Advancing Equity in Machine Learning for Healthcare (52:54)

In this talk, we characterise and mitigate the impact of imperfect data on machine learning models deployed in healthcare. We address two ways in which data can be flawed: imperfect labels and coarse demographics. First, we develop a method to correct for imperfect labels in the form of underdiagnosis between demographic cohorts. We then show how coarse race data obscures disparities across more granular race groups, suggesting existing algorithmic audits may significantly underestimate racial disparities in performance. The talk concluded with a prospective discussion of how we can better select problems to promote equity in healthcare.