The rapid evolution of artificial intelligence (AI) has ushered in a new era of healthcare tools, offering promising solutions to a myriad of medical challenges. However, to prevent the exacerbation of existing health disparities, experts emphasize the imperative of utilizing more inclusive and representative data in the development of these tools.
A collaborative effort led by researchers from Oxford University's Nuffield Department of Orthopaedics, Rheumatology and Musculoskeletal Sciences (NDORMS), University College London, and the Center for Ethnic Health Research, supported by Health Data Research UK, marks a significant step towards addressing this crucial issue. Their pioneering study, published in Scientific Data, delves into the intricacies of ethnicity data within the National Health Service (NHS) for the first time, aiming to mitigate bias in AI health prediction models.
Accessing de-identified data on ethnicity and other patient characteristics from general practice and hospital health records, the researchers meticulously analyzed the ethnicity data of over 61 million individuals in England, encompassing over 250 different ethnic groups. This comprehensive dataset, compiled into a research-ready database, sheds light on the complexities of ethnicity recording within the healthcare system.
Key findings reveal significant gaps and discrepancies in ethnicity data recording, with approximately 1 in 10 patients lacking ethnicity records, and around 12% exhibiting conflicting ethnicity codes in their patient records. Such insights underscore the pressing need for more accurate and representative data to drive equitable healthcare outcomes.
Sara Khalid, Associate Professor of Health Informatics and Biomedical Data Science at NDORMS, underscores the critical role of representative data in mitigating health inequities, particularly in the wake of the COVID-19 pandemic. Khalid emphasizes that AI-based healthcare technologies heavily rely on the quality of input data, and a lack of representativeness can lead to biased models and erroneous health assessments.
Moving forward, the researchers plan to leverage these findings to develop more equitable AI and machine learning tools tailored to diverse patient populations. By harnessing the power of detailed ethnicity data, the aim is to foster healthcare technologies that prioritize inclusivity and accuracy, ultimately advancing the health and well-being of all individuals.
Professor Cathie Sudlow, Chief Scientist at Health Data Research UK, underscores the transformative potential of these insights, affirming their relevance to healthcare stakeholders across diverse demographics. Sudlow emphasizes that leveraging insights from comprehensive health data analysis can empower informed decision-making and drive positive health outcomes for individuals of all backgrounds.
As the healthcare landscape continues to evolve, initiatives like this serve as catalysts for innovation, paving the way towards a more equitable and inclusive healthcare ecosystem.
More: https://medicalxpress.com/news/2024-02-bias-health-ai-tools.html
