Evaluating openEHR for defining machine readable electronic health record phenotypes: lessons from the CALIBER resource

Linked Electronic Health Records (EHR) are increasingly being used for research. The process of defining and validating EHR phenotypes poses significant challenges as data have been collected for care, auditing or administrative purposes and not research. Challenges are further amplified by the lack of common standards for describing EHR phenotypes, making their reproducibility problematic. While phenotype components are often controlled clinical terminology terms, definitions and algorithmic logic are expressed in unstructured textual and/or graphical form which is not machine-readable. Open-source clinical data specifications, such as openEHR, can potentially address some of these challenges and used to define computable EHR-derived phenotypes. The semantic interoperability between openEHR archetypes is based on a binding mechanism between internal archetype codes and controlled clinical terminologies. Using type- 2 diabetes as a case-study, the aim of our research was to evaluate openEHR for defining EHR-derived phenotypes in a large-scale linked EHR resource in the UK.