This document summarizes a study that assessed the validity of diabetes prevalence data derived from health records by comparing it to a primary care database called PREDICT. The study found: 1) The diabetes prevalence rates from the algorithm and PREDICT were similar at 20.1% and 20.9% respectively. 2) There was good agreement between the two data sources, with the algorithm showing 86% sensitivity and 96% specificity compared to PREDICT. 3) Using capture-recapture analysis, the algorithm undercounted diabetes prevalence by around 15% compared to 10% for PREDICT, suggesting the algorithm provides an accurate and cost-effective method for determining diabetes prevalence.