# Development and validation of mortality prediction models for heart transplantation using nutrition-related indicators: a single-center study from China

## TL;DR

This study developed a model to predict mortality after heart transplantation using nutrition-related indicators to help identify high-risk patients.

## Contribution

A novel mortality prediction model for heart transplant patients using nutrition-related indicators was developed and validated in a Chinese cohort.

## Key findings

- The model included age, nutritional risk index, serum creatine, and triglyceride as key predictors.
- The model showed improved discrimination and reclassification compared to simpler models using only age or nutritional risk index.
- The model's AUC was 0.76 in the derivation cohort and 0.71 in the validation cohort.

## Abstract

We sought to develop and validate a mortality prediction model for heart transplantation (HT) using nutrition-related indicators, which clinicians could use to identify patients at high risk of death after HT.

The model was developed for and validated in adult participants in China who received HT between 1 January 2015 and 31 December 2020. 428 subjects were enrolled in the study and randomly divided into derivation and validation cohorts at a ratio of 7:3. The likelihood-ratio test based on Akaike information was used to select indicators and develop the prediction model. The performance of models was assessed and validated by area under the curve (AUC), C-index, calibration curves, net reclassification index, and integrated discrimination improvement.

The mean (SD) age was 48.67 (12.33) years and mean (SD) nutritional risk index (NRI) was 100.47 (11.89) in the derivation cohort. Mortality after HT developed in 66 of 299 patients in the derivation cohort and 28 of 129 in the validation cohort. Age, NRI, serum creatine, and triglyceride were included in the full model. The AUC of this model was 0.76 and the C statistics was 0.72 (95% CI, 0.67–0.78) in the derivation cohort and 0.71 (95% CI, 0.62–0.81) in the validation cohort. The multivariable model improved integrated discrimination compared with the reduced model that included age and NRI (6.9%; 95% CI, 1.8%–15.1%) and the model which only included variable NRI (14.7%; 95% CI, 7.4%–26.2%) in the derivation cohort. Compared with the model that only included variable NRI, the full model improved categorical net reclassification index both in the derivation cohort (41.8%; 95% CI, 9.9%–58.8%) and validation cohort (60.7%; 95% CI, 9.0%–100.5%).

The proposed model was able to predict mortality after HT and estimate individualized risk of postoperative death. Clinicians could use this model to identify patients at high risk of postoperative death before HT surgery, which would help with targeted preventative therapy to reduce the mortality risk.