Medicare's prospective payment system for hospitals (PPS), introduced in the USA in 1983, replaced cost reimbursement with a system of fixed rates which created incentives for hospitals to control costs. Previous studies found that elderly patients were discharged from hospital ‘quicker and sicker’ under PPS and concluded that families were coping at home. We analyse a national longitudinal survey, the first National Health and Nutrition Examination Survey and its Epidemiologic Followup Study, which includes data on more outcomes over a longer period than earlier studies. We find that the rate of admission to nursing homes from the community in the first weeks after a hospital discharge more than tripled under PPS, suggesting that families were not always able to cope. As another response to sicker patients, discharges directly to nursing homes from hospitals, which jumped initially under PPS, may have risen further when payment rates were tightened in the early 1990s. Hospital readmissions fell after the first few years. Our findings are strengthened by the fact that we control for patients’ health using health information collected independently of hospital admission.