Development and validation of machine learning-based model for hospital length of stay in patients undergoing endovascular interventional embolization for intracranial aneurysms.

Objective: This study was to explore the factors associated with prolonged hospital length of stay (LOS) in patients with intracranial aneurysms (IAs) undergoing endovascular interventional embolization and construct prediction model machine learning algorithms.

Methods: Employing a retrospective cohort study design, this study collected patients with ruptured IA who received endovascular treatment at Jingzhou First People's Hospital during the inclusion period from September 2022 to December 2023. The entire dataset was randomly split into training and testing dataset with a 7:3 ratio. Six machine learning models including Logistic regression (LR) support vector machine (SVM), random forest (RF), extreme gradient boosting (XGBoost), K nearest neighbors (KNN), and Naive Bayes (NB) were constructed. Each model was assessed using sensitivity with 95% confidence interval (CI), specificity, positive predictive value (PPV), negative predictive value (NPV), area under the curve (AUC), accuracy, and F1-Score. The performance of the optimal model was compared against other models using the net reclassification index (NRI), and the integrated discrimination improvement (IDI).

Results: In this study, 325 patients were enrolled, with 227 assigned to the training set and 98 to the testing set. The training set comprised 163 patients with LOS below the third quartile and 64 patients with LOS at or above the third quartile. Age, Hunt-Hess grade, National Institutes of Health and Stroke Scale (NIHSS), white blood cell (WBC) count, Fisher grade above II, moderate aneurysm size, preoperative dexmedetomidine administration, and postoperative complications including electrolyte imbalance correction, encephaledema, and respiratory system disease were identified as predictive factors. The RF model exhibited the best predictive performance with AUC of 0.928 (95% CI: 0.895 to 0.961) in the training set. This high performance was consistent in the testing set, where the AUC remained strong at 0.912 (95% CI: 0.851 to 0.973).

Conclusion: This study comprehensively identified key predictive factors for prolonged LOS in patients with IA undergoing interventional embolization and confirmed the efficacy of an RF model for predicting prolonged LOS in patients with IA undergoing interventional embolization. The construction of LOS prediction model may effectively optimize healthcare resource utilization, inform better clinical decision-making, and offer valuable prognostic insights.