LAUSR

Suboptimal racial categorization potentially introduces bias in epidemiological analysis and interpretation, making it difficult to appropriately measure factors leading to racial health disparities. As part of an analysis focused on predictors of experiencing HIV-related stigma among men who have sex with men living with HIV in San Francisco, we struggled with the most appropriate ways to categorize people who reported more than one racial identity, and aimed to explore the implications of different methodological choices in this analysis. We performed three different multivariable linear regression models each utilizing a different approach to racial categorization: the "Multiracial", "Othering", and "Hypodescent" models. We estimate an adjusted risk difference in mean score for reported frequency of experiencing HIV-related stigma on a 4-point scale, adjusting for age, race, gender identity, injection history, housing, mental health concerns, and viral load. Use of a hypodescent model for racial categorization led to a shift in the point estimate through the null for Blacks/African Americans, and improved precision for that group. However, it obscured the association of increased stigma and race for multiracial people, compared to monoracial counterparts. We conclude that methodological decisions related to racial categorization of participants can dramatically affect race-related study findings in predictor regression models.