Blind/Disability and Intersectional Biases in E-Health Records (EHRs) of Diabetes Patients: Building a Dialogue on Equity of AI/ML Models in Clinical Care

使用AI/ML分析工具来预测疾病风险、发病和进展以及治疗结果

The use of AI/ML analytical tools to predict disease risk, onset and progression, and treatment outcomes is growing and holds promise for improving health outcomes for marginalized health disparities population. Yet, there is indication that people with disabilities—the largest health disparities group in the US—will not be able to reap the benefits of these scientific advancements. In the Parent R01, we explore the views of adults with vision, hearing, and mobility disabilities on trust in and trustworthiness of precision medicine research (PMR), a major training dataset for AI/ML applications. Community members in this R01 and the PI’s prior work identified disability bias in clinical and research settings as a key barrier to trust and participation in PMR. These findings are prominent for blind adults who both express the highest interest in participating in PMR and concern about disability bias in medical interactions. Studies also show that clinicians view blind patients as incompetent, regardless of abilities, and as difficult patients, despite structural issues that compromise the health outcomes of blind patients (e.g., inaccessible drug labels). Insofar as disability bias is presented in the medical documentation of blind patients, the use of such data in AI/ML models can affect care and reproduce, even worsen, existing health disparities. The worry is amplified for blind patients encountering intersectional marginalization, for whom health disparities are compounded. The prevalence of preventable blindness (e.g., diabetic retinopathy, a common and leading cause of blindness) is disproportionately high among women and marginalized racial/ethnic communities, especially Black/African American individuals, but also that gender and racial biases exist in electronic medical records (EHRs). Assessing whether disability bias—as an independent and intersectional factor—is presented in EHRs is thus crucial for AI/ML models to develop equitable analytical tools to improve health outcomes for all. Yet, no study has explored disability bias in EHRs, major training dataset for AI/ML models, or assessed how disability bias compounds racial and gender biases that are embedded in EHRs. The proposed study is led by a new interdisciplinary research team and uses an intersectionality framework and disability community-engaged model to begin closing the gaps. We will: 1) Develop, validate, and disseminate reproducible phenotype definitions for diabetes-related blindness and create cohorts for analyses using the EHRs of diabetes patients (2016-22) from a large urban medical center serving highly diverse racial/ethnic populations; 2) Identify and evaluate a list of blind/disability-related negative patient descriptors in clinical documentation; and 3) Assess the use of disability biased language in EHRs of diabetes patients (blind, nonblind) and if negative descriptors in EHRs varied intersectionally (men/women, Black/White). This project has the potential to inform equitable AI/ML models in clinical care, improve health outcomes of an often invisible but large and growing health disparity population, and build a dialogue on disability ethics and equity of AI/ML among clinicians, data scientists, blind adults, and ELSI researchers.