An explainable machine learning approach using contemporary UNOS data to identify patients who fail to bridge to heart transplantation.

Autor:	Mardini MT; Department of Health Outcomes and Biomedical Informatics, College of Medicine, University of Florida, Gainesville, FL, United States., Bai C; Department of Health Outcomes and Biomedical Informatics, College of Medicine, University of Florida, Gainesville, FL, United States., Bledsoe M; Division of Cardiovascular Medicine, University of Florida, Gainesville, FL, United States., Shickel B; Department of Health Outcomes and Biomedical Informatics, College of Medicine, University of Florida, Gainesville, FL, United States., Al-Ani MA; Division of Cardiovascular Medicine, University of Florida, Gainesville, FL, United States.
Jazyk:	angličtina
Zdroj:	Frontiers in cardiovascular medicine [Front Cardiovasc Med] 2024 May 20; Vol. 11, pp. 1383800. Date of Electronic Publication: 2024 May 20 (Print Publication: 2024).
DOI:	10.3389/fcvm.2024.1383800
Abstrakt:	Background: The use of Intra-aortic Balloon Pump (IABP) and Impella devices as a bridge to heart transplantation (HTx) has increased significantly in recent times. This study aimed to create and validate an explainable machine learning (ML) model that can predict the failure of status two listings and identify the clinical features that significantly impact this outcome. Methods: We used the UNOS registry database to identify HTx candidates listed as UNOS Status 2 between 2018 and 2022 and supported with either Impella (5.0 or 5.5) or IABP. We used the eXtreme Gradient Boosting (XGBoost) algorithm to build and validate ML models. We developed two models: (1) a comprehensive model that included all patients in our cohort and (2) separate models designed for each of the 11 UNOS regions. Results: We analyzed data from 4,178 patients listed as Status 2. Out of them, 12% had primary outcomes indicating Status 2 failure. Our ML models were based on 19 variables from the UNOS data. The comprehensive model had an area under the curve (AUC) of 0.71 (±0.03), with a range between 0.44 (±0.08) and 0.74 (±0.01) across different regions. The models' specificity ranged from 0.75 to 0.96. The top five most important predictors were the number of inotropes, creatinine, sodium, BMI, and blood group. Conclusion: Using ML is clinically valuable for highlighting patients at risk, enabling healthcare providers to offer intensified monitoring, optimization, and care escalation selectively. Competing Interests: The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest. (© 2024 Mardini, Bai, Bledsoe, Shickel and Al-Ani.)
Databáze:	MEDLINE
Externí odkaz:	Zobrazit plný text záznamu