Reflections on the Role of Open Source in Health Information System Interoperability

S. Sfakianakis1, C.E. Chronaki1, F. Chiarugi1, F. Conforti2, D.G. Katehakis1 1Institute of Computer Science, Foundation for Research and Technology–Hellas, Heraklion, Crete, Greece 2Institute of Clinical Physiology, National Research Council, Pisa, Italy

Summary 1 Introduction General Public License (GPL) [3]. Objectives: This paper reflects on the role of open source in health GNU GPL requires copies and deriva- information system interoperability. Open source is a driving force in Open source is a philosophy, a devel- tives of the source code to be made computer science research and the development of information opment model, but above all a license available on terms no more restrictive systems. It facilitates the sharing of information and ideas, enables model. According to the Berne conven- than those of the original license. evolutionary development and open collaborative testing of code, tion without a software license, a pro- Computers exchange health data in en- and broadens the adoption of interoperability standards. In health gram cannot be copied or modified coded formats. Software needs to care, information systems have been developed largely ad hoc without the explicit permission of the implement correctly the data formats, following proprietary specifications and customized design. authors [1]. Most software licenses are otherwise massive data corruption may However, the wide deployment of integrated services such as proprietary. Proprietary licences do not occur. Customers that save their data Electronic Health Records (EHRs) over regional health information networks (RHINs) relies on interoperability of the underlying allow users to access the source code of in the format of a particular vendor find information systems and medical devices. the program. Open source licenses, on themselves unable to choose a differ- Methods: This reflection is built on the experiences of the PICNIC the other hand, allow users to study, ent vendor who does not implement that project that developed shared software infrastructure components in copy, modify, and in some cases release format even though the corresponding open source for RHINs and the OpenECG network that offers open the modified program for the benefit software may be superior (vendor lock- source components to lower the implementation cost of of the community. They may be out). In healthcare, it is not perceived interoperability standards such as SCP-ECG, in electrocardiography. copyleft or non-copyleft. A non- as sustainable to allow a single vendor Results: Open source components implementing standards and a copyleft license is a non-persistent open to dominate the market and the wide use community providing feedback from real-world use are key enablers source license. A non-copylefted pro- of open standards is a requirement for of health care information system interoperability. gram may be redistributed, and users interoperability and continuity of care. Conclusions: Investing in open source is investing in interoperability and a vital aspect of a long term strategy towards comprehensive are allowed to make proprietary ver- The International Telecommunication health services and clinical research. sions of the program. The BSD license Union defines open standards as “stan- [2], a non-copyleft licence, allows in- dards made available to the general Keywords corporating the open source software public that are developed, approved, and Open source, interoperability, standards, health information systems, (OSS) in proprietary commercial prod- maintained via a collaborative and con- electrocardiography ucts to be released under different li- sensus-driven process.” [4] Thus, a key censes. That is particularly important element of open standards is a collabo- Geissbuhler A, Haux R, Kulikowski C, editors. IMIA Yearbook of for interoperability as it supports the rative process that supports their vol- Medical Informatics 2007. Methods Inf Med 2007; 46 Suppl 1: 51-61 consistent implementation of open stan- untary and market-driven development dards and the easy integration of het- following a transparent consensus- erogeneous information systems. In driven process that is reasonably open contrast, a copyleft license is persistent. to all interested parties and is reason- Redistribution of a copylefted program, ably balanced to ensure that it is not either in modified or unmodified form, dominated by any one interest group. gives the recipients the same rights as Intellectual property rights to imple- the original program. A prerequisite for ment the open standard are either for this is that the source code is distrib- free or on non-discriminatory terms uted with the program. The most wide- that may include reasonable monetary spread copyleft license is the GNU compensation. Implementation of stan-

dards should be free of economical, (led by the American College of Radi- health professionals, and healthcare fa- political, or legal implications on their ologists (ACR) and the National Elec- cilities), as well as clinical data access development and use. Open standards trical Manufacturers Association through open interfaces, are considered should facilitate interoperability among (NEMA)) and SNOMED [11] (led by necessary to establish regional services products and services, and their inte- the College of American Pathologists). and enable collaboration among health gration in realistic processes. Empha- Note that open standards doesn’t always care facilities. Using open standard in- sis on quality and sufficient level of mean cost-free standards. In some cases terfaces and open source technologies detail should permit the development like SNOMED, agreements are formed to implement such components as well of a variety of competing implementa- at the country level to maximise stan- as creating an open source community tions of interoperable products or ser- dardization benefits and manage the to support the evolution, customization, vices. Standardized interfaces should be costs involved. and maintenance of these components public and maintained by the Standard With Regional Health Information Net- was a key objective for the PICNIC Developing Organization (SDO) re- works (RHINs), the pressing need for project [17]. The PICNIC experience sponsible for the standard. wide expertise and multi-stakeholder with the development of OSS compo- Specifically in health care, open source consensus in addressing heterogeneous nents is presented in the next section. is linked interoperability which is a data exchange and complex workflows, With electrocardiographs and other low patient safety issue, as it minimizes er- became evident. Industry consortia like cost medical devices the situation was ror-prone processes and suboptimal care OMG CORBAmed [12] and indepen- different. The best that patients and within and across health organizations dent organizations like HL7 [13] de- carers could hope for is to get the world wide [5]. Interoperability, the velop “de facto” standards. Finally, the patient’s old ECG traces in paper or ability of two or more systems or com- need to collaboratively validate standards digitized. Despite the existence of SCP- ponents to exchange information (func- on the basis of integration profiles is ECG [18], an official interoperability tional interoperability) and to use the addressed by IHE [14]. IHE integra- standard in Europe (EN1064:2005, information (semantic interoperability) tion profiles promote the coordinated EN1064:2007), most digital electro- that has been exchanged [6], is a pre- implementation of standards in the con- cardiographs implement proprietary requisite to the establishment of an ac- text of concrete integration scenarios, data formats, as OpenECG pointed out tive life-long Electronic Health Record supplementing the vendors’ statements [19]. Section 3 reports on ECG inter- (EHR). Open standards are important of conformance to standards such as operability standards, conformance when heterogeneous systems need to co- DICOM and HL7. Validation exercises testing, and experiences from the operate to support clinical workflows or Connectathons are carried out in OpenECG open source repository ar- and exchange information to facilitate HIMMS [15] and IHE meetings every guing that OSS, converters, and con- availability of patient information when year, testing new integration profiles formance testing services are vital for and where needed, to reduce medical and identifying pitfalls and possible the seamless integration of digital ECGs errors, and to protect patient safety. limitations of existing standards. in the EHR and their effective use in Interoperability can undoubtedly only Meanwhile, the development of a ro- eHealth services. Then, section 4 re- be achieved with open standards and bust middleware infrastructure to sup- flects on the role of open source in an OSS community that enables their port IHE-type integration scenarios interoperability of health information consistent implementation and wide- became of paramount importance for systems, whereas section 5 presents the spread adoption. RHINs aiming to support registries such main conclusions. Standards in health care are “de jure” as those for immunization and disease (in principle) and “de facto” (in prac- surveillance, integrated to health infor- tice). Besides the ISO TC215 [7], CEN mation systems in daily use [16]. Start- TC251 [8] in Europe and ANSI [9] in ing from OMG CORBAmed compo- 2 PICNIC: Open Source the USA, which are the formal SDOs nent specifications, a number of Components for RHINs developing “de jure” standards, there common components in open source exist independent organizations or pro- were developed. In particular, common PICNIC stands for Professionals and fessional societies supporting the devel- infrastructure components for security Citizens Network for Integrated Care opment of eventually accredited “de- and confidentiality, resource identifi- and was a European Commission co- facto” standards such as DICOM [10] cation & access (including patients, funded research and development

project, established under the 5th Frame- involved Surveillance Information, code was necessary for PICNIC to de- work of European Research “Informa- Yellow Pages, Professional Guide- liver harmonized interoperable services tion Societies Technology Programme”. lines, Disease Quality Manage- throughout Europe by means of com- PICNIC designed, developed, validated, ment, Public Health Information, mon enabling OSS components. All and Continuing Professional Devel- and delivered a number of certified specifications relevant to the PICNIC opment OSS infrastructure components for • Administrative Services and Elec- architecture and associated standards, eHealth, in close co-operation with the tronic Commerce that involved scenarios and models for patient-centred industry [20]. Having eighteen RHIN Reimbursement, Electronic Com- care, and common components devel- services as a starting point, PICNIC merce, Patient Identification (ID), opment were made available in the pub- through a carefully planned storyboard and Resource Management. lic domain. Common components de- articulated the future healthcare re- A common goal for all three groups veloped by PICNIC were released in sponse with the empowered citizen at has been the provision of reliable com- open source, under the three clause BSD the centre. Professionals provide ser- munication, sharing and access to one’s licence [20]. vices and support through a RHIN that EHR [21] to enable safe and shared aims to maintain the citizens’ quality clinical care by means of an enabling of life despite clinical illness which Health Information Infrastructure (HII). 2.1 PICNIC OSS Components impacts not only their family life, but Component development concentrated also their ability to sustain an occupa- in three groups: An integrated EHR requires an enabling infrastructure that provides ac- tion. The initial set of RHIN services • Clinical Messaging consisting of had been divided into three groups: highly structured patient-related cess to heterogeneous health informa- • Clinical Services and Telemedicine information tion systems under appropriate security that involved Clinical Messages, • Access to Patient Data by means of constraints. The relevant common

Clinical e-Mail, Clinical Booking, accessing common information ser- components were delivered as follows Shared Records, Care Protocols, vices (Figure 1): Mobile and Emergency, Home-care • Collaboration through immediate • Clinical Observation Access Ser- Monitoring, and Telemedicine access to specialized expertise. vice (COAS) for obtaining patient • Health Information Services that Sharing of know-how as well as source data directly from Health Informa-

Fig 1 PICNIC IT services (PICNIC architecture)

tion Systems and updating patient cational resources were deemed neces- terprise integration tools/ methods only information based on OMG COAS sary to extend the group knowledge of becomes apparent at a later stage of interface specification [22] TAO’s intrinsics [24]. Finally, the cre- RHIN development, when integration • Patient Identification Service ation of the PICNIC Open Community across organisational/ regional bound- (PIDS) for enabling unique patient was encouraged to establish a commu- aries is attempted. identification across a RHIN and for the provision of a master pa- nity of OSS developers who were com- Therefore, PICNIC customers were tient index based on OMG PIDS mitted to further development, en- members of the healthcare ICT vendor interface specification [23] hancement, and re-use of the common community who were helping to build • Shared Records Indexing Service PICNIC components. and operate RHINs (i.e. the policy (SRIS) for the management and in- makers and technology procurement dexing of meta-information on the experts). PICNIC suppliers were a com- content of the patient data in health 2.2 PICNIC Results munity of OSS component developers information systems throughout the and health telematics researchers with RHIN PICNIC developed a new architectural shared interests and objectives. • Shared Records Update Broker paradigm for healthcare ICT based on The business case for using the PICNIC (SRUB) serving as a mediation tool for maintaining EHR consistency an information infrastructure compris- architecture and components was based across the RHIN by propagating to ing a set of OSS components that sup- on the economies of scale of using stan- SRIS modifications pertaining to port the architecture, a proactive assess- dards and components which are ac- patient data. ment model, multiple case studies with cepted across Europe and enable health- The collaboration component group de- independent evaluation reports and a care ICT vendors to develop software livered the following OSS components: certification model. incorporating components that operate • Collaboration Service (COLS) for PICNIC used many existing OSS com- in any EC country/ region and support the provision of the platform that ponents, enhanced or extended some of both the interoperability of products and allows general practitioners and these, and developed others. The result- their integration with existing legacy medical experts to share patient-re- ing prototype subsystems were inte- systems. Expected savings arise from lated information in the context of grated successfully to pre-existing adopting common standards and sharing a teleconsultation episode both in- health information systems. However, the costs of software development. side and across RHINs undertaking socio-economic assess- • Resource Service (RESS) for identification of healthcare related ments of individual components, when agents (e.g. organizations, devices, they were embedded in an application software, etc) and the means for ac- or service comprised of many compo- 3 OpenECG: Open Source cessing them and determining their nents, turned out to be a complex task. for ECG Interoperability availability. It turned out that European healthcare For the development of these compo- ICT service suppliers will not specula- Electrocardiography is the most fre- nents a large number of OSS tools, li- tively adopt an architectural framework quently applied non-invasive examina- braries, compilers, etc. were used for standards, no matter how robust the tion for early detection and follow-up mainly originating GNU. Development case for its use, without a clear demand of heart disease, a leading cause of support was provided by OSS mailing for such a framework coming from the morbidity and mortality in western lists and newsgroups. This kind of sup- buyers of ICT services. Thus, the lead countries. Electrocardiograms (ECGs) port is certainly “unpredictable”, since in adopting a RHIN architecture comes should be exchanged between health there is no guarantee that the problem from the ICT policymakers in the re- care providers and stored in EHRs in a encountered will be answered soon if gional health authorities. Many authori- device-independent standard ECG for- at all. Nevertheless, PICNIC found the ties, however, do not appreciate the mat, to support continuity of care. Se- OSS community to be highly respon- need for an inter-enterprise integration rial comparison of ECGs in the con- sive and very helpful while developing architecture, as they are still in the busi- text of an individual’s EHR can provide the components. The majority of issues ness of trying to integrate islands of valuable information and early warn- faced in PICNIC proved to be minor information within their organisations ing on the progress of heart disease. problems that other OSS developers had following “traditional” enterprise ap- However, due to inconsistent data for- already encountered. In addition, edu- proaches. Thus, the need for inter-en- mats and incompatible protocols, ECGs

Fig. 2 OpenECG network in January 2006. At the end of 2006 there are more than 700 members from 58 countries worldwide cannot be exchanged in a way that al- procedure for ECG device-to-computer on SCP-ECG section 1), to meet the lows visualization in high precision, communication and uploading ECG requirements of wearable devices. comparison with previous ECGs, and records from the computer to the ECG Research centers use mainly raw data interpretation in the EHR context. device. Since March 2005, SCP-ECG originating from laboratory instru- This problem led to the creation of the (EN1064:2005) is the European stan- ments, which typically generate inter- OpenECG network in 2002 (see Fig- dard for high quality diagnostic ECG laced channel data. If signals have to ure 2) to encourage and promote the exchange, and if consistently imple- be exchanged a simple signal-oriented use of ECG interoperability standards, mented, it ensures interoperability. An format like EDF [29] or the MIT for- to consolidate expertise and best prac- overview of the SCP-ECG data format mat [30] is typically used. Optimiza- tice in ECGs interoperability standards, appears in Figure 3. tions of data space and transmission to provide assistance in implementation Beyond SCP-ECG there are several time, as well as diagnostic quality, with open source tools & data sets, and open and numerous proprietary ECG which are issues of primary relevance to offer interoperability testing services. standards. The most well-known ones to the SCP-ECG standard, are not of are the HL7v3 Annotated ECG format major concern. [25] created to support full disclosure Nevertheless, it is rather difficult for 3.1 ECG Interoperability Standards in clinical trials and DICOM Supple- integrators to implement SCP-ECG SCP-ECG is the European standard ment 30 [26]. Both of them are appro- correctly and there are variations in [18] for ECG-specific data exchange priate for the storage of waveform data. implementations, which can be a bar- to achieve interoperability between dif- There is also MFER, a Japanese stan- rier to interoperability. Several ECG ferent ECG devices and eHealth sys- dard for the storage of waveforms in- device manufacturers and integrators tems. SCP-ECG is a standard commu- cluding ECGs [27]. Moreover, the have implemented the SCP-ECG stan- nication protocol that specifies the IMEX project is developing MSD [28], dard, but that alone does not attain the interchange format and a messaging a new micro-system data format (based interoperability objective. This is due

receive a list of errors and warnings. If no errors are detected, the submitter may request a certificate that is granted after a more thorough manual review of the file. The conformance testing service has been used extensively by members and an interoperable solution has been achieved in many occasions with support from the OpenECG Helpdesk. In the period 2003-2006, more than 1700 ECGs were submitted for conformance testing by more than 20 coun- ties worldwide (see Figure 5). Leading is Italy with 11 members, who have submitted 37.3% of the tests. After Italy, most ECGs have been submitted Fig. 3 The SCP-ECG data structure by Greece (17.74%) and Hungary (11.89%). As SCP-ECG is a European standard, 92% of the ECGs were sub- partly to misconceptions and partly to online certification service and the mitted by Europe, 4% by Asia (Japan, the lack of widely publicised conform- OpenECG open source repository. Taiwan, China, India, Hong Kong), and ance statements that exist for other stan- just 3% by USA. ECG devices and dards such as DICOM. Secondly, al- eHealth services have been tested, im- though ECG viewers and converters 3.2 OpenECG Certification and proved, and validated using online tools exist for alternative ECG formats, ter- Helpdesk and support from the OpenECG minology issues are not adequately re- An SCP file may include demograph- Helpdesk. In addition, based on mem- solved. Thirdly, achieving inter- ics, ECG channels, measurements, and ber feedback and analysis of tests car- operability is a very costly procedure, diagnostic statements as dictated by the ried out, the robustness of the service since a non-disclosure agreement is still conformance levels of SCP-ECG (Fig- is constantly improving. Several inte- required to integrate a closed ECG de- ure 4). In 2003, OpenECG established grators have received an OpenECG vice. OpenECG advocates that data for- an on-line conformance testing service interoperability certificate. mat and protocol specifications, open to support the OpenECG community source tools, and certification services at large in implementing interoperable should be freely available to reduce the eHealth systems with SCP-ECG sup- cost of device integration. For that pur- port. A member may submit an ECG 3.3 OpenECG Open Source Repository pose, OpenECG launched the first file in an alleged SCP-ECG format and The lack of publicly available OSS for ECG acquisition, storage, viewing, and serial comparison has been identified as one of the main reasons for the poor diffusion or the inaccurate implementation of ECG interoperability standards. OpenECG adopted an OSS strategy to complement the interoperability certification service and further assist developers in the correct implementation of ECG standards. Main manufacturers and research cen- tres were contacted for contributions to Fig 4 Conformance types for the SCP-ECG standard the OpenECG open source repository.

Fig 5 Use of the online conformance testing service for the SCP-ECG standard

Physionet [31] at MIT was very sup- are available in the OpenECG open source from different models of electrocardio- portive and contributed a reader (i.e. repository and serve as the basis for fur- graphs and stores them in a specific file parser) for a particular SCP-ECG dia- ther development. For example in 2005, folder monitored by a software daemon. lect. Other major contributions to the a web service variant of the conform- Each new ECG is processed by the dae- repository were the submissions to the ance testing service was integrated to the mon and stored in the cardiology infor- OpenECG programming contest and an ECG viewer in C++ and in that process mation system maintaining the EHR of SCP-ECG library in JAVA to read and certain limitations of the viewer were the patient. Secondly, an eMail SCP- write ECG records. identified and amended [35]. ECG parsing service build on the same The OpenECG programming contest The SCP-ECG library in JAVA provides library provides visual and textual feed- was organized to stimulate the creation methods for reading and writing an SCP- back on the form and contents of SCP- of testing tools, converters, and SCP- ECG record covering signal redundancy ECG records. OpenECG members that ECG viewers. The contest evaluation suppression (2nd derivative) and Huffman have registered to receive the service, may criteria were SCP-ECG feature cover- encoding. The library was tested with send email to [email protected] with age, accuracy, ergonomics, readability normative SCP-ECG samples from the attached SCP-ECG files. Each email and portability (cross-platform based OpenECG portal and ECG records from with an SCP-ECG attachment is auto- only on OSS software tools). Winning six different ECG device vendors. matically processed. A textual report contributions were: (a) SCP-ECG data OSS in the OpenECG repository can be with the contents of the SCP-ECG reader/writer and C++ viewer with anti- freely downloaded, redistributed and used record and an image of the raw data aliasing signal representation [32]), (b) as the basis for further development, signals is automatically sent back to the DICOM Toolkit extended with SCP- under the terms of GNU General Public sender. The ECG device printout can ECG viewing capabilities [33], and (c) License [3]. be compared with these reports to pro- SCP-ECG reader in Matlab/Octave Already the SCP-ECG library has provide feedback on the interoperability [34]. The submissions to the OpenECG vided SCP-ECG support to several of the recorded ECG. Finally, in coop- programming contest were tested with eHealth applications. Firstly, an ECG eration with a small Italian ECG de- multiple SCP-ECG data sets. All these management system built in the IFC- vice manufacturer an eHealth applica- submissions to the OpenECG contest CNR hospital receives ECG records tion for the real-time acquisition,

storage, processing, and communica- ceived by a specific cardiograph model been the case in the past, to have dif- tion of ECGs in the SCP-ECG format and is stored as BLOB in an ECG table ferent OSS products that are not able has been developed. ECGs are displayed of an Oracle database to be later re- to exchange and use information from in real-time, and diagnostic codes can trieved and overread by a cardiologist. each other. In principle, interoperability be added to the SCP-ECG record, while It was necessary to extend the SCP- is a goal that can be achieved by fol- customizable textual and graphic ECG ECG library with a Java method able lowing open standards and widely ac- reports can be printed. to receive as input two SCP-ECG files cepted practices and methodologies. Other notable cases include an ECG and to produce as output the result of Nevertheless OSS can foster inter- manufacturer and two Italian regional the identity check between the signal operability and the development of stan- health authorities in Italy (ASL6 sections of the SCP-ECG files. In the dards that promote it. The existence of Livorno and ASL4 Chiavarese) that next days the software authors were in an open source reference implementa- expressed interest in implementing contact directly with ASL4 Chiavarese tion for an open standard and special- SCP-ECG-compliant eHealth services and in less than one day the Java method ized tools lower the cost of implement- based on OSS components from the was successfully implemented. Since ing standards and facilitate inter- OpenECG repository. then, ASL4 Chiavarese is using the soft- operability. From the developer’s or the An experimental study was designed in ware in production [37]. integrator’s point of view, a large col- collaboration with ASL6 Livorno to In general, the response of the com- lection of OSS components is vital. establish clinical data exchange includ- munity to the OSS software developed With a reference implementation in ing signals and images among a selected and maintained by the OpenECG re- open source it is easier to build a com- group of general practitioners’ (GPs) pository is positive. However, deploy- munity around an open standard and practices, cardiologists and outpatient ment of the OpenECG OSS by users in assist in the standardization process with clinics. Such services are not particu- most cases required support by experts feedback from implementation. larly popular among GPs and that is not in the OpenECG community or the This case was clearly demonstrated in only due to political, ergonomic or tech- authors of the software, indicating a the case of the OpenECG repository. nical reasons, but also to the lack of tools basic trend and common observation in The role of OSS in the initial develop- for easy management of ECG signals and OSS development. ment, evolution and adoption of the patient data. The objective was to de- DICOM standard is another illustrative ploy a friendly network-integrated ECG example for this case. The development service, using SCP-ECG compliant elec- of the “central test node” (CTN [39]) trocardiographs and ECG e-mail, as well 4 Open Source as a Driver software by the Electronic Radiology as software for ECG acquisition, display, for Interoperability Laboratory of the Mallinckrodt Insti- and communication [36]. tute was the first implementation of the Occasionally, ECG manufacturers down- Today the need for interoperable written technical specification of the load the OSS software from the healthcare applications and tools is more DICOM standard. This implementation OpenECG repository to validate the imperative than ever. The complex served for many years both as a refer- implementations of SCP-ECG in their workflows for the provision of health ence implementation of the DICOM products. Sometimes they encounter care in a modern clinical environment standard and as a conformance test suite problems and request assistance from requires the presence of many poten- which vendors could use in order to test the management of the OpenECG re- tially heterogeneous software entities the compliance of their implementations pository, but, with the assistance of the that work in tandem and are able to un- to the DICOM standard. Since 2000, developers, the manufacturers are able derstand each other. Open source as a the OFFIS DICOM DCMT offers a col- to successfully use the tools at the methodology and development process lection of libraries and applications OpenECG repository can be also used in order to leverage implementing large parts of the At some time, another ECG manufac- interoperability and integration of het- DICOM standard for medical image turer requested assistance to support erogeneous software components. communication and serves as a refer- ASL4 Chiavarese in assuring the con- Open source software development ence implementation in Europe [40]. formance of ECGs to the SCP-ECG alone does not guarantee the inter- DICOM open source developments are standard, in the context of an ECG operability of the final products [38]: complemented with the IHE Medical overreading workflow. An ECG is re- It is undoubtedly feasible, and it has Enterprise Simulators and Analysis

(MESA) tools which provide support acknowledged. Currently there are sev- ity, to take control of modifications, and for conformance testing and eventually eral successful initiatives around the to obtain better security capabilities. interoperability [41]. Thus, OSS re- world that promote open source as a Commercial software packages are de- duces the entry costs for the adoption of facilitator for the diffusion of standards. veloped by skilled engineers inside the standards and provides insight in the The OpenEHR foundation is an inter- company under strict deadlines and development of standards-compliant national non-profit group collabora- rigid specifications. Frequently inter- software. Moreover, OSS supports the tively working to bring about compre- operability and conformance to stan- development and evolution of standards hensive electronic health records [44]. dards are low on the list of features to specifications by providing freely The openEHR foundation is commit- be supported. Often the main company available implementations that can be ted to supporting legislative and indus- is supported by partners, small compa- studied, inspected, modified, and experi- try standards and works closely with nies or group of users, generally for mented with. This was the experience standardization bodies as well as na- importing specific knowledge and for reported in PICNIC for the develop- tional and international project teams. experimental product evaluation. The ment of services for RHINs and Likewise, the Open Source Health Care relationships between the company and OpenECG in integrating electrocardio- Alliance [45] is an international collabo- its partners are always fixed by a legal graphs to EHR and eHealth applications ratory promoting the development and agreement and the product is generally using the SCP-ECG standard. use of OSS in the health care domain. protected by proprietary licence. No- On a related note the espousal of open The International Medical Informatics body can manipulate the software and standards and open source implemen- Association (IMIA) and the American the source files, when available, are tations of these standards can avoid Medical Informatics Association protected by a non-disclosure agree- what is commonly referred as “ven- (AMIA) have established open source ment. Once the product specifications dor lock-in”. The use of proprietary working groups that aim to provide edu- are defined, they are very difficult to software that does not adhere to open cation about the benefits of open source change. Any bug report and any request formats and standards can lead to a and raise awareness on OSS in general. for software modification or customi- situation where a customer is depen- Under the auspices of these organizations zation has to be addressed to the com- dent on a specific vendor’s offerings and the active members of the relevant pany; and the user has to wait for the and switching to another product en- user and developer communities, a sub- new release. Doug Schmidt, who leads tails significant costs. In the public sec- stantial number of OSS packages have the long term project on ACE+TAO, a tor, it is not perceived as sustainable to been developed. These OSS packages successful open source middleware bro- allow a single vendor to dominate the cover many different software needs in ker [55], notes that open source should market and the wide use of open data the healthcare domain ranging from be a facilitator to a business model to formats is a requirement for inter- image visualization, storage, and re- leverage expertise and improve software operability and the key for hampering trieval systems such as dsm4che -a quality. He also notes that while a core single vendor monopolies. An example DICOM implementation in Java [46], group of ten good programmers can of such an open standard is the Open DVTK -a DICOM validation toolkit develop high quality software, for soft- Document Format (ODF) which is a [47], and OSIRIS -a multiplatform ware testing you need a large number format for the storage and exchange of DICOM viewer [48], to EHR software of programmers. Moreover, the turn- “office” documents such as text docu- such as GNUmed [49], FreeMed [50], around time for correcting an error in ments, spreadsheets, and presentations. MirrorMed [51], OpenEMed [52], OSS can be a lot shorter, as the person ODF has been ratified as an OASIS OpenEHR [53], and Care2X [54]. identifying the bug may also provide Standard and also an ISO and IEC In- the fix (Figure 6). This was actually ternational Standard (ISO/IEC reflected in the experience of the 26300:2006) and there are many open 4.1 OSS in Retrospect: Advantages OpenECG repository discussed in sec- source and closed source software and Disadvantages tion 3.3 where all requests on OpenECG packages that support it, like the software customizations received a OpenOffice.org [42], Google Docs Frequently, health care organizations rapid response. [43], StarOffice, etc. adopt open source projects to lower the In the case where software vendors go The advantages offered by the open total cost of ownership for technology, out of business, health care institutions source methodology have been widely to access source code, to gain flexibil- are left with a system that they spent a

mented they have to be well-docu- mented and their full specification should be publicly available. OSS has several intrinsic characteristics that en- hance software quality and may act in support of interoperability. However, an aspect of OSS that can be a critical point to its wide adoption is the ability to form an OSS community with adequate capacity and knowledge to provide development support, sustain high software quality, and support OSS evolution. In retrospect, OSS in health care con- Fig. 6 OSS life cycle tributes to the standards adoption process supporting interoperability. Health information systems benefit from an lot of money to acquire and they can- abandonment. In reality, as Henning open source community which, in co- not upgrade or maintain because there states [57] “procedural failures are the operation with SDOs, supports the is no access to the source code. Secu- root cause of OSS technical failures”. maintenance of reference implementa- rity and reliability are enhanced when Procedures to guarantee a successful tions to establish and further develop the source code can be inspected for OSS implementation include standard- eHealth standards. In the long run, how- bugs. OSS has the potential for greater ization of best practice, technical vali- ever, it is the commercial marketplace reliability because anyone can study the dation through reference implementa- that acts as the test bed and evolution- source code, and “all bugs are shallow tions of realistic complexity, ary filter. to a million eyes” [56], but this poten- certification, and last but not least tiality does not necessarily translate into building a community to support its reality. There is a huge controversy be- evolution. References tween proprietary and OSS, which is Although, as far as interoperability in 1. Templeton B. 10 Big Myths about copyright linked to open standards and the fal- health care is concerned, technical vali- explained, 1994. Available on-line and continuously lacy that medical software or hardware dation and official certification are key updated: http://www.templetons.com/brad/ vendors may control one’s data and to achieving interoperability in prac- copymyths.html (accessed April 17, 2007). 2. Berkeley Software Distribution (BSD) License, lock-out competitors at will. tice, our experience with PICNIC and http://www.opensource.org/licenses/bsd-license.php At the same time, OSS has received a OpenECG suggests that there is still a (accessed April 17, 2007). 3. Free Software Foundation. GNU General Public lot of criticism over time, as not all OSS long way to go before it becomes a License. In: DiBona C, Ockman S, Stone M, exhibits the same level of quality and reality. An optimistic message can be editors. Open sources: voices from the open source support. A close look at medical soft- found in the interoperability strategy revolution. Sebastapol, CA: O’Reilly and Associates, 1999, www.gnu.org (accessed April ware available at SourceForge, the of ETSI [58], where technical valida- 17, 2007) and http://www.gnu.org/copyleft/ single largest repository of OSS on the tion, conformance testing and certifica- gpl.html (accessed April 17, 2007). 4. Open Standards Definition, International Internet, clearly illustrates this point. tion are recognised as important parts of Telecommunications Union, http://www.itu.int/ There are a few highly popular medi- the standards developing process [59]. ITU-T/othergroups/ipr-adhoc/openstandards.html cal software packages with high level (accessed April 17, 2007). 5. Chronaki CE, Chiarugi F. Interoperability as a of support and astounding number of Quality Label for Portable & Wearable Health downloads. However, the largest num- Monitoring Systems. In: Nugent CD, McCullagh 5 Conclusions PJ,McAdams ET, Lymberis A, editors. ber of OSS projects is incomplete, full Personalized Health Management Systems: the of errors, and with limited support. Health information Systems interoper- Integration of Innovative Sensor, Textile Informa- Serious technical shortcomings, com- ability requires the consistent imple- tion and Communication Technologies. IOS Press, 117/2005. plexity and lack of core features, like mentation of open standards. For open 6. Institute of Electrical and Electronics Engineers. security, eventually lead to scorn and standards to be consistently imple- IEEE Standard Computer Dictionary: A

Compilation of IEEE Standard Computer _access_service.htm (accessed April 17, 2007). 40. DCMTK - DICOM Toolkit, http://dicom.offis.de/ Glossaries. New York, NY; 1990. 23. OMG Person Identification Service (PIDS), ver- dcmtk.php.en (accessed April 17, 2007). 7. International Standardization Organization for sion 1.1, http://www.omg.org/technology/docu- 41. MESA tools, http://ihedoc.wustl.edu/mesasoftware/ Standardization Technical Committee 215 (ISO TC ments/formal/person_identification_service.htm index.htm (accessed April 17, 2007). 215), http://www.iso.org/iso/en/stdsdevelopment/ (accessed April 17, 2007). 42. OpenOffice.org suite, http://www.openoffice.org/ tc/TC.html (accessed April 17, 2007). 24. Schmidt D. TAO Developer’s Guide Verison 1.1a, (accessed April 17, 2007). 8. CEN Technical Committee TC251, http:// Object Computing Inc.; 1.1a edition ;2000. 43. Google Docs, http://docs.google.com/ (accessed www.centc251.org/ (accessed April 17, 2007). 25. Badilini F. Freeware ECG Viewer for the XML April 17, 2007). 9. American National Standards Institute (ANSI), http:/ FDA Format, Proceedings of the 2nd OpenECG 44. Kalra D. Electronic health record standards. In: Haux, /www.ansi.org (accessed April 17, 2007). Workshop, Berlin 1-4 April 2004. R, Kulikowski C, editors. IMIA Yearbook of Medi- 10. Digital Imaging and Communications in Medicine 26. Digital Imaging and Communication in Medicine cal Informatics 2006. Stuttgart: Schattauer; 2006. (DICOM), http://medical.nema.org/ (accessed April (DICOM) Supplement 30 waveform interchange, p. 136-44. ISBN 139783794525256 17, 2007). DICOM Standards Committee, Working Group 1 45. Open Source Health Care Alliance, http:// 11.Systematized Nomenclature of Medicine –Cardiac and Vascular Information, 1 Nov 1999, www.oshca.org/ (accessed April 17, 2007). (SNOMED), http://www.snomed.org (accessed http://medical.nema.org/Dicom/supps/ 46. dcm4che: Open Source Clinical Image and Object April 17, 2007). sup30_lb.pdf (accessed April 17, 2007). Management http://www.dcm4che.org/ (accessed 12. Object Management Group – Healthcare Domain 27. Hirai M, Kawamoto K. MFER - a Japanese Ap- April 17, 2007). Task Force, http://healthcare.omg.org/Roadmap/ proach for Medical Waveform Encoding Rules for 47. DICOM Validation Toolkit, http://dvtk.org/ (ac- corbamed_roadmap.htm (accessed April 17, 2007). Viewer Design, Proceedings of the 2nd OpenECG cessed April 17, 2007). 13. Health Level Seven (HL7), http://www.hl7.org (ac- Workshop, Berlin 1-4 April 2004. 48. OSIRIS DICOM viewer, http://www.sim.hcuge.ch/ cessed April 17, 2007). 28. Jettkant B, Dultgen P. Micro System Data Format osiris/01_Osiris_Presentation_EN.htm (accessed 14. Integrating the Healthcare Enterprise Initiative (IHE), (MSD) for Interoperability between Wearable De- April 17, 2007). http://www.ihe.net/ (accessed April 17, 2007). vices, In Proceedings of the 2nd OpenECG Work- 49. GNUmed, http://www.gnumed.org/ (accessed April 15. Healthcare Information and Management Systems shop, Berlin 1-4 April 2004. 17, 2007). Society (HIMSS), http://www.himss.org/ASP/ 29.European Data Format (EDF): http:// 50. FreeMed, http://www.freemed.org/ (accessed April index.asp (accessed April 17, 2007). www.edfplus.info/ (last accessed April 17, 2007). 17, 2007). 16. Karla D and Forslund D. Open Source Health 30. PhysioToolkit, open source software for biomedi- 51. MirrorMed, http://www.mirrormed.org/ (accessed Systems. In: Demetriades JE, Kolodner RM, cal science and engineering, http:// April 17, 2007). Christopherson GA, editors. Person-Centered physionet.ph.biu.ac.il/physiotools/index.shtml (ac- 52. OpenEMed, http://openmed.sourceforge.net/ (ac- Health Records - Toward HealthePeople, Series: cessed April 17, 2007). cessed April 17, 2007). Health Informatics; 2005. XXIV, ISBN-10: 0- 31. PhysioNet, the research resource for complex physi- 53. OpenEMR, http://www.openemr.net/ (accessed April 387-23282-6. ological signals: http://www.physionet.org/ (ac- 17, 2007). 17. Katehakis DG. The PICNIC Story. In: Regional cessed April 17, 2007). 54. Care2X, http://www.care2x.org/ (accessed April 17, Health Economies and ICT Services: The PIC- 32. Cerversato E, D’ Odorico G. Standard Communi- 2007). NIC Experience. IOS Press, Volume 115 Studies cations Protocol for Computer-Assisted Electrocar- 55. Schmidt D. Rationale for the ACE+TAO Commer- in Health Technology and Informatics. p. 4-36, diography Advanced Viewer, Proceedings of the 2nd cial support Model, http://www.cs.wustl.edu/ hardcover, ISBN: 1-58603-538-X, August 2005. OpenECG Workshop, Berlin 1-3 April 2004. ~schmidt/TAO-support.html (accessed April 17, On-line version: http://www.ics.forth.gr/ 33. Clunie D. Extension of an Open Source DICOM 2007). ~katehaki/publications/picnic_story2005.pdf (ac- Toolkit to Support SCP-ECG Waveforms, Proceed- 56. Raymond ES. The cathedral and the bazaar. First cessed April 17, 2007). ings of the 2nd OpenECG Workshop, Berlin 1-3 Monday, March 1998, vol 3, no 3. 18. CEN TC251. Health informatics - Standard com- April 2004. 57. Henning M. The Rise and Fall of CORBA, ACM munication protocol - Computer-assisted electro- 34. Schloegl A. SCP-Converter for Octave and Matlab, Queue 4:5, June 2006. http://acmqueue.com/ cardiography. ICS: 35.240.80 IT applications in Proceedings of the 2nd OpenECG Workshop, Berlin modules.php?name=Content&pa=showpage&pid=396 health care technology, Reference number EN 1-3 April 2004. (accessed April 17, 2007). 1064:2005, www.cenorm.org (accessed April 17, 35. Chronaki CE, Chiarugi F, Sfakianakis S, Zywietz 58. European Telecommunications Standards Institute 2007). C. A Web Service for Conformance Testing of ECG (ETSI): http://www.etsi.org (accessed April 17, 19. The OpenECG portal, http://www.openecg.net (ac- Records to the SCP-ECG Standard. Comput Cardiol 2007). cessed April 17, 2007). 2005;32:961-4. 59. van der Veer H. Achieving Technical Interoperability 20. Katehakis DG, Bruun-Rasmussen M, Pakarinen V, 36. Conforti F, Micalizzi M, Conte R, Ciani G, Ghione - the ETSI Approach, ETSI White paper, Oct 2006. Piggott D, Saranummi N. PICNIC Technology. S, Marabotti C, et al. Open Source Software to http://www.etsi.org/etsi_radar/whitepaper/ Included in “Regional Health Economies and ICT make the Telematic ECG Exchange popular among wp_3.htm (accessed April 17, 2007). Services: The PICNIC Experience”, IOS Press, physicians. Proceedings of MEDICON 2004, Is- Volume 115 Studies in Health Technology and chia, Naples, Italy, August 2004. Informatics, pp. 61-91, hardcover, ISBN: 1-58603- 37. Lightwood S, Greco F, Lizzio V, Brignole M, Gigli 538-X, August 2005. G. NET4ECG: An ECG Digital Network. Proceed- Correspondence to: 21. Katehakis D, Tsiknakis M. Electronic Health ings of Computers in Cardiology 2005;32:439- Catherine E. Chronaki Record. In: Akay M, editor. Wiley Encyclopedia of 442. Lyon, France, September 2005. Institute of Computer Science, FORTH Biomedical Engineering. Hoboken: John Wiley & 38. Gates B. Building Software That Is Interoperable PO 1385, Vassilika Vouton Sons, Inc. dx.doi.org/10.1002/9780471740360. By Design, http://www.microsoft.com/mscorp/ Heraklion, Crete ebs1440; 2006. execmail/2005/02-03interoperability.mspx (accessed 71110 Greece 22. OMG Clinical Observations Access Service April 17, 2007). Tel: +302810391691 (COAS), version 1.0, http://www.omg.org/tech- 39. CTS, http://erl.wustl.edu/research/dicom/ctn.html Fax: +302810391428 nology/documents/formal/clinical_observation (accessed April 17, 2007). E-mail: [email protected]

IMIA Yearbook of Medical Informatics 2007