Coverage that Goes the Distance:Extending Distance Metrics and Clustering Methods to Assess Access to HIV Preventative Care 330 views

In this work, we extend clustering methodologies to find groups in healthcare plan data, assessing the extent to which specific coverage practices facilitate or restrict effective preventative care for HIV. In doing so, we rely on Gower’s underexplored ideas about weighting in distance metrices to create a procedure that handles nested dependencies. Overall, the distance metric chosen effectively translates our priorities based on theory into a form that works well with common clustering algorithms. Running trials with several algorithms, we find convergent structures, settling on a hierarchical approach with three distinct clusters. The clusters exhibit distinct contrasts on how plans cover specific benefits that allow for ordinal interpretations in terms of plan restrictiveness. Broad level interpretation of the output suggests that, across the United States, monthly premiums are not related to plan restrictiveness, prior authorization is less likely among plans where individuals accept higher out of pocket payments for care, state-wide approaches affect what care residents can access, and variability in markets within states, specifically Texas, reflect patterns of discrimination toward individuals at risk of HIV.

MA (Master of Arts)

Clustering Algorithms; Health Equity; Affordable Care Act

English

2021-05-13