Towards Achieving Semantic Interoperability In eHealth Services

Introduction

Data sharing in the health domain is increasingly being recognised as a critical factor for improving healthcare and facilitating clinical research (Stead W. et al (2000), Garde, S. et al (2007), Taweel, A. et al (2006), Taweel, A. et al (2004)). The need for integrated systems that can provide up-to-date information about patients and their healthcare is not only critical to the day-to-day running and delivery of health functions but also potentially means saving lives. It is increasingly expected that patient health information is available at the point of care irrespective of its location. This expectation is driven by rapid developments in information technology, alongside their increasing dominance at the heart of individual healthcare institutions, through the use of data-intensive electronic health records systems (EHRs) that drive the health care process. One of the primary visions of eHealth has been to achieve a technology supported, seamlessly integrated health service that ensures the delivery of the right information about the right individual to the right person at the right time (Stead W. et al (2000), Taweel, A. et al (2006)). This vision, however, present many challenges, including:

• •

  The heterogeneity of eHealth systems and their data representations across health institutions, not just across the eHealth domain as whole but even within individual clinical domains, e.g. secondary care, primary care etc.

• •

  Delivery of complete clinical information at the point of care, potentially requires health information exchange across geographically distributed health centres, thus creating a massive scalability challenge that puts demands on networking and distributed system architectures.

• •

  Legacy and propriety systems and the lack of, or at best, slow pace adoption of standards make exchanging clinical data more difficult and hampers the chances of integrated healthcare. This puts demands on knowledge representation methodologies that must cope with the complexity of (heterogeneous) eHealth information, whilst remaining easy to deploy and use by system providers.

• •

  The requirements of patient privacy and confidentiality that require strict data governance policies. These policies vary across institutions to include a variation of access constraints that observe strict ethical rules to maintain patient confidentiality at all times. This requires secure yet configurable independent data sharing policies that reflect the needs of the participating organizations within the domain.

However, the most serious challenges to achieving the eHealth vision are not only technological but also ethical and structural. The large number of system providers and their business models and variations in ethical and regulations for accessing clinical information across health organisations (let alone individual health centres) implies an exponential explosion in the number and type of stakeholders that need to be involved in achieving the seamless interoperability required to achieve the above vision. To achieve the need is to enable interoperation between data centres not just at the system or data levels but also at the governance level. However, consider the complexity involved in reaching agreements on or enforcing conformance to interoperability standards at these different levels within a single enterprise let alone multiple countries. Therefore, the approach taken here is motivated by the practical implications of not being able to completely rely on the use of common interoperability standards to address this issue. Instead, this needs to be addressed at the granularity of the application, data and/or centre levels, in which an interoperability framework should overcome heterogeneity and adapt at run-time and dynamically interoperate with data providers. Heterogeneity between eHealth systems and data, however, takes several dimensions; the most challenging of which is the semantic one, which the approach proposed in this paper attempt to address.