Table D.1.
Independent variable | Dependent variable | |||||||
---|---|---|---|---|---|---|---|---|
| ||||||||
Post-discharge outpatient addiction treatment (Mediator) | Post-discharge mutual help (Mediator) | Risky drinking days (Outcome) | Indirect effect | |||||
|
|
|
|
|||||
b (95% CI) | p | b (95% CI) | p | b (95% CI) | p | b (95% CI) | p | |
| ||||||||
Concurrent associations - b (95% CI) | ||||||||
Between-person associations | ||||||||
A-CHESS (study arm) | 0.11 (0.04–0.17) | 0.002 | 0.07 (−0.01–0.15) | 0.101 | −0.60 (−1.04–−0.17) | 0.007 | -- | -- |
Post-discharge outpatient addiction treatment | -- | -- | -- | -- | −1.16 (−1.85–−0.41) | 0.002 | −0.12 (−0.24–−0.01) | 0.035 |
Post-discharge mutual help | -- | -- | -- | -- | −1.43 (−2.03–−0.83) | <0.001 | −0.01 (−0.22–−0.03) | 0.119 |
Within-person associations | ||||||||
Post-discharge outpatient addiction treatment | -- | -- | -- | -- | −1.16 (−1.97–−0.35) | 0.005 | -- | -- |
Post-discharge mutual help | -- | -- | -- | -- | −1.48 (−2.19–−0.76) | <0.001 | -- | -- |
| ||||||||
Lagged associations - b (95% CI) | ||||||||
| ||||||||
Between-person associations | ||||||||
A-CHESS (study arm) | 0.10 (0.02–0.18) | 0.018 | 0.08 (−0.05–0.17) | 0.064 | −0.63 (−1.17–−0.09) | 0.021 | -- | -- |
Post-discharge outpatient addiction treatment | -- | -- | -- | -- | −1.34 (−2.22–−0.45) | 0.003 | −0.13 (−0.27–0.01) | 0.071 |
Post-discharge mutual help | -- | -- | -- | -- | −1.49 (−2.19–−0.79) | <0.001 | −0.12 (−0.27–0.02) | 0.094 |
Within-person associations | ||||||||
Post-discharge outpatient addiction treatment | -- | -- | -- | -- | −1.34 (−2.23–−0.45) | 0.003 | -- | -- |
Post-discharge mutual help | -- | -- | -- | -- | −0.70 (−1.60–−0.25) | 0.154 | -- | -- |
Coefficients are displayed from Mplus models computed with the maximum likelihood robust estimator. Linear regression coefficients are shown for post-discharge outpatient addiction treatment and mutual help (the dependent variable was the person-level mean of service use follow-up periods) and Poisson regression coefficients are displayed for risky drinking days. Bolded values are statistically significant (p < 0.05). Models controlled for treatment site, interview month, age, gender, race, presence of mental health problems, prior addiction treatment, and reasons for entering treatment. Race was dichotomized to white vs. other (collapsing black and other categories) due to empty bivariate cells. The multilevel structural equation model provided separate estimates for between-person estimates (i.e., person-level means of risky drinking days and addiction treatment or mutual help) and within-person estimates (i.e., within-person observations of risky drinking days and group-mean centered addiction treatment or mutual help). Between-person estimates were used to calculate the indirect effects. Within-person estimates are shown for completeness.