Early Mortality in Incident Hemodialysis Patients: Risk Factors and Nomogram Prediction

Globally, The number of end-stage renal disease patients on hemodialysis is rising, but most of the research related to death focusing on the maintenance hemodialysis period, while there is few analyses related to the risk of death in the initial hemodialysis and lack of the risk prediction model for the early death. Hence, this study aimed to epidemiologically survey these patients and establish a death risk prediction model.

From January 1, 2015 to November 30, 2021, the Shenzhen Dialysis Database registered a total of 7,767 dialysis patients. We used the multiple imputation method to handle missing data and employed LASSO regression to screen variables, constructing univariate and multivariate Cox regression models to predict the risk of premature death. Finally, we drew Nomogram diagrams and ROC curves for visual presentation.

This study included 5841 patients with a median follow-up of 1.0 year. While the overall mortality rate across the entire observation period was 2.1 per 100 patient-years (208 total deaths), the mortality risk was disproportionately concentrated in the early phase, with 54 deaths occurring during the first 90 days, representing a cumulative incidence of 0.92% (54/5841). Multivariate Cox regression analysis identified advanced age, central venous catheter (CVC) use, coexisting cerebrovascular disease, and elevated C-reactive protein (CRP) as independent risk factors for 90-day mortality. Conversely, hemodialysis combined with hemodiafiltration (HD+HDF) was identified as a strong independent protective factor (HR=0.504). A single-chart model was successfully constructed. This model demonstrated excellent discrimination throughout the entire follow-up period (optimism-corrected Harrell's C-index = 0.888), and had strong predictive accuracy for the 90-day mortality parameter (AUC = 0.906).

This study found that being 75 years old or above was a key risk factor for 90-day mortality among newly admitted dialysis patients, with a 16-fold increase in risk. The established risk model provides preliminary clinical benefits for early individualized risk stratification. However, large-scale, multi-center prospective studies are still needed to externally validate the model and confirm its wide clinical applicability.