Abstract. Objective: To evaluate patient outcome and the efficiency of stays in intensive care units (ICUs). Design: Prospective study. Setting: Seven ICUs of teaching hospitals in the Paris area. Patients: Two hundred eleven stays including one in three consecutive patients admitted from September to November 1996. Measurements and main results: For each patient, the following information was collected during the ICU stay: diagnosis, severity scores, organ failures, workload, cost and mortality. A cost-effectiveness ratio was computed for 176stays with at least one organ failure, at hospital discharge and 6months later. Quality of life was measured with EuroQol questionnaires 6months after discharge in 64 patients representing 62% of the patients contacted. The mean total ICU cost per stay was US$14,130 (6,550) (higher for non-survivors US$19,060, median 10,590 than for survivors US$12,370, median 5,780). The incremental cost-effectiveness ratio was US$1,150 per life-year saved and the incremental cost-utility ratio was US$4,100 per quality-adjusted life-year (QALY) saved, without discounting. These results compare favourably with other health-care options. However substantial variations were observed according to age, severity, diagnosis, number of organ failures and discount rate. Intoxication had the lowest ratio (US$620/QALY) and acute renal insufficiency the highest (US$30,625/QALY). Conclusions: This work provides medical and economic information on ICU stays in teaching hospitals and enables comparisons with other health-care options.