Chris Smith and Alex Newsham
Policies that determine the data captured by clinicians and healthcare professionals in
Electronic Health Records (EHRs) are subject to change over time. Changes may arise
for a variety of reasons, including updated clinical practice, improved diagnostic tests,
and the introduction or cessation of specic public health initiatives. EHRs may cap-
ture dierent clinical concepts, or use dierent representations for clinical concepts, as
a result of these changes.
Longitudinal analysis of EHRs aims to identify patterns in health and healthcare over
time to inform the design of interventions. Analysis is predicated on the ability to
robustly identify the specic clinical concepts by which patients, interventions and out-
comes are to be characterised. Due to policy changes, the presence and representation
of these concepts may vary over the period of analysis. Moreover, the period of analysis
may be relative to a specic health-related event, and therefore dier for each patient. To
ensure that patients, interventions and outcomes are robustly characterised, dynamism
in clinical concepts needs to be integrated into the longitudinal analysis of EHRs.
In this talk, we illustrate dynamism in clinical concepts using denitions provided in
six successive versions of the Quality Outcomes Framework (QOF) – a set of policies
that determine recording in Primary Care. We show changes in the inclusion of clinical
concepts over time, and use Jaccard Similarity Coecients to show the extent of changes
in the clinical codes that are considered to represent particular clinical concepts over
time. Eects on longitudinal analysis of EHRs are discussed and potential approaches
to mitigate these eects are introduced. Finally, we motivate the need for tools and
techniques that can integrate concept dynamism into the longitudinal analysis of EHRs.
Abstract Submission to the UK Ontology Network (UKON 2016)
1
