Failure to use linked health records may lead to biased disease estimates

Failure to use linked electronic health records may lead to biased estimates of disease incidence and outcome, researchers warn in a paper published online today in the British Medical Journal.

The study looked specifically at heart attack reporting across four national health record sources in England, finding that up to 50 per cent of all heart attack cases are missed when just one data source is used to estimate the annual incidence.

These findings may be relevant to other common conditions, such as stroke, and support the wider use of linked multiple record sources by clinicians, policy makers and researchers.

Electronic health records are increasingly used to measure outcomes of healthcare and health policy, and for research in observational studies and randomised clinical trials. However, records from one part of the health service, for example primary care, may not capture health events occurring in other parts, such as hospital care.

A team of researchers from the UK and the Netherlands compared electronic health records for one major disease event, heart attack, across four different national health record sources: primary care records, hospital care records, disease registry and death records. Previous studies have typically compared only one or two electronic sources.

The study identified 21,482 patients with a record of heart attack, or acute myocardial infarction, in one or more of the four data sources. However, each data source individually missed a substantial proportion of heart attack cases. For example, only one-third of non-fatal myocardial infarctions were recorded in all three applicable data sources (primary care, hospital care and disease registry), with another third recorded in two of the three sources.

Primary care records were the single most complete source of non-fatal myocardial infarction records, missing only a quarter of cases; hospital records missed one-third and the disease registry nearly half. In other words, acute myocardial infarction was underestimated by 25-50 per cent using one source compared to using all three.

Dr Anoop Shah, a Wellcome Trust Clinical Research Fellow at UCL and co-author of the study, said: "With the current emphasis on measuring clinical outcomes in health systems and recent plans to use linked data to drive improvements in the care of patients with cardiovascular disease, our study has important implications for practice and policy."

The researchers call for future research to focus on areas such as improving how data are coded, understanding how linkages with primary care, admission to hospital and mortality data compare, and evaluating the quality of the data available from these linkages.

The study made use of the Clinical Practice Research Datalink, a research service launched in 2012 to maximise the way anonymised NHS clinical data can be linked. The service is designed to enable many types of observational research and deliver research outputs that are beneficial to improving and safeguarding public health.

Nicola Perrin, Head of Policy at the Wellcome Trust, said: "This study demonstrates the need for the highest-quality data to inform research and healthcare. It is striking that, because of the way health events are currently recorded across the healthcare system, there can be such variation between information from different data sources. The results highlight the importance of the Clinical Practice Research Datalink, which, by linking different datasets, enables researchers to access the best possible data."