Seman/cs is not a luxury
Building the new genera/on of Clinical Informa/on Systems
“Intra‐ and Inter‐operability”
Paolo Ciccarese*, PhD
Co‐founder and knowledge architect
*Assistant in Neurology @MassachuseHs General Hospital, Instructor in Bioinforma/cs @Harvard Medical School,
Health Care Life Science Interest Group @W3C, Consultant in SoOware innova/on
Timeline
Mass General/Harvard Medical School
Visitor at Stanford Univ.*
* Decision Support/ Workflow and Ontology
Based SoOware Dev. (Constant innova/on)
Online Scien/fic Communi/es
(Ontologies and Interoperability)
HCLS IG W3C
(Seman/c Web Technologies)
Simile Project – MIT**
HMS Faculty
(Seman/c Interoperability)
University of Pavia and Consultant
Clinical Workflow, Decision Support and Knowledge Management (Guide Project)
Temporal Abstrac/ons in Medicine
Regional and Na/onal Networks
2000 2004 2005 2006 2007 2008 2009
*Visi/ng Samson Tu (SAGE author) **Volunteer (coding)
Medicognos Research background
P.Ciccarese Phd ( Medicognos architect)
Mass General/Harvard Medical School
Online Scien/fic Communi/es
(Ontologies and Interoperability)
HCLS IG W3C
(Seman<c Web Technologies)
Simile Project – MIT**
HMS Faculty
(Seman<c Interoperability)
University of Pavia and Consultant
Clinical Workflow, Decision Support and Knowledge Management (Guide Project)
Temporal Abstrac<ons in Medicine
Regional and Na<onal Networks
New Genera/on Knowledge & Process Based EHR
2000 2004 2005 2006 2007 2008 2009
Decision Support/ Workflow and Ontology Based So:ware Dev.
Original GPs requirements 2005
1. Usability
2. Process Management/Op/miza/on/Improvement
3. Collabora/on ‐ care networks
4. Medical Knowledge Management ‐ formal mul/lingual
terminologies, evidence‐based recommenda/ons
5. Communica/on ‐ interoperability with labs and hospitals
Transform the problem oriented EPR in a distributed clinical process
management system with embedded clinical decision support
Con/nuity of Care and Disease Management
Why it is so hard?
Diminishing the seman/c impedance
physician mental model soOware
Usability
Evidence‐based
Workflow op/miza/on
medicine
Cost Containment Economies by Outcome
(Disease Management) Improvement
Interoperability
Quality of data informa/on knowledge
encoded in the plaporm and
exchanged with the external world
eHealth for Safety 2007*
New genera/on of advanced, user‐friendly and
ubiquitous tools for:
• Integra/on of decision support and workflow
support systems with pa/ent record
• Knowledge representa/on
• Advanced terminology‐driven eHealth tools for data
entry and retrieval
• Clinical informa/on systems integra/on of pa/ent
data across the con/nuum of care
* eHealth for Safety – Impact of ICT on Pa/ent Safety and Risk Management,
October 2007 European commission, Informa/on Society and Media
New organiza/onal architecture
Health Insurance
Data 2x anonym
P4P Quality Indicators
Primes
(global)
Feedback
GP Data
anonymized
P4P Primes
Cl Experts
QI PH Experts SP
Coaches MD
Pharmacy
Pa/ent
Nurse Hospital
New informa/on architecture
Knowledge based Clinical Workflow Management with Decision Support
Patient Status Patient Status Patient Status
Data Data Data
Data Data Data
KPI KPI KPI
Clinical Guideline
CDS Operational Process Management
ex: lab tests
Care flow execution Alerts
Agreed Care Plan
Quality /Safety /Efficiency
Indicators Ac/vity 1 D Ac/vity 2 Ac/vity n
Process Reenginering Process Mining
Strategic Process Management
CDS :clinical decision support
The key role of clinical decision support
• All currently known large studies concluded that: EPR/
EHR are not enough for quality improvement!
• As EHR use broadens, one should not assume an
automa/c diffusion of improved quality of care
• In selec/ng an EHR, physician prac/ces should carefully
consider the inclusion of: clinical decision support to
facilitate quality care
1.EMR SophisBcaBon Correlates to Hospital Quality Data, Comparing EMR AdopBon to Care Outcomes, HIMSS 2007
2.Electronic Medical Records and Diabetes Quality of Care: Results From a Sample of Family Medicine PracBces Jesse C. Crosson, PhD ANNALS OF FAMILY
MEDICINE MAY/JUNE 2007
3.Electronic Health Record Use and the Quality of Ambulatory Care in the United States, Jeffrey A. Linder ARCH INTERN MED JULY 2007
4.The Value of InformaBon Technology‐Enabled Diabetes Management Davis Bu,Center for InformaBon Technology Leadership. 2007
Ontology‐based plaporm
• Every aspect in the system is built on ontologies
defined using a subset of OWL (proprietary
solu/on at run/me)*
• More expressive OWL for analysts (knowledge
modeling, model checking, maintenance, mining)
with enterprise tools
• Referent tracking** (unique IDs)
* Ontology‐Based Integra/on of Medical Coding Systems and Electronic Pa/ent Records
Ceusters W, Smith B, De Moor G
** Strategies for referent tracking in electronic health records. Ceusters W,
Smith B. J Biomed Inform. 2006 Jun;39(3):362‐78. Epub 2005 Sep 9.
Unified Seman/c Model
Occurrence
How it occurred
When?
Who reported it?
What were circumstances?
…
Process (Workflow)
How it was acquired
Who?
Biomedical
Disease
When? Sign
What
Data
Within which Symptom
careplan task? …
…
Classifiers
‐ Ac/ve problems/cau/ons/…
Journal
‐ Encounter/sub‐encounter Medicognos
‐ Observa/on/Query templates Onthotypes®
How it organized for the users
Clinical Info Management
Medicognos onthotypes®
• Defined through expressions
• Decoupling meaning from presenta/on
• Extensive usage of classifica/on
• Mul/ple presenta/on forms can have the same
meaning
Medicognos architecture
20%
80%
Onto‐Terminology Info. Models Executable Drug Wf Executable CDS logic
“ontotypes” KB Templates KB Drug Therapy Wf KB Rules KB
Concept Template Workflow Drug KB Queries &Rules Builder
Builder Builder Builder Builder Clinical Domain Language
Clinical Knowledge Management Studio
Domain Clinical SituaBon Care OrganisaBon
Biomedical
Ontologies Care Plan
OBO?
BFO? OrganisaBon
Time/Space Process
FoundaBonal layer
2 cents on interoperability
• Need for formally defined seman/cs
• Intra‐operability and Inter‐operability go together
• Being able to communicate with other systems
depends on the expressiveness of the “internals” first
and of the communica/on protocols later on
• The seman/c models have to be produced
incrementally and tested con/nuously in seman/cally
oriented soOware (full stack)
• The seman/c is not limited to pa/ent data but is a
mix‐up of several clinical aspects
• All the aspects have to evolve in lockstep