Using inpatient data for surveillance: a study of cerebrovascular accidents in American Indians.

The authors examined hospitalization rates for stroke to evaluate the potential for using Indian Health Service (IHS) hospitalization data to perform population based disease surveillance. by investigating disease rate differences in two administrative areas. The methodology included a retrospective analysis of multi-year IHS hospital discharge data, with record linkage and chart review in the Cherokee and Albuquerque IHS area hospitals. The IHS inpatient data was complete and reliable for care provided in the two areas. The authors confirmed all stroke hospitalization records. Multiple hospitalizations for the same event and multiple events per person accounted for 8.0% and 11.8% of the 2.6 fold differences in crude hospitalization rates for stroke. Complete omission of events in the computer records was rare. Errors in patient information were low, ranging from 0 to 3.4%. The IHS data provide reliable reporting of hospitalization information making these data potentially useful for disease surveillance. Caution should be exercised in interpreting small differences in disease rates that might be due to factors other than disease incidence.