Validity of Discharge Diagnoses in the Surveillance of Stroke

BACKGROUND: Hospital administrative data have been suggested as a valuable cost-effective tool for providing information about the stroke burden. Nevertheless, the choice of the diagnosis codes has been a critical issue in the development of case ascertainment algorithms. METHODS: In this study, the Minimum Basic Data Set administrative database was used to analyze the accuracy of different ICD-9-CM algorithms based on the neurologist's clinical judgement as the 'gold standard'. RESULTS: The most accurate algorithm observed in our study involved the selection of ICD-9-CM codes 430-438 in the primary diagnosis. It yielded a sensitivity of 96.1%, a specificity of 87.5% and a positive predictive value of 82.5%. CONCLUSIONS: The Minimum Basic Data Set is a valuable source to evaluate stroke frequency when using an accurate algorithm to select events.