Hospital volume-outcome relationships among medical admissions to ICUs.

BACKGROUND Positive relationships between hospital volume and outcomes have been demonstrated for several surgeries and medical conditions. However, little is known about the volume-outcome relationship in patients admitted to medical ICUs. OBJECTIVE To determine the relationship between hospital volume and risk-adjusted in-hospital mortality for patients admitted to ICUs with respiratory, neurologic, and GI disorders. DESIGN Retrospective cohort study. SETTING Twenty-nine hospitals in a single metropolitan area. PATIENTS Adult ICU admissions from 1991 through 1997. METHODS Using Cox proportional hazards models, we compared in-hospital mortality between tertiles of hospital volume (high, medium, and low) for respiratory (n = 16,949), neurologic (n = 13,805), and GI (n = 12,881) diseases after adjusting for age, gender, admission severity of illness, admitting diagnosis, and source. Severity of illness was measured using the APACHE (acute physiology and chronic health evaluation) III methodology. RESULTS Among respiratory and neurologic ICU admissions, hazard ratios were similar (p > or = 0.05) in patients in low-, medium-, and high-volume hospitals. However, among GI diagnoses, risk of mortality was lower in high-volume hospitals, relative to low-volume hospitals (hazard ratio, 0.68; 95% confidence interval [CI], 0.54 to 0.85; p < 0.001), and was somewhat lower in medium-volume hospitals (hazard ratio, 0.83; 95% CI, 0.68 to 1.01; p = 0.06). Among subgroups based on severity of illness, high-volume hospitals had lower mortality, relative to low-volume hospitals, among sicker patients (APACHE III score > 57) in the respiratory cohort (hazard ratio, 0.77; 95% CI, 0.59 to 0.99) and the GI cohort (hazard ratio, 0.67; 95% CI, 0.53 to 0.85). CONCLUSIONS Associations between ICU volume and risk-adjusted mortality were significant for patients with GI diagnoses and for sicker patients with respiratory diagnoses. However, associations were not significant for patients with neurologic diagnoses. The lack of a consistent volume-outcome relationship may reflect unmeasured patient complexity in higher-volume hospitals, relative standardization of care across ICUs, or lack of efficacy of some accepted ICU processes of care.