PICU Length of Stay: Factors Associated With Bed Utilization and Development of a Benchmarking Model

Objectives: ICU length of stay is an important measure of resource use and economic performance. Our primary aims were to characterize the utilization of PICU beds and to develop a new model for PICU length of stay. Design: Prospective cohort. The main outcomes were factors associated with PICU length of stay and the performance of a regression model for length of stay. Setting: Eight PICUs. Patients: Randomly selected patients (newborn to 18 yr) from eight PICUs were enrolled from December 4, 2011, to April 7, 2013. Data consisted of descriptive, diagnostic, physiologic, and therapeutic information. Interventions: None. Measurements and Main Results: The mean length of stay for was 5.0 days (sd, 11.1), with a median of 2.0 days. The 50.6% of patients with length of stay less than 2 days consumed only 11.1% of the days of care, whereas the 19.6% of patients with length of stay 4.9-19 days and the 4.6% with length of stay greater than or equal to 19 days consumed 35.7% and 37.6% of the days of care, respectively. Longer length of stay was observed in younger children, those with cardiorespiratory disease, postintervention cardiac patients, and those who were sicker assessed by Pediatric Risk of Mortality scores receiving more intensive therapies. Patients in the cardiac ICU stayed longer than those in the medical ICU. The length of stay model using descriptive, diagnostic, severity, and therapeutic factors performed well (patient-level R-squared of 0.42 and institution-level R-squared of 0.76). Standardized (observed divided by expected) length of stay ratios at the individual sites ranged from 0.87 to 1.09. Conclusions: PICU bed utilization was dominated by a minority of patients. The 5% of patients staying the longest used almost 40% of the bed days. The multivariate length of stay model used descriptive, diagnostic, therapeutic, and severity factors and has potential applicability for internal and external benchmarking.