Online Program

WITHDRAWN - Deriving County-Level Measures of Implicit Racial Bias and Examining its Association with Healthcare Disparities (307871)

Keywords: health disparity, implicit bias, Project Implicit, county-level estimation

Disparities in health care are of great concern, with the potential for unconscious, implicit bias to influence disparity. In this paper, we estimate the contribution of racial implicit bias in explaining healthcare disparities, with the hypothesis that counties with greater negative attitudes towards Blacks than Whites also exhibit greater healthcare disparities. Since 2002, Project Implicit has collected over 2 million responses over the internet using an Implicit Association Test (IAT), which is a computer-based measure that relies on differences in response latency to reveal implicit bias. To obtain county-level IAT estimates, we needed to address issues of small or zero sample sizes, and bias stemming from non-representative internet samples. We applied multi-level modeling approach involving re-weighting and shrinkage to leverage county-level predictors (e.g. percent Black, percent below poverty limit), and to borrow information across counties to produce representative IAT estimates. We have produced estimates of implicit bias for every U.S. county, and examine its validity and association with healthcare disparities.