An Objective Risk Score Predicts Waitlist Mortality with Higher Accuracy Than 6-Status Ranking at Listing

Abstract

<h3>Purpose</h3> The US has a new 6-status ranking system for heart transplant candidates that, in contrast to kidney, liver and lung allocation, relies primarily on subjective treatment decisions as a surrogate for medical urgency. Incorporating objective patient characteristics could create a more equitable system and save more lives. <h3>Methods</h3> We utilized Scientific Registry of Transplant Recipients data on 32,567 US adult heart-only transplant candidates divided into pre-policy training (2010 - 2017) and post-policy test (11/1/2018 - 3/1/2020) datasets. We generated risk scores from a Cox Proportional Hazards (CPH) model fit to the outcome of death without a transplant. In addition to treatment variables, we included age, year of listing, height, weight, GFR, cardiac index, pulmonary artery mean pressure, pulmonary capillary wedge pressure, peak oxygen capacity, diagnosis, blood type, diabetes, and history of tobacco use, malignancy, or stroke. We compared the performance of this model to the performance of a model that includes only a candidate's 6-status ranking at listing using 1) Uno's c-index and 2) Kaplan-Meier of risk score groups. <h3>Results</h3> In the post-policy test dataset, the risk score model had higher discrimination than the 6-status based model (c-index 0.71 vs. 0.63, p < 0.01). Status 5 candidates had significantly lower survival without transplant than Status 4 candidates (p<0.01 by log-rank test, <b>Fig</b>). The risk score identified a highest risk group with greater urgency than Status 1 candidates (median survival 96 days vs 129 days, p = 0.05). An entirely objective model without treatment variables also outperformed the 6-status rankings (c-index 0.69 vs. 0.63, p < 0.01). <h3>Conclusion</h3> A risk score incorporating objective patient characteristics can rank order heart transplant candidates more effectively than the 6-status ranking at listing. Heart allocation systems should use objective measurements of physiological function to identify medically urgent candidates.