Utility of illness severity scoring for prediction of prolonged surgical critical care.

OBJECTIVE: To determine whether APACHE III and multiple organ dysfunction syndrome scores can predict a prolonged length of stay for critically ill surgical patients in the intensive care unit.
DESIGN: Prospective, inception-cohort study.
SETTING: Surgical intensive care unit (SICU) of an urban, tertiary care hospital. PATIENTS: 2,295 consecutive admissions for critical surgical illness, postoperative complications, or postoperative monitoring in 2,058 patients.
INTERVENTIONS: Calculation of Acute Physiology and Chronic Health Evaluation (APACHE) II and APACHE III scores 24 hours after admission to the SICU. Serial quantitation of organ dysfunction for the duration of hospitalization according to the multiple organ dysfunction score. Patients were stratified by survival and time intervals for the duration of critical care, and followed until discharge or death. MAIN OUTCOME MEASURES: Hospital mortality and length of stay in the SICU.
RESULTS: The mean APACHE II and APACHE III scores were 14.0 +/- 0.2 and 45.2 +/- 0.6 points, respectively (mean +/- SEM). The incidence of organ dysfunction was 43%, and the hospital mortality was 9.7%. The mean ICU length of stay was 6.1 +/- 0.2 days, but decreased progressively from 6.8 +/- 0.5 days in 1991 to 5.3 +/- 0.6 days in 1995 (p < 0.01) with no change in either illness severity or the number of admissions. By univariate analysis, increased length of stay in the ICU was associated with increasing APACHE scores, an increased incidence of emergency admissions, and the incidence and magnitude of organ dysfunction (all p < 0.01). Severity indices appeared to plateau in magnitude in patients whose ICU stay ultimately exceeded 21 days. By multivariate analysis of variance (MANOVA), independent predictors of a prolonged stay in the SICU were APACHE III (p = 0.0023), emergency admission (p = 0.0007), and the magnitude of organ dysfunction (p < 0.00001), but not APACHE II. Only an emergency admission (p = 0.0005) and the magnitude of organ dysfunction (p < 0.00001) predicted a prolonged stay independently in survivors. In contrast, only the admission APACHE III score(p = < 0.0001) and the magnitude of organ dysfunction (p = 0.0001) were independently predictive of mortality by MANOVA.
CONCLUSIONS: The development of multiple organ dysfunction syndrome is a powerful predictor of a prolonged ICU course in critical surgical illness, even in survivors. Increased risk of a prolonged stay in the ICU plateaued at 21 days, making 21 days an appropriate definition of prolonged care for future studies. Predictive models should account for organ dysfunction and very long stays in future estimations. The combined use of APACHE III and the multiple organ dysfunction score may provide improved prediction of a prolonged stay in the ICU, but further enhancements are needed before prediction of outcome in individual patients is reliable.