Objective To evaluate the performance of the National Health Service (NHS) Health Check in identifying people at high risk of having or developing type 2 diabetes.

Design Retrospective evaluation of the performance of the NHS Health Check diabetes filter (based on ethnicity, body mass index and blood pressure) in identifying people at risk for type 2 diabetes (glycated haemoglobin (HbA1c) ≥42 mmol/mol recorded within 3 months of their NHS Health Check).

Setting Heart of Birmingham Primary Care Trust (HoB PCT).

Subjects 34 022 patients with a Read code in the general practitioners’ (GP) clinical record indicating that they had attended an NHS Health Check over the period April 2009 – February 2012.

Outcome measures Primary outcome measure: proportion (%) of patients at risk of diabetes or non-diabetes hyperglycaemia not identified by a simple application of the NHS Health Check diabetes filter. Secondary outcome measures included sensitivity, positive predictive value (PPV) and specificity of the NHS Health Check diabetes filter.

Results In HoB PCT, the simple application of the NHS Health Check diabetes filter led to a failure to identify 1990/5968 (33.3% (95% CI, 31.2% to 35.4%)) of patients of known ethnicity at risk of having or developing diabetes (HbA1c≥42 mmol/mol). The NHS Health Check diabetes filter has a sensitivity of 66.8% (95% CI 65.7% to 68.0%), and the PPV was 41.1% (95% CI 40.1% to 42.1%). Specificity was 34.7% (95%CI 33.9% to 35.6%). Sensitivity and PPV of the NHS Health Check diabetes filter in the HoB PCT population are significantly greater for patients of Asian ethnic origin than for those of other ethnic backgrounds.

Conclusions This evaluation, which was based on a large population sample, demonstrates that the NHS Health Check programme diabetes filter failed to identify a third of people at high risk of having or developing diabetes.