A Predictive Model for Secondary Posttonsillectomy Hemorrhage in Pediatric Patients: An 8-Year Retrospective Study

Abstract

Objectives

Posttonsillectomy hemorrhage (PTH) is a common and potentially life-threatening complication in pediatric tonsillectomy. Early identification and prediction of PTH are of great significance. Currently, there are very few tools available for clinicians to accurately assess the risk of PTH. This study aimed to develop and validate a predictive model for secondary PTH.

Methods

A retrospective analysis was conducted on 492 individuals who underwent tonsillectomy or tonsillotomy in Children's Hospital of Soochow University from July 1st, 2015 to December 31th, 2023. The study population was randomly divided into the training set and the validation set at a ratio of 7:3. Univariate logistic regression analysis was used to screen features. Multivariate logistic regression and seven machine learning algorithms were used to construct predictive models. Discrimination, calibration, and clinical utility were used to compare the predictive performance. The SHapley Additive exPlanation (SHAP) method was used to interpret the results of the best-performing model.

Results

One multivariate logistic regression model and seven machine learning models were constructed. The XGBoost model yielded the best performance in the validation set. The SHAP method ranked the features of the XGBoost model based on their importance and provided both global and local explanations of the model.

Conclusion

This study established a machine learning-based predictive model for secondary PTH, which may enable clinicians to accurately assess the risk of secondary PTH in children.

Level of Evidence

4