Development and Validation of a Nomogram Model for Accurately Predicting Depression in Maintenance Hemodialysis Patients: A Multicenter Cross-Sectional Study in China

Abstract

Purpose

Depression is a major concern in maintenance hemodialysis. However, given the elusive nature of its risk factors and the redundant nature of existing assessment forms for judging depression, further research is necessary. Therefore, this study was devoted to exploring the risk factors for depression in maintenance hemodialysis patients and to developing and validating a predictive model for assessing depression in maintenance hemodialysis patients.

Patients and Methods

This cross-sectional study was conducted from May 2022 to December 2022, and we recruited maintenance hemodialysis patients from a multicentre hemodialysis centre. Risk factors were identified by Lasso regression analysis and a Nomogram model was developed to predict depressed patients on maintenance hemodialysis. The predictive accuracy of the model was assessed by ROC curves, area under the ROC (AUC), consistency index (C-index), and calibration curves, and its applicability in clinical practice was evaluated using decision curves (DCA).

Results

A total of 175 maintenance hemodialysis patients were included in this study, and cases were randomised into a training set of 148 and a validation set of 27 (split ratio 8.5:1.5), with a depression prevalence of 29.1%. Based on age, employment, albumin, and blood uric acid, a predictive map of depression was created, and in the training set, the nomogram had an AUC of 0.7918, a sensitivity of 61.9%, and a specificity of 89.2%. In the validation set, the nomogram had an AUC of 0.810, a sensitivity of 100%, and a specificity of 61.1%. The bootstrap-based internal validation showed a c-index of 0.792, while the calibration curve showed a strong correlation between actual and predicted depression risk. Decision curve analysis (DCA) results indicated that the predictive model was clinically useful.

Conclusion

The nomogram constructed in this study can be used to identify depression conditions in vulnerable groups quickly, practically and reliably.