Table 1.

Summary of publications describing artificial intelligence application in predicting heart transplant outcomes.

Author	Title	Data source used	Primary outcome	Best model performance	Study conclusion
Oztekin (16)	Predicting the graft survival for heart-lung transplantation patients: An integrated data mining methodology	Unified Network for Organ Sharing (UNOS), 1987–2008	Combined heart-lung transplantation survival, and methods of UNOS database mining	The neural network achieved the highest accuracy of 0.859.	Model uncovered relationships among the survival-related variables. Domain-expert input improved performance.
Delen (7)	A machine learning-based approach to prognostic analysis of thoracic transplantations	United Network for Organ Sharing (UNOS), 1987–2009	Survival time	The support vector machine model with a radial basis Kernel function produced the best fit with anR2value of 0.879 and MSE of 0.023.	Integrated machine learning is more effective in developing the Cox survival models than the traditional methods.
Nilsson (10)	The international heart transplant survival algorithm (IHTSA): A new model to improve organ sharing and survival	International Society for Heart and Lung Transplantation (ISHLT) registry, 1994–2010	1-year mortality	Artificial neural network model had anAUC of 0.650, 95% CI: 0.640–0.655.	The model predicts mortality and also estimates the expected benefit to the individual patient, and donor-recipient compatibility.
Medved (17)	Improving prediction of heart transplantation outcome using deep learning techniques	United Network for Organ Sharing (UNOS), 1997–2011	1-year mortality	International Heart Transplantation Survival Algorithm (IHTSA), based on deep learning, had AUC of 0.654, 95% CI: 0.629–0.679.	The IHTSA model was superior to Donor Risk Index (DRI), Risk Stratification Score (RSS), and Index for Mortality Prediction After Cardiac Transplantation (IMPACT).
Miller (5)	Predictive abilities of machine learning techniques may be limited by dataset characteristics: insights from the UNOS database	Unified Network for Organ Sharing (UNOS), 1987–2014	1-year mortality	The neural network model had the highest AUC of 0.66.	The prognostic abilities of machine learning techniques may be limited by quality of the clinical dataset.
Agasthi (11)	Machine learning helps predict long-term mortality and graft failure in patients undergoing heart transplant	International Society of Heart and Lung Transplant (ISHLT) registry data, 2000–2017	All-cause mortality and graft failure at 5 years after HTx	A gradient-boosted machine model had an AUC of 0.717, 95% CI: 0.696–0.737 for 5-year mortality and 0.716, 95% CI: 0.696–0.736 for graft failure.	This model would likely function as a predictive algorithm to estimate the risk of 5-year mortality and graft failure in each donor–recipient match.
Hsich (15)	Heart transplantation: An in-depth survival analysis	Scientific Registry of Transplant Recipients (SRTR), 2004–2018	Factors that determined survival	AI was used to identify variables that are associated with mortality, classified into early, late, and constant phases.	Transplantation from ECMO should consider end-organ function to reduce early post-transplantation mortality.
Ayers (18)	Using machine learning to improve survival prediction after heart transplantation	United Network for Organ Sharing (UNOS), 2000–2019	1-year survival	The final ensemble model had an AUC of 0.764, 95% CI: 0.745–0.782	Modern ML techniques can improve risk prediction in OHT compared to traditional approaches.
Zhou (13)	Prediction of 1-year mortality after heart transplantation using machine learning approaches: A single-center study from China	Local dataset from Wuhan union hospital, 2015–2018	1-year mortality	Random Forest model achieved the best AUC of 0.801 and gradient boosting machine showed the best sensitivity of 0.271	The model identifies high-risk HTx recipients, informs a personalized therapeutic plan, and reduces organ wastage
Kainuma (19)	Predictors of 1-year outcome after cardiac re-transplantation: Machine learning analysis	United Network for Organ Sharing (UNOS), 2000–2009	1-year survival predictors post heart re-transplantation	Random survival forests-ranked variable importance to evaluate the association with mortality	Short-term survival was related to liver function, and long-term survival was related to obesity and mechanical ventilation.
Mete (14)	Predicting post-heart transplant composite renal outcome risk in adults: A machine learning decision tool	Organ Procurement and Transplantation Network (OPTN), 2000–2019	Dependence on chronic dialysis, GFR < 20 ml/min per 1.73 m2, or having received a kidney transplant	The Random Forest model had AUC of 0.70, 95% CI 0.67–0.74 for the composite primary outcome.	The Model was used to create a validated web-based decision tool for assessing renal outcomes post HTx.
Miller (20)	Temporal shift and predictive performance of machine learning for heart transplant outcomes	United Network of Organ Sharing (UNOS), 1994–2016	1-year all-cause mortality	Random Forest model had an AUC of 0.893, CI: 0.889–0.897.	While AI models can predict transplant mortality, they are limited by temporal shifts in patient and donor selection.
Wang (12)	Comparison of four machine learning techniques for prediction of intensive care unit length of stay in heart transplantation patients	Local data from Wuhan Union Hospital, 2017–2020	Length of ICU stay post heart transplantation	The eXtreme Gradient Boosting (XGBoost) algorithm presented significantly better predictive performance (AUC 0.88).	Using the XGBoost classifier with HTx patients can facilitate precision medicine and best allocation of medical resources.

strong recommendation, weak recommendation/promising but not ready to implement, and recommended against AI usage, as suggested by the respective paper.