Development and external validation of a multivariate model for predicting pneumonia in patients receiving maintenance hemodialysis: a retrospective study

TL;DR

This study developed a model to predict pneumonia risk in patients on hemodialysis, helping identify high-risk individuals for better care.

Contribution

A validated multivariate model for predicting pneumonia in hemodialysis patients using clinical and laboratory variables.

Findings

The model achieved a C-index of 0.753 in external validation and 0.772 in the modeling set.

Key predictors included diabetes, coronary heart disease, and lab values like white blood cell count and albumin-globulin ratio.

The model showed excellent calibration and clinical utility for risk stratification in MHD patients.

Abstract

Patients receiving maintenance hemodialysis (MHD) who develop pneumonia experience substantially elevated risks of hospitalization and mortality, while also incurring significantly heightened healthcare-related financial burdens. Our goal is to establish a forecasting model to assess the individual risk of pneumonia in patients undergoing MHD.

A retrospective analysis was carried out between January 2018 and November 2024, involving 405 MHD patients from two medical centers. The variables underwent adjustment through multivariate Cox regression analysis, and the forecasting model was created and verified.

The median follow-up time of the external validation set was 35 months (interquartile range: 20–43), and the median follow-up time of the modeling set was 22 months (12–24). The event happened in 101 (34.83%) out of 290 patients in the modeling set and 45 (39.13%) out of 115 patients in the external validation set. The model predictors included history of diabetes and coronary heart disease; serous effusion; white blood cell; albumin-globulin ratio; left ventricular mass index, and age. The C-index was 0.753 (0.684, 0.822) for the external validation set and 0.772 (95% CI [0.724–0.821]) for the modeling set. The model showed excellent calibration ability throughout the risk spectrum, and decision curve analysis showed that it could maximize the prognosis of patients.

The created predictive model provided a precise, individualized evaluation of pneumonia risk in patients with MHD. It can be used to identify individuals at high risk of pulmonary infection in patients undergoing MHD and guide their treatment and prognosis follow-up.

Figures