We studied 432 admissions to two Danish ICUs by using a standard severity of illness classification system to assess utilization and outcome. Substantial differences in utilization were found. The patients in Hospital 2 were younger, had better previous health records, and were admitted significantly more often for active treatment as opposed to monitoring than the patients in Hospital 1. Although their measured severity of illness was similar, patients at Hospital 2 received significantly more therapy and their mortality exceeded that of the patients at Hospital 1. The mortality rate of Hospital 2 also exceeded that predicted from a recent survey of U.S. hospitals. We found, however, that 35% of the patients at Hospital 2 had been transferred to the ICU from other ICUs. This created the possibility of an adverse selection and lead-time bias for the patients at Hospital 2. These findings indicate that although national and international comparisons of intensive care are now possible using common classification systems, this progress has created a new need for more precise measurement of potential confounding biases, such as the duration of intensive care services received before formal ICU admission.