Development and temporal validation of an interpretable prediction model for 30-day meningitis or ventriculitis after pediatric brain tumor surgery: An 895-patient cohort study.
AI interpretation is pending for this paper.
Open original publication →What the AI sees
Not AI summarized yet.
Research significance
Pending deeper interpretation.
Source abstract
BACKGROUND: Meningitis or ventriculitis (MEN) is an important complication after pediatric brain tumor surgery, yet tools for early risk stratification remain limited. We developed and temporally validated prediction models for postoperative MEN. METHODS: We retrospectively included pediatric patients who underwent brain tumor resection at a single center during 2015-2025. Operations performed before January 1, 2023 constituted training cohort, and subsequent operations constituted temporal validation cohort. Elastic net-selected predictors were entered into five models: logistic regression (LR), support vector machine, decision tree, random forest (RF), and XGBoost. Model performance was assessed using the area under the receiver operating characteristic curve (AUC), calibration, Brier score, and decision curve analysis. SHapley Additive exPlanations (SHAP) analysis was used to interpret model predictions. RESULTS: Among 895 patients, 274 (30.6%) developed postoperative MEN, training and validation cohorts comprised 568 and 327 patients. Ten predictors were retained. AUCs in temporal validation ranged from 0.771 to 0.808. LR achieved the highest AUC (0.808), whereas XGBoost remained stable (training 0.823, validation 0.802), with low prediction error, and net benefit. SHAP analysis identified postoperative intraventricular hemorrhage (IVH) as the largest contributor to XGBoost predictions. Among 167 patients with intraoperative external-drain placement, external drain duration ≥7 days was independently associated with postoperative MEN (OR = 2.34; 95% CI: 1.04-5.28). CONCLUSION: Routinely available perioperative variables enabled temporal prediction of postoperative MEN after pediatric brain tumor surgery. The interpretable XGBoost model may assist postoperative risk stratification and external-drain management, although multicenter validation is required before clinical implementation.