Abstract: Objective To construct a visual risk prediction model for malnutrition in maintenance hemodialysis (MHD) patients, and to predict the risk of malnutrition individually. Methods A total of 457 patients receiving MHD in dialysis center of three tertiary grade-A hospitals in Henan Province from March to October 2022 were selected. Patients from the first two hospitals were included as the modeling group (n=320), and those from the third hospital as the validation group (n=137). Univariate and multivariate logistic regression analysis was used to construct the model. ROC curve and Hosmer-Lemeshow calibration curve were used to test the differentiation and calibration degree of the prediction models. The software packages shiny, shinydashboard, survival and regplot in R software were used to develop an online calculation tool for risk prediction of malnutrition in MHD patients. Results Age (OR=2.550,95%CI:1.390~4.724),dialysis age (OR=1.029,95%CI:1.016~1.043),dialysis adequacy (OR=4.041,95%CI:2.226~7.519), serum albumin(OR=0.860,95%CI:0.796~0.922), hemoglobin(OR=0.979,95%CI:0.963~0.996), depression score(OR=1.101,95%CI:1.057~1.148) were finally included in this study. Internal evaluation and external verification of the model showed that the area under the curve was 0.881 and 0.875, the P-value of Hosmer-Lemeshow goodness of fit test was 0.687 and 0.621, and the average absolute error was 0.019 and 0.012, respectively. Conclusion The risk prediction model of malnutrition in MHD patients and the online calculator constructed in this study are effective and easy to use, providing reference for the early screening of malnutrition risk in MHD patients.

Key words: maintenance hemodialysis, malnutrition, prediction model, nomogram, online calculator