Diabetes is closely linked to cardiovascular disease, but researchers using electronic health records have not always defined diabetes in the same way. Phenotype definitions are the instructions researchers use to translate information from health records into meaningful characteristics, like if a person has a diagnosis of a health condition. Different datasets, coding systems and study designs can lead to different definitions, making it harder to compare findings or reuse research with confidence.
The BHF Data Science Centre team helped bring the cardiometabolic research community together to develop a standardised, rule-based diabetes phenotyping algorithm. Built with clinical and health data expertise, our algorithm uses routinely collected diagnosis, prescribing and laboratory data to identify diabetes, define diabetes type and determine date of diagnosis.
An openly available, reusable diabetes definition supports more transparent and comparable research. To accompany the algorithm, we developed tools and documentation, and ran validation and benchmarking so teams can understand how definitions differ and apply our approach in future studies.
Consistent phenotyping makes linked health data more useful. For researchers, it reduces duplication and strengthens reproducibility. For people affected by diabetes and cardiovascular disease, it helps generate more reliable evidence about risks, treatments and outcomes.
Explore our diabetes phenotype and find out how BHF Data Science Centre services can help researchers develop, validate and share reusable phenotyping algorithms.