Skip to main content
. Author manuscript; available in PMC: 2023 Sep 5.
Published in final edited form as: Drug Alcohol Depend. 2023 Apr 25;247:109898. doi: 10.1016/j.drugalcdep.2023.109898

Table 1.

Multinomial logistic regression of workplace policy on the first positive drug test, National Survey on Drug Use and Health 2015–2019.

Fired (n = 17,615;
47.51%)
Referred for Treatment/Counseling (n = 9519; 18.2%)

RRR 95% CI RRR 95% CI
Gender
 Male 1.176 1.126–1.227 1.216 1.131–1.308
 Female 1.000 - 1.000 -
Age
 18–25 1.410 1.317–1.509 0.512 0.457–0.573
 26–34 1.304 1.223–1.390 0.758 0.706–0.813
 35–64 1.000 - 1.000 -
Ethnoracial identity
 Black 1.563 1.445–1.689 1.419 1.256–1.603
 Hispanic 1.372 1.258–1.496 0.797 0.727–0.879
 White 1.000 - 1.000 -
Educational attainment
 Less than high school 2.458 2.251–2.683 1.239 1.074–1.430
 High school 2.110 1.982–2.246 1.479 1.352–1.618
 Some college 1.708 1.598–1.827 1.401 1.277–1.537
 College or higher 1.000 - 1.000 -
Annual household income
 <$20,000 1.255 1.136–1.386 0.560 0.474–0.663
 $20,000–39,999 1.203 1.122–1.290 0.622 0.561–0.690
 $40,000–74,999 1.114 1.049–1.184 0.785 0.720–0.856
 $75,000+ 1.000 - 1.000 -
Employment status
 Full-time 1.000 - 1.000 -
 Part-time 0.814 0.756–0.876 0.675 0.608–0.749
Marriage status
 Married 1.000 -
 S/D/W 0.991 0.904–1.085 1.114 1.006–1.234
 Never married 0.887 0.828–0.951 0.911 0.818–1.013
Urbanicity
 CBSA w/pop. > 1 mil 1.000 - 1.000 -
 CBSA w/pop. < 1 mil 1.199 1.135–1.267 1.104 1.032–1.180
 Non-CBSA 1.498 1.343–1.671 1.254 1.076–1.461
Illicit drug use disorder
 No 1.000 - 1.000 -
 Yes 0.947 0.801–1.119 0.798 0.628–1.014

Notes. The reference category is other (n = 24,926; 41.2%) which includes no policy exists, handled on an individual basis, and nothing happens. Relative risk ratios were adjusted for gender, age, ethnoracial identity, educational attainment, annual household income, employment status, marital status, urbanicity, and past-year illicit drug use disorder status. Relative risk ratios in bold are statistically significant at p <0.05. All prevalence estimates were weighted to account for NSDUH’s complex sampling design.