# A time-dependent predictive model for cardiocerebral vascular events in chronic hemodialysis patients: insights from a prospective study

## TL;DR

This study creates a model to predict cardiovascular and cerebrovascular events in hemodialysis patients using routine lab data.

## Contribution

A novel time-dependent predictive model for CVCs in HD patients using clinical and lab data.

## Key findings

- The model included age, sex, hemoglobin, albumin, phosphate, WBC, blood flow, and ultrafiltration volume.
- The model showed good accuracy with a C index of 0.704 in development and 0.775 in validation cohorts.
- It outperformed models based solely on p-values from regression analysis.

## Abstract

The conventional risk factors for cardiocerebral vascular events (CVCs) in non-Hemodialysis (HD) patients cannot be directly applied to HD patients due to the unique characteristics of this population. More accurate information on the risk of progression to CVCs is needed for clinical decisions.

To develop and validate time-dependent predictive models for the progression of CVCs in HD patients.

Development and validation of time-dependent predictive models using demographic, clinical, and laboratory data from 3 dialysis centers between 2017 and 2021. These models were developed using time-dependent Cox proportional hazards regression and assessed for discrimination using the concordance index, goodness of fit using the Akaike information criterion and net reclassification improvement.

CVCs included acute heart failure, acute hematencephalon, cardiac or brain-derived death, acute myocardial infarction, acute cerebral infarction, ischemic cardiomyopathy, unstable angina pectoris, and stable angina pectoris.

The development and validation cohorts included 233 and 215 patients, respectively. The most accurate model included age, sex, hemoglobin, serum albumin, serum phosphate, white blood cell count, blood flow rate and ultrafiltration volume during HD (C index, 0.704; 95% CI, 0.639–0.768 in the development cohort and 0.775; 95% CI, 0.706–0.843 in the validation cohort). In the validation cohort, this model was more accurate than a model containing variables whose p value in the Cox proportional hazards regression was less than 0.05 (NRI: 0.351, 95% CI: −0.115–0.565).

A time-dependent model using routinely obtained laboratory tests can accurately predict progression to CVCs in HD patients.