Health care professionals' ability to make accurate prognostic judgments for long-term care patients was tested in a study employing quarterly assessments and prognoses for more than 700 patients. Prognoses were made for patients to 'improve', 'decline', or 'remain the same', in physical functioning ability for the next 90 days, after which reassessments took place. Despite substantial information on their patients, these teams of physicians, nurses, physical therapists, and social workers were not very accurate in making prognoses. They were right only about half the time. In general, they tended to be too optimistic. The study has profound implications for policy makers considering reimbursement on the basis of performance related to patient outcome goals. The goals are not likely to have much validity. Health care professionals may also be disturbed by the poor rate of correct guesses about patients' future courses of recovery.