Development of an interpretable machine learning model for predicting intestinal necrosis in older patients with incarcerated hernia: A retrospective cohort study.

ObjectiveTo develop a machine learning model for predicting intestinal necrosis in older patients with incarcerated hernia and interpret the model using SHapley additive exPlanations (SHAP) to assist clinicians in early identification of intestinal necrosis and formulating more appropriate treatment strategies.MethodsThis retrospective cohort study analyzed 324 older patients with incarcerated hernia who were admitted to the Department of Emergency Surgery and Department of Hernia and Abdominal Wall Surgery at the Affiliated Lu'an Hospital of Anhui Medical University between March 2015 and February 2024. Clinical data were collected, and patients were divided into training and testing sets in a 7:3 ratio using stratified random sampling. The least absolute shrinkage and selection operator was applied for feature selection, and nine machine learning models were constructed. The optimal model was further analyzed using SHAP for interpretability.ResultsAmong the 324 patients, 81 developed intestinal necrosis, while 243 did not. Least absolute shrinkage and selection operator regression identified nine key features. Nine machine learning algorithms were evaluated: (a) K-nearest neighbors; (b) support vector machine; (c) extreme gradient boosting; (d) approximate nearest neighbor; (e) decision tree; (f) light gradient boosting machine; (g) random forest; (h) extremely randomized trees; and (i) gradient boosting machine. The extremely randomized trees model demonstrated the best performance on the test set, with a sensitivity of 0.755, specificity of 0.932, F1-score of 0.714, accuracy of 0.796, and receiver operating characteristic area under the curve of 0.785. SHAP analysis revealed that the top nine contributing variables were procalcitonin, absolute neutrophil count, D-dimer, neutrophil-to-lymphocyte ratio, hernia disease time, duration of incarceration, femoral hernia incarceration, lymphocyte count, and C-reactive protein.ConclusionThis study developed an efficient and interpretable extremely randomized trees model for predicting intestinal necrosis risk in older incarcerated hernia patients, which can aid clinicians in identifying high-risk patients and optimizing clinical decision-making.