Between October 1992 and July 1993, 11 Northern Italian geriatric departments systematically screened 1386 inpatients hospitalized for at least 10 days. Of those screened, 74% (N = 1019) were excluded by one or more of 7 exclusion criteria; 32% (N = 118) of the 367 remaining subjects failed to meet more than 1 of 8 inclusion criteria. The 11 Geriatric Evaluation Units (GEUs) examined the remaining eligible 249 inpatients with a uniform comprehensive geriatric assessment (CGA) plan, which included a number of validated assessment scales. Of those evaluated, 39% (N = 97) were ineligible for the study because of being unwilling, noncompliant, too sick or "not truly frail", and the remaining 152 (11% of all patients screened) were randomly enrolled in two groups; 79 were assigned to the GEU (experimental group), and 73 to standard care in the National Health Care System (control group). At entry there was no statistically significant difference between the two groups. At 6 months, preliminary results are encouraging; GEU patients had a lower mortality than controls (2 vs 8, 2.5% vs 10.9%; p < 0.05). Slight differences were seen in both clinical-cognitive-affective-functional status and the use of health and social care resources. Only 3 subjects dropped out, 8 refused further follow-up, and 3 went to a nursing-home. We conclude that a standardized selection plan can recognize frail elderly inpatients and that GEU care seems to achieve good results.