Abstract: Objective  To construct a predictive model based on LASSO regression for maintenance hemodialysis (MHD) patients with cognitive decline.  Methods  100 patients treated with MHD in Yixing People's Hospital from March 2016 to March 2021 were prospectively included. The patients were classified into a cognitive impairment (CI) group (MocA＜26) and a non-CI group (MocA≥26) using the Montreal cognitive assessment (MocA) scale. General and clinical data were collected. Characteristic factors related to CI in MHD patients with non-zero coefficients were screened by LASSO regression. Lasso-Logistics regression model was constructed and receiver operating characteristic curve (ROC) was drawn to evaluate the model effectiveness. Data rows of 71 MHD patients were selected for external verification.  Results  There were 62 patients in CI group and 38 patients in non-CI group, respectively. LASSO regression showed that age, dialysis vintage, diabetes mellitus (OR=1.673, 95%CI:0.150～18.661,P=0.676), higher sensitive C-reactive protein(hs-CRP), triglyceride, total cholesterol, serum albumin and serum calcium were significantly correlated with the occurrence of CI in MHD patients. Logistic regression analysis showed that age (OR=1.135, 95%CI:1.041～1.238, P=0.004), dialysis vintage (OR=1.138, 95%CI :1.012～1.279, P=0.031), hs-CRP (OR=2.394, 95%CI:1.158～4.950, P=2.394), serum albumin (OR=0.759, 95%CI: 0.593～0.971,P=0.028) were risk factors for CI in MHD patients. Lasso-logistics regression model predicted that the AUC of CI occurrence in MHD patients was 0.892 (95%CI: 0.813～0.945). External verification showed that the area under the curve (AUC) was 0.899 (95%CI: 0.0.804～0.958), and the Youden index J was 0.713.   Conclusion  Age, dialysis vintage, hs-CRP and lower albumin were significantly correlated with the incidence of CI in MHD patients, and the established prediction model could better predict the incidence of CI in MHD patients. 

Key words: LASSO regression, Maintenance hemodialysis, Cognitive function, Predictive