Abstract
The mortality rates for patients undergoing hemodialysis (HD) remain unacceptably high compared to the general population, and more specific information about the causes of death is not known. The study aimed to develop and validate a risk prediction model that uses common clinical factors to predict the probability of cardiovascular events in maintenance hemodialysis (MHD) patients. The study involved 3488 adult patients who received regular scheduled hemodialysis treatment at 20 hemodialysis centers in southwest China between June 2015 and August 2020, with follow-up until August 2021. The optimal parameter set was identified by multivariable Cox regression analyses and Cross-LASSO regression analyses and was used to establish a nomogram for predicting the risk of cardiovascular events in maintenance hemodialysis patients at 3 and 5 years. The performance of the model was evaluated using the consistency index (Harrell's C-index), the area under the receiver operating characteristic (ROC) curve, and calibration plots. The model was validated by tenfold cross-validation and bootstrapping with 1000 resamples. In the derivation cohort, the model yields an AUC of 0.764 [95% confidence interval (CI), 0.737-0.790] and 0.793 [CI, 0.757-0.829] for predicting the risk of cardiovascular events of MHD patients at 3 and 5 years. In the internal validation cohort AUC of 0.803 [95% CI, 0.756-0.849], AUC of 0.766 [95% CI, 0.686-0.846], and the external validation cohort AUC of 0.826 [95% CI, 0.765-0.888], AUC of 0.817 [95% CI, 0.745-0.889] at 3 and 5 years. The model's calibration curve is close to the ideal diagonal. By tenfold cross-validation analyses, the 3- and 5-year risk of cardiovascular events (AUC 0.732 and 0.771, respectively). By the bootstrap resampling method, the derivation cohort and validation cohort (Harrell's C-index 0.695 and 0.667, respectively) showed good uniformity with the model. The constructed model accurately predicted cardiovascular events of MHD patients in the 3rd and 5th years after dialysis. And the further research is needed to determine whether use of the risk prediction tool improves clinical outcomes.