53 © 2017 IMIA and Schattauer GmbH

The Role of Free/Libre and Open Source Software in Learning Health Systems C. Paton1, T. Karopka2 1 Group Head for Global , Centre for Tropical Medicine and Global Health, , UK 2 Chair of IMIA OS WG, Chair of EFMI LIFOSS WG, Project Manager, BioCon Valley GmbH, Greifswald, Germany

LHSs, examining the academic literature, Summary Introduction the MedFLOSS database that catalogues Objective: To give an overview of the role of Free/Libre and As more patient data are collected electron- FLOSS in use in healthcare [6], and the Open Source Software (FLOSS) in the context of secondary use of ically through Electronic Health Records grey literature from websites, reports, and patient data to enable Learning Health Systems (LHSs). (EHRs) and other information systems used personal communications with experts in Methods: We conducted an environmental scan of the academic in healthcare organisations, the opportunity the area of FLOSS adoption in healthcare. and grey literature utilising the MedFLOSS database of open to reuse the data these systems collect for Although some EHRs and research source systems in healthcare to inform a discussion of the role research and quality improvement becomes systems are fully open source, such as the of open source in developing LHSs that reuse patient data for more apparent. There is now a wide range of Veterans Information Systems and Technology research and quality improvement. large-scale research projects [1, 2] that are Architecture (VistA) [7], modern healthcare Results: A wide range of FLOSS is identified that contributes to reusing routinely collected data for analysis infrastructure is often a combination of the information technology (IT) infrastructure of LHSs including and the concept of a Learning Health Sys- open source and proprietary systems. In our operating systems, databases, frameworks, interoperability tem (LHS) has been coined to summarise analysis, we will therefore examine how the software, and mobile and web apps. The recent literature around this cycle of data collection, analysis, and open source (OS) ‘stack’ from infrastructure the development and use of key clinical data management tools health service improvement [3–5]. to user interfaces is used in healthcare appli- is also reviewed. There is a plethora of different information cations. We will then examine the range of Conclusions: FLOSS already plays a critical role in modern systems in use across healthcare systems OS data analysis tools and look at how new health IT infrastructure for the collection, storage, and analysis around the world. Although many user-facing modular EHR “apps” that utilise open source of patient data. The nature of FLOSS systems to be collaborative, systems are proprietary in nature, meaning software and open standards can close the modular, and modifiable may make open source approaches that source code for the software is protected loop to create a LHS that reuses patient data appropriate for building the digital infrastructure for a LHS. and cannot be modified, many systems use in- to enable better clinical decision making. frastructure components that are open source Keywords (meaning that the source code is licenced in a FLOSS; Open Source; Learning Health Systems; Secondary Use way that allows for modification and reuse). This paper aims to discuss the role of Free/ Yearb Med Inform 2017:53-8 Libre and Open Source Software (FLOSS) in Results http://dx.doi.org/10.15265/IY-2017-006 modern healthcare information technology Our environmental scan has resulted in the Published online May 8, 2017 (IT) infrastructure and will highlight some identification of a wide range of FLOSS that of the major open source projects that are contribute to health IT projects critical for contributing to the development of LHSs LHSs. We have therefore divided our find- that reuse patient data for research and health ings into general purpose FLOSS building service improvement. blocks such as operating systems, frame- works, and databases (Core Infrastructure FLOSS); FLOSS projects used to analyse patient data but not specifically designed for clinical purposes such as general purpose Methodology statistical software; FLOSS projects that We have undertaken a broad environmental are used in healthcare but may or may not scan utilising a snowballing approach to contribute to LHS development depending the discovery of FLOSS that contributes to on how they are implemented, such as open

IMIA Yearbook of Medical Informatics 2017 54 Paton et al.

source EHR systems (Clinical FLOSS); and Table 1 General categories of FLOSS Clinical Data Management FLOSS that, when combined with EHR systems and core Categories Examples technologies, can form the digital infrastruc- Core Infrastructure FLOSS Operating Systems, Frameworks, Databases, Core Applications ture for LHSs. Table 1 displays the set of the (See Table 2) general categories identified along with a list Data Analysis FLOSS Statistical Software, Big Data Software, AI/Machine Learning of examples in each category. (Table 4) Clinical FLOSS EHRs, HIS, LIMS, PACS, Interoperability Engines (See Table 3) Clinical Data Management FLOSS OpenClinica, TranSMART, i2b2, CohortExplorer, SEMCARE, OMOP Core Infrastructure FLOSS Open source operating systems are now used Table 2 Example of FLOSS used in Core Infrastructure in a wide variety of clinical environments. Servers running often power in-house Operating Systems Frameworks Databases Applications data-centres and cloud-based EHR systems. Databases such as MySQL, PostgresQL and Linux AngularJS MySQL Apache newer NoSQL languages are now storing Android Bootstrap PostgresQL Chromium clinical data for a wide range of EHR sys- Apache Spark Riak Firefox tems including the NHS ‘spine’ that uses Java Spring Eclipse the OS Riak NoSQL system at its core [8]. Django Most modern EHR systems allow clinicians to input and retrieve data through mobile devices or web interfaces (or web interfaces Table 3 Examples of FLOSS used in Data Analysis to run complex analyses over large datasets on mobile devices). The two major mobile by breaking the dataset down into chunks operating systems are based on OS projects: Statistical Software Big Data/AI which are analysed in parallel using open Android is a fully open source system and source systems such as Apache Hadoop. HDFS iOS is based on the BSD version of Unix. Table 3 details some of the core FLOSS DataMelt MapReduce Clinical apps use a variety of OS frameworks projects designed for data analysis that are including AngularJS, the Spring framework, GNU Octave Weka used as building blocks for LHSs. and Django to deliver the user interface. Table Mondrian Torch 2 shows a few examples of core infrastructure OpenNN FLOSS (although the number of projects in use in various healthcare systems around the Clinical FLOSS world is too numerous to list in full here). As or levels of adherence to clinical guidelines. There are now a number of large scale an example, the development of a nursing in- A simple analysis can often be performed FLOSS projects that enable the collection telligence system in a large university hospital by exporting information from an EHR of patient data in clinical settings. Most of [9] illustrates how these building blocks can database into a CSV file for analysis in a these are not specifically designed for data be put together to build a LHS. The authors spread sheet. As more data are collected and reuse as part of a LHS but, as with the core use different open source tools for data ware- analyses become more automated as part of infrastructure tools and statistical software housing and business intelligence. They used a formal LHS, this approach becomes more described above, can form part of a LHS the Talend Open Studio for data integration difficult and organisations and researchers when combined with other software. Some (www.talend.com), a PostgreSQL (www.post- use more sophisticated systems for manag- high-profile examples from the MedFLOSS gresql.org) database for data management, ing and analysing the clinical data. The R database are detailed in Table 4. and for data analysis, business intelligence, open source statistical analysis package is and result , they used the BIRT often used as part of the infrastructure of a (Business Intelligence and Reporting Toolkit, LHS as there is now a large library of open www.eclipse.org/birt) toolkit. source tools that allow data to be import- Clinical Data Management FLOSS ed, analysed, and exported to dashboards At the heart of a LHS is the ability to trans- or other visualisation systems often in an form clinical data into new knowledge. This automated or semi-automated fashion. As knowledge may be generalizable and may Data Analysis FLOSS datasets grow larger and analyses become contribute to new scientific understanding or In order to reuse data from clinical sys- more complex, the limits of statistical tools it may be specific to a particular healthcare tems, they must first be analysed to create such as R mean that new “Big Data” tools organisation and contribute to local quality actionable insights such as mortality rates are required. These tools enable researchers improvement. Over recent years, a number

IMIA Yearbook of Medical Informatics 2017 55 The Role of Free/Libre and Open Source Software in Learning Health Systems

Table 4 Examples of Clinical FLOSS TranSMART: TranSMART is another open source clinical and basic science re- Application Types Examples search platform, originally developed by Johnson and Johnson and then established (EHR) WorldVistA, OpenMRS, GNU Health, OpenMAXIMS, LibreHealth EHR, as an open source platform supported by the tranSMART Foundation from 2013 [22]. Picture Archive and Communication System (PACS) MRIdb, , DICoogle, Xebra, OSPACS, Both the tranSMART and OpenClinica OpenSourcePACS, ClearCanvas, Conquest DICOM software, CDMEDIC PACS WEB, DCMTK - DICOM Toolkit, dcm4che platforms have been extended, connected, and enhanced by the international open Health Information System (HIS) DHIS2 source community. For example, Cama- Laboratory Information Management System (LIMS) OMERO, FreeLIMS, OpenELIS cho Rodriguez et al. implemented Mirth Interoperability Mirth, OpenXDS, IHE open source, OpenARNA, IHE Gazelle Connect as a Communication Server Tools, Open eHealth Integration Platform (IPF) (CS) to convert HL7 messages either to Operational Data Model (ODM) data for the automatic import in OpenClinica or of FLOSS projects have been developed profiles generated by Next Generation tabular-delimited text format files, whose to assist in the analysis of clinical data. Sequencing [15]. Westra et al. were able data is uploaded in tranSMART using the Many of these have been developed for to demonstrate the feasibility of creating tMDataLoader tool [23]. Firnkorn et al. the purpose of large scale observational a hierarchical flowsheet ontology in i2b2 have addressed the problem of extracting trials but can also be used for local quality using data-derived information models data from data warehouses like i2b2 and improvement. In this section, we highlight and determined the underlying informatics tranSMART. They have implemented a some of the recent literature around FLOSS and technical issues [16]. Rance et al. have Generic Case Extractor for clinical items clinical data tools to give a broad overview built a platform for translational research in out of these two databases with the aim of of some of the current major projects and cancer research using i2b2 and tranSMART extracting data for more complex statisti- how they can work together and support [17, 18] (described further below). cal analysis [24]. Satagopam et al. have data and terminology standards: OpenEHR: OpenEHR is an open data combined tranSMART with two other i2b2: i2b2 (Informatics for Integrating specification supported by an internation- server-based platforms addressing different Biology & the Bedside) [10] is an NIH-fund- al community that have also developed a aspects of data processing in translational ed suite of open source tools for supporting number of open source tools for modelling medicine research: data integration and clinical and basic biological science research clinical data. Haarbrandt et al. have imple- exploration, workflow projects and is now used in several major mented a solution for the automated pop- construction, and interpretation of anal- academic medical centres and research insti- ulation of an i2b2 clinical data warehouse ysis results in the disease context [25]. In tutes in the USA. The system has also been from an openEHR-based data repository particular, they have integrated MINERVA used for secondary use of claims data from [19]. With this approach openEHR-based (Molecular Interaction NEtwoRk VisuAl- the Austrian Health Insurance system [11] data can be made available to researchers ization) [26], with tranSMART and Galaxy, and has undergone successful viability tests for secondary use. a web-based platform for computational for implementation in networked medical OpenClinica: OpenClinica is an open biomedical research [27]. research in Germany although some chal- source clinical research data management Weasis: A remaining challenge to using lenges were identified when importing from platform that is now in use in over 100 electronic data capture systems (EDCSs) heterogeneous data sources [12]. countries [20]. The system has a number of in the context of secondary use of patient The open source nature of the i2b2 features for conducting clinical trials includ- data is in respect to the use of image data. platform has resulted in multiple projects ing randomization and support for patient-re- Although this is usually provided in a extending and adding new capabilities ported outcome measures (PROMs) and it standardized form (DICOM), there are still to the existing tools. For example, Bauer can connect with other clinical systems such challenges in connecting PACS systems et al. reported on work where they have as Electronic Health Records and Picture Ar- with EDCS. Haak et al. have implemented developed a suite of tools for loading chive and Communications Systems (PACS). a web-based solution integrating EDCS, clinical data from various formats such as REDCap: REDCap is another free PACS, and a DICOM viewer using the CSV, SQL, CDISC ODM or biobanks [13]. platform for data capture (although not open source projects OpenClinica, dcm- Gabetta et al. have developed a flexible licensed as open source). This platform has 4che, and Weasis [28,29]. The system is extension of i2b2 able to exploit different been adopted for many clinical studies in used to perform decentralized computed statistical engines [14]. The same group low-resource settings such as the Clinical tomography (CT) screening. has developed BigQ, an extension of the Information Network in Kenya where it is CohortExplorer: Another challenge is i2b2 framework, which integrates patient integrated with a suite of open source data the underlying database model of EDCSs clinical phenotypes with genomic variant capture, processing, and analysis tools [21]. which is usually based on generic entity-at-

IMIA Yearbook of Medical Informatics 2017 56 Paton et al.

tribute-value (EAV) schemas. This struc- i2b2 and called the solution C3-PRO [38]. business sustainability and growth. This ture is cumbersome to query with SQL and C3-PRO stands for Consent, Contact, and method of building on an ‘open source even more challenging when combining Community framework for Patient Report- stack’ may encourage more entrepreneurial distinct sources with different underlying ed Outcomes and aims to collect personal activity and be a cost effective way for new schemas. The application CohortExplorer health data and patient reported outcomes start-ups to enter the market. [30] provides a solution to this issue. It is (PRO) from distributed populations. For healthcare providers, utilizing available under a GPL 3+ and available OMOP: One of the core challenges FLOSS may also offer a more sustain- with several APIs for connecting to other to implementing LHSs in practice is the able approach. Due to the complex and systems (e.g. for Opal and REDCap). Fu- standardisation and adoption of common interconnected nature of LHSs, it may be ture versions will also provide an API for data models to ensure consistency between beneficial for healthcare providers to adopt OpenClinica. data used in day-to-day clinical practice and open source systems rather than purchasing SMART: The Strategic Healthcare IT data analysed to generate recommendations proprietary systems that may be difficult to Advanced Research Projects (SHARP) for quality improvement or research. The integrate with other systems (for example, program is an initiative from the Office of Observational Medical Outcomes Partner- if an EHR does not allow for integration the National Coordinator for Health Infor- ship (OMOP) [39] developed a common with automated data analysis systems). mation Technology (ONC) in the USA that data model and suite of open source tools Various commercial business models, supports breakthrough health IT research. to facilitate the adoption of the model and implemented for reasons of business sus- One area they are supporting is secondary analysis of patient data gathered from clin- tainability and growth, may work against use of EHR data. In one of the SHARP ical IT systems. the development of a LHS. For example, projects, Klann et al. are using Substitut- The open source nature of these tools encouraging customers to purchase a sin- able Medical Apps Reusable Technologies has fostered a range of collaborations and gle system across an organisation may be (SMART) to build a patient summarization extensions enabling them to work with each problematic if the system doesn’t include app which displays a problem-oriented other, adopt data standards, and support the data analysis functionality and does not view of medications and presents a line- integration of clinical systems such as EHRs offer a way to easily integrate research and graph display of laboratory results [31, 32]. with data management tools to build digital data management systems. The summarization app can run in any EHR infrastructures that can support LHSs. The use of open standards (such as HL7 environment that either supports SMART FHIR) may mitigate some of these issues if or runs SMART-enabled i2b2. In this way they are widely adopted by commercial ven- data can be extracted from an i2b2 repos- dors capable of meeting the needs of both itory reusing clinical data extracted from interoperability and data analysis. However, EHRs. Area four of SHARP (SHARPn) Discussion if the proprietary systems on offer focus is aimed at developing open-source tools In order to promote the adoption of LHSs only on more basic messaging standards that could be used for the normalization and the reuse of patient data, a number of (such as HL7 v2), it may be preferable to of EHR data for secondary use using high new clinical solutions will be required and adopt an open source infrastructure that throughput phenotyping [33]. One of the existing systems will need to be adapted and allows for modification and collaboration outcomes of this project is that clinical el- extended to allow for LHS functionality. between institutions to develop APIs, meta- ement models proved capable of capturing The role of FLOSS in these projects will data, and modular extensions (such as seen data originating from a variety of sources range from building on top of core FLOSS with the OpenMRS system [40]). within the normalization pipeline and are infrastructure (such as operating systems and Although the use of patient data for capable of serving as suitable normaliza- databases) to fully open source clinical and research has been underway now for many tion targets [34]. research systems. years in academic medical centres, large FHIR: Fast Healthcare Interoperability For software developers, the adoption healthcare providers are now starting to Resources (FHIR) is an emerging open or not of open source software in clinical gather data from across multiple healthcare HL7 standard [35]. Mandel et al. have solutions may depend on the types of busi- facilities (sometimes at a national level) to adopted the emerging FHIR standard for ness models necessary to support business build larger scale LHSs. The use of this data use on the SMART framework and called or organisational sustainability. It may be for research and commercial exploitation is the platform SMART-on-FHIR [36]. cost-effective for a clinical software sup- generating interesting discussions on own- Wagholikar et al. have implemented an plier to leverage open source software for ership and privacy rights for the individuals interface for an i2b2 installation using core infrastructure (such as the operating that data reference. In the UK, a large scale SMART-on-FHIR and have transformed system and database) and then use an open clinical data aggregation project called i2b2 into an apps platform [37]. Authors source programming language to create Care.Data made headlines about whether from the same group have implemented copyrighted code for the user interface citizens should have to opt-in and perhaps an open source toolchain to connect any layer which is released as a commercial explicitly consent to such data collection, app developed with Apple’s ResearchKit to product with the proceeds supporting to what degree the data should be made

IMIA Yearbook of Medical Informatics 2017 57 The Role of Free/Libre and Open Source Software in Learning Health Systems

anonymous, and whether or not such data 2015 Oct 11 [cited 2017 Mar 24];2015. Avail- 17. Rance B, Canuel V, Countouris H, Laurent-Puig should be sold to researchers and commer- able from: https://www.hindawi.com/journals/ P, Burgun A. Integrating Heterogeneous Biomed- bmri/2015/961526/ ical Data for Cancer Research: the CARPEM cial organisations [41]. A recent report by 3. Friedman C, Rubin J, Brown J, Buntin M, Corn M, infrastructure. Appl Clin Inform 2016 May Dame Caldicott offers some useful guid- Etheredge L, et al. Toward a science of learning 4;7(2):260–74. ance on how we should communicate with systems: a research agenda for the high-function- 18. Dunn W Jr, Burgun A, Krebs M-O, Rance B. Ex- patients about data reuse and when and ing Learning Health System. J Am Med Inform ploring and visualizing multidimensional data in translational research platforms. Brief Bioinform where we need to obtain explicit consent Assoc 2015 Jan;22(1):43–50. 4. Roundtable on Value & Science-Driven Health [Internet]. 2016 Sep 1; Available from: http:// [42]. Collaboratively developed FLOSS for Care, The Learning Health, Institute of Medicine. dx.doi.org/10.1093/bib/bbw080 19. Haarbrandt B, Tute E, Marschollek M. Automated helping manage consent and data security Digital Infrastructure for the Learning Health population of an i2b2 clinical data warehouse from may help to mitigate some of these issues. System: The Foundation for Continuous Improve- an openEHR-based data repository. J Biomed ment in Health and Health Care: Workshop Series Inform 2016 Oct;63:277–94. Summary. Grossmann C, Powers B, McGinnis JM, 20. Cavelaars M, Rousseau J, Parlayan C, de Ridder editors. Washington (DC): National Academies S, Verburg A, Ross R, et al. OpenClinica. J Clin Press; 2011. 320 p. Bioinforma 2015;5(Suppl 1):S2. Conclusions 5. English M, Irimu G, Agweyu A, Gathara D, Oliwa J, 21. Tuti T, Bitok M, Paton C, Makone B, Malla L, Ayieko P, et al. Building Learning Health Systems Muinga N, et al. Innovating to enhance clinical FLOSS infrastructure including operating to Accelerate Research and Improve Outcomes of data management using non-commercial and systems, databases, and software frame- Clinical Care in Low- and Middle-Income Coun- open source solutions across a multi-center works have enabled much of the technology tries. PLoS Med 2016 Apr;13(4):e1001991. network supporting inpatient pediatric care and 6. Medical Free/Libre and Open Source Software research in Kenya. J Am Med Inform Assoc 2016 we use every day as we access information | Structured project database, service providers, Jan;23(1):184–92. on our smartphone apps or through the in- publications, events, [Internet]. [cited 2017 Mar 22. Athey BD, Braxenthaler M, Haas M, Guo Y. ternet. Many of these tools also power the 24]. Available from: http://medfloss.org tranSMART: An Open Source and Communi- 7. Brown SH, Lincoln MJ, Groen PJ, Kolodner RM. ty-Driven Informatics and Data Sharing Platform core infrastructure behind current health for Clinical and Translational Research. AMIA Jt IT systems and will likely form the digital VistA—U.S. Department of Veterans Affairs national-scale HIS. Int J Med Inform 2003/3;69(2– Summits Transl Sci Proc 2013 Mar 18;2013:6–8. infrastructure for the learning health sys- 3):135–56. 23. Camacho Rodriguez JC, Stäubert S, Löbe M. Automated Import of Clinical Data from HL7 tems of the future. FLOSS can help new 8. Todd R. Spine2 built on Riak database | Digital Messages into OpenClinica and tranSMART Us- clinical software companies rapidly ‘spin Health [Internet]. Digital Health. 2013 [cited ing Mirth Connect. Stud Health Technol Inform. up’ new solutions and, on a larger scale, 2017 Mar 24]. Available from: https://www. 2016;228:317–21. healthcare providers, research organisa- digitalhealth.net/2013/10/spine2-built-on-riak- 24. Firnkorn D, Merker S, Ganzinger M, Muley T, database/ tions, and governments can take advantage Knaup P. Unlocking Data for Statistical Analyses 9. Hackl WO, Rauchegger F, Ammenwerth E. A and Data Mining: Generic Case Extraction of of the collaborative and extensible nature Nursing Intelligence System to Support Secondary Clinical Items from i2b2 and tranSMART. Stud of FLOSS to implement clinical and data Use of Nursing Routine Data. Appl Clin Inform Health Technol Inform 2016;228:567–71. management systems that can speed up 2015 Jun 24;6(2):418–28. 25. Satagopam V, Gu W, Eifes S, Gawron P, Ostasze- quality improvement cycles and enable 10. i2b2: Informatics for Integrating Biology & the wski M, Gebel S, et al. Integration and Visual- Bedside [Internet]. [cited 2017 Mar 24]. Available ization of Translational Medicine Data for Better large scale clinical research projects that from: https://www.i2b2.org/ Understanding of Human Diseases. Big Data 2016 utilize learning healthcare systems. 11. Endel F, Duftschmid G. Secondary Use of Claims Jun;4(2):97–108. Data from the Austrian Health Insurance System 26. Gawron P, Ostaszewski M, Satagopam V, Gebel S, with i2b2: A Pilot Study. Stud Health Technol Mazein A, Kuzma M, et al. MINERVA—a plat- Funding Statement Inform 2016;223:245–52. form for visualization and curation of molecular CP is funded by the Health Systems Re- 12. Ganslandt T, Mate S, Helbing K, Sax U, Prokosch interaction networks. NPJ Syst Biol Appl 2016 Sep search Initiative grant (MR/N005600/1) HU. Unlocking Data for Clinical Research – The 22;2:16020. jointly supported by the Department for German i2b2 Experience. Appl Clin Inform 2011 27. Goecks J, Nekrutenko A, Taylor J, Galaxy Team. Mar 30;2(1):116–27. Galaxy: a comprehensive approach for supporting International Development (DFID), the Eco- 13. Bauer CRKD, Ganslandt T, Baum B, Christoph J, accessible, reproducible, and transparent computa- nomic and Social Research Council (ESRC), Engel I, Löbe M, et al. Integrated Data Repository tional research in the life sciences. Genome Biol the Medical Research Council (MRC) and Toolkit (IDRT). A Suite of Programs to Facilitate 2010 Aug 25;11(8):R86. the Wellcome Trust (WT). Health Analytics on Heterogeneous Medical Data. 28. Home - Weasis - dcm4che.org Wiki [Internet]. [cit- Methods Inf Med 2016;55(2):125–35. ed 2017 Mar 24]. Available from: https://dcm4che. 14. Gabetta M, Malovini A, Bucalo M, Zini E, Tibollo atlassian.net/wiki/display/WEA/Home V, Priori SG, et al. Beyond Cohort Selection: An 29. Haak D, Page C-E, Reinartz S, Krüger T, Deserno References Analytics-Enabled i2b2. Stud Health Technol TM. DICOM for Clinical Research: PACS-Inte- Inform 2016;228:572–6. grated Electronic Data Capture in Multi-Center 1. About PCORnet - PCORnet [Internet]. PCORnet. 15. Gabetta M, Limongelli I, Rizzo E, Riva A, Segagni Trials. J Digit Imaging 2015 Oct;28(5):558–66. [cited 2017 Mar 24]. Available from: http://www. D, Bellazzi R. BigQ: a NoSQL based framework 30. Dixit A, Dobson RJB. CohortExplorer: A Generic pcornet.org/about-pcornet/ to handle genomic variants in i2b2. BMC Bioin- Application Programming Interface for Entity 2. Delaney BC, Curcin V, Andreasson A, Arvanitis formatics 2015 Dec 29;16:415. Attribute Value Database Schemas. JMIR Med TN, Bastiaens H, Corrigan D, et al. Translational 16. Westra BL, Christie B, Johnson SG, Pruinelli L, Inform 2014 Dec 1;2(2):e32. Medicine and Patient Safety in Europe: TRANS- LaFlamme A, Park JI, et al. Expanding Interpro- 31. Klann JG, McCoy AB, Wright A, Wattanasin N, FoRm—Architecture for the Learning Health fessional EHR Data in i2b2. AMIA Jt Summits Sittig DF, Murphy SN. Health care transformation System in Europe. Biomed Res Int [Internet]. Transl Sci Proc 2016 Jul 20;2016:260–8. through collaboration on open-source informat-

IMIA Yearbook of Medical Informatics 2017 58 Paton et al.

ics projects: integrating a medical applications dards-based, interoperable apps platform for developing countries. AMIA Annu Symp Proc platform, research data repository, and patient electronic health records. J Am Med Inform Assoc 2006;1146. summarization. Interact J Med Res 2013 May 2016 Sep;23(5):899–908. 41. Carter P, Laurie GT, Dixon-Woods M. The social 30;2(1):e11. 37. Wagholikar KB, Mandel JC, Klann JG, Wattanasin licence for research: why care.data ran into trouble. 32. Klann JG, Abend A, Raghavan VA, Mandl KD, N, Mendis M, Chute CG, et al. SMART-on-FHIR J Med Ethics 2015 May;41(5):404–9. Murphy SN. Data interchange using i2b2. J Am implemented over i2b2. J Am Med Inform Assoc 42. Review of data security, consent and opt-outs Med Inform Assoc 2016 Sep;23(5):909–15. [Internet] 2016 Jun 6; Available from: http://dx.doi. - GOV.UK [Internet]. [cited 2017 Mar 24]. 33. Rea S, Pathak J, Savova G, Oniki TA, Westberg org/10.1093/jamia/ocw079 Available from: https://www.gov.uk/government/ L, Beebe CE, et al. Building a robust, scalable 38. Pfiffner PB, Pinyol I, Natter MD, Mandl KD. publications/review-of-data-security-consent-and- and standards-driven infrastructure for secondary C3-PRO: Connecting ResearchKit to the Health opt-outs use of EHR data: the SHARPn project. J Biomed System Using i2b2 and FHIR. PLoS One 2016 Inform 2012 Aug;45(4):763–71. Mar 31;11(3):e0152722. 34. Oniki TA, Zhuo N, Beebe CE, Liu H, Coyle JF, 39. Stang PE, Ryan PB, Racoosin JA, Overhage JM, Parker CG, et al. Clinical element models in the Hartzema AG, Reich C, et al. Advancing the sci- Correspondence to : SHARPn consortium. J Am Med Inform Assoc ence for active surveillance: rationale and design Dr Chris Paton 2016 Mar;23(2):248–56. for the Observational Medical Outcomes Partner- Centre for Tropical Medicine and Global Health 35. Index - FHIR v3.0.0 [Internet]. [cited 2017 Mar ship. Ann Intern Med 2010 Nov 2;153(9):600–6. Peter Medawar Building for Pathogen Research 24]. Available from: http://hl7.org/fhir/index.html 40. Wolfe BA, Mamlin BW, Biondich PG, Fraser HSF, South Parks Road 36. Mandel JC, Kreda DA, Mandl KD, Kohane Jazayeri D, Allen C, et al. The OpenMRS system: Oxford OX1 3SY, UK IS, Ramoni RB. SMART on FHIR: a stan- collaborating toward an open source EMR for E-mail: [email protected]

IMIA Yearbook of Medical Informatics 2017