Search

Introduction:Predicting post-operative ICU length of stay (LOS) after congenital heart surgery (CHS) could inform scheduling, improve capacity management and model resource use variation.Hypothesis:A prediction model including patient characteristics and operative complexity at post-operative ICU admission can accurately predict ICU LOS >7 days. The model could be used to explore institutional differences in LOS.Methods:CHS admissions to 31 Pediatric Cardiac Critical Care Consortium (PC4) centers from April 2014 - September 2018 were identified. Hierarchical logistic regression with center-specific random intercepts identified variables associated with ICU LOS >7 days. The final model was used to compare risk-adjusted rates of ICU LOS>7 days across centers.Results:Of 29,113 CHS ICU admissions, 22.9% of the cohort had ICU LOS >7 days (range 1.8 - 41.2% by center). Important predictors of ICU LOS>7 days were STAT category (C=0.796) and pre-operative status (C=0.803), a descriptive composite of age, pre-operative mechanical ventilation, and pre-operative admission. Together, STAT category and pre-operative status were powerful predictors of ICU LOS>7 days (C=0.848, Table). Incorporation of post-operative characteristics of illness severity did not improve the model. Risk-adjusted rates for ICU LOS >7 days ranged from 10.8% to 40.5% (Figure).Conclusions:We developed a model to predict ICU LOS >7 days at postoperative ICU admission with excellent discrimination. We observed 4-fold variability in case-mix adjusted rates of ICU LOS>7 days across 31 centers. Further investigation is required to understand best practices and inform future quality initiatives.