INTRODUCTION: We developed a decision support tool that can guide the development of heart disease prevention programs to focus on the interventions that have the most potential to benefit populations. To use it, however, users need to know the prevalence of heart disease in the population that they wish to help. We sought to determine the accuracy with which the prevalence of heart disease can be estimated from health care claims data. METHODS: We compared estimates of disease prevalence based on insurance claims to estimates derived from manual health records in a stratified random sample of 480 patients aged 30 years or older who were enrolled at any time from August 1, 2007, through July 31, 2008 (N = 474,089) in HealthPartners insurance and had a HealthPartners Medical Group electronic record. We compared randomly selected development and validation samples to a subsample that was also enrolled on August 1, 2005 (n = 272,348). We also compared the records of patients who had a gap in enrollment of more than 31 days with those who did not, and compared patients who had no visits, only 1 visit, or 2 or more visits more than 31 days apart for heart disease. RESULTS: Agreement between claims data and manual review was best in both the development and the validation samples (Cohen's kappa, 0.92, 95% confidence interval [CI], 0.87-0.97; and Cohen's kappa, 0.94, 95% CI, 0.89-0.98, respectively) when patients with only 1 visit were considered to have heart disease. CONCLUSION: In this population, prevalence of heart disease can be estimated from claims data with acceptable accuracy.