Outcome variables over time
presents the findings for the second-level predictor variables modeling the wave trajectories for each of the four primary dependent variables. At baseline, 93.3% used alcohol or drugs in the past 6 months. Five individuals in each condition had not used substances because they were either in a treatment facility or jail during the entire 6-month period before baseline, and all participants were included in all analyses.
Main and interaction effects for primary outcome variables over time.
Four analyses requiring logistic variations of HLM were conducted with the four primary dependent variables: reported substance use, employment status, self-regulation and awaiting criminal charges. Each analysis, with time as a first-level predictor, included an equivalent set of second-level predictors, including predictors as control variables, and variables representing main effects (i.e. individual predictors) and interactions. In each case, the predictor variables included the following: gender as a predictor variable, experimental condition (Oxford House versus usual after-care), age (under 36 years, at or above 37 years), psychiatric comorbidity and interactions representing condition × age and condition × psychiatric comorbidity.
For each of the four models, significantly more positive outcomes were found in the Oxford House condition compared to the usual after-care condition (see ), as noted by the gammas for the condition effect of wave trajectory between time and outcome. That is, the condition effect was significant for any reported substance use [gamma = −0.34, odds ratio (OR) = 0.71, confidence interval (CI) = (0.58, 0.87), P < 0.01], employment status [gamma = 0.34, OR = 1.40, CI = (1.11, 1.76), P < 0.005], self-regulation tendencies [gamma = −0.08, SE = 0.03, t = −2.69, P < 0.01] and awaiting criminal charges [gamma = −1.07, OR = 0.34, CI = (0.02, 0.51), P < 0.001].
For the variable awaiting criminal charges, significant interaction effects included condition × age [gamma = 1.61, OR = 4.99, CI = (2.06, 12.09), P < 0.01] and condition × psychiatric comorbidity [gamma = 0.79, OR = 2.20, CI = (1.06, 4.61), P < 0.05]. As indicated in , in the usual after-care condition participants who were younger and participants with no psychiatric comorbidity were more likely to be awaiting charges compared to participants in the Oxford House condition. This result appeared most prominent by the final assessment wave (wave 4).
Length-of-stay outcome findings
Because a stay of 6 months or more in Oxford House might be needed for residents to obtain the most benefits from this recovery home experience, as was found in a national sample of Oxford House residents [9
], we examined those Oxford House residents who had lived in an Oxford House for 6 or more months (45.2% of the present sample) versus those residents who had been in an Oxford House for less than 6 months (54.8% of the sample). We again used HLM to model whether the dichotomous variable related to length-of-stay (i.e. less than 6 months or 6 or more months in Oxford House) predicted the wave trajectory for the four major outcome variables, including the same second-level control variables and moderators as the prior analyses (except, of course, condition).
There was a significant length-of-stay effect for substance use [gamma for the length-of-stay variable = −0.45, OR = 0.63, CI = (0.41, 0.99), P < 0.05] as well a length-of-stay × age interaction [gamma = 0.90, OR = 2.46, CI = (1.02, 5.92), P < 0.05]. shows that residents of Oxford House who remained in the house for at least 6 months had significantly better outcomes over time. In fact, by the 24-month assessment the differences were rather large (15.6% versus 45.7%, respectively).When examining younger versus older residents, younger residents who stayed in Oxford Houses for at least 6 months had extremely low substance use (6.7%), whereas those younger residents who stayed for less than 6 months had much higher use (62.5%; see ).
Main and interaction effects by length of time in Oxford House.
For the other three outcome variables, there was a significant length-of-stay × age interaction. For employment [gamma = −1.47, OR = 0.23, CI = (0.06, 0.89), P < 0.05], younger individuals who resided in the Oxford Houses for over 6 months had substantially better employment status than those who stayed in the house for less than 6 months (at the 24-month assessment employment rates were 93.8% versus 56.3%, respectively). The length-of-stay × age interaction was also significant for reported self-regulation tendencies [gamma = 0.18, SE = 0.08, t = 2.28, P < 0.05], such that younger Oxford House members who left their houses in less than 6 months had the least improvement in self-regulation. Finally, there was a significant length-of-stay × age interaction effect for awaiting criminal charges [gamma = 1.26, OR = 3.52, CI = (1.18, 10.50), P < 0.05], but all groups had zero levels of this outcome variable by the last assessment (see ).
At the 24-month follow-up, information on the living situation or current status was also recorded for all participants. Percentages for Oxford House participants are reported first and usual after-care participants next: living in their own home or apartment (40% versus 13%), living with a sexual partner or spouse (21% versus 31%), living with relatives (19% versus 24%), living with friends (5% versus 16%), incarcerated (3% versus 9%), living in an Oxford House (5% versus 0%), living in a homeless shelter (3% versus 3%), living in a staffed recovery home (1% versus 1%), undergoing in-patient treatment (1% versus 1%) and deceased (1% versus 1%). In addition, in the Oxford House condition, 14 mothers were able to obtain custody of their children while one mother lost custody; in contrast, in the usual care condition, six mothers gained custody of their children and two mothers lost custody of their children.