The Relationship of Oxford Houses Across Heterogeneous House and Setting Characteristics

Abstract

Oxford Houses (OH) are democratically run, self-funded, substance-use recovery homes that operate across the United States and internationally. Previous research shows the OHs are present in diverse neighborhoods. The current study examined the neighborhoods of 42 OHs located in Oregon, Texas, and North Carolina to better quantify and understand house and neighborhood characteristics that are related to relapse rates. Independent variables were participant’s length of stay in OH, wages earned from employment, and income/education neighborhood characteristics. Neighborhood characteristics were related to relapse rates, with higher relapse rates occurring in neighborhoods with lower income and education levels. This finding supports the OH organization’s premise that while OHs may work across community settings, they perform better in neighborhoods with higher average income and education levels.

Recovery home residences span from low to high service intensity, with levels of support ranging from level 1 involving peer-operated residences to level 4 residences that offer a wide variety of treatment and recovery support services. (Jason, Mericle, Polcin, & White, 2013). An example of level 1 residence is Oxford Houses (OHs), which are democratically-run, sober living houses with no limit on length of stay. Members are expected to remain abstinent from drugs and alcohol, pay their portion of the rent and utilities, and attend weekly house meetings (Oxford House, 2012). OHs exist in a variety of environments and neighborhoods across the country and even internationally. The contexts and environments of where OHs are located may have implications for the success of residents living in these settings.

Neighborhood context may affect people in these recovery settings. Social Disorganization Theory (Shaw & Mckay, 1969) suggests that neighborhood disorganization can affect a neighborhood’s residents because of factors that ultimately lead to a lack of social cohesion. Indicators for social disorganization include poverty rates, socioeconomic status, perceived safety, number of alcohol outlets, the physical appearance of the neighborhood, ethnic heterogeneity, crime rates, and residential mobility (Winstanley et al., 2007). Previous research has found that social disorganization predicts drug use, dependence, and mortality. Winstanley et al. (2007), for example, found that social disorganization predicts adolescent alcohol and drug use and dependence even after controlling for individual and family-level factors. Similarly, Boardman et al. (2001) found that neighborhood disadvantage is moderately related to licit and illicit substance use, especially for individuals with low incomes. Elliot and Lowman (2013) also found that socioeconomic status (SES), measured by education level and income, was negatively associated with alcohol misuse via locus of control. Finally, Hannon and Cuddy (2006) found that poverty is a predictor of drug dependence mortality, while neighborhood homeownership rates and drug dependence mortality have a moderately negative relationship independent of poverty. Overall, these studies indicate that neighborhood context has implications in relation to drug use and dependence despite an individual’s specific intrapersonal characteristics.

Although everyone residing in an OH is required to remain abstinent from drugs and alcohol, there are variables other than abstinence that predict success in substance use recovery. For example, a predictor of positive treatment outcomes is length of stay in OHs. A study by Jason et al. (2007) found that residents who stayed in an OH at least 6 months reported less substance use at a two year follow-up than those who stayed less than 6 months. Research has also shown employment to be a significant predictor of traditional substance use treatment completion (Melvin et al., 2012) and could even be a protective factor against relapse (Walton, 2016). Additionally, the Substance Abuse and Mental Health Services Administration (SAMHSA, 2012), incorporated employment in their working definition of recovery under the dimension of “purpose.” Specifically, wages from employment are an important consideration in regard to long-term abstinence; research indicates that low SES is associated with increased substance use. In other words, the higher an individual’s wages, the less likely he/she is to use substances (Karriker-Jaffe, 2013). Such findings make wages from employment an important possible indicator of success in recovery homes.

The OH organization (Oxford House, Inc., 2010) refers to conditions that are optimal for starting new recovery houses. They indicate that new recovery houses are best located in good locations, as some low-income areas have high levels of drug use and crime, which can be deterrents to remaining abstinent. The focus of the current study is to ascertain how wages, length of stay in a recovery home, and house-level neighborhood characteristics are associated with treatment outcome. We predicted that higher wages while living in the OH, longer length of time in an OH, and better neighborhood SES characteristics would be related to less relapse.

Method

Participants

The sample for this study consisted of 42 recruited OHs located in Texas, North Carolina, and Oregon. In the current study, we used data collected from a baseline assessment and four months later to assess relapse. Three field recruiters associated with the OH organization approached OHs in their geographic area to enlist them for the study. Those participants who were current residents of houses during the baseline assessment were then approached to participate in the study. DePaul University granted the approval of its Institutional Review Board.

The houses contained 229 participants with an average of 6.2 participating residents per OH. The sample was 45.5% women, 55% men, and 0.4% other. The race of participants included Caucasians (82.1%), African American (9.2%), Hispanic (6.6%), American Indian (1.3%), Alaskan Native (0.4%), and Pacific Islander (0.4%). Regarding levels of education, 53.3% attended and/or graduated college, 18.8% had a GED, 18.3% of resident had a high school diploma, and 9.6% were other. Regarding employment, 67.7% were employed full-time, 11.3% were employed part-time, 10.9% were unemployed, and 10% other. The length of stay for residents in the house ranged from 0 months to 6.8 years with the mean length of stay 10.3 months.

Measures

For the sample of 42 houses used in this study, demographics varied widely by neighborhood. The U.S. Census American Fact Finder (https://factfinder.census.gov/faces/nav/jsf/pages/index.xhtml) provided the demographic data for the houses. Poverty rates among the areas where the 42 houses were located ranged from 7.0% to 30.7% (N = 26, M = 20.0%, SD = 6.4%), home ownership rates ranged from 33.4% to 78.4% (N = 26, M = 55.7%, SD = 9.6%), and unemployment rates ranged from 3.2% to 9.7% (N = 26, M = 5.6%, SD = 1.6%). As for education, the 42 houses were located in areas where between 60.7% and 95.6% (N = 26, M = 83.5%, SD = 8.4%) graduated with a high school diploma or higher and between 8.3% and 53.4% (N = 26, M = 25.0%, SD = 11.4%) graduated with a bachelor’s degree or higher. The racial breakdown of the 42 areas where houses were located consisted of a mean 70.3% were Caucasian, 12.5% were African American, 8.6% were Other, 4.0% were Asian, 3.6% were two or more races, and 0.9% were American Indian/Alaska Native. Within these geographic areas, 29.2% identified with a Hispanic or Latino/a background. These demographic statistics highlight the heterogeneity of neighborhoods in our sample.

Our measure of SES used the Washington Post’s Super Zip, which ranks zip codes using income and education data. Neighborhoods were determined using zip code boundaries. Education level and income were captured by a single variable: Super Zip (Mellnik & Morello, 2013). The information is converted to a score from 0-99 based on average percentile ranks in income and college education data with a higher score indicating higher income and higher level of education (N = 42, μ = 37.02, SD = 17.22). There was considerable variability between the Super Zip rankings in the 42 neighborhoods that were part of this study, ranging from scores of 2 to 96.

We used the Addiction Severity Index (ASI) (Mclellen et al., 1992) to collect information on the OHs, including the length of stay of residents, wages, and the number of employed individuals in the house. Items from the ASI used in this study included “What is your current employment status?”, “How much money did you receive from the following sources in the past 30 days: Employment?” and “How long have you lived at your current address?” The median wages per month for each OH were calculated using the median total amount of money the house received via employment (μ = $1,463.76, SD = 542.63) per the number of employed individuals in the house (μ = 4.38, SD = 1.83). Length of stay was calculated as a log transformation of the median data in months (μ = 1.79, SD = 0.80). We used log transformation in order to make the highly skewed distributions less skewed, and this transformation helped meet the inferential statistic’s assumptions. There was a significant relationship between length of stay and house wages (r = 0.47, p < 0.05).

We used the Form-90 Timeline Follow-back (TLFB; Miller, 1996), which assesses alcohol and drug use over time. TLFB is a self-report that captures frequency of alcohol and drug use between wave 1 and wave 2. This retrospective assessment of alcohol and drug use provides reliable substance use data (Del Boca, Babor, & McRee, 1994) and has demonstrated convergent validity (Project Match, 1993). For this study, participants were asked how many days between wave 1 and wave 2 (collected 4-months apart) they had at least one drink containing alcohol and number of days they used any drugs. This was coded into a dichotomous relapse variable, with participants reporting any drug or alcohol use were coded as “1” and participants that remained abstinent between wave 1 and wave 2 were coded as “0.” Relapse rates for each house were calculated by generating the percentages of all wave 1 residents that reported any alcohol or drug use between wave 1 and wave 2.

Results

A beta regression was used to predict the effect of length of stay, wages from employment, and Super Zip on the rate of relapse (See Table 1). Heteroscedasticity was assumed based upon a studentized Breusch-Pagan test (df=3, 8.1796, p=0.04). The pseudo R² of the estimated regression was .18, and the model significantly predicted relapse rates (ϕ = 3.80, p < 0.01). This was due to the significant and negative relationship of Super Zip and relapse rates (β = −0.02, p=0.03). No other effects were significance.

Table 1.

Beta regression of super zip scores, house wages, and length of stay predicting relapse rates

	Estimate	Std. error	Z stat	p-value
Intercept	−0.76	0.53	−1.43	0.15
Super Zip	−0.02	0.00	−2.20	0.03*
Wages	0.00	0.00	0.63	0.53
Length of stay	−0.28	0.22	−1.28	0.20
Φ	3.80	0.87	4.34	0.001**

Note:

^*p<0.05

^**p<0.005.

Discussion

This study found that lower relapse rates occurred when houses were located in neighborhoods with higher education and income (i.e., higher Super Zip scores). Length of stay and wages from employment, known markers of long-term sobriety success, were not significantly predictive of relapse rate. The findings suggest that relapse could potentially be affected by the location of the recovery houses in terms of neighborhood-level of education and income.

Social disorganization theory stipulates that context can negatively affect individuals despite individual protective factors. As mentioned in the introduction, social disorganization predicts drug use, dependence, and mortality (Winstanley et al. 2007). Because the current study found relapse rates to be negatively related to Super Zip, it is possible that the risk of relapse could potentially increase in a lower SES, more socially disorganized neighborhoods. A previous study by Callahan et al. (2016) found that OHs closure rates were positively related to community unemployment rates, which is another indicator of social disorganization. Their results are compatible with the current study’s finding of an association between relapse rates and the SES characteristics of the community in which OHs are located. Overall, this finding supports the OH organization’s suggestion that living in a good neighborhood motivates individuals to remain active in recovery from substance use disorder (Oxford House, Inc., 2010). While OHs are currently located in different neighborhood settings, less relapse appears to occur when they are located in less disorganized and higher SES communities.

This exploratory study had several of limitations. First, the sample size was small. In addition, we were not able to randomize houses to different environments, which would have provided a stronger test of the study’s hypothesis. Further, this sample was collected from predominantly white participants in only three geographical regions of the United States. The skewed white sample in the current study is a significant limitation both for generalizability, and because this is a consistent oversight in research and how that translates to how programs and services are delivered, from an equity standpoint as well as cultural competence. For generalization, there is a need for a more diverse sample of participants and houses. Another limiting factor is that residents of OHs are not necessarily from the neighborhood where their recovery home is located. Also, while the education level of the participants is not significantly different from that of the general US population, it does appear quite different than that of the homeless population (28% vs. 53%). Finally, it is unclear how long a person has to live in a neighborhood to be affected by the environment, especially in regards to social disorganization theory. Previous research does suggest that OH members easily adapt to their OH environments and are seen favorably by their neighbors (Jason, Roberts, & Olson, 2005).

Our research did not answer questions about whether different ethnic populations fare better in neighborhoods where they have access to people from similar races/ethnicities/cultures for social and other kinds of support. With income often being a proxy for race, acknowledging nuance is important so that OH's are not situated only in affluent neighborhoods, which would often be skewed as Caucasian. Other factors will be important to understand and consider, depending on the population(s) being served.

There are attractive features that have allowed the OH model to expand over the past 40 years to over 2,000 recovery houses. As this continuing expansion occurs, it might be useful to consider the findings of the current study, which found that lower rates of relapse may occur in higher SES communities. There is clearly a need for more research to better understand these findings so that we have more insights into residents’ interactions with their neighborhoods and how these may affect relapse. Finally, larger studies involving more houses in different parts of the US would possibly increase the generalizability of the study’s findings.

Public Policy Relevance:

Recovery homes are an inexpensive option for many who are seeking post treatment community support. Our study found that less relapse may occur when these homes are situated in higher SES neighborhoods.