To verify and elucidate the results of a previous study that documented four-fold variations in small area admission rates for diabetes mellitus in Washington state, the authors examined the hospital records of a sample of diabetes discharges from hospitals in high-, medium-, and low-rate counties. Hospitals in high-rate counties were generally smaller than those in medium- and low-rate counties and admitted a greater proportion of patients with mild illness. Also, physicians in hospitals in high-rate counties tended to be less vigorous in the diagnosis and treatment of diabetes. Variations in admission rates were not explained by coding errors or by differences in selected sociodemographic characteristics of admitted patients. The authors conclude that small area variations in hospitalization rates for diabetes were related to clinical differences in admission criteria and may also reflect differences in the medical management of hospitalized patients.