A SHAP-interpretable nomogram for predicting postoperative pneumonia in elderly hip fracture patients with preoperative anemia: a single-center retrospective study with internal validation

This is the authors' abstract. We don't add key points for this paper.

Background Postoperative pneumonia is a frequent and serious complication in elderly patients undergoing surgery for hip fractures, particularly in those with preoperative anemia. Early identification of high-risk individuals is crucial for improving clinical outcomes. Methods This retrospective study included 339 elderly patients with hip fractures and preoperative anemia. A multivariable logistic regression model, incorporating variables selected via the Boruta algorithm, was used to identify independent predictors of postoperative pneumonia. A clinical nomogram was constructed to facilitate individualized risk stratification. The model's performance was evaluated using the area under the receiver operating characteristic curve (AUC), calibration plots, and decision curve analysis (DCA). The SHapley Additive exPlanations (SHAP) method was utilized to interpret the model and quantify the contribution of each predictor. Results Four variables were identified as independent predictors of postoperative pneumonia: chronic obstructive pulmonary disease (COPD), preoperative planned intensive care unit (ICU) admission, preoperative serum albumin (ALB) levels, and preoperative non-steroidal anti-inflammatory drug (NSAID) use. The developed nomogram showed moderate discrimination, with an apparent AUC of 0.746 and a nested bootstrap-validated AUC of 0.751 (95% CI 0.691–0.808). The calibration plot demonstrated acceptable agreement, and DCA confirmed the model's clinical utility across a broad threshold probability range (0.13–0.87). SHAP analysis revealed that COPD and hypoalbuminemia were the most significant drivers of risk. Conclusion This study provides an internally validated nomogram for predicting postoperative pneumonia in elderly, anemic patients with hip fractures. By integrating clinical and laboratory data, this tool may support early risk stratification and inform the implementation of targeted perioperative preventive strategies for high-risk patients.