A recent study conducted by Yale University has raised concerns about racial bias in artificial intelligence (AI) when it comes to interpreting radiology reports. The research, published in the journal Clinical Imaging, focused on the use of Open AI’s GPT Chat 3.5 and GPT 4.0 in simplifying medical language from radiology reports.
The study involved providing over 700 radiology reports to ChatGPT to be simplified for easier understanding. What researchers found was that when the AI model was informed of the patient’s race along with the text, it adjusted the reading grade level of the simplified response based on the shared race information.
Dr. Melissa Davis, co-author of the study and vice chair of Medical Informatics at Yale, noted that white and Asian patients typically received responses with a higher reading grade level compared to Black, American Indian, or Alaskan Native patients. This disparity in response readability based on race highlights the potential risks of including racial information as a socio-economic determinant in AI healthcare models.
While AI technology has the potential to revolutionize healthcare delivery, experts caution against disclosing patients’ race as a factor in determining how medical information is presented. Instead, factors such as education level or age could be more appropriate considerations for tailoring the communication of medical information.
In the realm of healthcare innovation, AI is being increasingly utilized. Hartford HealthCare recently launched the Center for AI Innovation in Healthcare in collaboration with MIT and Oxford University, aiming to transform healthcare delivery for improved access, affordability, equity, and excellence. Similarly, researchers at UConn Health are leveraging AI to diagnose lung cancer at earlier stages through innovations like the Virtual Nodule Clinic.
As AI continues to expand its presence in healthcare, it is crucial to address issues of bias and equity to ensure that these technologies serve all patients effectively and ethically. The findings from the Yale study underscore the importance of thoughtful consideration when integrating AI into healthcare practices.
