Prediction Model for Infrared Thermotherapy Responsiveness in Mandibular Osteomyelitis and Its Application in Precision Nursing

This study aimed to develop and validate a prediction model for infrared thermotherapy response in mandibular osteomyelitis and evaluate its value for nursing decision support. In the retrospective phase, data, including a prediction model for infrared thermotherapy responsiveness, were established using Elastic Net regularized logistic regression. Cross-validation was used to assess the discriminative ability [receiver operating characteristic (ROC) curve and area under the ROC curve (AUC)] of the model. In the prospective phase, the model was applied to newly enrolled patients to stratify them by risk based on predicted probabilities. The efficacy of infrared thermotherapy and nursing-related outcomes was then compared across different responsiveness groups. The prediction model exhibited excellent discriminative ability (AUC=0.947) and good calibration performance in internal validation. According to the SHapley Additive exPlanations (SHAP) analysis, pretreatment C-reactive protein served as the most important negative predictor, while body mass index (BMI) functioned as the primary positive predictor. The results of prospective nursing intervention revealed that the high responsiveness group displayed a significantly higher response rate to infrared thermotherapy than the low responsiveness group (94.4% versus 18.2%, P<0.001), as well as better nursing-related outcomes. The prediction model using readily available clinical indicators can effectively assess infrared thermotherapy response in mandibular osteomyelitis, providing an objective basis for risk stratification and individualized nursing decisions.