Despite advances in artificial intelligence (AI) in glaucoma prediction, most models do not consider biases and fairness regarding sex, race, or ethnicity. Our latest study titled "The Impact of Race, Ethnicity, and Sex on Fairness in Artificial Intelligence for Glaucoma Prediction Models", aims to address this gap by examining the impact of these sensitive attributes on developing fair AI models that predict the need for incisional glaucoma surgery. We found AI models predicting glaucoma surgery progression demonstrated bias related to sex, race, and ethnicity.  The choice to include or exclude sensitive attributes had varying impacts on performance and fairness depending on the test population. Before deploying AI models in healthcare, it’s critical to evaluate fairness in the target population.

In recent years, calls for greater diversity in genetic data have come from multiple sources. While the importance of diverse data is recognized, there has often been ambiguity around what types of diversity matter and how to achieve them. In a commentary published in Nature Genetics in Sept. 2024, Dr. Hernandez-Boussard, and others from The Global Alliance for Genomics and Health (GA4GH), have developed a policy framework that guides researchers on addressing diversity in their data collection and research processes. The framework encourages researchers to focus on how diversity can help achieve specific goals, emphasizing that diversity should be a means to further scientific equity and improve health outcomes. It also stresses that diversity is not necessarily about statistical representativeness, but rather about including necessary variations (genetic, geographical, social) that align with the objectives of the research. Researchers are encouraged to consult bioethics experts and the communities impacted by their work, ensuring that diversity is pursued in a meaningful, ethically guided manner.