Summary year 2 report
During the first two years, SemanticHealthNet has analysed the information recommended by best practice clinical guidelines to support shared clinical care in the management of heart failure in order to propose a candidate heart failure patient summary, and studied examples of public health interventions relevant to the prevention of cardiovascular disease. These clinically-driven specifications have been formally modelled using multiple interoperability standards and specifications, and formalisms have been developed to enable semantic brokering between these different specifications. The project has begun to develop best practice guidance on how clinical models and term value sets should be developed, and assessed for their quality. Importantly, recognising the value of a Network of Excellence but also the time and scale limitations of the project, a proposal for sustaining this effort has been developed through multi-stakeholder engagement as a not for profit Institute. These areas of progress are described in more detail below, together with a reflection on the insights gained.
Our early work has shown how difficult it is for the conceptions of clinical information and interoperability amongst clinicians to be mapped into the formalisms and standards that have been built up by the health informatics community. In particular, clinicians tend to underestimate the difficulties that computers may have in robustly interpreting what seems to them to be common sense understandings, and the informatics community has difficulty in being proportionate in the level of specificity and detailed formality that clinicians and patients could realistically provide when capturing data, and what extent of this is genuinely needed in order to deliver some near term benefits. Industry seems to be caught in the middle, partly by seeking to make good use of the informatics standards on offer but partly struggling to convince purchasers of why the interoperability they ideally wish for remains elusive despite substantial ICT spends.
Despite these challenges, the consortium has developed an initial specification for a shared care heart failure summary, derived from European guidelines, which has largely been formally represented using clinical models and terminology value lists. A formalism known as “clinical patterns” has been developed which uses description logic to represent the knowledge within clinical models, and therefore allows for mappings to be inferred between heterogeneous models that represent the same concepts. This kind of semantic brokering will be necessary in the foreseeable future, when multiple standards and specifications will be used by different vendors and sub-systems. In order to enlarge the network internationally, SemanticHealthNet held a joint three-day workshop with the Clinical Information Modeling Initiative (CIMI) in March 2014, which was attended by experts from the US, Korea, Brazil as well as many from Europe. The project has also contributed to workshops and conference sessions in support of the EU-US Memorandum of Understanding on the digital agenda for health. Interactions have also taken place with CEN, and will continue during year 3 by examining in more depth the standards that collectively support continuity of care, and how semantic interoperability may be more formally enhanced between them.
A series of population health vignettes for cardiovascular disease prevention have been developed to depict how information in electronic health records, if consistently captured and combined across systems, can be used to support strategic health planning decisions at a population level. Simple examples such as smoking history have helped to reveal how the clinical care and public health needs for information – granularity and precision - differ.
The project has also started to look at the needs and wishes of patients, in particular to ask patients what information they personally might wish to see incorporated within a heart failure summary to inform clinicians about their own concerns about their condition and functioning. This work will be taken forward next year, including an examination of how well InterRAI and the International Classification of Function can formally represent such patient-focused information.
The development of these assets has been iterative and exploratory, and it is now possible to bring together the experiences of developing the heart failure summary and population health vignettes, in order to develop good practice guidance for future clinical and population health communities undertaking similar tasks. In parallel the project has undertaken international surveys of clinical modelling practices in order to understand the key success strategies that have been learned though experience. These have been collated and, when combined with in-project experiences, will enable the development of consensus good practice that can be disseminated.
Not only must semantic interoperability assets be developed though sound, well evidenced and multi-stakeholder engaged processes, but the eventual assets themselves need to be quality assured. Work has commenced this year to collate the existing approaches to how the quality of informatics assets can be defined, and the next steps will be to propose more specific quality labelling criteria for interoperability assets. Taking advantage also of work in the UK by the Professional Record Standards Body to define a clinically-driven information specification for shared care between hospitals and GPs, and the European Commission sponsored epSOS Patient Summary specification, the project will develop a set of conformance criteria for shared care summaries, including implementation independent criteria and specific conformance criteria for an implementation that has used HL7 CDA, openEHR, EN ISO 13606 or IHE XDS.
Now over half of the way through its funded period, the project has become clearer about the complexity of its initial challenge, including the culture gaps between the stakeholders that need to influence the design of interoperability assets, in their understandings of why health information needs to be captured systematically and consistently, why and how it needs to be communicated and used for different purposes, and therefore why – and to what extent – interoperability and the computable interpretation of health information adds value over and above the effort required to capture and manage it. It has therefore become more apparent that one of the most important objectives of the project must be to devise and put into place a sustainability approach that can grow the scale, support and resources for these stakeholder communities – who inevitably must work together to align their interests and capabilities - to design, implement, adopt and productively use interoperable capability. During year 2 the project has formed a Sustainable Semantic Interoperability Task Force (SSI-TF) to examine the challenges and opportunities for sustaining the network and activities that SemanticHealthNet has started. This group, comprising a rich range of stakeholder representatives, has used formal business modelling methods to propose a future two-tier organisational structure: a SemanticHEALTH Institute (SHI) that will be established as a not-for-profit organisation to promote and co-ordinate future multi-stakeholder co-operation in semantic interoperability, and act as a catalyst and champion for scaling up initiatives amongst a wider network of stakeholders: the SemanticHEALTH Alliance (SHA). The SHA will in effect be co-ordinated and supported operationally by the SHI. The SHA will be a continuation and progressive enrichment of the Network of Excellence being established by SemanticHealthNet. The eHealth Governance Initiative has been closely involved in this proposal, through key representatives.
Because of the challenges we have faced in bringing heterogeneous groups together to tackle these complex goals, and despite substantial progress in many areas of the work plan, the project has experienced some delays in certain key milestones. A no-cost extension of six months has been agreed with the EC, with the result that the project will now finish in May 2016.