Published in Science · View the paper (DOI)
Clinical prediction models for schizophrenia treatment outcomes perform poorly on independent trial data, despite high accuracy within their training dataset. The study highlights three key reasons for this limitation, emphasizing the need for more stringent validation methods in machine learning approaches.
Coverage from 2 institutions
- Machine learning clinical prediction models fail to generalize across trial data American Association for the Advancement of Science (AAAS) · Jan 11, 2024 · first to report
- Clinical predictive models created by AI are accurate but study-specific University of Cologne · Jan 12, 2024