Machine Learning for Enhanced Prognostication: Predicting 30-Day Outcomes Following Posterior Fossa Decompression Surgery in a Pediatric Cohort with Chiari Malformations

Introduction: Chiari Malformation (CM) affects 0.1% of the population, and in pediatric cases, surgical treatment with posterior fossa decompression (PFD) is linked to notable short-term adverse outcomes. Predictive tools to assess the risk of such events are needed to aid patient care and cost management.

Methods: This retrospective study used data from the ACS NSQIP-P Database for pediatric patients undergoing PFD (CPT code 61343) between 2012 and 2021, excluding those treated for tumors or vascular lesions. The primary outcomes were 30-day unplanned readmission and reoperation rates, with additional outcomes such as hospital stay length, 30-day complications, discharge status, and mortality. Multivariable logistic regression and several machine learning (ML) models, including random forest and decision tree, were used. Data were split 80:20 for training/testing, with SMOTE for class imbalance, and validated with 10-fold cross-validation, using AUC-ROC and F1 scores for model evaluation.

Results: In a cohort of 7,106 pediatric patients (median age 9.2 years, 56% female), 30-day readmission and reoperation rates were 7.5% and 3.4%, respectively. CNS-related (32%) and wound-related complications (30%) were primary readmission reasons, while wound revisions and fluid/blood evacuation (61%) and CSF diversion procedures (28%) led to reoperations. Random forest models showed the highest predictive accuracy for readmissions (AUC=0.960) and reoperations (AUC=0.990).

Conclusion : ML models demonstrated high accuracy in predicting 30-day readmissions and reoperations after PFD in pediatric CM cases, underscoring the value of ML in risk assessment and surgical planning.