Health Disparities
Unveiling Disparities: Examining Differential item functioning’s Impact on Racial Health Equity Among White and Black Populations Ester Villalonga-Olives* Ester Villalonga-Olives Yun-Yi Pan Yusuf Ransome Ester Villalonga-Olives Abdolvahab Khademi
Introduction: Due to persistent structural oppression against Black individuals in the United States, the relationship between social capital and health for Blacks differs compared to Whites. Results from cross-sectional data analyses suggested that social capital indicators are not psychometrically equivalent by race. Our aim is to test longitudinal measurement invariance (MI) and differential item functioning (DIF) of social capital indicators by race and education.
Methods: Longitudinal data were collected from the Midlife in the United States study (n= 7700; years 1995-2016). Scales included social capital indicators measuring Contributions to Community and Community Involvement. We used structural equation modeling (SEM) and item response theory (IRT) to test longitudinal MI and DIF by race and education.
Results: We found violation of longitudinal and multi-group MI at configural and metric levels in both scales, where factor structure and indicator loadings failed to sustain over time. IRT results showed DIF in both scales in specific indicators, such as ‘Many people come for advice,’ indicating a consistent advance of one racial group over the other even if both groups had the same levels of social capital (P(χ2,2) = 0.00). When we investigated race and education interaction, all items in the Contributions to Community and some of the Community Involvement items showed DIF.
Conclusion: We found lack of invariance in several social capital items. These findings suggest that attempts to compare social capital between Black and White people may contain biases that should be acknowledged in research. To correct the issue, future work on social capital measures should involve processes that evaluate the assumptions behind each question (for existing data) and involve key stakeholders from racial and underrepresented communities when creating new items.