Identifying patients who will need long-term care may improve the efficiency and effectiveness of acute hospital care. This prospective study evaluated clinicians' ability to identify patients requiring nursing home care. The study had two principal objectives. The first objective was to measure whether registered nurses, physicians, and social workers made similar estimates of the probability of nursing home placement early in an acute care hospitalization. The second objective was to identify the clinical characteristics of patients for whom the clinicians incorrectly predicted that they would return home. The study subjects were 342 patients older than age 55 who were admitted to the medicine, surgery, and neurology services of two university-affiliated Veterans Affairs hospitals. Fifteen percent were discharged to nursing homes. The nurses, physicians, and social workers had high agreement in their estimates of the probability of nursing home placement for each patient. However, each of the provider groups assigned low probability estimates to more than 20% of the patients discharged to nursing homes. Examination of the characteristics of patients assigned low probability estimates revealed that mental impairment and functional disability were higher in those patients who ultimately were discharged to nursing homes than in those patients who returned to their homes. These findings suggest that better assessment and interpretation of patient characteristics early in the hospital stay may improve discharge planning. Some clinicians appear to underestimate mental and functional impairment as risk factors for long-term care needs.