Abstract: Objective To analyze the risk factors for cognitive impairment (CI) in middle-aged and elderly peritoneal dialysis (PD) patients and to develop a risk prediction model. Methods Middle-aged and elderly PD patients from the Peritoneal Dialysis Center of Beijing Luhe Hospital, Capital Medical University between October 1, 2024 and December 31, 2024 were enrolled for group analysis and prediction model construction. Results A total of 138 PD patients were included in this study, with 96 in the training set. Univariate analysis revealed that the CI group had significantly lower serum potassium levels (t=2.150, P=0.034), lower albumin levels (t=2.310, P=0.023), lower Activities of Daily Living (ADL) scores [90.00 (60.00, 100.00) vs. 100.00 (100.00, 100.00), Z=-4.170, P＜0.010), higher prevalence of diabetes (71.19% vs. 48.65%, χ²=4.930, P=0.026), greater proportion of alcohol drinkers (67.80% vs. 35.14%, χ²=9.810, P=0.002), poorer health literacy (88.14% vs. 54.05%, χ² =4.930, P=0.026), and higher proportion of low income (67.80% vs. 32.43%, χ²=11.460, P＜0.001), as compared those with the non-CI group. Binary logistic regression identified decreased ADL score (OR=0.960, 95% CI: 0.930～0.990, P=0.003), history of alcohol consumption (OR= 16.310, 95% CI: 3.740～71.080, P＜0.001), annual family income ＜100,000 yuan (OR=9.620, 95% CI: 2.450～37.750, P=0.001), and poor medication adherence (OR=0.050, 95% CI:0.010～0.430, P=0.005) as the independent risk factors for CI in middle-aged and elderly PD patients. A prediction model was constructed based on these factors, and a nomogram was drawn. The model demonstrated an area under the curve (AUC) of 0.884 (95% CI: 0.819～0.948) in the training set, with an accuracy of 0.781 (95% CI:0.685～0.859) and a sensitivity of 0.973 (95%CI: 0.921～1.000). The validation set showed a predictive accuracy of 0.839 (95% CI: 0.716～0.961), a precision of 0.762 (95% CI: 0.605～0.879), and a F1 score of 84.2% (95% CI: 67.8～ 100.0%), which indicated that the model had higher predictability, accuracy and stability, and a better generalization ability. Conclusion The risk prediction model developed in this study demonstrates excellent performance for early identification of CI in middle-aged and elderly PD patients, having potential value for clinical application and promotion.

Key words: Peritoneal dialysis, Cognitive, Risk factor, Model construction