497 Developing Predictive Models to Anticipate Shunt Complications in 33,248 Pediatric Patients with Shunted Hydrocephalus Utilizing Machine Learning

Abstract

INTRODUCTION: Hydrocephalus is a common pediatric neurosurgical pathology and is typically treated with a ventricular shunt. Shunt complications are common in pediatric hydrocephalus, and ∼30% of patients experience shunt failure during the first year after shunt placement. METHODS: We queried the Healthcare Cost and Utilization Project (HCUP) National Readmissions Database (NRD) from 2016 to 2017 using ICD-10 codes and found 33,248 pediatric patients who underwent shunt placement. Elixhauser comorbidity index (ECI) scores, Johns Hopkins Adjusted Clinical Groups (JHACG) frailty-defining criteria, and Major Diagnostic Category (MDC) classifications were obtained. To determine predictors of shunt complication, the database was divided into a training dataset (n=19,948), a validation dataset (n=6,650), and a testing dataset (n=6,650). Multivariable analysis was performed using the training dataset to identify significant independent predictors of shunt complications in the other datasets. Statistically significant variables were used to develop logistic regression models, and post-hoc receiver operating characteristic (ROC) curves were created. RESULTS: Average age of the 33,248 pediatric patients was 6.9±5.7 years. Variables found to be positively correlated with shunt complications included: number of diagnoses during primary admission (OR: 1.05, 95% CI: 1.04-1.07) and neurological MDCs (OR: 3.83, 95% CI: 3.33-4.42). Variables found to be negatively correlated with shunt complications included: ECI (OR: 0.82, 95%CI: 0.77-0.86), Female sex (OR: 0.87, 95%CI: 0.76-0.99), and elective admissions (OR: 0.62, 95%CI: 0.53-0.72). The ROC curve for the regression model utilizing all significant predictors of readmission in multivariate analysis demonstrated an area under the curve (AUC) of 0.733. CONCLUSION: Expeditious and safe treatment of pediatric hydrocephalus is of paramount importance. Our machine learning algorithm established numerous variables predictive of shunt complications in the pediatric population. Our AUC is suggestive of excellent predictive value. These risk factors may be of clinical importance when evaluating pediatric patients with hydrocephalus.