Novel Machine Learning Models to Predict Pneumonia Events in Supratentorial Intracerebral Hemorrhage Populations: An Analysis of the Risa-MIS-ICH Study

Background: Stroke-associated pneumonia (SAP) contributes to high mortality rates in spontaneous intracerebral hemorrhage (sICH) populations. The accurate prediction and early intervention of SAP are associated with prognosis. Although various predictive scoring systems have been previously developed, none are widely accepted. We aimed to derive and validate novel supervised machine learning (ML) models to predict SAP events in supratentorial sICH populations. Methods: : In this work, the data of eligible supratentorial sICH individuals were extracted from the database of the Risa-MIS-ICH study, and the participants were split into training, internal validation, and external validation datasets. The primary outcome was SAP during hospitalization. Univariate and multivariate analyses were used for variable filtrations, and logistic regression (LR), Gaussian naïve Bayes (GNB), random forest (RF), K-nearest neighbor (KNN), support vector machine (SVM), extreme gradient boosting (XGB), and ensemble soft voting model (ESVM) were adopted for ML model derivations. The metrics of accuracy, sensitivity, specificity, and area under the curve (AUC) were adopted to evaluate the predictive value of each model with internal/cross-/external validations. Results: : After screening 909 individuals with sICH, a total of 468 were included in this work. Six independent variables [nasogastric feeding, airway support, unconscious onset, surgery of external ventricular drainage (EVD), sICH volume, and intensive care unit (ICU) stay] for SAP were identified and selected for seven ML prediction model derivations and validations. The internal and cross-validations revealed the superior and robust performance of the GNB model with the highest AUC value (0.861, 95% CI 0.793-0.930), while the LR model had the highest AUC value (0.867, 95% CI 0.812-0.923) in external validation. The ESVM method combining the other six methods had moderate but robust abilities in both cross- and external validations and achieved an AUC of 0.843 ( 95% CI 0.784, 0.902) in the external validation. Conclusion: The ML models could effectively predict SAP events in sICH populations, and our novel ensemble models demonstrated reliable robust performance outcomes despite the populational and algorithmic differences. Registration: URL: https://www.clinicaltrials.gov. Unique Identifier: NCT03862729