Frailty and metabolic dysfunction as key components of an individualized prediction model and web-based tool for postoperative residual numbness after posterior lumbar interbody fusion: a retrospective study

To develop and validate an individualized prediction model incorporating frailty and metabolic dysfunction-related parameters and establish a web-based clinical tool for predicting postoperative residual numbness (PRN) after posterior lumbar interbody fusion (PLIF). This retrospective cohort study included patients undergoing PLIF for lumbar degenerative disorders between January 2017 and January 2022. Patients were classified into PRN and non-PRN groups according to 12-month postoperative numbness outcomes. Candidate predictors, including demographic, metabolic, frailty-related, and surgical/radiological factors, were collected. Variables with P < 0.05 in univariate logistic regression analysis were subsequently entered into multivariable logistic regression to identify independent predictors of PRN. The triglyceride-glucose (TyG) index and frailty index (FI) were calculated to assess metabolic dysfunction and physiological vulnerability, respectively. A nomogram was developed based on independent predictors and evaluated using receiver operating characteristic (ROC) analysis, calibration curves, and decision curve analysis (DCA). A web-based tool was established for individualized PRN risk prediction. A total of 411 patients undergoing PLIF were included, and independent predictors of PRN were identified using multivariable logistic regression. An individualized nomogram was developed based on these predictors, demonstrating excellent discrimination (area under the receiver operating characteristic curve (AUC) = 0.929 in the training cohort and 0.918 in the validation cohort), good calibration, and favorable clinical utility by DCA. Cost-sensitive threshold analysis further supported its applicability under different clinical preferences. The nomogram was subsequently transformed into a web-based application for individualized PRN risk prediction. This study developed and validated an individualized prediction model for PRN after PLIF by integrating frailty-related, metabolic, and surgical factors. The proposed nomogram demonstrated favorable predictive performance and clinical utility, providing a practical approach for individualized risk stratification. The accompanying web-based tool further enhances the clinical accessibility of this model, facilitating early postoperative identification of high-risk patients and supporting personalized perioperative management after PLIF.