OBJECTIVE: To determine the prognosis of delirium in elderly patients. DATA SOURCES: MEDLINE was searched for relevant articles published from January 1980 to March 1992. The bibliographies of identified articles were searched for additional references. STUDY SELECTION: Eight reports (involving 573 patients with delirium) met the following inclusion criteria: original research, published in English or French, prospective study design, diagnosis based on acute deterioration in mental state, sample of at least 20 patients, patients aged 60 years or over and follow-up of at least 1 week. The validity of the studies was independently assessed according to the criteria for prognostic studies established by McMaster University, Hamilton, Ont. No study met all the criteria. DATA EXTRACTION: Information about the patient sample, length of follow-up and results was systematically abstracted from each report and tabulated. DATA SYNTHESIS: A meta-analysis of the outcomes indicated that elderly patients with delirium had a mean length of stay of 20.7 days. One month after admission 46.5% were in institutions, and 14.2% had died; only 54.9% had improved mentally. Six months after admission 43.2% were in institutions. Compared with unmatched control subjects they had longer hospital stays, higher mortality rates at 1 month and higher rates of institutional care at 1 and 6 months. The presence of severe physical illness or dementia may have been related to some outcomes. CONCLUSIONS: Delirium in the elderly appears to have a poor prognosis. However, this finding may have been confounded by the presence of concomitant dementia or severe physical illness. Future studies must pay attention to methods and design, particularly the composition of study populations and the control of extraneous prognostic factors.