Orthopedic perioperative nursing under navigation nurse management: Machine learning-based risk prediction models for postoperative recovery quality and explainable artificial intelligence analysis

This study aimed to evaluate the effectiveness of navigation nurse management (NNM) in orthopedic perioperative care and develop machine learning (ML) models to predict postoperative recovery quality. We sought to identify key factors influencing recovery outcomes in patients undergoing hip surgery and assess whether NNM intervention improves patient outcomes compared to standard care. This retrospective study included 216 patients who underwent orthopedic surgery for femoral neck fractures, femoral head necrosis, or hip joint disorders at Zhangjiagang Hospital between November 2023 and February 2025. The NNM model, comprising 6 core elements, guided nursing care. Data were analyzed using SPSS 26.0 and R 4.4.2. The dataset was randomly split into training (70%) and validation (30%) cohorts. In addition to logistic regression (LR) and nomogram construction, we applied 6 ML algorithms including random forest (RF), eXtreme gradient boosting, support vector machine, decision tree, Naïve Bayes, and LR. We evaluated model performance using area under the curve (AUC), sensitivity, specificity, precision, and F1 scores. SHapley Additive exPlanations (SHAP) analysis was employed to enhance model interpretability and identify key contributing factors. Among the 216 patients, 122 were classified as the high-quality recovery group and 94 as the poor recovery group. Multivariate LR identified postoperative first meal time, time to first ambulation (postoperative), Final Visual Analogue Scale (at discharge), and receipt of NNM as independent predictors. The nomogram achieved AUCs of 0.983 and 0.992 in training and validation sets, respectively. Among ML models, RF demonstrated the best performance with perfect scores across all metrics (AUC = 1.000, sensitivity = 100%, specificity = 100%, precision = 100%, F1 = 100%), followed by eXtreme gradient boosting (AUC = 0.998). SHAP analysis revealed that Final Visual Analogue Scale (at discharge) was the most influential factor, while NNM significantly reduced the risk of poor recovery quality. Patients managed under the NNM model demonstrated significantly better postoperative recovery quality compared to those who did not receive NNM. NNM improves postoperative recovery quality in orthopedic patients. RF algorithms showed better predictive accuracy than traditional methods for identifying high-risk patients. SHAP analysis improved model interpretability, supporting personalized care decisions.