Study Reveals AI Bias Evaluation Gaps in U.S. Hospitals
Healthcare IT News
|
Contributed by: Kate Gamble
Summary
A study from the University of Minnesota School of Public Health reveals significant disparities in the evaluation of AI predictive models among U.S. hospitals, with only 44% assessing these tools for bias despite two-thirds of hospitals utilizing them. The research indicates that hospitals with more financial resources and technical expertise are better equipped to develop and evaluate AI tools, creating potential inequities in patient care. Many hospitals rely on AI for predicting health outcomes and optimizing services, yet under-resourced institutions face challenges in effectively evaluating these tools. Assistant Professor Paige Nong urges the adoption of predictive model labels from the Assistant Secretary for Technology Policy to enhance bias assessments and calls for localized evaluations to ensure AI tools serve the needs of diverse patient populations.
