APACHE II (an acronym formed from acute physiology score and chronic health evaluation) has been proposed to limit intensive care unit (ICU) admissions ('cost containment') and to judge outcome ('quality assurance') of surgical patients. To judge its performance, a 6-month study of 372 surgical ICU patients was performed. When patients were divided by mean duration of stay, mortality rates rose from 1% (short stay) to 19% (long stay) (p less than 0.001) for patients with APACHE II scores less than 10, but decreased from 94% (short stay) to 60% (long stay) (p less than 0.01) for patients with APACHE II scores more than 24. Exclusion of patients by high or low APACHE scores would 'save' 6% of ICU days but risk increasing morbidity, hospital costs, and deaths. Grouped APACHE II scores did not correlate with total hospital charges (r = 0.05, p = 0.89) or ICU days used (r = 0.42, p = 0.17). Grouping by APACHE II score and duration of ICU stay showed neither symmetry nor uniformity of mortality rates. Surgical patients would not be well served by APACHE II for quality assurance or cost containment.