Estimating diabetes prevalence in South Auckland: how accurate is a method that combines lists of linked health datasets?

AIMS: To assess the accuracy of a method for estimating adult diabetes prevalence that combines linked, routine health datasets in South Auckland, New Zealand. METHODS: We used a simple algorithm that combined records of laboratory testing, drug dispensing and hospital diagnoses applied to National Health Index-linked health data in South Auckland to estimate the prevalence of diabetes in 2007. We investigated the sensitivity of this 'combined list' algorithm against a gold standard of individuals with diagnosed diabetes enrolled in a Chronic Care Management programme (CCMP). We also assessed the sensitivity of this algorithm against an estimated diabetes population generated using capture-recapture methods. RESULTS: From the combined-list algorithm, 25,797 (7.2%) South Aucklanders aged 15 years and over had diabetes. During this period, 10,725 patients were enrolled in the CCMP. The combined list algorithm correctly identified (sensitivity) 10,351/10,725 (96.5%) of those enrolled. When we used the capture-recapture estimated diabetes population as an alternative gold standard, 34,418 [9.5%] of South Aucklanders 15 years and over had diabetes, with the sensitivity of the combined list method falling to about 75% (25,797/34,418). CONCLUSION: Linked health data provide reasonably accurate estimates of diabetes prevalence in a New Zealand population; particularly for cases with longstanding or complicated disease.