Real World Performance of the Model for End-Stage Liver Disease Score Across Different Races and Ethnicities

Abstract

Background and aims: The MELD score assesses liver disease severity and guides liver allocation in the United States. This study aims to address the knowledge gap regarding potential disparities in the performance of MELD-Na and MELD 3.0 scores across races and ethnicities.

Methods: A retrospective cohort study of Asian, White, African American, and Hispanic patients with cirrhosis from MIMIC-IV (Beth Israel Deaconess Medical Center) and Emory University databases. The performance of the MELD scores in predicting a composite outcome of death or liver transplant (LT) during the hospital stay or within 90 days was analyzed using AUROC for discrimination, and calibration curves for calibration.

Results: A total of 4,306 patients with cirrhosis from MIMIC-IV and 5801 patients from Emory were included. Across both datasets, no significant differences were found in the discriminative ability of both scores to predict in-hospital death or LT across races and ethnicities (p > 0.05). A small group of Asian patients influenced score performance differences at 90 days in MIMIC-IV. Calibration plots showed both scores tend to overpredict outcomes at higher values, particularly MELD-Na in MIMIC-IV among African Americans.

Conclusions: Both scores demonstrated consistent discriminative ability across major ethnic/racial groups. Both scores predicted outcomes better in Emory, a high-volume LT center, than in MIMIC-IV, suggesting site-specific differences and the Emory cohort's closer resemblance to populations used for MELD derivation. These findings suggest disparities in cirrhosis care may relate to socioeconomic factors rather than score performance.

Keywords: Cirrhosis; Disparities; Equity; MELD.