Table 4.

Association of heavy adolescent marijuana use with adult outcomes: regression estimates and p-values for the propensity score matched sample.

Midlife outcome	Total effect B 95% confidence interval (CI)	Percent of the effect mediated by high school dropout, p-valuea	Percent of the effect mediated by adult drug trajectories, p-valuea
Incomec	−1.429 (−2.611 to –0.246)	45.08%, p = 0.060b	42.61%, p = 0.005
	Odds ratio (OR) 95% CI
Poverty
No	Reference	44.65%, p = 0.066 b	34.38%, p = 0.047c
Yes	1.965 (1.059–3.646)
Marital status
Married	Reference
Unmarried	1.785 (1.005–3.171)	32.00% p=0.180
Living with partner			N/A
Anxious mood	1.304 (0.488–3.614)
Low	Reference	N/A	28.45%, p = 0.107
High	2.118 (1.002–4.476)

Note: All models adjust for gender; household structure; poverty status; maternal education; maternal depressed mood; maternal drug and alcohol use; maternal school aspirations for her child; childhood family punishment; IQ; reading scores; teacher ratings of shy behavior, aggressive behavior, and classroom conduct; and adolescent self-reports of family conflict, violent and property offenses, alcohol use, and smoking. The anxious mood model also controls for young adult anxious mood (age 32), which is the same variable as the outcome.

N/A = Mediation model not tested because there was not a statistically significant association between the mediator and the outcome.

^ap-Value based on Sobel test (1982).

^bControl for adult drug trajectories.

^cControls for high school dropout.