Related Experiment Video
Updated: Jun 5, 2025
Author Spotlight: Enhancing Graft Viability Assessment Through Quantitative Metrics and Innovative Reservoir Systems
Published on: August 2, 2024
Predictive Performance of Artificial intelligence Models on Heart and Lung Posttransplant Health Outcomes: A
George C Sargiotis1, Theodoros N Sergentanis, Elpida Pavi
1From the Department of Public Health Policy, University of West Attica, Athens, Greece.
Objectives:
The efficacy and capacity of artificial intelligence models to predict posttransplant health complications have been disputed over the past few years. In this systematic review, we assessed the performance of different artificial intelligence models in predicting health outcomes after heart and lung transplantations.
Materials And Methods:
We researched online databases. We gathered and analyzed data on performance metrics of artificial intelligence applications in heart and lung transplantations. In addition, we conducted a risk of bias assessment.
Results:
Of the 122 initial studies that we gathered, 15 were included in the analyses. The artificial intelligence models showed high performance, with metrics for discrimination such as the area under the receiver operating curve ranging from 0.620 to 0.921 and good calibration for long-term outcomes. Random forest and extreme gradient boosting models outperformed other models, particularly traditional linear models. North American, White people were the predominant subsample, and pediatric populations were excluded from the analysis. Most studies demonstrated a high overall risk of bias, whereas applicability to research questions showed a low risk.
Conclusions:
Supervised machine learning models performed well in predicting posttransplant health outcomes. However, biases and ethical concerns on the application of artificial intelligence models in transplantation must be considered to draw safe conclusions.
