Assessing the potential for racial bias in hair analysis for cocaine: Examining the relative risk of positive outcomes when comparing urine samples to hair samples

This article examines the conjecture that hair analysis, performed to detect cocaine use or exposure, is biased against African Americans. It does so by comparing the outcomes of 33,928 hair and 105,792 urine samples collected from both African American and white subjects. In making this comparison the analysis seeks to determine if there is a departure in rates of positive and negative outcomes when comparing the results of hair analysis for cocaine to the results from urinalysis for cocaine by racial group. It treats urine as an unbiased test. It compares both the relative ratios of positive outcomes when comparing the two groups and it calculates the relative risk of outcomes for each group for having positive or negative outcomes. The findings show that the ratios of each racial group are effectively same for hair and urine assays, and they also show that the relative risk and risk estimates for positive and negative outcomes are the same for both racial groups. Considering all samples, the cocaine positive risk estimate for the hair samples comparing the two racial groups is 3.28 and for urinalysis the risk estimate is 3.10 (Breslow-Day χ2 .250, 1 df, p = 0.617) a non-significant difference in risk. For pre-employment samples, the cocaine positive risk estimate for the hair samples comparing the two racial groups is 3.10 and for urinalysis the risk estimate is 2.90 (Breslow-Day χ2 .281, df = 1, p = 0.595), also a non-significant difference in risk.