REVIEW article
Front. Digit. Health
Sec. Health Informatics
Volume 7 - 2025 | doi: 10.3389/fdgth.2025.1633458
Applications of generative artificial intelligence in outcome prediction in intensive care medicine - a scoping review
Provisionally accepted
- 1Medical University of Vienna, Vienna, Austria
- 2University of Portsmouth, Portsmouth, United Kingdom
- 3Ludwig Boltzmann Institute for Arthritis and Rehabilitation, Vienna, Austria
- 4Queen Alexandra Hospital, Portsmouth, United Kingdom
- 5Uniersity of Portsmouth, Portsmouth, United Kingdom
Select one of your emails
You have multiple emails registered with Frontiers:
Notify me on publication
Please enter your email address:
If you already have an account, please login
You don't have a Frontiers account ? You can register here
When a patient survives the first 24 hours in intensive care, outcome prediction is crucial for further treatment decisions. As recent advances have shown that Artificial Intelligence (AI) outperforms clinicians in prognostication, and especially generative AI has developed rapidly in the past ten years, this scoping review aimed to explore the use of generative AI models for outcome prediction in intensive care medicine. Of the 481 records found in the search, 119 studies were subjected to abstract screening and, when necessary, full-text review for eligibility assessment.Twenty-two studies and two review articles were finally included. The studies were categorized into three prototypical use cases for generative AI in outcome prediction in intensive care: (i) data augmentation, (ii) feature generation from unstructured data, and (iii) prediction by the generative model. In the first two use cases, the generative models worked together with downstream predictive models. In the third use case, the generative models made the predictions themselves. The studies within data augmentation either fell into the area of compensation for class imbalances by producing additional synthetic cases or imputation of missing values. Overall, Generative Adversarial Network (GAN) was the most frequently used technology (8/22 studies; 36%), followed by Generative Pretrained Transformer (GPT) (7/22 studies; 32%). All publications except one were from the last four years. This review shows that generative AI has immense potential in the future, and continuous monitoring of new technologies is necessary to ensure that patients receive the best possible care.
Keywords: Large Language Model, generative adversarial network, Critical Care, Survival, Mortality, Comorbidity
Received: 22 May 2025; Accepted: 15 Jul 2025.
Copyright: © 2025 Stamm, Bader-El-Den, McNicholas, Briggs and Zhao. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.
* Correspondence: Tanja Stamm, Medical University of Vienna, Vienna, Austria
Disclaimer: All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article or claim that may be made by its manufacturer is not guaranteed or endorsed by the publisher.