Development and internal validation of risk scores to predict survival in the pediatric population following out-of-hospital cardiac arrest.

IntroductionOut-of-hospital cardiac arrest (OHCA) in the pediatric population is associated with poor survival and neurological outcomes. We aimed to develop and internally validate risk scores to predict survival to discharge in pediatric OHCA patients. MethodsWe included pediatric OHCA patients in the Cardiac Arrest Registry to Enhance Survival from 2013 to 2023. We used logistic regression (LR) and classification and regression trees (CART) to develop risk scores using 70% of the data and validated them using the remaining 30% of the data. We used discrimination and predictive accuracy to evaluate model performance. ResultsWe included 26895 pediatric patients, of whom 27.8% survived to hospital admission and 11.9% survived to hospital discharge. We developed three separate risk scores for i) infants ii) preschool and school-age children, and iii) adolescents. The most important predictor variables were gender, arrest location, witness status, etiology of arrest, bystander CPR, use of prehospital AED, and first rhythm. Models developed using both LR and CART approaches had good classification accuracy ( AUROCC for LR: 0.81, 0.77 and 0.82, AUROCC for CART: 0.78, 0.76, 0.82 for the three age groups) However, we found CART models to be the most useful because they are simpler to follow, classified more patients in low and high survival categories using a smaller number of predictors. The average survival probability in infants for each risk group was 3%, 13%, and 33%, for preschool and school-age children it was 4%, 11%, and 29%, whereas for adolescents it was 4%, 13%, and 48% for the low, moderate, and high survival categories. ConclusionPediatric patients experiencing OHCA can be classified into low, moderate and high survival categories using a simple risk score and easily identified prehospital variables. These risk scores can facilitate research, monitor the quality of medical services and can potentially support clinical decision-making.