721: MULTICENTER VALIDATION AND REFINEMENT OF THE CRITICAL BRONCHIOLITIS SCORE

Introduction: We previously created the “Critical Bronchiolitis Score” (CBS) to predict outcomes for this very common, low-mortality Pediatric Intensive Care Unit (PICU) disease. We created 4 models to evaluate 2 outcomes, duration of ICU-level respiratory support (D-ICU-LRS) and PICU length of stay (LOS), in 2 groups: children requiring or not requiring early endotracheal intubation within first 12 hours of PICU stay. The original CBS performed better than mortality based PICU severity scores. However, relevant variables were unavailable in the original dataset. This study aims to improve CBS performance by adding variables in a second, large multi-center dataset. Methods: With IRB approval, we retrospectively studied children < 2 years of age admitted to 37 PICUs with bronchiolitis. We collected demographics, original CBS and several novel variables: age, specific co-morbidities, SpO2, encephalopathy, effort of breathing, and specific ICU-LRS settings (e.g., FiO2). As previously, missing variables were imputed with age-specific norms for the 12 centers with incomplete chart review. The newly assessed variables were tested for inclusion and retained if performance improved. CBS performance was assessed by root mean square error (RMSE) and mean absolute error (MAE). Data shown as median (IQR). Results: In 1163 subjects, median age was 5 (2-11) months, 60% had respiratory syncytial virus, 16% required early intubation. Median D-ICU-LRS was 2.3 (1.4-4.2) days and median PICU LOS was 2.8 (1.8-5.0) days. In all models, adding FiO2 improved performance. In the 2 models for children not requiring early intubation, adding minimum SpO2 in the first 12hrs improved performance for both D-ICU-LRS and PICU LOS. In early intubated patients, RMSE improved from the original CBS to the revised CBS by 6% for D-ICU-LRS and 18% for PICU LOS, and MAE improved by 2% and 16%. Among non-intubated patients, RMSE and MAE improved by 55% for both outcomes. Similar improvements were seen in centers with (n = 667) and without (n = 496) complete chart review. Conclusions: Compared to the original CBS, adding FiO2 and minimum SpO2 substantially improved performance and expanded its ability to aide providers in predicting clinically relevant outcomes in bronchiolitis.