Act Together and Connect for Patient-Centered Outcome Research- A concept paper submitted to PCORI by Chistopher Yang at Drexel University *Specific AimsDaily, patients are facing complex health care decisions. They are given many different treatmentpaths and often do not understand or know which one is the best path to walk down. These patientsoften turn to the internet to explore which option is the best for them. In their exploration of differenttreatment options, they often notice gaps in research. Even though patients are concerned andinterested in research, they often do not have a way to connect to researchers.We propose an online platform called Act Together and Connect for Patient-Centered OutcomeResearch or ACTONNECT. ACTONNECT was created so that its foundation would be grounded on thetheory of action-based matchmaking and engagement. Meaning, if there are established actions fromboth partners, the matchmaking process is more likely to be successful. The established partnershipwill continue to prosper due to its sustainable grounded foundation.ACTONNECT is more than just an idea. It is a strong team of five institutes that work in clinical careand research. This team has been assembled to demonstrate the feasibility of ACTONNECT. Threespecific aims have been established to turn ACTONNECT into a reality.Aim 1: Assemble a Stakeholder Advisory Council for the development of ACTONNECTThe council will contain five members inclusive of patients, caregivers, and researchers. Both rural andmetropolitan perspectives will be represented. This advisory council will be recruited and supported byMindy Gribble, RN who has extensive experience in working with patient advisory councils. Monthlymeetings will be scheduled between the Stakeholder Advisory Council and researchers.Aim 2: Fully develop and test ACTONNECTIn collaboration with the Stakeholder Advisory Council, the development team will fully develop and testACTONNECT. It will have the following five components: 1. MipiDAT: The Marshfield Index of Patient Interests in Drugs and Treatments (MipiDAT) is a metasearch engine for the voice of patients 2. Natural Language Processing (NLP): negative statement analysis and sentiment analysis can be conducted using NLP. 3. Ontology Annotation: ontologies will be used to solve the problem that patients and researchers often use different vocabularies to describe the same thing. 4. Visualization: An ACTONNECTION is an info graph that summarizes MipiDAT search results. It can be used to visualize the related concepts. The ACTONNECTION will reveal the current gaps and knowledge of the searcher; this information will also allow patients to make informed decisions. 5. ACTONNECT: a website to match patients, care givers, researchers, physicians, and other stakeholders based on their search activities using MipiDAT.Aim 3: Evaluate the impact of ACTONNECTWith inputs from the Stakeholder Advisory Council, key metrics of success will be developed. Thosemetrics will be automatically calculated and tracked to demonstrate the impact of ACTONNECT. * A team consisting of Drs. Chistopher Yang2, Eneida Mendonça3, Umberto Tachinardi3, Andrew Su4, Chunlei Wu4, Tian Xia5, and Simon Lin1 will be assembled for ACTONNECT. 1Marshfield Clinic Research Foundation, 2Drexel University, 3University of Wisconsin‐Madison, 4Scripps Research Institute, 5Huazhong University of Science & Technology. 1
Team and EnvironmentThe team, resources, infrastructure and environment are well-suited for developing the programACTONNECT.Team: Under the leadership of Dr. Simon Lin, Director of the Biomedical Informatics Research Centerat Marshfield Clinic Research Foundation, a team of experts from five patient care and researchinstitutes will be assembled to implement ACTONNECT. The team has demonstrated its productivitywith previous collaborations and current prototyping activities (Table 1). The Marshfield team wasrecognized by the ONC in 2013 for their innovations in creating the patient-centered mobile applicationof “Heart Health Mobile”.Patient-Centered Development: A patient-centered approach will be used to actively engage patients inevery step from the design, development, and testing to the evaluation of ACTONNECT. Thedevelopment team has access to both rural and metropolitan areas. In the rural communities,Marshfield Clinic Research Foundation can access advisors though the Community-AcademicPartnerships Core (CAP). While in metropolitan areas, the team is able to access through the UW-Madison Institute for Clinical and Translational Research.Technical Development Resources: With more than 40 full time employees, the Biomedical InformaticsResearch Center (BIRC) at the Marshfield Clinic Research Foundation (MCRF) is home to one of thelargest clinical informatics research centers in the United States. Eye-tracking equipment is available atMarshfield Clinic. This technology has been used to improve the usability of designs by tracking the eyemovement of subjects when looking at a particular device. MCRF is also part of the consortium for theInstitute for Clinical andTranslational Research(ICTR) at University ofWisconsin Madison thatoffers significantresources to translateinnovations into clinicalpractice. Eachcollaborator is adequatelysupported by outstandingcomputational andscientific infrastructure attheir home institutes.What isACTONNECTACTONNECT providesan innovative way toconnect patients andresearchers when theyare looking for answers totheir own questions.ACTONNECT is a Figure 1. ACTONNECT connects patients and researchers by engaging them in common “playground” for searching MipiDAT. The search results can be summarized into an ACTONNECTION – both patients and an infograph to intuitively depict the concepts and knowledge gaps. Each author of researchers. an ACTONNECTION is recognized. Throughout the process of creating of ACTONNECT users will ACTONNECTIONs, patients and researchers are able to know each other and form research partnerships. ❶Engaging in search activities ❷Results summarized into 2 ACTIONNECT ❸❹ ❻❼Four different ways to find new connections ❺ Identify research and knowledge gaps.
be able to view the searches of other patients, caregivers, physicians and researchers.To develop a foundation for the search, the team will sample the voice of patients from social mediawebsites like the patient forums at MedHelp and WebMD, and then index them into a database as wellas a metasearch engine called MipiDAT- the Marshfield Index of Drugs And Treatments.Patients will be able to search MipiDAT and find out the experience of other patients on different drugand treatment options. Researchers can also search MipiDAT to verify the patient’s concerns. Aftersearching MipiDAT, both patients and researchers are able to create an info graph calledACTONNECTION (Figure 1). We use the info graph as an intuitive way to present the results.ACTONNECTIONs also serve as an important function for patients and researchers to communicate.With an ACTONNECTION, the concerns and knowledge gaps that are needed to help patients makedecisions can be visualized and identified.Throughout the process of creating each ACTONNECTION, patients and researchers are able to knoweach other and form research partnerships by four different ways (❸❹ ❻❼ in Figure 1). The initialintroduction of the users of ACTONNECT is handled by LinkedIn or MedHelp. By this matching method,not only is the complexity of developing a matchmaking system is reduced, but it also preserves thehuman touch, for example John and Lori might have a common friend from LinkedIn.ACTONNECT is based on the grounded theory of action-based matchmakingACTONNECT is based on the grounded theory of action-based matchmaking and engagement(Sawhney, Verona et al. 2005; Gasson and Waters 2011). There are three levels of engagements inthe matchmaking process: from the initial willingness, to the actions, to the 1 +1 >2 partnerships (Figure2). ACTONNECT allows the matchmaking at the highest level by allowing patients and researchers toshare a common ground. This common ground is found through their investigation and creation anACTONNECTION info graph. With the action–based matchmaking, the partnership between patientsand researchers can be more sustainable. Engaged by Willingness Engaged by Actions Engaged by Partnership •Shared vision •Shared activities •Shared investigation •Find each other at conferences or •Utilize the platform to derive •Partnership in actions online research questions •1 + 1 > 2 •Join the same campaign •Stronger commitmentFigure 2. Three levels of engagements in the matchmaking process. The key is to engage the partners in the matchmaking process into shared activities. 3
Innovations in ACTONNECTMipiDAT: To our knowledge, MipiDAT is the first metasearch engine used for searching the voice ofpatients. As of now, Google and other search engines cannot effectively search active discussions bypatients which are found in different forums located in various health social media websites. Nor canGoogle or other search engines visually present the results using an info graph. One of the co-PIs ofACTONNECT, Dr. Christopher Yang, at Drexel University, has demonstrated the feasibility of creatingMipiDAT. This was done by the same web crawling approach used by Google and other searchengines. The web crawler can be reconfigured to reach multiple forums from different sources. As such,other kinds of social media such as Twitter or Facebook can be added. With MipiDAT, the voice ofpatients can be visualized and heard by anyone who searches at ACTONNECT.Figure 3. Under the hood of ACTONNECTION Natural Language Processing (NLP) and Ontology Annotation: Patient postings often containmisspelling or the use of alternative names for drugs. With NLP and ontology annotation, ACTONNECTis able to map these differences, for instance, “venlafaxine”, “efexor” and “effexor” all into the“venlafaxine” category. More interestingly, the ontology technology knows both Effexor and Zoloft areantidepressants. This advanced technology is used in the search engine making it more effective andintellectually advanced (Figure 3). To demonstrate the feasibility of these innovations, the co-PI’s ofACTONNECT, Dr. Eneida Mendonça and Dr. Umberto Tachinardi, at University of Wisconsin Madison,have conducted sentiment analysis of patient postings in MipiDAT. Moreover, Dr. Andrew Su and Dr.Chunlei Wu, who are the co-PIs of ACTONNECT at Scripps Research Institute, established the pipelineof using the NCBO annotator to annotate MipiDAT postings using all 339 ontologies maintained byNCBO.ACTONNECTION: An ACTONNECTION is an innovative info graph to depict different concepts in thesearch result and their relationships to each other. This graph can be used to facilitate critical thinking.When combined with a short written text, an ACTONNECTION will be a powerful way to summarize the 4
existing knowledge gaps. Dr. Tian Xia, co-PI at Huazhong University of Science & Technology, hasdemonstrated the feasibility of ACTONNECTION using Cytoscape. Cytoscape is a Web 2.0 componentto visualize networks (Mei, Xia et al. 2012).Reliance on LinkedIn and other Infrastructure: We do not want to reinvent the wheel of socialnetworking to make the connections. This is due to the inconvenience of having to create an account,locating others that you may know and establishing that account. ACTONNECT is built upon LinkedInand MedHelp. Users of ACTONNECT can use their existing, established LinkedIn or MedHelp accountsto publish their ACTONNECTIONs.Why MipiDAT is a magnet for both patients and researchersWith MipiDAT, patients and care givers are able to search dozens of patient forum websites and findout the experience of other patients. Moreover, patients are able to cross reference the professionalliterature in PubMed (❺ in Figure 1). As such, some patients might be able to identify gaps in currentevidence that is needed for them to make informed decisions. Patients will be able to publish theirfindings in an ACTONNECTION info graph and share it with others. The intellectual contribution ofpatients is fully acknowledged at ACTONNECT. One critical mission of ACTONNECT is to promote thesense of ownership by different stakeholders in the research process. With ACTONNECTIONs, theorigin of the idea becomes more transparent and traceable.Researchers, on the other hand, are constantly seeking funding for patient-centered outcomesresearch. To support the PCORI mission of working on “what matters most to patients”, researchersneed to quantify what patients want and need. MipiDAT offers a one-stop search engine forresearchers to quantify the voice of patients. For instance, a researcher who is planning on a study ofthe comparative effectiveness of Zoloft vs. Effexor can search MipiDAT to find out to what degreepatients are concerned about the side effects of constipation or orgasms.See more in the following YouTube video. http://youtu.be/1ZSdK6rfwKc How to turn ACTONNECT into reality?The idea of ACTONNECT is highly innovative. To manage the technical complexity of theimplementation, the PI of this project, Dr. Simon Lin, has identified co-PIs at four other institutes to forma multidisciplinary development team. The team has demonstrated productivity either through previouscollaborations or though the prototyping of key ACTONNECT components (†).Table 1. The idea of ACTONNECT is solidly supported by the prototype of its key components and a strong team of collaboration between patients, clinical providers, and research institutes. Components Prototyped To be developed Team Leader Established Collaborations The Marshfield Index The feasibility of Create a distributed Dr. Christopher Dr. Yang invited Dr. Lin of Patient Interests in creating MipiDAT by web crawling approach Yang from Drexel to be on the health Drugs and Treatments web crawling was to index the patient University and social media and (MipiDAT): MipiDAT is established; An voices over the Dr. Simon Lin patient‐centered a database of patient example data set of Internet. MipiDAT will from Marshfield research panel in the voices. It indexes consumer postings become the PubMed Clinic Research IEEE 2013 conference on patient opinions from in MipiDAT was equivalent of Patient Foundation Healthcare Informatics. various social media, created (Marshfield‐ voices. Different such as MedHelp and 300) to drive the adaptors will be † The participation of each institute to develop ACTONNECT further is subject to the terms and conditions of future funding announcements. 5
WebMD. prototyping of other developed for MipiDAT components. to index different social media sites. Natural Language The Apache cTAKES Fully develop Natural Dr. Eneida A. Drs. Mendonça,Processing (NLP): NLP was utilized to Language Processing Mendonça and Tachinardi and Lin are all is critical to analyze analyze Marshfield‐ (NLP) approaches Dr. Umberto key leaders at the patient voices in 300 data set. toward all data in Tachinardi from Biomedical Informatics MipiDAT. MipiDAT; Include University of Core of the University of critical NLP techniques Wisconsin ‐ Wisconsin Institute for such as negation Madison Clinical and Translational analysis and sentiment Research. They have analysis. established collaborations. Ontology Annotation: The Marshfield‐300 Fully implement the Dr. Andrew Su Dr. Su and Dr. Lin Each posting by data set has been annotation pipeline for and Dr. Chunlei worked together on the patients in MipiDAT used to test the the full MipiDAT data Wu from Scripps annotation of genes in will be automatically NCBO annotation set. Research social media and mapped to pipeline. Institute published a paper in standardized terms. BMC journals (Good, Howe et al. 2011). Visualization: An Cytoscape was used Build an interactive Dr. Tian Xia at Dr. Xia and Dr. Lin ACTONNECTION is an to create the process to slice and Huazhong from published three papers intuitive infograph to ACTONNECTION. dice the University of together on using visualize and to ACTONNECTION using Science & Cytoscape to communicate the ontologies, for Technology visualization complex ideas between instance, showing biomedical data (Mei, patients and gastrointestinal side Xia et al. 2012; Xu, Wang researchers. effects only. et al. 2012; Peng, Xu et al. 2013). ACTONNECT: An Wireframes of the Fully implement and Dr. Simon Lin Biweekly conference online website to ACTONNECT program the website to from Marshfield calls and shared online connect patients and website were connect all the four Clinic Research workspace for file researchers based on established. components above; Foundation and exchange have been the grounded theory ACTONNECT builds fully engage Dr. Tian Xia from established between five of “action‐to‐ upon general‐ stakeholder advisory Huazhong institutes to prepare the connection” for long‐ purpose networking board throughout the University of prototypes for this term collaborative tools such as design, implementation Science & proposal. engagement. LinkedIn, MedHelp and test phases. Technology and others. Design and Project Management ConsiderationsManagement of Complexity and Scalability: The complexity of this project is controlled by properlyscoping the core development and leveraging existing services. For instances, we utilize thespecialized social media companies in healthcare found on pages such as MedHelp or WebMD. Thesecompanies are used to deal with false postings or junk postings. We utilize LinkedIn or MedHelp todefine user identities; otherwise, the user can post their ideas with the ability to stay anonymous. TheLinkedIn and MedHelp accounts are used to handle login/password, and to manage introductoryemails. Outsourcing the basic profile process allows the system to be highly scalable and thedevelopment process more manageable.Usability: The development team has access to the usability lab at Marshfield Clinic ResearchFoundation. The usability lab has eye-tracking devices and focus group facilities. The usability lab 6
technology is used to observe, record, and analyze how people interact with computer and Internetresources. A clean and simple interface is a top priority of the ACTONNECT design.Section 508 Accessibility: The ACTONNECT website will be built with the awareness of Section 508requirements. As such, the website can be more accessible to people with disabilities. We will useFunctional Accessibility Evaluator tool developed by University of Illinois at Urbana-Champaign to testthe accessibility of the final design.Engagement of rural population: The Marshfield Clinic is located in rural Wisconsin and serves a ruralpopulation of more than 0.5 million. The Marshfield Clinic Research Foundation has experience ofengaging rural patients in healthcare. At least one member of the Stakeholder Advisory Council ofACTONNECT will represent rural population. Patients in rural and remote areas can easily get accessto ACTONNECT through internet, the same as patients in metropolitan areas. ACTONNECT can alsobe used as a tool for ambassadors to rural areas to engage patients in conversations of comparativehealthcare decisions.Timeline: Timeline and deadline concerns will be managed by regular scheduled meetings amongcollaborators on a monthly basis. Meetings will be used to track project progress, address impedimentsto progress and evaluate whether identified benchmarks are on track. Problem solving and correctiveplanning of actions will be devised for tasks which are brought to attention outside meeting timelineexpectations.Figure 4. Timeline of 18‐month development of ACTONNECT Literature CitationsGasson, S. and J. Waters (2011). "Using a grounded theory approach to study online collaboration behaviors†." European Journal of Information Systems 22(1): 95‐118. 7
Good, B., D. Howe, et al. (2011). "Mining the Gene Wiki for functional genomic knowledge." BMC genomics 12(1): 603. Mei, H., T. Xia, et al. (2012). "Opportunities in systems biology to discover mechanisms and repurpose drugs for CNS diseases." Drug discovery today. Peng, K., W. Xu, et al. (2013). "The disease and gene annotations (DGA): an annotation resource for human disease." Nucleic acids research 41(D1): D553‐D560. Sawhney, M., G. Verona, et al. (2005). "Collaborating to create: The Internet as a platform for customer engagement in product innovation." Journal of interactive marketing 19(4): 4‐17. Xu, W., H. Wang, et al. (2012). "A Framework for Annotating Human Genome in Disease Context." PloS one 7(12): e49686. This Concept paper is submitted by Dr. Christopher Yang at Drexel University to the PCORI Patient-Researcher Matching Challenge.MARSHFIELD CLINIC RESEARCH FOUNDATION, A DIVISION OF MARSHFIELDCLINIC (“MC”) IS EXPRESSLY CONDITIONED ON SPONSOR AND ADMINISTRATORLIMITING THE USE OF MC’S NAME, LIKENESS, IMAGE, AND BIOGRAPHICALINFORMATION FOR THE SOLE PURPOSE OF ADVERTISING AND PROMOTING THECHALLENGE. ANY OTHER USE OF MC’S NAME, LIKENESS, IMAGE, OR BIOGRAPHICALINFORMATION IS STRICTLY PROHIBITED WITHOUT THE EXPRESS WRITTEN CONSENTOF MC. IF PCORI AWARDS MC WITH A PRIZE, THIS PROVISION SHALL SUPERSEDE ANYCONFLICTING CONTEST TERM OR CONDITION. 8