The impact of pediatric intensive care unit volume on mortality: a hierarchical instrumental variable analysis.

OBJECTIVE: To evaluate the relation between annual pediatric intensive care unit (PICU) admission volume and mortality. DESIGN: Nonconcurrent cohort design. SETTING: Pediatric patients included in the most currently available research database from the Pediatric Intensive Care Unit Evaluations (PICUEs). PATIENTS: A total of 34,880 consecutive pediatric admissions to a contemporary volunteer sample of 15 U.S. PICUs. MEASUREMENTS AND MAIN RESULTS: We conducted an instrumental variable analysis and adjusted for similarities between patients admitted to different PICUs using mixed-effects, hierarchical techniques. Case mix and severity of illness was adjusted for using patient-level data and the Pediatric Risk of Mortality, version III (PRISM III). On average, admission to higher-volume PICUs was associated with lower severity-adjusted mortality (odds ratio = 0.68 per 100 patient increase in volume; 95% confidence interval: 0.52-0.89) when volume was analyzed as a linear term; however, when PICU volume was analyzed as a quadratic term, we found the lowest severity-adjusted mortality rates among PICUs with annual admission volumes between 992 and 1,491. Furthermore, lower severity-adjusted mortality rates were primarily found among patients with less than a 10% PRISM III predicted risk of mortality. CONCLUSIONS: Although there is an association between lower severity-adjusted mortality among higher volume PICUs, our data suggest that best outcomes are among mid- to large-sized PICUs. These data support minimum annual admission criteria for PICUs but raise the concern that PICUs with very high annual admission volumes may operate beyond an ideal capacity.