Reducing health disparities among black individuals in the post-treatment environment

Abstract

An important step in reducing health disparities among racial and ethnic minorities with substance use disorders involves identifying interventions that lead to successful recovery outcomes for this population. The current study evaluated outcomes of a community-based recovery support program for those with substance use disorders. Participants included 632 residents of recovery homes in three states in the US. A multi-item recovery factor was found to increase over time for these residents. However, rates of improvement among Black individuals were higher than for other racial/ethnic groups. Black Americans perhaps place a higher value on communal relationships relative to all other racial/ethnic groups, and by adopting such a communitarian perspective, they might be even more receptive to living in a house that values participation and involvement. The implications of these findings for health disparities research are discussed.

Keywords:

• Ethnic/racial health disparities
• recovery homes
• Oxford Houses
• substance use disorders

Acknowledgments

The authors appreciate the financial support from the National Institute on Alcohol Abuse and Alcoholism (grant number AA022763). The authors appreciate the social network help of Ed Stevens and the editorial work of Meghan Salomon-Amend. We also acknowledge the help of several members of the Oxford House organization, and in particular Paul Molloy, Alex Snowden, Casey Longan, and Howard Wilkins.

Disclosure statement

The authors have no conflicts of interest.

Correction Statement

This article has been republished with minor changes. These changes do not impact the academic content of the article.

Notes

1 Although resident level variation in quadratic trends appeared inconsequential (not unexpected given the limited number of residents with four or more recovery factor scores), we retained the random quadratic effect to avoid convergence problems (divergence of the independent simulations) using the Bayesian estimation algorithm. Preliminary analyses also indicated that house level differences in linear and quadratic trends were absent, most likely due to the modest number of recovery homes, so only a random intercept was included at the house level.

2 In a quadratic growth model, the choice of contrasts for the time metric affects the interpretation of both the intercept and the linear growth component. This can be easily visualized by imagining the level and slope one would observe while traversing from left to right all points on a U shaped (quadratic) curve. The slope starts out negative on the left leg but eventually flips and becomes positive on the right leg. We suspect that the recovery factor changes more quickly immediately following house entry, than the change that takes place after a longer period of residence, and this initial change is important for resident success. Accordingly, contrasts for resident wave, the time metric, were specified so that the intercept represented initial status at wave 1 and the linear component represented initial slope at wave 1.

3 Each potential predictor of growth was initially tested singly and included effects on all the growth components at level 1 (initial status, initial slope and quadratic trend) and level 2 (initial status). Predictors with no significant effects were not considered further. For the predictors with at least one significant effect, we tested a set of models that combined all the significant resident level predictors with one significant house level predictor at a time because of the modest number of recovery homes in the study.

4 No Predictor Model: The main findings from the two-level growth curve model before any predictors were entered included the significant positive initial slope mean and significant negative quadratic mean for the recovery factor. Resident level variation was only substantial for the initial status and initial slope. House level variation was only substantial for the initial status. The random growth effects accounted for between 58–70% of the variance in the repeated measures of the recovery factor. The Bayesian posterior predictive p value was 0.48 indicating the two-level model fit well (i.e., lack of fit was non-significant).

5 Two of the three resident-level predictors, age at wave 1 and prior time in residence at wave 1, had significant intra-class correlations and so had effects at both resident and house levels.

6 For house gender and house savings, the effects were not significant, mostly because of correlation with the competing predictor, house level prior time in residence. In other words, residents in female houses tended to have lower average prior time in residence compared to residents in male houses (r = −0.31, p = 0.14) and residents of more affluent houses had longer average prior time in residence compared to residents of poorer houses (r = 0.40, p = 0.06). House average resident poverty, however, was not as highly correlated with house level prior time in residence (r = −0.12, p = 0.56) and the effect on the house level recovery factor initial status was negative and significant (the correlations involve latent variables so the df is not available or relevant for computing p-values). Because the recovery factor includes monthly wages, this relation could be due primarily to lower wages. Accordingly, gender and house level savings were not considered further.