To examine the estimated prevalence of the four SUDs across ages 18–90, we used intercept-only logistic TVEMs using the weighted TVEM macro (Dziak et al., 2017). The weighted TVEM analyses includes options for examining models by group while retaining full sample survey weighting (see Dziak, et al., for additional information). Thus, for each SUD outcome we estimated population-weighted age trends for male and female and White, Black and Latino subgroups (the three largest groups with adequate data for this analysis). We estimated models with between 1 and 5 knots (splitting points) and the best-fitting models were selected for each group using AIC and BIC. Results were plotted together; non-overlapping confidence intervals indicate ages where the estimated prevalence of a SUD differed by group. Results are presented as figures because the coefficients are estimated as a function of continuous time.
