction to Distributed Artificial Intelligence: Morbiducci, U., Tura, A., Grigioni, M. (2005).Addison-Wesley. Genetic algorithms for parameter estimation in mathematical modeling of glucose metabolismGeorgopoulos, V. C., Malandraki, G. A. (2005). Computers in Biology and Medicine, 35(10),A Fuzzy Cognitive Map Hierarchical Model 862-874.for Differential Diagnosis of Dysarthrias andApraxia of Speech. Paper presented at the 27th Pomerol, J.-C. (1997). Artificial Intelligence andAnnual International Conference of the Engineer- human decision making. European Journal ofing in Medicine and Biology Society. Operational Research, 99(1), 3-25.Greenes, R. A. (Ed.). (2007). Clinical Decision Pop, D., Negru, V., Sandru, C. (2006). Multi-Support. The Road Ahead: Elsevier. Agent Architecture for Knowledge Discovery. Paper presented at the Eighth International Sym-Haas, O. C. L., Burnham, K. J. (Eds.). (2008). posium on Symbolic and Numeric Algorthms forIntelligent and Adaptive Systems in Medicine (1st Scientific Computing, Timisoara.ed.): Taylor Francis. Porter, R. (Ed.). (2006). The Cambridge History of Medicine (1st ed.): Cambridge University Press. 109
  • 127. A Distributed Approach of a Clinical Decision Support System Based on CooperationRakus-Anderson, E. (2007). Fuzzy and Rough van Bemmel, J. H., Musen, M. A. (Eds.).Techniques in Medical Diagnosis and Medication (1997). Handbook of Medical Informatics (1st ed.).(1st ed.): Springer. Houten, the Netherlands: Springer-Verlag.Ramnarayan, P., Roberts, G. C., Coren, M., Vihinen, M., Samarghitean, C. (2008). Medi-Nanduri, V., Tomlinson, A., Taylor, P. M., et al. cal Expert Systems. Current Bioinformatics,(2006). Assessment of the potential impact of a 3(1), 56-65.reminder system on the reduction of diagnostic Vitenzon, A. S., Mironov, E. M., Petrushans-errors: a quasi-experimental study. BMC Medical kaya, K. A. (2005). Functional ElectrostimulationInformatics and Decision Making, 6(22). of Muscles as a Method for Restoring Motor Func-Ramnarayan, P., Tomlinson, A., Rao, A., Coren, tions Neuroscience and Behavioral Physiology,M., Winrow, A., Britto, J. (2003). ISABEL: 35(7), 709-714.a web-based differential diagnostic aid for Waraporn, N. (2007). Confidence levels for medi-paediatrics: results from an initial performance cal diagnosis on distributed medical knowledgeevaluation. Archives of diseases in childhood, nodes. Paper presented at the International Confer-88(5), 408-413. ence on Computer Engineering and Applications,Rangayyan, R. M. (2004). Biomedical Image Gold Coast, Queensland, Australia.Analysis (1st ed.): CRC. Weisz, G. (2005). Divide and Conquer: A Com-Roberts, L. M. (2000). MammoNet: a bayesian parative History of Medical Specialization:network diagnosing breast cancer. Machine Per- Oxford University Press (USA).ception and Artificial Intelligence, 39, 101-148. White, D. (2006). Decision Theory: AldineRuiz Fernández, D., Garcia Chamizo, J. M., Maciá Transaction.Pérez, F., Soriano Payá, A. (2005). Modelling WHO (Ed.). (2005). The International Statisticalof dysfunctions in the neuronal control of the Classification of Diseases and Related Problemslower urinary tract. Lecture Notes in Computer (2nd ed.). Geneva: World Health Organization.Science, 3561, 203-212. Wootton, R., Craig, J., Patterson, V. (Eds.).Taylor, A., Manatunga, A., Garcia, E. V. (2006). Introduction to Telemedicine (2nd ed.):(2007). Decision Support Systems in Diuresis Rittenhouse Book Distributors.Renography. Seminars in Nuclear Medicine,38(1), 67-81.110
  • 128. 111 Chapter VI Managing Mobile Healthcare Knowledge: Physicians’ Perceptions on Knowledge Creation and Reuse Teppo Räisänen University of Oulu, Finland Harri Oinas-Kukkonen University of Oulu, Finland Katja Leiviskä University of Oulu, Finland Matti Seppänen The Finnish Medical Society Duodecim, Finland Markku Kallio The Finnish Medical Society Duodecim, Finland ABSTRACT Incorporating healthcare information systems into clinical settings has been shown to reduce medica- tion errors and improve the quality of work in general by improving medical decision making and by saving time. This chapter aims to demonstrate that mobile healthcare information system may also help physicians to communicate and collaborate as well as learn and share their experiences within their work community. Physicians’ usage of a mobile system is analyzed through a knowledge management framework known as the 7C model. The data was collected through the Internet among all of the 352 users of the mobile system. The results indicate that frequent use of the system seemed to improve individual physicians’ knowledge work as well as the collective intelligence of a work community. The guide for Copyright © 2009, IGI Global, distributing in print or electronic forms without written permission of IGI Global is prohibited.a
  • 129. Managing Mobile Healthcare Knowledgeacute care, evidence-based medicine guidelines and information related to drug interactions supportedthe knowledge creation to a large extent. As such, mobile healthcare information systems may be capableof supporting the different sub-processes of knowledge creation and the knowledge work of individualphysicians, and through this also improving the collective intelligence of the work community. Overall,knowledge management seems to be a prominent approach for studying healthcare information systemsand their impact on the work of physicians.InTROoduion conceptual framework known as the 7C model (Oinas-Kukkonen, 2004). It suggests that the sevenPersonal digital assistants and mobile applica- Cs or knowledge creation sub-processes, namelytions are promising tools for managing medical Connection, Concurrency, Comprehension, Com-information and accessing it at the point of care munication, Conceptualization, Collaboration,(Ebell et al., 1997). They have been shown to as- and Collective intelligence, play a central rolesist in evidence-based practice in a clinical setting in knowledge management. According to the 7Cand support the educational needs of physicians, model, going through the key phases of Com-nurses and other clinical staff, while drug informa- prehension, Communication, Conceptualizationtion, medical calculations, guideline information and Collaboration repeatedly, in a seamless andand administrative tasks have been identified as spiral-like way leads into the growth of the or-the most useful resources (Honeybourne et al., ganizational or social knowledge, i.e. Collective2006). Topics such as e-prescribing (Kushniruk intelligence.et al., 2005) and patient tracking (Lapinsky et This paper focuses on mobile access to medicalal., 2001) have gained a lot of attention recently. literature and electronic pharmacopoeias, aimingOn the other hand, access to medical literature to demonstrate that these may help physicians toand electronic pharmacopoeias, i.e. drug infor- communicate and collaborate as well as learn andmation, seem to be excellent tools for providing share their experiences within their user com-physicians with knowledge at the point of care munity. There are relatively few scientific studies(Fischer et al., 2003). on managing knowledge with mobile healthcare Incorporating healthcare information systems information systems. Moreover, only a smallinto clinical settings has been also shown to re- number of articles provide knowledge about theduce medication errors (Grasso Genest, 2001) actual use of mobile medical applications (Fischerand improve the quality of work in general by et al., 2003). We will present a case study amongimproving medical decision making and by sav- the users of Duodecim mobile healthcare informa-ing time. Mobile versions of these systems are tion system. The data was collected through therelatively easily incorporated into the workflow Internet among all of the 352 physicians (with theof the physicians (Rothschild et al., 2002) as they response rate of 66.5%, n=234), who were userscan be carried around and used when ever needed, the case system.for example during home visits or ward rounds. The article is organized as follows. Chapter II In the information systems field the topic of describes the background for the study. Chapterknowledge management has received a lot of III presents the 7C model for knowledge creationattention recently (for an excellent review on and management, Chapter IV introduces the casethe subject, see (Alavi Leidner, 2001)). Our system, Chapter V discusses the results, and finallyview on knowledge management is through a Chapter VI concludes the article.112
  • 130. Managing Mobile Healthcare Knowledge Bakg Fischer et al. (2003) classify the main uses of mobile applications for medicine as: accessing Systematic processing of health-related data, in- medical literature, electronic pharmacopoeias, formation and knowledge focusing on the study patient tracking, medical education, research of information processing principles and solutions data collection, e-prescribing, business manage- in healthcare is referred to as health informatics, ment and specialty-specific applications. Other while the scientific discipline related to it is called classifications have also been suggested, such medical informatics (Hasman et al., 1995). The as those ones by Adatia and Bedard (2003) and terms are often used synonymously, even though Embi (2001). some differences exist in their use between coun- Providing access to medical literature increas- tries. For instance, medical informatics in Ger- es the extent to which evidence will be sought and many also includes nursing informatics and dental incorporated into patient care decisions (Sackett informatics, while in other countries medical et al., 1998). The access to medical literature informatics primarily focuses on solutions from through a mobile application will allow decisions the physicians’ viewpoint (Hasman et al., 1996). to be made by the bedside or at the point of care. In general, health informatics is often examined This could further improve the decisions made by from different perspectives such as information physicians (Sackett et al., 1998). Mobile devices technology or user needs. containing decision-making tools and summaries A healthcare information system (HIS) is of evidence may also improve learning in evi- defined as an application of information technol- dence-based medicine (Honeybourne et al., 2006) ogy in healthcare, encompassing a wide range of and reduce patients’ length of stay in hospitals applications from disciplines such as medicine, (Sintchenko et al., 2005). Usually, evidence-based computer science, management science and statis- information is presented in a guideline form to tics (Raghupathi, 1997). Based on the interaction further support decision-making. between the medical personnel and the patient, Pharmacopoeias are drug information data- HISs may be categorized as customer support- bases and drug interaction guides. Drug infor- ing systems, interaction supporting systems, mation refers to information such as adult and consultation systems, decision support systems, paediatric dosing guidelines or common side process supporting systems, economic systems, effects, while drug interaction guides contain preparation tools and administrative tools (Suomi, information about possible interactions that two 2001). or more drugs used together can have (e.g. drug A According to Siau (2003), mobile healthcare may influence the absorption of drug B). Access information systems are among the basic tools to drug information may reduce medication errors employed in the healthcare industry, the other two (Grasso and Genest, 2001) as it is impossible in being Internet applications and enterprise systems. practise to know all conceivable drug interactions Mobile HISs offer two distinct advantages for the by heart. Thus providing an easy manner to dou- healthcare sector: firstly, they are important for the ble-check these interactions should indeed help success of telemedicine, and secondly they enable the work of physicians at the point of care. The physicians to access information whenever and survey conducted by Rothschild et al. (2002) with wherever needed. Access to real-time information palmtop drug information guide users suggests is important for physicians and hospitals because that mobile systems may save time in informa- information is often needed immediately to enable tion retrieval and improve drug-related decision accurate decision making (Siau, 2003). making and can be relatively easily incorporated 113ound
  • 131. Managing Mobile Healthcare Knowledgeinto the workflow of physicians. This is important, approach aims to “educating physicians to helpas it could improve technology acceptance and them bring more research and evidence into theirsave time. individual decisions about individual patients” Patient tracking systems aid medical staff in (Eddy, 2005).ward rounds by keeping them informed about Knowledge management is “a business con-the patient’s condition. Medical education ap- cept which includes concerted, coordinated andplications are designed to help medical students deliberate efforts to manage the organization’sin their studies, e.g. by monitoring their clinical knowledge […] and applying it to enhance organi-experience or by finding gaps in their education. zational performance” (Bose, 2003). Because ofMobile applications used for data collection have the growing costs of healthcare various knowledgealso been found promising for research purposes management solutions have been applied in hospi-(Fischer et al., 2003). Using mobile applications for tals and medical centers to enhance performanceelectronic prescribing has been found to decrease and e.g. provide better care.medication error rates (Grasso Genest, 2001) Yet the healthcare sector has been said toand business management applications help the be data rich but knowledge poor (Abidi, 2001).efficiency of hospitals, e.g. by improving coor- That is to say different healthcare organizationsdination and billing (Fischer et al, 2003). Finally, generate huge amounts of data from hospitaldifferent specialties (for example family medicine) reports to clinical trials but not much of the datahave their own specific mobile applications. is transformed into strategic decision-support Usually medical literature includes not only de- knowledge (Abidi, 2001).scriptions of treatment methods but also evidence Thus most of the knowledge managementsupporting each method. This form of decision solutions in healthcare have been concentratingmaking is referred as evidence-based medicine on transferring data into knowledge. One wayand is defined as “the conscientious, explicit and of doing this is to utilize data mining (Fayyadjudicious use of current best evidence in making et al., 1996). By using data mining we can finddecisions about the care of individual patients” e.g. correlations or dependencies from the vast(Sackett et al., 1996). For example, mobile decision amounts of data available to us. This way thesupport systems have contributed a significant data could be transformed into a more usablereduction in antibiotic prescribing (Sintchenko form (i.e. knowledge) for e.g. evidence-basedet al., 2005), i.e. physicians’ have been able to medicine. An example could be to use data min-see better when to prescribe antibiotics and when ing to predict patients’ length of stay in hospitalnot to. (Kraft et al., 2003). There are two approaches to applying evidence Clinical decision support systems (Teich to medicine (Eddy, 2005). The first approach is Wrinn, 2000) have also been used to utilize theto use evidence-based guidelines. Twenty years data. Usually, they combine population statisticsago medical guidelines were based on experts’ with existing knowledge to offer real-time infor-consensus but over the years most of the guide- mation to support physicians’ daily work (Teich lines have changed into evidence-based. Quite Wrinn, 2000). They can also facilitate evidence-interestingly the new guidelines have been dra- based medicine (Jadad et al., 2000).matically different than the previous ones (Eddy, Recently, information systems focusing on the2005). The second approach is to apply evidence knowledge and relationships between patients andin individual decision making (Evidence-based hospitals have also been introduced to healthcareMedicine Working Group, 1992). The differ- settings. This solution is called patient relationshipence to evidence-based guidelines is that this management (Siau, 2003) and through it healthcare114
  • 132. Managing Mobile Healthcare Knowledge unit can provide better care to patients by allow- (the 2nd C) is provided in a technologically sound ing the unit to get an increased understanding of manner, e.g. through the Web, Internet, wireless, patients’ needs. mobile and other technologies. These may pro- Besides the abovementioned the use of knowl- mote options and allow freedom of choice with edge management based solutions on healthcare contextual support, providing users with a rich offers other benefits, too. For example if we had environment for comprehending (the 3rd C) and knowledge management based healthcare sys- communicating (the 4th C) the information they tems we could better analyze different types of find. Knowledge is conceptualized (the 5th C) as knowledge found in healthcare organizations (e.g. artifacts, which serve as a vehicle for collaboration clinical knowledge stored in repositories) as well (the 6th C) through interaction between informa- as achieve significant organizational productivity tion producers and consumers, within a team of improvements (Bose, 2003). co‑workers or among other stakeholders. All of Our goal is not to define new ways to facilitate these six preceding Cs contribute to the growth knowledge creation through healthcare informa- of “togetherness” or collective intelligence (the 7th tion systems but rather to investigate knowledge C). The creation of organizational knowledge is reuse. Using the 7C model of knowledge creation not a linear process, but rather a multi-cycle spiral and management we argue that over time through process (Oinas-Kukkonen, 2004). See Figure 1. knowledge reuse healthcare organization do not The four central sub-processes in knowledge only generate new knowledge but get better at creation are comprehension, communication, their work, too. The next chapter will present the conceptualization and collaboration (Oinas-Kuk- 7C model used in this study. konen, 2004). Comprehension is a process of embodying explicit knowledge in tacit knowledge by surveying and interacting with the external 7C model of KNOWLEDGE environment on an ongoing basis in order to CREea and managemen identify problems, needs and opportunities (e.g. learn by doing or re-experiencing). The 7C model suggests that the following seven Cs Communication is a process of sharing experi- play a critical role in the creation of organizational ences between people and thereby creating tacit or social knowledge: Connectivity, Concurrency, knowledge in the form of mental models and Comprehension, Communication, Conceptualiza- technical skills, producing dialogue records which tion, Collaboration, and Collective intelligence emphasize needs and opportunities, and integrat- (Oinas-Kukkonen, 2004). The 7C model follows ing the dialogue and resulting decisions with other Nonaka and Takeuchi (Nonaka Takeuchi, 1995) project knowledge on an ongoing basis. in that the integration of individual and social Conceptualization is a collective reflection orientations (individual and organizational in their process articulating tacit knowledge to form terminology) is emphasized, and that knowledge explicit concepts and rationales and systematiz- is assumed to be created through interaction be- ing these into a knowledge system, which are tween tacit and explicit knowledge. The model iteratively and collaboratively developed, possibly follows Engelbart (1992) in the outcomes of the including proposals, specifications, descriptions, Comprehension, Communication and Conceptu- work breakdown structures, etc., but rarely a alization sub-processes. one-shot effort. The framework assumes that connectivity of Collaboration is a team interaction process all stakeholders with the joint information space of using the resulting conceptualizations within (the 1st C) and with people potentially concurrently teamwork and other organizational and social processes. 115Cnowledgeion
  • 133. Managing Mobile Healthcare Knowledge Each of the sub-processes may also be regarded their colleagues (Communication). As they shareas the building of an artifact and reasoning over they can collectively add to the knowledge of thewhy it has been built in the way it has, i.e. captur- group and create e.g. best practice guidelinesing the knowledge rationale. Repeatedly going (Conceptualization) to help them perform theirthrough these phases in a seamless and spiral-like work better in the future (Collaboration). Overway leads to the growth of collective intelligence. time, as these processes go around over and overSupport for capturing deep individual thinking again the hospital unit could get better at provid-and recording the dialogue between team members ing care for its patients (Collective intelligence).may help create truly innovative knowledge prod- In spite of receiving a lot of attention recentlyucts. The learning involved in the comprehension among practitioners, relatively little knowledgeand communication processes is closely related management research has discussed the evalua-to the attitudes of the participants, i.e. whether tion of the solutions suggested (Oinas-Kukkonen,they understand their weak points in the sense of 2005). This kind of evaluation may be carriedindividual learning styles, for example. out at the individual, work unit (group, team, It is important to notice that the 7C model or department), or overall organizational levels.does not try to define how information systems The increase in the sharing and dissemination ofshould manage knowledge. Rather, it models information and the increase in varied interpreta-the processes of how individuals interact with tions are obvious and in any case by no meansinformation and knowledge (and with each other) the most important measures of the success ofto increase the collective intelligence of the orga- knowledge management solutions. The trulynization. In a hospital physicians and nurses can important measure is the identification of under-learn and understand new things (Comprehension) lying non-obvious, complex problems and issueswhile they perform their daily work. They can (Oinas-Kukkonen, 2005).then share their work related experiences withFigure 1. Knowledge creation sub-processes (Oinas-Kukkonen 2004) Tacit Communication Individual Collective Concept- Comprehension ualization Intelligence Social Collaboration Explicit116
  • 134. Managing Mobile Healthcare Knowledge Evaluation of the Comprehension and Com- collection of clinical guidelines for primary care munication sub-processes means, for instance, combined with the best available evidence. The assessment of whether the following goals are collection includes almost 1,000 concise primary achieved: better understanding of current and care practice guidelines covering a wide range of potential future customers, the key organizational medical conditions and including both diagnosis business processes, the product portfolio, product and treatment, over 2,700 high-quality evidence features and potential future products and mar- summaries supporting the recommendations, a kets in general. Quite naturally, an improvement library of 900 high-quality photographs and im- in any of these will lead to either an increase in ages of all common and many rare dermatological new ideas or the achievement of better ideas for conditions, electrocardiograms and eye pictures future business, and it may also help solve some as well as abstracts from the Cochrane Library of the problems that the organizations are faced (which is a collection of databases in medicine with over time (by being more capable of defining and other healthcare specialties). the core processes and their key challenges) or The system also contains the pharmacology even avoid some of the pitfalls they might suffer database Pharmaca Fennica with a wireless from. (Oinas-Kukkonen, 2004). update service for a complete medicine price list, a drug interaction database for drug-related decision making, the international diagnosis code Ca DUODECIM guide known as The International Classification of Diseases ICD-10, an acute care guide, a medical The mobile healthcare information system un- dictionary of over 57,000 terms, and a compre- der investigation is evaluated at the individual hensive database of healthcare-related addresses and work unit levels. The system comprises a and contact information for pharmacies, hospitals set of medical information and knowledge da- and health centres. The system is described in tabases developed by Duodecim Publications Table 1. It is typically used through smartphones Ltd (the Finnish Medical Society Duodecim is and it is delivered to users as a memory card that a scientific society with over 18,000 of Finnish includes a search engine, user interface software doctors and medical students - almost 90% - as and the core databases. its members.). The system emphasizes the role of The knowledge base is planned to support evidence-based medical guidelines (EBMG), i.e. a physicians in their day-to-day activities. It can Table 1. The Duodecim mobile HIS under study Duodecim database Description / functionality Evidence-based medical guidelines Search for evidence-based guidelines including literature references and abstracts from the Cochrane Library. Pharmaca Fennica Drug lists, adult and paediatric dosing guidelines, common side effects. ICD-10 International Statistical Classification of Diseases and Related Health Problems. Codes for classifying diseases and a wide variety of signs and symptoms. Acute Care Guide Pathogenesis, causes, symptoms, differential diagnosis. Drug Interaction Database Possible interaction effects of selected drugs. Medical Picture Database Descriptions of symptoms and pictures. Contact Information Search for contact information on pharmacies, hospitals and health centres. 117euodeim
  • 135. Managing Mobile Healthcare Knowledge be carried around and applied in the bedside or not actually use the system, and one was deleted at the point of care. As physicians apply and re- because the respondent did not answer any of the use the knowledge they may be better equipped main questions related to this study. Thus the final with the tasks at hand. This knowledge may also data set consisted of 231 replies. support their medical decision making as well as help them to learn new things. Some earlier studies of the system (see Han et al., 2004a, Han Re et al., 2005) have demonstrated that physicians have a positive perception of it and intend to use 62.3% (n=144) of the respondents were men and it, and that the most frequently requested mobile 37.2% (n=86) women. 61.9% (n=143) were special- content entities were EBMG, Pharmaca Fennica ists, 27.3% (n=63) general practitioners, and 10.4% and ICD-10. (n=24) researchers or working in administrative positions. Most of the respondents had more than Data Collection 20 years of experience of working as a physician (55.8%, n=129), while 32.0% (n=74) had over ten The data were collected through the Internet years of experience and only 12.1% (n=28) had less. during a two-week period from January 23 to The majority of the physicians worked daily with February 7, 2007. The key knowledge creation patients (80.5%, n=186), nurses (86.6%, n=200) issues under investigation were Comprehension, and other physicians (85.3%, n=197). Communication, Conceptualization, Collabora- 45.9% (n=106) of the physicians used the tion and Collective Intelligence. The technologi- information system daily, 37.7% (n=87) several cal viewpoints of the 7C model (Connection and times a week, 11.7% (n=27) once a week, 3.9% Concurrency) were omitted as they are beyond the (n=9) once a month and only two used it less often scope of this research. Medical performance was than once a month. The two most frequently used not measured either. The questionnaire contained parts were Pharmaca Fennica drug information 18 questions. See Appendix 1. Five-point Likert (96.5%) and EBMG (88.7%). The least used was scale from “Completely disagree” to “Completely the Picture Database (n=46, 19.9%). It was the agree” with the choice “I don’t know” in the middle most recent addition to the system and not all was utilized. Physicians were very familiar with physicians had access to it yet which at least to this scale, as it had already been used in previous some extent explains its low usage (the medical studies of the same system (cf. Han et al., 2004b, society estimated that about half of the users had Han et al., 2006). the picture database installed). Besides using the The respondents were approached by email mobile HIS, 27.7% (n=64) of the physicians read with a link to the online questionnaire. The ques- emails through the mobile device and 36.4% tionnaire was sent to all of the 352 users of the (n=84) used it for other Internet services. Quite mobile system. They were all physicians who had a naturally, the less experienced physicians more smartphone of their own and the software installed often felt that the system helped them to learn in it, donated by a large international medical new things, and they also found it more useful company. It should be mentioned that while all to some extent than did the more experienced of the users were members of the Finnish Medi- physicians (see Table 2). cal Society Duodecim, they were not necessarily Interestingly, there was a slight difference working at the same hospital. The response rate in how specialists and general practitioners felt was 66.5% (n=234). Two responses were deleted about the system’s ability to support learning of from the data set because the respondents did new things. 62.9% (n=39) of the general practi- 118Cul
  • 136. Managing Mobile Healthcare KnowledgeTable 2. Experience affected the perceived usefulness and learning Experience Learning Usefulness (Chi-Square=15.445, p=0.000) (Chi-Square=7.459, p=0.024) under 10 y (n=28) 81.5% 92.9% 10-20 y (n=74) 58.1% 86.3% over 20 y (n=129) 42.2% 74.2%Table 3. A mobile HIS may improve all key knowledge creation sub-processes Independent s amples t est Lev en es T est for 95% C on fide nce Interv al o f the M ean S td. E rr or C o llec tiv e Intelligenc e E qual v arianc es ,001 ,972 -4,63 0 229 ,000 -,576 ,124 -,821 -,331 as s ume d E qual v arianc es -4,60 2 216 ,8 01 ,000 -,576 ,125 -,823 -,329 not as su me d C o mpr ehen sion E qual v arianc es 3,246 ,073 -5,50 4 226 ,000 -,692 ,126 -,939 -,444 as s ume d E qual v arianc es -5,55 8 224 ,0 10 ,000 -,692 ,124 -,937 -,447 not as su me d C o mm unic atio n E qual v arianc es ,074 ,786 -3,30 6 227 ,001 -,427 ,129 -,682 -,173 as s ume d E qual v arianc es -3,28 1 211 ,9 89 ,001 -,427 ,130 -,684 -,171 not as su me d C o nce ptualiz atio n E qual v arianc es 6,197 ,014 -2,77 5 229 ,006 -,334 ,120 -,571 -,097 as s ume d E qual v arianc es -2,73 6 205 ,8 83 ,007 -,334 ,122 -,574 -,093 not as su me d C o llabo ratio n E qual v arianc es ,035 ,851 -2,84 0 229 ,005 -,347 ,122 -,587 -,106 as s ume d E qual v arianc es -2,83 0 219 ,7 74 ,005 -,347 ,122 -,588 -,105 not as su me d S um V ar ia ble E qual v arianc es ,672 ,413 -4,42 4 229 ,000 -,445 45 ,1006 9 -,643 86 -,247 05 as s ume d E qual v arianc es -4,39 3 215 ,6 31 ,000 -,445 45 ,1014 0 -,645 31 -,245 60 not as su me dtioners felt that the mobile HIS helped them to dichotomy in the knowledge creation model. Thelearn, while 47.9% (n=68) of the specialists felt next strongest correlation were between Com-that way (Chi-Square=3.902, p=0.048). The ex- munication and Conceptualization (r=0.554),planation may be simple fact that the specialists’ Comprehension and Conceptualization (r=0.538),area of expertise is more focused while general and Communication and Collaboration (r=0.534)practitioners have to treat patients with wide while the lowest correlation was between Com-variety of symptoms. prehension and Collaboration (r=0.514). The four Cs of the knowledge creation spiral To investigate the knowledge creation spiralcorrelated with each other strongly. Interestingly, a sum variable was constructed, representing thethe highest correlations were between Compre- Comprehension, Communication, Conceptualiza-hension and Communication, i.e. the individual tion and Collaboration sub-processes (referred toside of the model (r=0.626) and Conceptualization later simply as the “sum variable”). We used theand Collaboration, i.e. the social side of the model sum approach as each of the 7C sub-processes may(r=0.675). This supports the individual-social be treated equally important. Since five responses 119
  • 137. Managing Mobile Healthcare Knowledgehad one or more missing data items related to (Honeybourne et al., 2006) but for the professionalthese, the missing data were replaced by means skills of the physicians as well.from similar respondents. The sum variable has We also analysed the different parts of thea high reliability (Cronbach’s α=0.841) and corre- system to find out which functionalities hadlates strongly with Collective Intelligence (0.629). on effect on different knowledge creation sub-This seems to confirm the interplay among the processes. As 96.5% of the physicians used thefour Cs, i.e. the spiral, indeed leads to the growth drug information (only eight did not use it) weof Collective Intelligence. could not compare users and non-users with it. A comparison between those who used the The Independent Samples T-tests between thosesystem daily and those who used it less frequently who used EBMG and those who did not suggestindicates that the daily use improves all knowl- that EMBG use improves user perception on theedge creation sub-processes as well as the sum mobile system’s help to Communicate (F=1.813,variable (see Table 3). This seems to indicate that p=0.001), Conceptualize (F=0.538, p=0.001)it actually helps physicians to perform their jobs and Collaborate (F=0.035, p=0.001). See Tablebetter and it eventually may increase the Collec- 4. It improves the spiral (F=2.195, p=0.000), andtive intelligence of the whole work community. to some extent it also affects ComprehensionThis is an important finding and provides some (F=4.949, p=0.022).empirical evidence for the usefulness of mobile The ICD-10 improved Collective Intelligenceinformation systems in healthcare in general. (F=1.550, p=0.000) and the spiral (F=0.084,Thus, a mobile healthcare information system p=0.003). Whereas ICD-10 is packed with four-would be of benefit not only for patient safety letter abbreviations of various diseases and itTable 4. Use of EBM Guidelines improves physicians’ Communication, Conceptualization and Col-laboration Independent samples test Levenes Test for Equality of Variances t-test for Equality of Means 95% Confidence Interval of the Mean Std. Error Difference Collective Intelligence Equal variances F Sig. t df Sig. (2-tailed) Difference Difference Lower Upper ,285 ,594 ,897 229 ,371 ,184 ,205 -,220 ,587 assumed Equal variances 1,003 34,055 ,323 ,184 ,183 -,189 ,556 not assumed Comprehension Equal variances 4,949 ,027 -2,731 226 ,007 -,563 ,206 -,970 -,157 assumed Equal variances -2,410 29,842 ,022 -,563 ,234 -1,040 -,086 not assumed Communication Equal variances 1,813 ,179 -3,357 227 ,001 -,681 ,203 -1,080 -,281 assumed Equal variances -2,996 29,968 ,005 -,681 ,227 -1,144 -,217 not assumed Conceptualization Equal variances ,538 ,464 -3,424 229 ,001 -,644 ,188 -1,014 -,273 assumed Equal variances -3,193 30,537 ,003 -,644 ,202 -1,055 -,232 not assumed Collaboration Equal variances ,035 ,852 -3,341 229 ,001 -,639 ,191 -1,016 -,262 assumed Equal variances -3,046 30,213 ,005 -,639 ,210 -1,067 -,211 not assumed SumVariable Equal variances 2,195 ,140 -3,941 229 ,000 -,63088 ,16007 -,94627 -,31548 assumed Equal variances -3,569 30,118 ,001 -,63088 ,17676 -,99182 -,26994 not assumed120
  • 138. Managing Mobile Healthcare Knowledge seems impossible for anyone to know all diseases little inconclusive due to its low usage. Contact and their codes by heart, it is surprising that there Information improved Collaboration (F=1.724, was no correlation with it and Comprehension. p=0.002) and the sum variable (F=0.025, p=0.004). It may well be that these abbreviations are really Interestingly it did not improve Communication, used only for healthcare management and they do even if it provided contact information. not involve diagnosing or modelling the groups Table 5 summarizes the correlations between of diseases. parts of the system and knowledge creation sub- Use of Acute Care Guide improved all processes. knowledge creation sub-processes: Collaboration (F=10.312, p=0.000), Comprehension (F=6.067, p=0.000), Collective Intelligence (F=0.339, DiSCUSSIion p=0.000), Communication (F=1.730, p=0.001), Conceptualization (F=0.001, p=0.008), as well as From the five Cs addressed in this study Compre- the sum variable (F=1.560, p=0.000). As such, the hension was improved by the use of the Acute Care use of the Acute Care Guide seems to improve Guide and Drug Interaction Database. The Acute each of the knowledge creation sub-processes. Care Guide was used slightly more often by the Use of Drug Interaction Database improved less experienced physicians, as 33.3% (n=43) of Collaboration (F=1.218, p=0.000), Conceptu- those who had more than 20 years of experience alization (F=0.979, p=0.001), Comprehension used it, compared with 48.0% (n=49) of the rest (F=0.095, p=0.001), Collective Intelligence of the physicians (Chi-Square=5.140, p=0.023). (F=0.073, p=0.010), as well as the sum variable Quite obviously, the less experienced physicians (F=0.922, p=0.000). Quite interestingly, it did not still have more to learn and comprehend. Maybe affect Communication. Maybe the drug interac- this is especially true in acute medical situations. tion information is useful in places where com- The fact that Drub Interaction Database improves munication is not required, e.g. the physicians Comprehension seems feasible too, since there has already decided that he will prescribe drug are a large number of different drugs and their A and he uses the system to check out possible combinations that it is practically impossible to interactions with existing medication. know all of their interactions. Thus an easy way Use of Medical Picture Database improved of checking these interactions should indeed help only the sum variable (F=0.000, p=0.009). The physicians and over time they may comprehend results concerning the Picture Database may be something new. Interestingly, EBMGs did not Table 5. Usage of the different parts of the system and their impact on the 7C processes Duodecim database Frequencies CI Comp Comm Conc Coll Sum var Acute Care Guide 39.8% X X X X X X Drug Interaction Database 54.5% X X X X X Evidence-based medical 88.7% X X X X guidelines Contact Information 74.5% X X ICD-10 57.6% X X Medical Picture Database 19.9% X Pharmaca Fennica 96.5% 121u
  • 139. Managing Mobile Healthcare Knowledge affect Comprehension. This might be because From the different subsystems the Acute most of the physicians were experienced and thus Care Guide improved all knowledge creation familiar with the guideline information. On the sub-processes. Mobile applications such as the other hand, most of the users (61.9%, n=143) had Acute Care Guide combine the “any time, any- specialized in certain medical domains, which im- where” possibilities of mobile applications with plies that their knowledge needs might have been actual needs in acute medical situations, where more specialized than what is provided through knowledge must be acquired and applied swiftly. the evidence-based medical guidelines. Thus, instead of concentrating on any time and Communication was improved by the EB- anywhere mobile applications in knowledge work MGs and the Acute Care Guide, which both are context might need to concentrate more on exact well-structured and evidence-based. Thus, they situations where knowledge is needed, e.g. in contain guideline information that is relatively healthcare at the point of care. easy to deliver. For example, all guidelines in Of the other subsystems Evidence-Based the Acute care guide are organized in the same Medical Guidelines and Drug Interactions seem format, i.e. pathogenesis, causes, symptoms and to support the knowledge creation sub-processes differential diagnosis. to a greater extent. Overall, guideline informa- Conceptualization was improved by the EB- tion seems to provide a good fit with knowledge MGs, Acute Care Guide and Drug Interaction creation. Guidelines contain information about Database. Indeed, evidence-based information diagnostic procedures that may be used with may help a group of physicians to reach a con- certain symptoms as well as suggestions for sensus in making medical decisions. which drugs might work best for different ill- Collaboration was improved by the EBMGs, nesses. Physicians may also find them helpful Acute Care Guide, Drug Interaction Database when consulting other physicians, as the guideline and Contact Information. It seems natural that information may provide a basis for communicat- guidelines help physicians to collaborate. Simi- ing and collaborating. A general practitioner may larly providing Contact Information helps find first check the information found in the mobile the right people. HIS, for example, and then use it as a reference Collective Intelligence was improved by ICD- when consulting a specialist. 10, Acute Care Guide and Drug Interactions. Interestingly, 65.0% (n=93) of the specialists used ICD-10, but only 39.7% (n=25) of the general prac- FuTURE RESEARCH titioners. This could mean that specialists have a greater need for the ICD-10 than general prac- Future research should be directed towards titioners, but as such it does not explain why the empirical testing of the knowledge processes use of ICD-10 improves Collective Intelligence. in more detail, e.g. what type of communica- One reason for this could be that hospitals are tion do the evidence-based medical guidelines very bureaucratic by nature and these classifica- really support and how can the transfer of tacit tions of diseases are needed in many situations, knowledge into explicit be better supported. Also e.g. when a patient checks in, when a patient’s multiple sources of data would help obtain deeper treatments are entered into hospital records, or understanding. when a patient is discharged. The use of a mobile We are also planning a longitudinal study on ICD-10 application can provide practical support the case system described in this paper. With the in these situations. longitudinal data we can see e.g. how the regu- 122ueeah
  • 140. Managing Mobile Healthcare Knowledge lar use of the system affects the 7C knowledge benefit not only for patient safety (Honeybourne creation processes. For example the collective et al., 2006) but for the professional skills of the intelligence of the hospital units should increase physicians as well. over time. A limitation of our study is that we were not We would also like to compare the case system able to go deeper with studying the differences to other ways of obtaining the same information between experienced and less experienced physi- and knowledge. For example, how does the us- cians. Another limitation would be that the picture age of the mobile system compare to e.g. books database was not in use by all of the physicians so or desktop information systems (Duodecim also the results concerning it are not conclusive. Also has the desktop version of the case system). We one limitation on our study could be that most are especially interested in finding out does the of the physicians participating in the study had case system really offer better support at the more than 20 years of work experience. It could point of care? be argued that the less experience the physician Another line of research we are interested in has, the more he has to learn and more he could is what kind of features of functionalities would benefit from the use of a mobile healthcare in- support the processes of the 7C model? We are formation system. especially interested in the comprehension and In sum, knowledge management seems to be conceptualization processes as they have received a prominent approach for studying healthcare less attention in the scientific literature than com- information systems and their impact on physi- munication and collaboration. cians’ work. Conlu Refeen This article discusses physicians’ usage of a Abidi, S. (2001). Knowledge management in mobile healthcare information system. This was healthcare: towards “knowledge-driven” deci- studied through the 7C knowledge management sion-support services. International Journal of framework. The usage of the system seemed to Medical Informatics, 63(1)2 , 5-18. improve the knowledge work of individual physi- Adatia, F., Bedard, P. (2003). Palm reading: cians as well as the collective intelligence of work 2. Handheld software for physicians. Canadian community. The easiest sub-process to support Medical Association Journal, 168(6), 727-734. through the system seemed to be collaboration between the physicians. Comprehension and, Alavi, M., Leidner, D. E. (2001). Review: quite surprisingly, Communication were the Knowledge management and Knowledge Man- most difficult aspects to support. All parts of agement Systems: Conceptual Frameworks and the case system helped improve the knowledge Research Issues. MIS Quarterly, 25(1), 107-136. creation spiral. Bose, R. (2003). Knowledge management-enabled These findings go hand in hand with some health care management systems: capabilities, of the previous findings (e.g. Ebell et al., 1997, infrastructure, and decision-support. Expert Honeybourne et al., 2006) of the usefulness of Systems with Applications, 24, 59-71. healthcare information systems, especially in acute medical situations where decisions have to be Ebell, M. H., Gaspar, D. L., Khurana, S. (1997). made swiftly. The findings also hint that the daily Family physicians’ preference for computerized use of such a system may indeed over time be of decision-support hardware and software. The Journal of Family Practice, 45(2), 127-128. 123ione
  • 141. Managing Mobile Healthcare KnowledgeEddy, D. M. (2005). Evidence-Based Medicine: A Han, S., Harkke, V., Mustonen, P., Seppänen, M.,Unified Approach. Health Affairs, 24(1), 9-17. Kallio, M. (2005). Understanding physician acceptance of mobile technology: insights fromEmbi, P. (2001). Information at hand: Using hand- two telephone interviews in Finland. Interna-held computers in medicine. Cleveland Clinic tional Journal of Electronic Healthcare, 1(4),Journal of Medicine, 68(10), 840-853. 380-395.Engelbart, D. (1992). Toward High-Performance Han, S., Mustonen, P., Seppänen, M., Kallio,Organizations: A Strategic Role for Groupware. M. (2006). Physician’s acceptance of mobileIn proceedings of the GroupWare ‘92 Confer- communication technology: an exploratory study.ence, San Jose, CA, August 3-5, 1992, Morgan International Journal of Mobile Communications,Kaufmann Publishers. 4(2), 210-230.Evidence-based Medicine Working Group (1992). Hasman, A., Albert, A., Wainwright, P., Klar, R., Evidence-based Medicine: A New Approach Sosa, M. (1995). Education and training in healthto Teaching the Practice of Medicine. Journal informatics in Europe, State of art, guidelines,of the American Medical Association, 268(17), applications. In Studies in Health Technology and2420–2425. Informatics, IOS Press, Amsterdam, 25.Fayyad, U. M., Shapiro, G. P., Smyth, P., Uthuru- Hasman, A., Haux, R., Albert, A. (1996). Asamy R. (Eds.) (1996). Advances in knowledge systematic view on medical informatics. Com-discovery and data mining. California: AAAi puter methods and Programs in Biomedicine,Press. 51, 131-139.Fischer, S., Stewart, T., Mehta, S., Wax, R., Honeybourne, C., Suttont, S., Ward, L. (2006).Lapinsky, S. E. (2003). Handheld computing Knowledge in the Palm of your hands: PDAs in thein medicine. Journal of the American Medical clinical setting. Health Information and LibrariesInformatics Association, 10(2), 139-149. Journal, 23, 51-59.Grasso, B., Genest, R. (2001). Clinical comput- Jadad, A. R., Haynes, R. B., Hunt, D., Browman,ing: Use of a personal digital assistant in reduc- G. P. (2000). The Internet and evidence-baseding medication error rates. Psychiatric Services, decision-making: a needed synergy for efficient52(7), 883–886. knowledge management in health care. CanadianHan, S., Harkke, V., Mustonen, P., Seppänen, M., Medical Association Journal, 162 (3), 362-365.Kallio, M. (2004a). Mobilizing medical informa- Kraft, M. R., Desouza, K. C., Androwich, I.tion and knowledge: some insights from a survey. (2003). Data Mining in Healthcare InformationProceedings of the 12th European Conference Systems: Case Study of a Veterans’ Administra-on Information Systems (ECIS), Turku, Finland. tion Spinal Cord Injury Population. ProceedingsHan, S., Mustonen, P., Seppänen, M., Kallio, of the 36th Hawaii International Conference onM. (2004b). Physicians’ behaviour intentions System Sciences (HICSS’03). Jan. 6-9.regarding the use of mobile technology: an ex- Kushniruk, A., Triola, M., Borycki, E., Stein, B.,ploratory study. Proceedings of the 8th Pacific Kannry, J. (2005). Technology induced errorAsia Conference on Information Systems (PACIS), and usability: the relationship between usabilitySanghai, China. problems and prescription errors when using a124
  • 142. Managing Mobile Healthcare Knowledgehandheld application. International Journal of reference guide. Journal of the American MedicalMedical Informatics, 74(7-8), 519-526. Informatics Association, 9(3), 223–229.Lapinsky, S. E., Weshler, J., Mehta, S., Varkul, Sackett, D .L., Rosenberg, W. M. C., Gray, J. AM., Hallett, D., Stewart, T. (2001). Handheld .M., Haynes, B.. Richardson, W. S. (1996).computers in critical care. Critical Care, 5(4), Evidence based medicine: what it is and what it227–231. isn’t. British Medical Journal, 312(13), 71-72.Nonaka, I., Takeuchi, H. (1995). The Knowl- Sackett, D. L., Straus, S. (1998) Finding andedge-Creating Company — How Japanese applying evidence during clinical rounds: theCompanies Create the Dynamics of Innovation. evidence cart. Journal of the American MedicalOxford University Press. Association, 280(15), 1336–1338.Oinas-Kukkonen, H. (2004). The 7C Model for Siau, K. (2003). Health Care Informatics. IEEEOrganizational Knowledge Sharing, Learning Transactions on Information Technology in Bio-and Management. Proceedings of the Fifth Eu- medicine, 7(1), 1-6.ropean Conference on Organizational Knowl- Sintchenko, V., Iredell, J., Gilbert, G., Coiera,edge, Learning and Capabilities (OKLC ‘ 04), E. (2005). Handheld Computer-based DecisionInnsbruck, Austria, April 2-3. Support Reduces Patient Length of Stay andOinas-Kukkonen, H. (2005). Towards Evaluating Antibiotic Prescribing in Critical Care. JournalKnowledge Management through the 7C Model. of the American Medical Informatics Association,Proceedings of the European Conference on In- 12(4), 398-402.formation Technology Evaluation, (ECITE ’05), Suomi, R. (2001). Streamlining Operations inTurku, Finland, September 29-30. Health Care with ICT. In T.A. Spil R.A Steg-Raghupathi, W. (1997). Health Care Information wee (Eds.), Strategies for Healthcare InformationSystems. Communications of the ACM, 40(8), Systems, Idea Group Publishing: Hershey, PA,81-82. 2001, 31-44.Rothschild, J. M., Lee, T. H., Bae, T., Bates, Teich, J. M., Wrinn, M. M. (2000). ClinicalD.W. (2002). Clinician use of a palmtop drug decision support systems come of age. MD Com- puting, 17(1), 43-46. 125
  • 143. Managing Mobile Healthcare KnowledgeAppendix ADemographics1. Gender Male / Female 2. Experience Less than 1 years / 1-5 years / 5-10 years / 10-20 years / over 20 years3. Occupation General practitioner / Specialist / Researcher / Management position4. I use the mobile databases Daily / A few times a week / Once a week / Once a month/ Less than once a month5. I use the following parts of the system EBM guidelines Pharmaca Fennica ICD-10 Acute care guide Drug interactions Picture database Connection information 6. I work with hospital management Daily/A few times a week/Once a week/Once a month/Less than once a month/Never7. I work with physicians Daily/A few times a week/Once a week/Once a month/Less than once a month/Never8. I work with nurses Daily/A few times a week/Once a week/Once a month/Less than once a month/Never9. I work with patients Daily/A few times a week/Once a week/Once a month/Less than once a month/NeverThe medical databasesPlease, answer using these criteria:1 = Completely disagree, 2 = Partially disagree, 3 = I don’t know4 = Partially agree, 5 = Completely agree10. The mobile medical databases increase the professional capabilities of my work community. 1 2 3 4 511. The mobile medical databases help me better comprehend issues relatedto work at hand. 1 2 3 4 512. The mobile medical databases help me communicate better. 1 2 3 4 513. The mobile medical databases help the working community to reach a consensus in issuesrelated to work at hand. 1 2 3 4 514. The mobile medical databases support collaboration. 1 2 3 4 515. This mobile service makes me to learn new things. 1 2 3 4 516. In my opinion, this is a useful mobile service. 1 2 3 4 5126
  • 144. Managing Mobile Healthcare KnowledgeThe use of mobile Internet17. Do you read email with you mobile phone Yes / no18. Do you use your mobile phone for other internet services. Yes / no 127
  • 145. Section IIPatient Monitoring and Wearable Devices
  • 146. 129 Chapter VII Patient Monitoring in Diverse Environments Yousef Jasemian Engineering College of Aarhus, Denmark ABSTRACT Recording of physiological vital signs in patients’ real-life environment could be especially useful in management of chronic disorders; for example for heart failure, hypertension, diabetes, anorexia nervosa, chronic pain, or severe obesity. Thus, monitoring patients in diverse environments, by a mobile health system, is one of the major benefits of this approach, however at the same time the demands and chal- lenges for improving safety, security and integrity increase. Top priorities for patients under recovery of health and elderly under care are the feeling of being cared securely and safely in there home and its surroundings. Solving these issues will elevate users’ compliance and trust to mobile health services. Most research activities have been focused on achieving common platform for medical records, moni- toring health status of the patients in a real-time manner, improving the concept of online diagnosis, developing or enhancing telemedicine solutions, which deals with remote delivery of health care services applying telecommunications, etc.This chapter intends to explore the issues and limitations concerning application of mobile health system in diverse environments, trying to emphasize the advantages and drawbacks, data security and integrity suggesting approaches for enhancements. These issues will be explored in successive subsections by introducing two studies which were undertaken by the author. INTRODUCTION public by taking use of information and communi- cation technologies. Most research activities have In recent years, initiatives have been taken both been focused on achieving common platform for from academia and by the industries with a view medical records, monitoring health status of the for improving the health care and safety of the patients in a real-time manner, improving the con- Copyright © 2009, IGI Global, distributing in print or electronic forms without written permission of IGI Global is prohibited.aTRCT
  • 147. Patient Monitoring in Diverse Environments cept of online diagnosis, enhancing security and bandwidth and relatively high loss rate. Finally, integrity of the patients, developing or enhancing the most fundamental challenge is the security telemedicine solutions, which deals with remote and privacy of sensitive patient data. Because delivery of health care services applying telecom- the data is transmitted wirelessly, it is easy for munications, etc. (Freedman, 1999; Shimizu, an eavesdropper with a properly tuned receiver 1999; Woodward, Istepanian Richards,2001; to intercept the data. Hence, mechanisms must Jasemian Arendt-Nielsen, 2005a; Bønes, exist for data authenticity and integrity. Moreover, Hasvold, Henriksen Strandenaes, 2006; Sax, patients’ compliance concerned in a mobile health Kohane Mandl, 2005). is an important issue in focus. Recent advances in embedded computing The present chapter intends to explore the systems have led to the evolution of wireless issues and limitations concerning application of and mobile health services, consisting of small mobile health system in diverse environments, try- battery-powered entities with computation and ing to emphasize the advantages and drawbacks, radio communication capabilities. This permits data security and integrity suggesting approaches data gathering and computation to be deeply in- for enhancements. These issues will be explored tegrated in the patients’ daily environment. The in successive subsections by introducing two technology has also the potential of automatically studies which were undertaken by the author of collecting vital signs to be fully integrated into the the present chapter. patient care record and used for real-time triage, correlation with hospital records, and long-term observation. BACKGROUND AND MOTIVATIONS During the past few years, advances in sen- sor technology have enabled the development of The number of people with chronic diseases such small, lightweight medical sensors such as pulse as heart arrhythmia, diabetes, cancer and chronic oximeters and electrocardiogram leads that can be obstructive pulmonary disease (COPD) is increas- worn by the patient while wirelessly transmitting ing in most Western countries, and the majority data. This frees the patient from the confinement are elderly. Chronic diseases are the leading causes of traditional wired sensors, allowing him or her of death and disability, and these accounts for 70 to move at leisure and increasing comfort in daily % of all deaths in the U.S., which is 1.7 million environment. It is foreseen that with the help of each year (National Centers for Chronic Disease these enhanced mobile health systems, better Control and Prevention, 2008). Almost 25 million health care and services can be delivered to users, people have major limitations in daily living in and hospitals can also benefit a better informa- the United States (National Centers for Chronic tion management and administration. Also, it Disease Control and Prevention, 2008). Chronic will provide the users the ability to access their disease is a growing problem in the United States. medical records anywhere, anytime. More than 125 million Americans had at least 1 As the patients in a mobile heath system are chronic care condition in 2000, and this number monitored in diverse environments, several chal- is expected to grow to 157 million by the year lenges are present. First, Mobile health network 2020 (Marchibroda, 2008).Some of the challenges may contain a large number of mobile sensors due associated with chronic care management ap- to the mobility of patients. Second, the timeliness proaches are the use of telemedicine and mobile and reliability of data delivery is crucial in mo- health services. bile health services. Third, wireless and mobile Nowadays, in USA, Canada, Australia and communication media generally have limited many European countries, the health authorities 130BCRTS
  • 148. Patient Monitoring in Diverse Environmentstend to optimise the resources most effectively. challenge. Not all patients show high complianceBy introducing e-health, telemedicine and mobile with the application of these home monitoring de-health services, it became possible to treat/moni- vices. Nor this monitoring arrangement providestor as many patients as possible at remote areas comfort to them. Because, these devices do not(BeWell Mobile Technology Inc, 2006; Farmer, function on-line and the arrangement requires aGibson, Hayton, Bryden, Dudley et al., 2005, number of hospital visits in order to deliver theFriedewald, Pion, 2001; Logan, McIsaac, Tisler, recorded data on a tape or memory card to theIrvine, Saunders et al., 2007). Thanks to emerging specialists, which is sometimes a stressful andtechnologies the elderly now have the opportunity time consuming process for the patients.to stay longer in their homes and manage everyday To address these drawbacks, some more ad-tasks without significant burden for their caregiv- vanced ECG monitoring systems are emerging. Aners. Improving the quality of life for patients is online mobile health service has been suggestedalso an essential task in these countries (Sneha as an alternative to the above mentioned monitor- Varshney, 2007; Cocosila, Coursaris Yuan, ing methods (Jasemian Arendt-Nielsen, 2005c;2004; Jasemian, 2006; Jasemian, 2008; McLean, Clarke, Bratan, Kulkarni Jones, 2007; Gostin,Mendis, Harris Canalese, 2007). 1965). Some can also detect and signal a warn- Cardiovascular disease is the main cause of ing in real-time if any adverse event is captureddeath in the UK and it accounts for 39% of all (Standing, Dent, Craig Glenville, 2001).death each year. Among patients who had heart The trend of providing more and more wire-attacks, about 30% of them died even before less health care solutions is especially visible,reaching to the hospital (Petersen, Peto Rayner, because going wireless is supported by the tel-2004). Although heart attack can happen suddenly ecommunications service providers as well as bywithout apparent indications, cardiac arrhythmia the end-users. For users, wireless means beingcan often be found before the event. They can free from inconvenient cables and thereby morepotentially be used as the precursor to major mobility plus easier and more flexible access tocardiac episodes (Panidis Morganroth, 1983). healthcare services. For operators and providers,In Aalborg hospital, Denmark, for instance, 40 % wireless means cheaper access, more users on theof the heart arrhythmia patients are hospitalized network and more benefit.for disease monitoring and control purpose. They In many situations different telecommunica-are all monitored by short range telemetry in the tion systems coexist thus forming a heterogeneoushospital for one to five days (patient registration telecommunications environment. This, however,section, Aalborg Hospital, Denmark). Non-risky does not exhaust the problem of heterogeneity. Theheart patients are referred for monitoring at home term applies also to the coexistence of differentby HOLTER or event recording devices. operators and service providers. Finally, different Currently, electrocardiogram (ECG) Holter users’ need can also be defined as heterogeneous.monitoring is the most widely used technique Due to these aspects, addressing variety of sys-for providing ambulatory cardiac monitoring tems, environments, services, technologies andfor capturing rhythm disturbances. A traditional needs has already now become a big problem forHolter monitor can record up to 24 hours of ECG the telecommunications technology and it is verysignals, and the recorded data is subsequently likely to continue gaining importance.retrieved and analyzed by a clinician. Due to the Wireless technology has its own advantagesshort duration involved and the unknown context and drawbacks. Among the advantages, mobil-within which the ECG signal is captured, reliable ity and flexibility are important characteristics.interpretation of the recorded data is always a Among the drawbacks, major problems with a 131
  • 149. Patient Monitoring in Diverse Environments mobile health system are safety, data security and approach, however at the same time the demands integrity. These concerns increase, as the patients and challenges for improving safety, security and are monitored in diverse environments. Research integrity increase. for optimization of security has been done and in Security relates to the means by which an this relation a number of security arrangements entity protects the privacy of any information, have been suggested (Elmufti, K., Weerasinghe, and it depends very much on the applied commu- D., Rajarajan, M., Rakocevic, V., Khan, S. 2008; nication technology and data processing. Privacy MacDonald, JA. 2008). refers to the individual’s right to keep certain data The remote monitoring of patients, for chronic or information private, unless that information diseases or to follow up elderly people at home, is will be used or disclosed with his/her permission a particular application of the promising mobile (Jasemian, 2006). healthcare services in home environment. The The rapidly emerging infrastructure of health follow up of patients at home must satisfy the care information and its relation to patient privacy same security standard as it does in the hospital. have been described in the literature (Gostin, The mobile healthcare device plays the role of the Brezina, Powers, Kozloff, Faden et al., 1993; human to machine interface. This communicates Gostin, Lazzarini, Neslund Osterholm, 1996; with a home care station that should contain the Boncella, 2002). Authentication, confidentiality following features: sufficient local data storage and integrity of the transferred information are (memory), a local processing facility (real time minimum requirements any patient will demand alarm agents), a communication mediator, an (Boncella, 2002). Security and privacy are very authentication agent (PIN), and a “fire wall” to much intertwined; indeed this is security that preserve data from piracy. assures the privacy. Thus, top priorities for patients under recovery of health and elderly under care are the feeling Security and Privacy in a Wireless of being cared securely and safely in there home Remote Medical System for Home and its surroundings. Solving these issues will Healthcare Purpose (Study 1) elevate users’ compliance and trust to mobile health services. The study explores data security and patients’ privacy in a wireless remote patient monitoring system which has been designed, implemented and SECURITY AND INTEGRITY tested in a clinical setup by the author (Jasemian ARRANGEMENTS IN M-HEALTH et al., 2005a; Jasemian et al., 2005c; Jasemian SOLUTIONS Arendt-Nielsen, 2005b). A wireless remote patient monitoring system Recording of physiological vital signs in patients’ (Figure 1) consisting of a patient-unit (an ECG real-life environment could be especially useful device, a Bluetooth module and a Mobile phone), in management of chronic disorders; e.g. for heart public GSM/GPRS network, a GSM/GPRS modem failure, hypertension, diabetes, anorexia nervosa, server, and a graphical monitoring station were ar- chronic pain, or severe obesity. This could also ranged and setup. To make the setup functioning in be used to provide feedback about someone’s a reliable manner with god performance a generic health in the form of behavioural feedback in communication platform based on Bluetooth and order to prevent diseases. Thus, monitoring GSM/GPRS protocols were designed, developed patients in diverse environments, by a mobile and integrated (Jasemian et al., 2005a; Jasemian health system, is one of the major benefits of this et al., 2005c; Jasemian et al., 2005b). 132SCRTRRTS
  • 150. Patient Monitoring in Diverse EnvironmentsS System Functionality (Study 1) Health care personal at the hospital have the possibility to communicate with the patient by The ECG signal is collected, via 4 disposable sending him/her text message (in packet format). electrodes, by an ECG device. The ECG device is In case an audio conversation is needed, either a connected to a Bluetooth module, which transfers mobile phone (an extra one) which is on the patient data via Bluetooth connection to a mobile phone outdoors, or a fixed telephone at home is used. (Figure 1). The Bluetooth module invokes the mobile phone as soon as the ECG device has de- The Benefit of the System (Study 1) tected any electric activity of the heart. The mobile phone establishes a GSM or a GPRS connection Most of the existing telemetry devices are off- to the public mobile network automatically. The line (Store and Forward Telemedicine) and rely transmission of data, from mobile phone to Modem on wired telecommunication network such as Server at the hospital, is carried out in real time Digital Subscriber Line, Public Switched Tele- and continuously in packet format. The Modem phone Network and Integrated Services Digital Server receives the data and converts it to a pre- Network. Even though, very few devices/systems defined format. The data are sent to the central applying wireless and cellular technologies, those monitoring station via a serial cable. Central are most off-line, and the majority use Wireless monitoring station interoperates and converts Local Access Network and Internet connection, the received data to graphical ECG (Jasemian which make these telemetry devices dependent et al., 2005a; Jasemian et al., 2005c; Jasemian et on a fixed access point and fixed infrastructure. al., 2005b). The mobile phone is connected in the Although, these telemedicine models have a course of the real-time monitoring period. In case reasonable performance but need a great deal the network connection fails or no GSM/GPRS of preparation from the network provider side network coverage, the Bluetooth module automati- before any application, as regards installation cally, via the mobile phone, repeatedly attempts and logistics. Moreover, these models limit users’ for connection reestablishment until a complete movement freedom and bound them only to their connection is established. When the connection is home environment and very close surroundings. established, the GSM mobile phone is functioning This telemedicine setup needs also sophisticated as a mobile modem to the ECG device. security management, as hackers can easily Figure 1. A principal sketch of the wireless remote monitoring system consisting of an ECG device, Bluetooth module attached to ECG device, mobile phone, GSM/GPRS network, mobile modem server and a central monitoring station. 133S
  • 151. Patient Monitoring in Diverse Environments intrude the internet and access vital patient’s wearing the patient unit, was not permitted. information. One of the main factors that make The ECG data was transferred anonymously a telemedicine system a success is the use of a via Bluetooth-GSM/GPRS connection in packet secure modern communication format, assigning each patient an id-number, start The designed and implemented telemedicine monitoring time and date. No name, personal id- model in the present study employs advanced number, age, or address, were transferred along wireless and mobile technologies (Bluetooth with the ECGs. protocol, and TCP/IP connection over GSM Only healthcare personal knew whom each and GPRS) utilising the existing public cellular ECG was belonging to. The data were collected network (Jasemian et al., 2005a). One the main and processed by a modem server at destination benefits of this model is that there is no need for any side. The ECGs in the graphical interface, on the preparation regarding installation and logistics server side, were identified by patients’ id-num- from the network provider side, and the patients bers. The server was assigned a user-name and a need only a short instruction in the employment password, which were known only by the in charge of the telemedicine device at the hospital/health health care personal. Data security from the tech- centre before using it (Jasemian et al., 2005c), and nological point of view was investigated (Jasemian application of a mobile phone is more common et al., 2005a). The applied telecommunication in these days. The second benefit is that the pa- technologies and services (Bluetooth, GSM and tients are not bounded to their home environment GPRS) offered Access Control, Authentication, and surroundings, and they can move wherever Data Encryption, and User Anonymity. there is network coverage, thus the telemedicine The privacy and security of the transferred device is not dependent on a fixed infrastructure. ECGs were judged by a committee consisting Moreover, the system takes advantage of using the of three competent persons who were blind to solid security arrangement build in the Bluetooth, the experiments. The wireless remote patient GSM and GPRS security protocols. Hence, the monitoring system was inspected and examined present model guarantees as well the Portability in order to explore any possible intrusion from as the Accountability of the system. unauthorised persons and to unveil any possible impersonalised ECG data. In this relation the Method and Material (Study 1) authentication, confidentiality and integrity of the data were tested for the risk of Insertion attacks, Fifteen non risky heart patients (n = 15), aged Client-to-client attacks and Misconfiguration. (49±14) years (6 males and 9 females) were re- cruited. The patients’ ECGs were continuously Results (Study 1) monitored (72 h), while they were performing their every day’s indoors and outdoors activities. Without knowing user-id and password of the pa- Following the instructions, the patients wore the tients on the patient side, no access was possible; patient-unit, mount the disposable electrodes by the Bluetooth module which controls connection own self. establishment and termination, data flow and For safety reasons, the patients had a fixed dial-up connection could not be accessed by any telephone line at home and were equipped with unauthorised person as well. No received ECG an extra mobile phone when they were outdoors, could be personalized at the server side and only and they were promptly contacted in case there the authorized healthcare personnel could access was any technical or health problem (Jasemian et the data on the server side. However, the achieved al., 2005c). Taking shower or swimming, while results could not be generalised, since the pres- 134SR
  • 152. Patient Monitoring in Diverse Environments ent remote patient monitoring system was tested A telemedicine system composed of a patient- on limited number of patients (n = 15), only few unit (an ECG device, a Bluetooth module and a health care providers were involved (n = 4), and Sony Ericsson T610 mobile phone), GSM/ the system was tested only within one specific GPRS network, a router, a data interpreter and healthcare environment with specific security and a monitoring system were used (Figure 2). The privacy policies. However, the applied approach system is designed implemented and tested by for security and privacy measurement and evalu- the author (Jasemian et al., 2005a; Jasemian et ation were basic and fundamental; hence the used al., 2005b). method is valid. System Functionality (Study 2) COMFORT, COMPLIANCE AND A telemetry device collects the ECGs from the TR A MOBILE HEALTH patient’s chest via 4 disposable electrodes. The SM telemetry device is connected to a Bluetooth mod- ule via a serial interface. The Bluetooth module is The success of integration and adaptation of a wirelessly connected to a mobile phone (Figure 2). mobile health technology depends on the pa- The Bluetooth module is designed to invoke the tients’ compliance and trust to the introduced mobile phone to establish either a GSM or a GPRS system. Trust is a composition of many different connection automatically. The transmission of the attributes; reliability, dependability, honesty, data, from the mobile phone to the server at the truthfulness, security, competence, timeliness hospital, is carried out via GSM/GPRS network. and comfort, which may have to be considered On hospital side the interpreter receives the data depending on the environment in which trust is through a router and converts it to pre-defined being specified. format. The data are sent to the monitoring system According to The Compact Oxford English via serial cable. The monitoring system converts Dictionary (Compact Oxford English Dictionary, the received data to graphical ECG (Jasemian et 2007), trust is “firm belief in the reliability, truth, al., 2005a; Jasemian et al., 2005b). The mobile ability, or strength of someone or something”. phone is connected in the course of the real-time A trustworthy entity will typically have a high monitoring period. In case of network failure or reliability and so will not fail during the course no GSM/GPRS network coverage, the Bluetooth of an interaction, will perform a service or action module via mobile phone, repeatedly attempts within a reasonable period of time, will tell the for connection reestablishment until a complete truth and be honest with respect to interactions connection is established. The system is equipped and will not disclose confidential information. with alarm procedure, and gets benefit of the in- tegrated data security arrangements in Bluetooth, Elderly Comfort and Compliance to GSM and GPRS (Jasemian, 2006). Modern Telemedicine System (Study 2) Method and Material (Study 2) The aims of the present study are to investigate, Twenty four non risky elderly heart patients, aged verify and evaluate elderly patients’ compliance, (60±5) years (12 males and 12 females), were in- trust and comfort in relation to a real-time wire- cluded. A week of continuous ECGs for each of less telemedicine system at home. the elderly was recorded. The experiments were carried out while the elderly were wearing the 135CRTSTTBS
  • 153. Patient Monitoring in Diverse Environments Figure 2. A principal sketch of the wireless remote patient monitoring system containing ECG device, Bluetooth module, mobile phone, GSM/GPRS network, router, data interpreter and a graphical ECG monitoring system. patient-unit, performing their every day’s indoors elderly scored the user friendly and usability of the and outdoors activities. They were instructed how system as good. The majority (92 %) could easily to mount the disposable electrodes, how to operate manage employing the system. Only 16 % of the the patient unit and how and when they should elderly sought help from the healthcare personnel change/ recharge the batteries. They were asked in relation to employment of the patient-unit, and not to shower while wearing the patient-unit. The few (4 % - 8%) had problem with changing or re- elderly had a fixed telephone line at home and charging batteries for the mobile phone and ECG were equipped with an extra mobile phone when device as well. 76 % of the elderly (n = 24) scored they were outdoors. They were contacted in case the reliability of the system as “reasonable”, 20 there was any problem. The patients were asked % as “only now and then”, and only 4 % scored to keep a diary of their daily activities. the reliability as “excellent”. To evaluate the elderly compliance, trust and The majority of the elderly believed that their comfort in respect to the present telemedicine expectation to privacy was in 84 % fulfilled. system, three questionnaires were designed and And only 4% believed that their expectation to prepared. The first one was for the evaluation of privacy was only now and then fulfilled. Almost the system’s degree of user-friendly, usability and 52 % of the elderly patients (n = 24) scored for a reliability, the second one was for the evaluation reasonable mobility and freedom, and the rest 20 of the patients’ privacy, freedom and mobility % scored for not complete fulfilment of freedom during monitoring period, the third one for the and mobility. evaluation of the patients’ degree of confidence Only 12 % of the elderly patients do not trust and trust in respect of using the present wireless the present wireless remote monitoring system remote monitoring system at home. at all, whereas 72 % trust the system. More than 60 % of the elderly are used to employ a mobile Results (Study 2) phone in daily life and 50 % have a reasonable understanding of the system application. 80 % The elderly spent (15 ± 3) minutes to learn how believe that their comfort is satisfied. Eighteen out to use the patient-unit. In average, 80 % of the of twenty four elderly patients (76 %) preferred to 136RS
  • 154. Patient Monitoring in Diverse Environments be monitored from their home (in more natural Data security, patients’ trust and compliance, environments). in relation to a mobile healthcare system were the central concern in these two studies. In this connection, the patient’s mobility, freedom, pri- DISCON AND CONCLUSION vacy, and comfort in addition to the user-friendly and reliability of the system, were verified and Ubiquitous computing environment is the fun- evaluated. damental of a mobile health care system. In The results in first study showed that the this environment multiple mobile devices and system was reliable, functioning with a clini- healthcare personal are combined to provide an cally acceptable performance, and transferred all-pervasive computing and communications ser- medical data with clinically acceptable quality, vice to end-users. Advanced medical technologies even though the system was tested under totally provide solutions for distant home care in form uncontrolled circumstances during the patients’ of specialist consultations and home monitoring. daily activities (Jasemian et al., 2005c). A number This requires automatic configuration of certain of important design techniques that were tightly aspects of some devices, since there is no global coupled with the real-time patients’ monitoring management infrastructure, yet. Ubiquitous were elaborated, in order to enhance the system computing is still at an early stage of research and performance. The ECG data were transferred development, and very few work environments anonymously via Bluetooth-GSM/GPRS connec- have been described. tion in packet format, assigning each patient an Many threats existing in a mobile health system id-number, time and date of monitoring onset. No are the same as those arising in a more conven- name, personal id-number, age, or address, were tional environment. However, there are also new transferred along with the ECGs. Only healthcare threats in an M-health setup, e.g. mobile devices personnel knew whom each ECG was belonging typically offer less physical security and it may to. On the server side on the graphical interface, need to communicate with other devices where a the ECGs were identified by patients’ id-numbers. single security infrastructure is not present, e.g. This was a secure way for providing anonymity, in a hospital. For instance, a mobile device may and was practical only because the number of capture personal/medical information without the patients was limited and the study was fully requiring user consent or knowledge. controlled. However, when the setup is applied in a While Information Technology (IT) enables larger scale in medical practice a very careful and the use of security arrangement in medical remote precise Id-number assignment system should be monitoring system to limit access to confidential designed and elaborated, otherwise a little mistake information, it also introduces some real vulner- can cause confusion, some data mismatches and ability. Unless proper controls and procedures lost of some data identification. are implemented, these kinds of applications also To evaluate the safety and security of the trans- invite unauthorized users to access the data. If the ferred data in the proposed system a number of concerns are not sufficiently addressed, the health tests and control were worked out with great care. care consumers will hesitate to share informa- Although in the first study the clinical application tion. Therefore, IT application development and of the system was implemented in a small scale, use of that in remote monitoring system must be the ECG data was secured and patients’ privacy done in the midst of maintaining confidentiality, was achieved, though was not guaranteed. privacy, and security. However, if the setup should be tested and evaluated in a large scale, where larger number 137SSCS
  • 155. Patient Monitoring in Diverse Environments of patients is involved, several health care pro- patients’ home and the elderly are reasonably viders are in charge, data magnitude is huge, and confident in using it. However, the patient-unit the setup is tested in several health care centres should be designed more user-friendly (small with deferent infrastructure and different security and lightweight), and the ECG sensors should policy, then there will be a need for development be enhanced. These subjects also need further of security standards for the management of au- investigation in a larger scale. thority access and coding structure. Although reasonable and clinical acceptable In the second study the patients’ compliance, results have been achieved, the studies had some trust and comfort in relation to a real-time wireless limitations. Some confounding factors such as telemedicine system at home were investigated. In age, social status, education and gender difference this connection, the patient’s mobility, freedom, were also presented and should have been treated trust and compliance in addition to the system’s and analyzed. But this was not possible because user friendly and reliability were verified and of the small number of patients in the study. evaluated. The results showed that the system Hence, the results of both studies could not be in second study was reliable, functioned with a generalized. However, the evaluated parameters clinically acceptable performance, even though are essential key issues in mobile health services the system was tested under totally uncontrolled hence these should be explored, investigated and circumstances while the patients’ were performing evaluated in a large scale and multidimensional their daily activities indoors and outdoors. The pa- environments. tients have expressed reasonable compliance and trust to the application of the system at home; the more natural environment. The majority believed FUT ON AN that their comfort was satisfied. However, a few M-HEALTH SYSTEM numbers of the elderly were not satisfied with the weight and user interface of the ECG device. These Before trying to decide how to provide and sup- issues bring up an important principal approach port privacy in a mobile health environment, in a system design and development, namely we need to explore the issues that privacy can patients’ satisfaction relies on a more user-driven arise. This requires identifying where Person- design and development. The ECG sensors and ally Identifiable Information is at risk of access the corresponding leads malfunction have been or disclosure. Disclosure of such vital informa- the cause of signal deterioration in some cases. tion can occur in a variety of ways (one way is This can be enhanced by introducing wireless e.g. linking of sensitive information to a unique body-sensor network. identifier, which may eventually be linked to a The results could not be generalized, as the particular individual). number (n = 24) of the recruited elderly patients It is important to distinguish between security were not representing heart patients’ among the and privacy. Privacy is not just a special case of se- elderly population. Furthermore, the present inves- curity – there are interesting interactions between tigation has been performed only in respect with security and privacy. Indeed the two sometimes one specific telemedicine system and should be conflict. For instance there is a conflict between applied on a number of similar systems in order accountability and anonymity e.g. conflict of to have a better picture of the general attitude of denial of service resistance versus anonymity. the elderly patients’ in the population. Finally, it It is nevertheless true that supporting privacy could be concluded, that the system is applicable requires the provision of security services, e.g. for patient monitoring and aftercare in elderly confidentiality for stored and transmitted data, and access control. 138RSRCTST
  • 156. Patient Monitoring in Diverse Environments In a mobile environment, for health informa- • Many security issues arise in establishingtion handling, there are a number of possible working relationships in such a network,point where the sensitive information is at risk. e.g.:The following are just a representative number ° Initial trust setting;of those listed. ° Managing collaborative activities (e.g. routing);• When a communication between two devices ° Authentication, authorisation, … are established.• At the point of capture of information on Identity both side of communicating side, e.g. by individuals, physicians, paramedics or au- A user may have many identities with associated tomatically when a sensor is used. identifiers for use when communicating with• When the information is stored/used in different third parties. For instance, we all have personal devices (e.g. mobile phone, PDA, a name (although this is not a unique identifier); smart card or sensor equipment). an employee may have an employee number for• When the information is stored in fixed use with his/her employer; a citizen has one or databases (e.g. in a hospital or in a network more numbers for interactions with government; provider agency). a health care user may have a government ID,• When the information is stored or used in and one or more health insurance identifiers. mobile third party devices (e.g. a healthcare These must be arranged or defined as uniquely mobile device belonging to a physician or as possible. paramedic). Credentials To secure the information and provide privacyand anonymity some action and arrangement are When a service provider wants to authenticate anecessary. The following are some key issues that user as a legitimate holder of an identity, the usermust be consider improving the quality of mobile may be required to provide one or more creden-healthcare devices or services. tials. Possible credentials include: a password; a biometric sample; a public key certificate; or a• Identity signature on a challenge provided by the service• Credentials provider. These must be assigned in a secure way• Authorisation and must be protected against intruders.• Anonymity• Consent AuthorisationGeneral Problem Once an entity has been authenticated, the pro- vider needs to decide whether or not to grant the• An ad hoc network is a collection of com- requested service. This is refereed to as authorisa- municating devices with no pre-existing tion. It means that the network provider has the relationships or infrastructure. responsibility to insure whether the holder of this• A typical scenario for use of such a network identity authorised to access this service. is an emergency situation, e.g. a major This could, for instance, be supported using transport accident. server-held Access Control Lists (ACLs). 139
  • 157. Patient Monitoring in Diverse Environments Anonymity or services one node should make available to another. Then the question is: Can another node A user may wish to be able to access a service in be trusted to provide a communications service an anonymous way. Anonymity means that no without eavesdropping, manipulating messages, party will learn any of the identities of the user. and/or selectively dropping packets? If a service requires using stored data, then some level of identification to the stored data might be required by the provider. If payment is needed, Diionfou then an anonymous payment system is needed, Reea e.g. cash or e-cash. However, absolute anonymity is difficult to achieve, since even revealing an IP In this section, the fundamental problems that address or a MAC layer address to some extent need to be solved to realise full potential of mobile compromises it. healthcare system, have been identified; these are: need for a ubiquitous security infrastructure to Consent support secure communications between mobile devices; need for one device to be able to verify In many medical scenarios, the subject may be the conditions under which data will be stored, required to give implicit or explicit consent for handled, and retransmitted by another device. treatment to be given, and information to be passed Thus, the questions are: can trusted comput- to a practitioner or insurance company. ing systems realise all the security infrastructure In case of treatment, some measure of non- needs of future pervasive computing environ- repudiability may be required; In case of passing ment? Who will be the trusted third parties to information, the information source will need support the trusted computing based security to authenticate the subject. This is potentially infrastructure? What if some mobile devices are problematic since the information source may be trusted computing enabled and others are not? remote and only communicating with a device What other solutions are there? belonging to the practitioner. Devices collecting, storing and/or using health data may need to share this data with other de- An Ad Hoc Network vices. For instance a wireless heart monitor may need to pass data to a portable device used by a An ad hoc network is a collection of communi- physician which integrates and displays the data cating devices with no pre-existing relationships to the physician. Such data transfers should not or infrastructure. Such network is one of the take place without restriction for privacy/security fundamental components in mobile healthcare reasons. What if the devices interacting do not systems/services. A typical scenario for use of all belong to the same individual? such a network is an emergency situation, e.g. a We need trusted computing (TC) technologies major transport accident. Many security issues to be implemented as a combination of hardware arise in establishing working relationships in such and software enhancements to a computing plat- a network, e.g.: initial trust setting; managing col- form such as PC, PDA, server, or mobile phone. laborative activities (e.g. routing); authentication, The root of most of these problems is the authorisation, etc. potential lack of a single pre-existing managed One fundamental issue for two devices in an security infrastructure. When such an infrastruc- ad hoc network is deciding whether to trust one ture is established, many of the problems become another and in what extend, and what resources much less fundamental. 140euh
  • 158. Patient Monitoring in Diverse Environments Thus, the future research direct towards devel- Proceedings, peer reviewed conference article, opment of trusted computing (TC) technologies IEEE Xplore. Retrieved June 15, 2008, from: http:// and security infrastructure. Providing a trusted www.google.dk/earch?hl=daq=Timestamp+Au computing platform can help to provide security thentication+Protocol+for+Remotemeta= infrastructure. In the other hand provision of Farmer, A., Gibson, O., Hayton, P., Bryden, K., a security infrastructure enables one device to Dudley, C., Neil A., et al. (2005). A real-time, determine its level of trust in another device. mobile phone-based telemedicine system to sup- port young adults with type 1 diabetes. Inform Prim Care, 13(3), 171-177. Refeen Freedman, S. B. (1999). Direct Transmission BeWell Mobile Technology Inc. (2006). Better of Electrocardiograms to a Mobile Phone for health well in hand: Cell phones have the capac- Management of a Patient with Acute Myocardial ity to more frequently and efficiently connect Infarction. Journal of Telemedicine and Telecare, chronically ill patients with caregivers. Healthc 5, 67-69. Inform. 23(4), 56-57. Friedewald, V. E., Pion, R. J. (2001). Telemedi- Boncella, R. J. (2002). Wireless Security: an cine/home care. Returning home. Health Manag overview. Communication of the Association for Technol, 22(9), 22-24, 26. Information System, 9, 269-282. Gostin, L. O. (1965). Health information privacy. Bønes, E., Hasvold, P., Henriksen, E., Stran- Cornell Law Review, 80, 451-528. denaes, T. (2007). Risk analysis of information Gostin, L. O., Brezina, J. T., Powers, M., Kozloff, security in a mobile instant messaging and pres- R., Faden, R., Steinauer, D. D. (1993). Privacy ence system for healthcare. Int J Med Inform, and security of personal information in a new Sep. 76(9), 677-87. Epub 2006 Aug 23. health care system. JAMA, 270, 2487-2493. Clarke, M., Bratan, T., Kulkarni, S., Jones, R. Gostin, L. O., Lazzarini, L., Neslund, V. S., (2007). The impact of remote patient monitoring Osterholm, M.T. (1996). The public health in- in managing silent myocardial infarction in a formation infrastructure: A national review of residential home setting. Anadolu Kardiyol Derg., the law on health information privacy. JAMA, 7(Suppl. 1), 86-88. 275, 1921-1927. Cocosila, M., Coursaris, C., Yuan, Y. (2004). M- Jasemian, Y., Arendt-Nielsen, L. (2005a). healthcare for patient self-management: a case for Design and implementation of a telemedicine diabetics. Int J Electron Healthc, 1(2), 221-41. system using BLUETOOTH and GSM/GPRS, Compact Oxford English Dictionary, free for real time remote patient monitoring. The In- internet access. Retrieved March 5, 2008, ternational Journal of Health Care Engineering, from: http://www.askoxford.com/concise_oed/ 13, 199-219. trust?view=uk. Jasemian, Y., Arendt-Nielsen, L. (2005b). Elmufti, K., Weerasinghe, D., Rajarajan, M., Evaluation of a real-time, remote monitoring Rakocevic V., Khan, S. (2008). Timestamp telemedicine system, using the Bluetooth protocol Authentication Protocol for Remote Monitoring and a mobile phone network. J Telemed Telecare, in eHealth. 2nd International Conference on Per- 11(5), 256- 260. vasive Computing Technologies for Healthcare, 141e
  • 159. Patient Monitoring in Diverse EnvironmentsJasemian, Y., Arendt-Nielsen, L. (2005c). 767. Retrieved April 15, 2008, from: http://www.Validation of a real-time wireless telemedicine rrh.org.au.system, using Bluetooth protocol and a mobile National Centers for Chronic Disease Control andphone, for remote monitoring patient in medical Prevention, opdated June 2008, Retrieved Aprilpractice. European Journal of Medical Research, 15, 2008, from: http://www.cdc.gov/nccdphp/1(6, 2)54-262. Panidis, I. P., Morganroth, J. (1983). SuddenJasemian, Y. (2008). Elderly Comfort and Compli- death in hospitalized patients: cardiac rhythmance to Modern Telemedicine System at home, disturbances detected by ambulatory electrocar-2nd International Conference on Pervasive Com- diographic monitoring. J Am Coll Cardiol, 2(5),puting Technologies for Healthcare, Proceedings, 798-805.peer reviewed conference article, ISBN 978-963-9799-15-8. Petersen, S., Peto, V.v Rayner, M. (2004). Coronary heart disease statistics, British HeartJasemian, Y. (2006). Security and privacy in a Foundation statistics website. Retrieved June 30,wireless remote medical system for home health- 2008, from http://www.heartstats.org/datapage.care purpose. 1st International Conference on asp?id=1652.Pervasive Computing Technologies for Healthcare[CD-ROM],. s. 3, Proceedings in IEEE Xplore, Sax, U., Kohane, I., Mandl, K. D. (2005). Wire-peer reviewed conference article, 29 November-1 less technology infrastructures for authenticationDecember 2006, Innsbruck, Austria. of patients: PKI that rings. J Am Med Inform Assoc. 2005 May-Jun, 12(3), 263-8. Epub 2005Logan, A.G., McIsaac, W.J., Tisler, A., Irvine, Jan 31.M.J., Saunders, A., Dunai, A., et al. (2007). Mobilephone-based remote patient monitoring system for Shimizu, K. (1999). Telemedicine by Mobilemanagement of hypertension in diabetic patients. Communication. IEEE Engineering in MedicineAm J Hypertens, 20(9), 942. and Biology, 32-44.MacDonald, J. A. (2008). Cellular Authentication Sneha, S., Varshney, U. (2007). A wireless Key Agreement for Service Providers. 2nd ECG monitoring system for pervasive healthcare.International Conference on Pervasive Comput- International Journal of Electronic Healthcare,ing Technologies for Healthcare, Proceedings, 3(1), 32-50.peer reviewed conference article, IEEE Xplore, Standing, P., Dent, M., Craig, A., Glenville,Retrieved June 25, 2008, from: http://ieeexplore. B. (2001). Changes in referral patterns to cardiacieee.org/Xplore/guesthome.jsp out-patient clinics with ambulatory ECG moni-Marchibroda, J. M. (2008). The impact of health toring in general practice. The British Journal ofinformation technology on collaborative chronic Cardiology, 8(6), 396-398.care management. J Manag Care Pharm., 14(2 Woodward, B., Istepanian, R. S. H., Richards,Suppl), S3-11. C. I. (2001). Design of a Telemedicine SystemMcLean, R., Mendis, K., Harris, B. Canalese, Using a Mobile Telephone, IEEE, 13-arch?hl=dJ. (2007). Retrospective bibliometric review of aq=Timestamp+Authentication+Protocol+forrural health research: Australia’s contribution +Remotemeta=and other trends. Rural and Remote Health 7:142
  • 160. 143 Chapter VIII Monitoring Hospital Patients Using Ambient Displays Monica Tentori CICESE, Mexico Daniela Segura CICESE, Mexico Jesus Favela CICESE, Mexico ABSTRACT Hospital work is characterized by intense mobility, a frequent switching between tasks, and the need to collaborate and coordinate activities among specialists. These working conditions impose important demands on hospital staff, whose attention becomes a limited and important resource to administer. Nurses in particular, need to constantly monitor the status of patients in order to assess their condi- tion, assist them and/or notify physicians or specialists. Given their work load, it is not rare for them to miss important events, such as a catheter being disconnected due to the patient movement or the need to change a urine bag that has been filled. Pervasive technologies by being able to continuously moni- tor patients could provide awareness of the patients’ health condition. This awareness must be subtle, expressive, and unobtrusive without intruding on hospital workers’ focal activity. In this chapter the authors explore the use of ambient displays to adequately monitor patient’s health status and promptly and opportunistically notify hospital workers of those changes. To show the feasibility and applicability of ambient displays in hospitals they designed and developed two ambient displays that can be used to provide awareness patients’ health status to hospital workers. The first display takes into account the mobility experienced by nurses during their work to supervise the activities of daily living (ADL) con- ducted by patients. The second display is a flower vase that notifies nurses the urine output of patients and the status of their urine bag. Copyright © 2009, IGI Global, distributing in print or electronic forms without written permission of IGI Global is prohibited.a
  • 161. Monitoring Hospital Patients Using Ambient DisplaysInTROoduion (Chin, T., 2005), RFID tags for patient tracking (O’Connor, M. C., 2006), voice-activated com-Hospital staff face working conditions that are munication devices (Stanford, V., 2003), andsubstantially different from those of office work- sensors for patient monitoring (Pentland, A.,ers, for which traditional desktop computers were 2004). Indeed, widespread adoption of sensorsdeveloped (Bardram, J. E., Bossen, C., 2003; that monitor the patients’ vital signs and otherBardram, J. E., Bossen, C., 2005; Moran, E. B., indicators promise to improve care and reduceTentori, M., González, V. M., Martinez-Garcia, medical costs.A. I., Favela, J., 2006). Most hospital staff need Thus, pervasive technologies for hospitals areto move continuously around the premises to ac- increasingly supporting heterogeneous devicescess people, knowledge, and resources in order that range from handheld computers that canto perform their work effectively (Bardram, J. E., be used to capture and access limited amountset al, 2005). Thus, mobility characterizes work of information, to PCs that can be used at fixedin these environments. For instance, physicians sites for longer periods of time, and semi-publicmake daily rounds to assess and diagnose pa- displays located at convenient places that can betients, changing their location to find colleagues used to share and discuss information with col-or locate artifacts (patient records, x-ray images, leagues (Favela, J., Rodríguez, M. D., Preciado,medications) placed in bed wards, laboratories A., Gonzalez, V. M., 2004; Markarian, A.,or offices. Therefore, information in hospitals Favela, J., Tentori, M., Castro, L. A., 2006).is not generally concentrated in a single place, Hence, hospital workers today need to interactbut distributed among a collection of artifacts with different devices with a wide range of func-in different locations. Consequently, hospitals tionality (Bardram, J. E., 2005). Consequently,can be seen as an information space and it is by carrying out a single activity typically involves“navigating” this space that hospital staff can the use of several systems that call for the user’saccess the information required to support their undivided attention where several informationgoal (Bossen, C., 2002). sources battle for a piece of space in the already Indeed, nowadays highly mobile hospital limited screen real state.workers spend more than 50% of their time on- One way to overcome such difficulties is tothe-move, making it difficult for them to be aware develop ambient displays that could be embeddedof the status of the patients they are responsible into the environment to provide a getaway for thatfor (Moran, E. B., et al., 2006). For instance, information that could be displayed by objectssometimes hospital workers have patients placed already placed in the physical space instead ofin different rooms or even in different areas of the traditional computer displays. Objects alreadythe hospital. Consequently, hospital workers have known and used by hospital workers could bebeen held liable for their failure to monitor and augmented with pervasive sensors making thempromptly respond to patients needs (Smith, K. S., capable of extending their capabilities beyond Ziel, S. E., 1997). This has motivated the intro- its primary role while still constituting a part ofduction of pervasive technologies in hospitals to the user’s environment. For instance, a mirrorallow hospital workers to closely monitor patients. augmented with infrared sensors and an acrylicFor instance, a hospital in Boston is testing an panel could detect human presence and act as aultrasound tracking system that can determine message board to display relevant informationthe location and vital signs of patients (O’Connor, when a user faces the mirror. Hence, hospitalM. C., 2006). These pervasive technologies being environments could be augmented with suchintroduced range from wireless networks, PDAs displays that unobtrusively convey information144
  • 162. Monitoring Hospital Patients Using Ambient Displays to hospital workers without requiring their full computing devices where hospital workers use attention, while at the same time, allowing an a set of specialized services that account for implicit and natural interaction. Indeed, the no- contextual information (Camacho, J., Galicia, tion of what constitutes a computer display is L., Gonzalez, V., Favela, J., 2008; Favela, J., et changing. No longer is a display confined to the al., 2006; Markarian, A., et al., 2006; Munoz, M., typical CRT monitor with a single user paying Rodriguez, M. D., Favela, J., Martinez-Garcia, focused attention while interacting with virtual A. I., Gonzalez, V. M.). To help realize this objects on the screen (Lund, A., Wilberg. M., vision, we have developed several ubiquitous and 2007). Rather, computer displays are found in such context-aware applications which provide support diverse forms as small screens in mobile phones for the following functionality: or handheld computers, to ambient displays that provide peripheral awareness to the presence and Providing Awareness of People and status of people, objects or information. Such Artifacts ambient displays could be located throughout hospital premises providing hospital workers In a hospital, artifacts and people are distributed awareness of relevant events associated to their in space and time. Hence, hospital workers must patients while they are on-the-move. navigate hospital premises in order to gather the In this chapter we explore the use of ambient information they need to conduct their work. Un- displays to adequately monitor patient’s health like others processes, gathering information is a status and promptly and opportunistically notify necessary task but adds overtime to the already hospital workers of those changes. To show the busy day of hospital workers. Consequently, the feasibility and applicability of ambient displays iHospital, to reduce the time hospital workers in hospitals we developed two ambient displays spend searching and gathering information, aimed at creating a wearable ambient connection provides the means for them to be aware of the between patients and hospital workers –particu- presence, location and status or artifacts and other larly nurses. The rest of the chapter is organized people by showing this information through a floor as follows: In Section II we describe our vision map or a list of users as reported in (Bardram, J. of a hospital as an interactive smart environment E., Bossen, C., 2005). saturated with heterogeneous computing devices and specialized services –the iHospital. Section Spporting Collaboration through III describes the results of a workplace study Context-Aware Communication and conducted in a public hospital to understand the the Seamless Interaction among way hospital workers monitor and assess patients. Heterogeneous Devices In section IV we describe the design of two am- bient displays in support of patient monitoring. Hospital staff can send messages that depend Finally, in section V we present our conclusions on environmental conditions. As an example, and directions for future work. a physician can send a message that will be de- livered to the doctor responsible for a patient in the next shift when laboratory results are ready. The ipial: TThe HOSPpi The sender does not need to know a-priori the a SMmaenvionmen identity of the doctor that will be attending the patient nor the time when the laboratory results The iHospital is our vision of a highly interactive will become available. In addition, hospital staff smart environment saturated with heterogeneous can transfer information from public spaces to 145oalaCS
  • 163. Monitoring Hospital Patients Using Ambient Displays personal devices, share information between when switching between them (Camacho, J., heterogeneous devices, remotely monitor other 2008). Using the mobileSJ application, the user computers, and share handheld applications. For can define each of his activities and associate to instance, two colleagues carrying their PDAs and them, information resources, contacts relevant discussing a clinical case using a public display, to the activity, emails related to the activity and could seamlessly transfer information between pending issues. When a user switches between their personal information space (PDA) and the spheres, each sphere is enabled to quickly gather shared space (public display). and retrieve its own workspace state (windows Mobility and collaboration create a need to positions, status and overlay order) and context contact colleagues within the hospital, either to information like open documents or idle time, discuss a case with a specialist or request help to in a silent manner. In addition, mobileSJ allows transfer a patient. Several mechanisms are used sharing activities and resources, as well as, com- for these purposes and technology has been de- municating with colleagues through either SMS veloped to assist in this task; such as the Vocera messages or phone calls. communication system which enables users to Although these pervasive technologies use contact a fellow hospital worker either by name, contextual information to provide opportunistic role or location using a hands-free voice com- services and information to hospital staff they munication system (Stanford, V., 2003). aren’t appropriate mechanisms to convey infor- mation about the health status of a patient. This Using Context to Adapt and type of information changes very fast and is Personalize the Information highly sensitive. Thus using this type of mecha- nisms of sending messages to a display or a PDA To provide relevant information to users, the iHos- each time an event occurs with a patient could pital takes into account contextual information, be extremely disruptive to the task the hospital such as the user’s identity, role, location, time, or worker is engaged in. For instance, suppose that status of an information artifact (e.g. availability a nurse is monitoring the urine and evacuation of laboratory results). Thus, when a physician, habits of the patient in room 240. While she is carrying a PDA, is near a patient, the system of- inserting a catheter to the patient in room 226 fers to display the clinical record of the patient. she doesn’t want to receive a sound or message Contextual information such as identity or role is alert in her PDA notifying to her each time such also taken into account to adapt and personalize patient has urinated or evacuated. Although the the presentation of information to the user. Thus, nurse wants to be aware of this information she when a physician approaches a public display, it wants to receive this information in a subtle and shows only the physician’s patients, personnel unobtrusive manner. One way to overcom