Data from the National Hospital Discharge Survey was used to estimate the admission rate and in-hospital mortality for US stroke patients from 1970 to 1987. Over this period the hospital admission rate increased from 5.0 per 1,000 population to 5.5 per 1,000, while the in-hospital mortality fell from 18.6 to 10.0% of discharges. When patients experiencing general or late effects or transient ischemic attacks were excluded, the hospital admission rate increased from 2.8 to 4.8 per 1,000 and in-hospital mortality fell from 26 to 12% of discharges. Admission rates increased from 1970 to 1987 for both hemorrhagic strokes and cerebral infarctions, but fell for ill-defined cerebrovascular events, presumably reflecting increased use of computed tomography and magnetic resonance imaging during this period. Mortality for hemorrhagic strokes decreased from 45 to 30% of discharges, mortality for infarctions decreased from 24 to 13% and mortality for ill-defined events fell from 24 to 17%. There are several limitations with the use of the National Hospital Discharge Survey data to estimate changes in hospital admission rates and in-hospital mortality. However, these data suggest a greater role for decreasing case fatality in explaining nationwide declines in stroke mortality rates than has generally been acknowledged.