An interpretable nomogram for predicting sleep disturbance risk in maintenance hemodialysis patients

Abstract Background This study aimed to develop and validate an interpretable nomogram to predict the risk of sleep disturbance in maintenance hemodialysis (MHD) patients. Methods In this single-center study, 208 MHD patients were enrolled. Sleep quality was assessed using the Pittsburgh Sleep Quality Index (PSQI), with a score > 5 defining the sleep disorder group. Univariate and multivariable logistic regression analyses identified independent predictors, which were used to construct a nomogram. The model’s performance was evaluated by the area under the receiver operating characteristic curve (AUC), calibration plot, and decision curve analysis (DCA). Explainable AI (SHapley Additive exPlanations, SHAP) was applied to interpret the model. Results Among 208 patients, 144 (69.2%) had sleep disturbance. Multivariable analysis identified restless legs syndrome (RLS) (OR = 4.52, 95% CI: 2.22–9.18), older age (OR = 1.04, 95% CI: 1.01–1.07), lower serum albumin (Alb) (OR = 0.86, 95% CI: 0.77–0.96), and lower parathyroid hormone (PTH) (OR = 0.99, 95% CI: 0.99–0.99) as independent predictors. Primary kidney disease etiology was retained for clinical comprehensiveness. The nomogram incorporating these five predictors demonstrated good discrimination (AUC = 0.80, 95% CI: 0.73–0.86), satisfactory calibration, and positive net benefit on DCA. SHAP analysis confirmed RLS as the most influential predictor and revealed complex, non-linear relationships for Alb and PTH. Conclusion This study presents a validated, interpretable nomogram that is associated with the individual risk of sleep disturbance in MHD patients using five readily available clinical parameters. The tool demonstrates good performance and clinical utility, potentially facilitating early identification and personalized management of high-risk individuals.