Table 3.
Exponentiated regression coefficients for models examining the association of cross-lagged weekly cannabis use on primary substance use.
| Predictors | Unadjusted Model IRR (95% CI) | Final Adjusted Model IRR (95% CI) |
|---|---|---|
| Status at Week 2 of Treatment: | ||
| Intercept | 7.81 (5.05, 12.07) *** | 7.12 (5.32, 9.53) *** |
| Sex (female) | -- | NS |
| Age | -- | 0.96 (0.93, 0.98) *** |
| Lagged Cannabis Use | 0.74 (0.67, 0.80) *** | 0.74 (0.69, 0.79) *** |
| Lagged PTSD | 1.00 (1.00, 1.01) | NS |
| Rate of Change: | ||
| Time (linear) | 0.91 (0.87, 0.95) *** | 0.91 (0.89, 0.93) |
| Lagged Cannabis Use X Time (linear) | 0.98 (0.96, 1.00) | NS |
| Lagged PTSD X Time (linear) | 1.00 (1.00, 1.00) | NS |
Note: This model was specified with a negative binomial distribution. Negative binomial regression coefficients are expected differences in log counts. All coefficients presented here are exponentiated. When exponentiated, the intercept represents an expected count of days of primary substance use (when time = 0), and the slope represents a ratio of expected counts or an incident rate ratio. All predictors are also interpreted as ratios of expected counts or incident rate ratios (IRRs). For example, the IRR of 0.74 for lagged cannabis use is interpreted to mean that a one-unit change in lagged cannabis use was associated with a 26% lower expected count of primary substance use (when all other predictors were held constant). Sex was coded dichotomously with males equal to 0, while age was mean centered. *** p < 0.001.