. Author manuscript; available in PMC: 2016 Apr 26.

Published in final edited form as: J Epidemiol Community Health. 2014 Apr 11;68(8):728–733. doi: 10.1136/jech-2013-203741

Table 3.

Social inequalities in mortality explained by biased and unbiased VIQ test scores

	HR	Excess risk explained by VIQ scores uncorrected for DIF	Excess risk explained by VIQ scores corrected for DIF
SES factors adjusted for Gallup-Thorndike test
Occupational prestige	1.66 (1.34 to 2.05)	0.29 (0.11 to 0.47)	0.07 (0.01 to 0.13)
Education	1.92 (1.47 to 2.52)	0.40 (0.14 to 0.66)	0.29 (0.11 to 0.48)
Income	2.05 (1.71 to 2.45)	0.15 (0.05 to 0.24)	0.04 (0.00 to 0.07)
Minority Race/ethnicity	1.74 (1.52 to 1.98)	0.11 (0.05 to 0.17)	0.02 (0.00 to 0.03)
SES factors adjusted for WAIS-R Similarities Test
Socioeconomic Index	1.94 (1.16 to 3.21)	0.13 (−0.27 to 0.72)	0.04 (−0.17 to 0.32)
Education	1.65 (0.90 to 3.01)	0.36 (−2.98 to 4.70)	0.09 (−1.30 to 1.52)
Income	2.34 (1.56 to 3.50)	0.04 (−0.23 to 0.27)	0.01 (−0.14 to 0.15)
Minority	1.44 (1.01 to 2.06)	0.09 (−0.26 to 0.59)	0.02 (−0.09 to 0.21)

Note. VIQ=verbal IQ. The HR is the Relative Index of Inequalities, or ratio of risk for those at the bottom of each distribution (ie, worst scoring or most disadvantaged) compared to those at the top. Percent of RII explained=excess risk explained by adjustment factor, computed as (HR_unadjusted−HR_adjusted)/(HR_unadjusted−1) with 95% CIs based on bootstrap percentiles over 1000 replicates of 1000 observations. Same n=9381 for Gallup Thorndike and 2444 for WAIS-R similarities test. Results from separate Cox proportional hazard models for each factor, with age as time scale, all adjusted for gender as time-varying covariate due to non-proportional hazard. Minority status is also time varying, with estimates presented at age 50.