Prediction of swallowing improvement after tracheal reconstruction in patients receiving LFES-containing rehabilitation: a retrospective cohort study.

Background: Dysphagia is a clinically important complication after tracheal reconstruction. Low-frequency electrical stimulation (LFES) supports postoperative swallowing rehabilitation, yet swallowing outcomes vary substantially among patients. We aimed to identify prognostic factors associated with observed swallowing improvement and to develop an internally validated clinical prediction model among patients receiving LFES-containing rehabilitation after tracheal reconstruction. Methods: This retrospective cohort study included 200 patients with postoperative dysphagia who received LFES-containing rehabilitation. Patients were grouped by prespecified criteria for swallowing improvement after rehabilitation. Demographic, perioperative, swallowing-related, and prespecified treatment-plan variables available before rehabilitation were evaluated. The cohort was divided by outcome-stratified random sampling into training and independent validation sets. Candidate predictors were examined using univariable logistic regression, LASSO, random forest, and XGBoost, followed by multivariable logistic regression. Model performance was assessed using ROC analysis, calibration, bootstrap internal validation, and exploratory decision-curve, subgroup, and sensitivity analyses. Results: Among 200 patients, 126 (63.0%) met the prespecified swallowing-improvement criteria. Older age (OR = 0.96, 95% CI: 0.93-0.98) and aspiration (OR = 0.41, 95% CI: 0.18-0.88) were independently associated with a lower probability of improvement, whereas a greater planned number of LFES sessions was associated with a higher probability of improvement (OR = 1.09, 95% CI: 1.04-1.16). The logistic regression model achieved AUCs of 0.776 in the training set and 0.767 in the independent validation set, with acceptable calibration. The optimism-corrected AUC after 500 bootstrap resamples was 0.721. Exploratory decision-curve analysis suggested potential net benefit across portions of the evaluated threshold-probability range. XGBoost achieved the numerically highest validation AUC (0.780), followed by logistic regression (0.765) and random forest (0.760), with only small absolute differences among these models. Main associations were generally consistent across sensitivity analyses. Conclusion: Among patients receiving LFES-containing rehabilitation after tracheal reconstruction, older age and aspiration were associated with a lower probability of observed swallowing improvement, whereas a greater planned number of LFES sessions was associated with a higher probability of improvement. The model demonstrated moderate discrimination and may provide a preliminary framework for prognostic risk stratification and identifying patients who warrant closer rehabilitation monitoring. External validation is required before routine clinical use.