Table 1.

Related work in the literature.

Study	Prediction methods	Sample	Sample size, n	Outcomes	AUROCa, best (95% CI)	AUROC, best (outcome)
Gupta et al [11], 2022	Stepwise logistic regression, gradient boosting, and random forest	Pediatric Heart Transplant Society database; aged <18 years; heart transplantation; discernible discharge date; transplanted between January 2005 to December 2018	4414	Prolonged length of stay (>30 days) after transplantation	0.750 (0.720-0.780)	N/Ab
Killian et al [15], 2021	Logistic regression, multilayer perceptron, sequential minimal optimization algorithm polynomial kernel, random forest, and deep learning	UNOSc data for a single transplant center; aged 0-18 years; heart transplant; transplanted between 1988 and May 31, 2017	193	Hospitalization owing to rejection over 1-, 3-, and 5-year posttransplant periods	N/A	0.740 (5‐year hospitalization)
Miller et al [12], 2019	Artificial neural networks, classification and regression trees, and random forest	UNOS data; aged <18 years; heart transplant; transplanted between January 2006 and December 2016	2802	Mortality over 1-, 3-, and 5-year posttransplant periods	N/A	0.720 (1‐year mortality)
Miller et al [22], 2022	Random forest, XGBoostd, and L2 regularized logistic regression	UNOS data; aged <18 years; heart transplant; transplanted between January 1994 and December 2016	8349	1-year and 90-day all-cause mortality	0.836 (0.823-0.849)	N/A

^aAUROC: area under the receiver operating characteristic curve.

^bN/A: not applicable.

^cUNOS: United Network for Organ Sharing.

^dXGBoost: extreme gradient boosting.