Assessment of the completeness and accuracy of case ascertainment in the Michigan Stroke Registry.

BACKGROUND Accurate case ascertainment is essential for clinical registries to be valid and representative. We assessed case ascertainment in the Michigan Stroke Registry by linking to a statewide hospital discharge database (Michigan Inpatient Database [MIDB]). METHODS AND RESULTS In 2009, all ischemic stroke cases submitted by 30 registry hospitals were linked to ischemic stroke discharges (International Classification of Diseases, Ninth Revision code 433.x1, 434.x1, or 436) in the MIDB. Databases were linked using hospital, age, sex, and admission date. The MIDB was regarded as the gold standard. To assess completeness, we calculated the percent difference between the number of cases entered in the registry relative to the MIDB. To quantify accuracy, we defined sensitivity as the proportion of cases identified in the MIDB that were matched to the registry and positive predictive value as the proportion of cases identified in the registry that were matched to the MIDB. Before data linkage, 4 hospitals were known to be using a case sampling approach. The remaining 26 registry hospitals submitted 21% fewer cases (n=3403) than were found in the MIDB (n=4340). The overall sensitivity was 68.8% (95% confidence interval, 76.4%-79.3%), and positive predictive value was 87.7% (95% confidence interval, 87.4%-89.8%). The sensitivity of case ascertainment was significantly lower in teaching hospitals and primary stroke centers but was higher in the sites that used prospective case ascertainment methods. CONCLUSIONS Among registry hospitals, these results revealed relatively high levels of completeness and accuracy. Matching registry data to hospital discharge data identified hospitals that changed their case ascertainment method to a case sampling approach. This study illustrates the value of monitoring case ascertainment in stroke registries using external data sources.