Emphasis on controlling health-care costs has led to many activities aimed at avoiding medically unnecessary hospitalizations. Much less attention has been given to patients hospitalized in advanced stages of illness and the impact of these late admissions on cost and quality of care. A panel of physicians developed criteria to categorize hospital admissions into one of three groups--early, timely, or late--based on the timing of the initial hospitalizations of patients admitted with any one of 14 diagnoses. Over a period of one year (fiscal year 1984) the criteria were applied retrospectively to 2,713 patients admitted to either of two hospitals. Twenty-one percent of the admissions studied in one hospital and 19% in the other were judged to occur later than was desirable. The mean length of stay for late hospitalizations exceeded that for timely hospitalizations by 11.1 days at one hospital and by 7.5 days at the other (p less than .01). Similar patterns were observed in analyzing the 14 diseases individually and in an analysis of hospital charges at the one hospital where charge data were available. In-hospital mortality rates for patients with a principal diagnosis of bacterial pneumonia were over ten times greater for those admitted late than for those whose admissions were timely (39.0% versus 3.8%, p less than .001, at one hospital; 28.9% versus 2.1%, p less than .001, at the other). While not all late hospitalizations are avoidable, the authors believe that the analysis of late hospitalization patterns is an important part of any effort that can be made to reduce them.