To formulate recommendations for the development of intensive care unit (ICU) admission policies.Literature review of published reports over the period 1966 to 1991 pertaining to admission criteria for intensive care or coronary care units (CCUs).Studies identifying patients least likely to benefit from ICU or CCU admission were analyzed. Patient populations of interest included adults (> or = 18 years of age) with medical conditions possibly requiring intensive care; trauma patients were excluded.Of 970 articles identified as being pertinent to intensive care, only two case-control studies used the direct method of measuring the effect of ICU intervention on mortality. No studies were found that compared outcomes of low-risk patients treated in a CCU vs those treated in alternative hospital locations, and none identified patients with a very high probability of a bad outcome.The use of decision-making models for ICU and CCU admissions must be tested in prospective, randomized clinical trials. Critical care units and ICUs should be studied separately. Existing studies of early discharge from CCUs need to be summarized and evaluated. The triaging of ICU patients to alternative hospital locations needs to be evaluated, as do existing predictive models for early triage decision-making.