Association of facility characteristics and substance use disorder outcomes at discharge from residential treatment

Abstract

Introduction:

Individuals with substance use disorder (SUD) may benefit from services and supports delivered in residential settings. Prior research in this area has primarily focused on individual-level factors that affect outcomes, with little focus on the relationship between facility-level characteristics and treatment outcomes.

Methods:

This study incorporated facility-level features into an analysis of the effectiveness of residential settings to treat SUD. Administrative data from 2,713 individuals with an index enrollment in publicly funded residential treatment in Massachusetts during 2015 were linked with facility-level survey data from 33 treatment providers. This study conducted multilevel regression analysis, adjusting for resident demographic, socioeconomic, and substance use history and severity, to test facility-level effects on treatment duration and completion, and housing and employment status at discharge.

Results:

Residents stayed longer when they made and enforced rules (β= 30.22, p = 0.006). Residents were less likely to complete treatment as the number of non-clinical services increased (aOR= 0.918, p= 0.029), or in facilities where residents ate together family style (aOR=0.485, p= 0.039). Being employed at discharge was more likely when house meetings were held less than once per week (aOR= 3.37, p= 0.005) and less likely when held more than once per week (aOR= 0.385, p= 0.038).

Conclusion:

Overall, increased resident governance and fewer contingencies for successful treatment participation were associated with positive treatment outcomes. Future research should examine the internal processes of residential settings, including peer-to-peer interactions, to better understand how within-residence features affect outcomes.

1. Introduction

Individuals with substance use disorder (SUD) often have comorbid psychosocial service needs that make it difficult to maintain recovery long term (Galea & Vlahov, 2002). Unemployment, lack of social connections that are supportive of recovery, and unstable housing all contribute to low treatment engagement and poor treatment outcomes (Brewer, Catalano, Haggerty, Gainey, & Fleming, 1998). Recognizing the importance of addressing these challenges, and in light of evidence that substance use is a chronic disease, recovery rather than abstinence is increasingly viewed as the primary measure of success. Recovery, as defined by the Substance Abuse and Mental Health Services Administration (SAMHSA), is “a process of change through which individuals improve their health and wellness, live a self-directed life, and strive to reach their full potential” (SAMHSA, 2012). Ideally, both clients’ clinical and non-clinical service needs would be addressed during treatment.

1.1. Residential treatment: Models of care and evidence base

Individuals whose biopsychosocial and environmental conditions are not supportive of recovery may benefit from receiving treatment or support in a residential setting. While residential settings all offer 24-hour support, the intensity of on-site services and the degree to which trained staff govern daily operations occurs along a continuum. Less intensive residential settings can range from those that are wholly peer-run (e.g., Oxford Houses; Jason & Ferrari, 2010), to those with a house manager (e.g., Sober Living Homes in California; Polcin & Henderson, 2008) or that require concurrent outpatient treatment as a contingency of residency (e.g., recovery homes in Philadelphia; Mericle, Miles, & Cacciola, 2015).

Higher intensity settings are typically licensed and correspond to the residential treatment level of care as defined by the American Society of Addiction Medicine (ASAM; Mee-Lee, Shulman, & Fishman, 2013). Placement into one of three types of residential care settings is generally determined based on an individual’s acute intoxication and/or withdrawal potential, comorbid general medical and behavioral health conditions, readiness to change, risk of continued use, and recovery environment (Mee-Lee, Shulman, & Fishman, 2013). Individuals with greater severity of biomedical and/or emotional/behavioral complications, who are at an earlier stage in their readiness to change, who are highly likely to relapse, and whose recovery environment outside of treatment is likely to encourage continued use, are appropriate for high intensity residential care (ASAM Level 3.5 and higher). Medium or low intensity settings are appropriate for individuals whose biopsychosocial profiles suggest they will be successful in a lower level of care (ASAM Levels 3.3 and 3.1).

Recent systematic literature reviews have found moderate evidence for the effectiveness of residential care settings. While these findings are promising, methodological challenges limit the generalizability of prior research such as small sample size, high rates of attrition, and weak study design (de Andrade, Elphinston, Quinn, Allan, & Hides, 2019; Malivert, Fatséas, Denis, Langlois, & Auriacombe, 2012; Reif et al., 2014). Additional research that examines associations between facility characteristics and treatment outcomes can help to inform decisions related to programming and resource distribution.

1.2. Role of facility characteristics

Organizational theory and management factors can help to guide the selection of facility-level features that may be relevant in multilevel analyses that incorporate both individual and facility-level characteristics. One prominent organizational theory, Resource Dependence Theory (Pfeffer & Salancik, 2003), posits that organizations seek to reduce scarcity through interdependence with external agencies, such as alignment with a parent organization or seeking public funding resources from a government licensing body. This interdependence may dictate how an organization operates, including the range of available services, financial resources, or minimum licensing requirements. Other management factors such as facility size and staffing caseload have also been observed to be associated with service provision. D’Aunno (2006) reported numerous studies that found being part of a larger parent organization, funding source, and client-to-staff ratio were associated with the delivery of evidence-based practices in addiction treatment settings.

Organizational characteristics may be associated with treatment outcomes, though we know less about the specific features that may be important in residential treatment settings. One study found that individuals receiving care in a residential setting were less likely to experience post-treatment symptom recurrence if they received care in an accredited treatment setting, and this was more likely in residential care settings with a greater proportion of funding from managed care (Ghose, 2008). Another study of less intensive residential settings found that residing in a house that was part of a larger parent organization or affiliated with an outpatient treatment provider was associated with reporting no substance use or being employed at follow-up (Mericle, Mahoney, Korcha, Delucchi, & Polcin, 2019).

In addition to organizational structure, program processes and treatment philosophy may also affect substance use outcomes (Jason et al., 2020). Program processes govern the extent to which the daily operations of a facility foster a recovery-oriented environment. For example, while not empirically tested, it is possible that holding meetings with staff and residents fosters a sense of community by providing an opportunity to identify and address resident concerns. Facilities where residents eat together family style may promote positive, meaningful interpersonal relationships that may facilitate positive outcomes (Stevens et al., 2015; Jason et al., 2014). Residents who gain exposure to 12-step principles, including cultivating a relationship with a sponsor during their stay, may be better able to maintain their recovery in the community after leaving residential treatment (Tonigan & Rice, 2010).

Treatment philosophy, i.e., social model versus medical model, may also have implications for resident outcomes. Medical model programs are those in which a trained clinician identifies a disease or illness (e.g., SUD), and directs patients to a specific course of medical intervention; the patient is often a passive participant in their care (Farre & Rapley, 2017). Social model programs emphasize experiential learning opportunities that encourage the formation of new habits through practice and by the example of peers and staff (Borkman, 1982; Kolb, 2014). Social model characteristics include whether residents (vs. staff) determine the rules of the care setting, if residents are allowed to come and go without permission, the nature of staff and resident interactions, and whether staff are themselves in recovery (Kaskutas, Greenfield, Borkman, & Room, 1998).

The current study sought to expand on prior research by leveraging a large administrative dataset on individuals utilizing publicly funded residential treatment in Massachusetts, linked with a survey of facility characteristics to examine facility-level predictors of outcomes at discharge.

2. Methods

This analysis is part of a larger study that examined the association of resident and organizational characteristics with outcomes at discharge and 12-months post-discharge among individuals utilizing residential treatment in Massachusetts. The current analysis focuses on the association between facility characteristics and outcomes at discharge for a sample of residents whose care setting completed an organizational survey. The Institutional Review Boards of both Brandeis University and the Massachusetts Department of Public Health approved the procedures and methods this study used.

2.1. Sites and residents

The Massachusetts Department of Public Health Bureau of Substance Addiction Services (BSAS) licenses and funds several types of residential treatment: medically managed intensive inpatient treatment (or detoxification; ASAM 3.7), clinically managed transitional support services (also ASAM 3.7), Therapeutic Communities (TC; ASAM 3.5), Recovery Homes (RH; ASAM 3.3), and Social Model Recovery Homes (SMRH; ASAM 3.1). This study excluded both ASAM 3.7 facility types because they represent acute inpatient levels of care and have a maximum length of stay <30 days (Commonwealth of Massachusetts, 2021).

The three remaining residential facility types are considered to be long-term programs in Massachusetts, and individuals are eligible for these programs if they are already medically stabilized (Commonwealth of Massachusetts, 2021). All types must comply with the same minimum building and safety standards; briefly, they must: operate 24-hours a day; prohibit substance use on the premises; adhere to rules regarding the physical structure of the building and meals/food handling; institute policies regarding the handling and storage of resident medications and personal effects, and resident transfers; and provide clinical services that assist residents with activities of daily living.

BSAS also regulates specific minimum staffing and program components required for each service type. All three must provide a daily schedule of services designed to develop and apply recovery skills, individual service plans, planned program activities, and case management. TCs and RHs must also offer daily program activities, (e.g., counseling, clinical monitoring by qualified staff), and TCs also provide planned community reinforcement activities designed to foster pro-social values and group living skills. All facility types are required to have a clinical supervisor and program manager; medical doctors or nurses are not required for ASAM 3.5, 3.3, or 3.1 facilities in Massachusetts. While each facility type has general requirements, individual facilities have flexibility in how they implement them. Services available on-site in 80% or more of the facilities surveyed included drug and alcohol services (e.g., drug/alcohol education groups, relapse prevention groups, group and individual counseling), social services (e.g., case management, life skills training); see online Appendix for a full list of services.

Residential treatment facilities were eligible for the study if they were a publicly funded, licensed long-term residential treatment program in Massachusetts (ASAM 3.5, 3.3, 3.1) and in operation during the observation period (N= 57). A total of 36 (63%) facilities completed a survey, with no direct refusals. The study excluded three of these facilities from the analyses because they did not have any resident admissions during the enrollment period, for a final facility sample size of 33.

The research team obtained administrative data on residents from BSAS and those that are routinely collected by all publicly funded treatment providers at admission and discharge. The intake assessment records the date of enrollment, and gathers information on client’s sociodemographic characteristics (e.g., age, gender, marital status, race/ethnicity, education level, housing status, employment status, income, health insurance), substance use history, prior substance use and mental health treatment utilization, referral source, and criminal justice involvement. Upon discharge, facilities record the date of and reason for discharge, employment and housing status, and what services the client received during their stay.

The study included residents if they were 18 or older and had an index admission to a residential treatment program during the enrollment period (July 1, 2015–June 30, 2016; N= 4,806). The study excluded residents if they were younger than 18 (N= 1; 0.02%), only attended the program for an assessment (N= 2; 0.04%), deceased as the reason for discharge (N= 6; 0.12%), or whose program did not complete a survey (N= 2,084; 43%). The final sample comprised 2,713 residents in 33 programs that completed an organizational survey and were operating during the observation period.

2.2. Primary data collection

The study obtained facility characteristics directly via an electronic survey comprising questions from existing organizational surveys for formal and informal treatment and residential settings. The survey collected information across three conceptual domains: organizational characteristics, house processes, and program treatment philosophy. Using a modified version of the Addiction Treatment Inventory (ATI; Carise, McLellan, & Gifford, 2000), the study collected organizational structure, patient profile, types of services that are available, staffing mix, and financing model for formal treatment services. The Recovery Housing Processes Questionnaire collected information on the facility’s physical features, characteristics of the surrounding area, and internal processes (RHPQ; Ferrari, Jason, Blake, Davis, & Olson, 2006; Ferrari, Jason, Davis, Olson, & Alvarez, 2004; Ferrari, Jason, Sasser, Davis, & Olson, 2006).

The study captured the extent to which a facility applied a social model treatment philosophy using the Social Model Philosophy Scale (SMPS; Kaskutas et al., 1998), which collects information on the following domains: 1) physical environment (e.g., whether the space allows for informal interaction, residents have a sense of ownership); 2) staff role (e.g., staff seen as role models, encouraging resident autonomy); 3) authority base (e.g., are persons with lived experience seen as having more authority than credentialed/clinical staff without lived experience); 4) responding to SUD (e.g., SUD treated as chronic condition, need for holistic supportive approach rather than immediate dismissal due to symptom recurrence); 5) governance (e.g., residents are involved in decisions related to daily operation); 6) community orientation (e.g., facilities support residents’ engagement with external community). The SMPS is scored from 0–100, and a facility with a score of 75 or higher is considered to be a “true” social model program. Facilities with scores<75 may still apply elements of the social model, but likely reflect a combination of a medical model and social model characteristics.

Study staff conducted three key informant interviews to ensure that the compiled survey items were relevant to residential treatment programs in Massachusetts. We sent electronic and paper surveys to all residential treatment programs between August 2017 and September 2018, and study staff attempted a second round of outreach by phone for programs that did not complete the survey on the initial round of outreach. Research staff linked resident-level administrative data by BSAS to maintain resident privacy using a random program identifier.

2.3. Measures

2.3.1. Resident outcome measures

The study measured resident outcomes at discharge from the index admission. The study operationalized treatment duration as length of stay (in days), calculated by BSAS using the enrollment and discharge dates. The remaining dependent variables were dichotomous and constructed from multicategorical variables: completed treatment (e.g., reason for disenrollment - completed versus administrative discharge, relapse, deceased, left against medical or clinical advice), employed (e.g., full time, part time, or unemployed-looking versus not in the labor force), and housed (e.g., living in a home/apartment, room/boarding house, institution, group home vs. shelter/mission, streets).

2.3.2. Resident-level explanatory variables

The models controlled for resident demographic characteristics at intake: race/ethnicity (White, Black or African American, Multi-Racial, Latino, Other), gender (male, female), and age (18–24, 25–30, 31–40, 41–50, 50+). We also controlled for socioeconomic characteristics at intake: marital status (married, partnered vs. divorced, separated, widowed, never married), education level (less than high school diploma, high school diploma/GED, more than high school), receiving income from any source, health insurance (public, private, none), housing status, and employment status. The study adjusted for SUD severity and history using the Deck Severity Index (McCamant, Zani, McFarland, & Gabriel, 2007; McFarland, Deck, McCamant, Gabriel, & Bigelow, 2005), age of first use, and primary/secondary substance at intake. We also controlled for co-occurring mental health disorder (i.e., having any prior mental health service), prior SUD treatment engagement (e.g., any prior SUD service, any prior residential treatment), referral source (e.g., SUD treatment provider, criminal justice system, or other), and receipt of any other state services at intake. The study included treatment duration as a covariate when it was not used as an outcome measure.

2.3.3. Facility-level independent variables

Given the number of facility survey respondents, each multilevel regression model could only accommodate a small number of facility-level characteristics. Therefore, the study retained variables in each conceptual domain in regression models if they were significantly associated with outcome measures in bivariate tests. The final set of organizational characteristics included whether the program required a minimum time abstinent at admission, number of non–clinical services offered, being part of parent organization, percent of services offered directly on-site, ratio of full-time to part-time staff, number of beds, and staff-to-client ratio. Variables related to house processes included the extent to which 12-step principles were applied, frequency of house meetings, and whether residents eat family style. The research study measured social model characteristics using the SMPS scale score, as well as individual items from the survey, e.g., whether residents made and enforced rules, if residents could leave without permission, if staff ate with residents, and the percent of staff in recovery.

2.4. Analysis plan

Study staff conducted descriptive statistics to describe the residents and programs in the study sample. We conducted multilevel models, which account for residents nested within programs (Luke, 2004), to examine the association of facility characteristics with outcomes at discharge, controlling for resident characteristics (e.g., sociodemographics, SUD history and severity). To test the need for multilevel modeling, study staff calculated an Intraclass Correlation Coefficient for each model’s program-level random-intercept, and conducted a Likelihood-Ratio test to confirm that model’s improved fit compared to a model that did not account for multilevel nesting (Rabe-Hesketh & Skrondal, 2008). The study conducted multilevel linear regression for the treatment duration outcome measure, and multilevel logistic regression for each of the dichotomous outcome measures (i.e., completed treatment, housed, and employed.)

The study ran three models for each outcome measure, one for each facility-level conceptual domain: organizational characteristics (model 1), house processes (model 2), and social model philosophy (model 3). As mentioned, each model included only those variables in the conceptual domain that we found to be significantly associated with each outcome measure in bivariate tests. For example, model 1 for the outcome “treatment duration” included the number of beds, number of non–clinical services offered, and percent of services available on-site. However, model 1 for the outcome “treatment completion” included those variables as well as being part of larger parent organization, minimum abstinence requirement at intake, and staff-to-client ratio. This paper does not show results from resident-level only regression models, as the primary focus of this analysis is the association between facility characteristics and resident outcomes. Study staff conducted all analyses using Stata v16.

3. Results

3.1. Resident sample description

Residents were primarily male (68%), White (81%), and between the ages of 25 and 40 (61%). Compared to statewide provider analytics for all public residential treatment facilities in 2015, more individuals in the current study sample entered treatment from a stable living environment (39% vs. 60%, respectively) and fewer were employed at intake (21% vs. 22%, respectively; Massachusetts Department of Public Health, 2015a, 2015b, 2015c). The residents in the study sample overall had biopsychosocial characteristics that indicated a need for residential treatment: the average SUD severity index score was 0.61 (SD= 0.13), which is consistent with prior research on individuals receiving publicly funded residential treatment (Deck & McFarland, 2002), and nearly three-quarters (73%) reported having received treatment for a mental health disorder in the past. See Table 1 for additional resident characteristics.

Table 1.

Resident sample characteristics

	Overall (N= 2,713)
Demographics	N	%
Race/Ethnicity
White	2,208	81.39
Black or African American	128	4.72
Multi-Racial	207	7.63
Latino	137	5.05
Other	13	0.48
Age
18–24	327	12.05
25–30	804	29.64
31–40	847	31.22
41–50	452	16.66
50+	283	10.43
Gender
Male	1,850	68.19
Female	862	31.77
SUD Service Needs and History at Intake
Deck Severity Index (Median, IQR)	0.64	0.17
Primary or Secondary Substance
Opioids	1,934	71.29
Alcohol	1,044	38.48
Stimulants	984	36.27
Marijuana	363	13.38
Other	256	9.44
Age at first use (alcohol or drugs; M, SD)	13.76	3.88
Any prior SUD service	2,597	95.72
Any prior residential treatment	1,774	65.39
Any prior MH treatment	1,992	73.42
Referral Source
SUD Services	1,668	61.48
CJ System	773	28.49
Other	265	9.77
Recovery Capital at Intake
Housed	2,088	76.96
Education
< High school	559	20.60
High School/GED	1,249	46.04
>HS/Other Credentials	872	32.14
Employed	57	2.10
Client receives income from any source	832	30.67
Health Insurance
Private/Other	264	9.73
Public	2,275	83.86
None	166	6.12
Client received any state services	1,377	50.76
Married	156	5.75
Receipt of Services
# of services received during RR stay (Mean, SD)	5.21	3.35

3.2. Facility description

More than half of the facility operators that responded to the organizational survey reported that their facility was part of a larger parent organization (61%), with an average resident capacity of 28 (SD= 10) and an average staff-to-client ratio of 0.52 (SD= 0.20; see Table 2). Although most programs fell below the minimum threshold to be considered a “true” social model program (i.e., did not meet a minimum scale score of 75%), many survey respondents reported that their program applied specific social model characteristics. For example, many respondents reported that their staff ate meals together with the residents (67%), residents could leave without permission (69%), and many of the staff were themselves in recovery (M= 70%, SD= 23%). While almost all survey respondents reported admitting individuals using medication to treat opioid use disorder (MOUD; N= 32, 97%), few indicated these medications were available on-site and only 16% of residents received MOUD during their stay. See Table 2 for additional information on the programs in this sample.

Table 2.

Program sample description

	Overall (N= 33)
Program Location	N	%
Region
Boston	5	15
Central	4	12
Western	6	18
Southeast	6	18
Metrowest	9	27
Northeast	3	9
Neighborhood economically depressed	5	15
Organizational Characteristics
# beds (M, SD)	27.70	9.5
# non-clinical services (M, SD)	22.76	6.13
Ownership
Parent organization	20	60.6
Independent	13	39.4
Minimum abstinence at admission
No requirement	19	57.6
Some requirement	14	42.4
Non-clinical services offered on-site (M%, SD)	9.88	5.66
Staff:Client ratio (M, SD)	0.54	0.19
Full-time:Part-time staff ratio (M, SD)	2.18	1.89
Medication for OUD
Acceptance
Admit Residents on MOUD	32	97.0
Accepting of Methadone	33	100
Accepting of Buprenorphine	33	100
Accepting of Oral naltrexone	33	100
Accepting of Vivitrol	32	96.97
Availability*
On-site
Methadone	4	12.12
Buprenorphine	5	15.15
Oral naltrexone	5	15.15
Vivitrol	4	12.12
Via referral
Methadone	30	90.91
Buprenorphine	30	90.91
Oral naltrexone	30	90.91
Vivitrol	29	87.88
House Processes
Residents eat family style	25	75.8
12-step principles applied
Very much/quite a bit	26	78.8
Somewhat/A little	7	21.2
Frequency of house meetings
< 1x per week	3	9.1
1x per week	18	54.6
> 1x per week	12	36.4
Social Model
SMPS Scale Score (M, SD)	61.6	10.3
Staff eat with residents	22	66.7
Rules are made and enforced by residents	14	42.4
Residents can leave without permission	22	66.67
Staff in recovery (M%, SD)	69.4	23.3

3.3. Unadjusted outcomes

The average treatment duration was 103 (SD= 83) days, and 38% of residents completed their index treatment episode. The unadjusted proportion of residents in the sample who were housed increased significantly from intake to discharge (77.0% vs. 83.9%, respectively; p< 0.001), as did the proportion who were employed (2.1%; vs. 43.8%, respectively; p< 0.001).

3.4. Multilevel regression results

Table 3 reports the results of multilevel linear and logistic analyses that examine the association between program characteristics and treatment duration and completion. Residents in programs where rules are made and enforced by residents rather than staff had significantly longer stays (β= 30.22, p = 0.006). Residents had lower odds of completing treatment as the number of non–clinical services increased (aOR= 0.918, p= 0.029), or in facilities where residents ate together family style (aOR=0.485, p= 0.039).

Table 3.

Facility predictors of treatment duration and completion

	Treatment Durationa (N= 2,604)				Treatment Completionb (N= 2,574)
	β	95% CI		P-value	aOR	95% CI		P-value
Organizational Characteristics	Model 1				Model 1
Number of beds	0.354	−0.816	1.524	0.553	1.030	0.995	1.066	0.091
Number of non-clinical services	−1.458	−4.293	1.378	0.314	0.918	0.851	0.991	0.029
% of services offered on-site	0.047	−1.232	1.325	0.943	1.021	0.987	1.058	0.221
Part of parent organization	--	--	--	--	0.998	0.514	1.936	0.995
Minimum abstinence at admission	--	--	--	--	0.842	0.455	1.562	0.587
Staff to Client ratio	--	--	--	--	0.946	0.158	5.666	0.587
Ratio of full-time to part-time staff	--	--	--	--	--	--	--	--
House Processes	Model 2				Model 2
12-step principles applied very much/quite a bit	−24.959	−51.052	1.133	0.061	--	--	--	--
Frequency of house meetings (Ref: Once/week)
Less than once/week	21.229	−17.203	59.660	0.279	1.604	0.537	4.792	0.397
Greater than once/week	18.376	−4.205	40.957	0.111	0.884	0.465	1.681	0.707
Residents eat family style	--	--	--	--	0.485	0.244	0.945	0.039
Social Model Philosophy	Model 3				Model 3
Overall SMPS Score	--	--	--	--	1.001	0.963	1.040	0.976
Residents can leave without permission	--	--	--	--	1.487	0.618	3.578	0.376
Staff eat with residents	−1.866	−23.940	20.209	0.868	0.782	0.314	1.949	0.597
% Staff in recovery	--	--	--	--	--	--	--	--
Rules made and enforced by residents	30.216	8.600	51.832	0.006	--	--	--	--

Notes:

^aMulti-level logistic regression; All models adjusted for the same resident-level characteristics at intake (e.g., demographics, socioeconomic characteristics including employment and housing status, SUD severity, age of first use, SUD treatment history, receipt of other state services); Treatment duration was included as a covariate when not used as an outcome measure; Cells with “—” indicate that variable was not included in a model; Each model only controlled for program-variables in the same conceptual category (i.e., “Organizational Characteristics”, “House Processes”, “Social Model Philosophy”)

Table 4 reports the results of multilevel logistic regression analyses that examine the association between facility characteristics and housing and employment status at discharge. None of the facility-level variables were statistically significant. Residents were more likely to be employed at discharge if their facility held house meetings less frequently than once per week (aOR= 3.37, p= 0.005), but less likely when house meetings were held more than once per week (aOR= 0.385, p= 0.038).

Table 4.

Facility predictors of housing and employment status at discharge

	Houseda (N= 2,476)				Employeda (N= 2,461)
	aOR	95% CI		P-value	aOR	95% CI		P-value
Organizational Characteristics	Model 1				Model 1
Number of beds	--	--	--	--	0.999	0.960	1.039	0.942
Number of non-clinical services	--	--	--	--	0.971	0.878	1.073	0.561
% of services offered on-site	0.990	0.972	1.009	0.296	1.002	0.958	1.048	0.922
Part of parent organization	1.022	0.646	1.616	0.927	0.851	0.363	1.995	0.710
Minimum abstinence at admission	0.729	0.478	1.111	0.142	1.561	0.714	3.412	0.264
Staff to client ratio	--	--	--	--	--	--	--	--
Ratio of full-time to part-time staff	0.951	0.845	1.070	0.401	--	--	--	--
House Processes	Model 2				Model 2
12-step principles applied very much/quite a bit	0.721	0.413	1.257	0.249	0.704	0.326	1.522	0.372
Frequency of house meetings (Ref: Once/week)
Less than once/week	2.402	0.893	6.458	0.082	3.374	1.070	10.638	0.038
More than once/week	1.073	0.649	1.774	0.784	0.385	0.198	0.751	0.005
Residents eat family style	0.722	0.438	1.191	0.202	0.986	0.482	2.020	0.970
Social Model Philosophy	Model 3				Model 3
Overall SMPS Score	--	--	--	--	--	--	--	--
Residents can leave without permission	1.315	0.788	2.193	0.295	1.417	0.609	3.300	0.419
Staff eat with residents	1.056	0.175	1.296	0.841	1.079	0.467	0.2.496	0.858
% of staff in recovery	0.476	0.175	1.296	0.146	4.522	0.903	22.641	0.066
Rules made and enforced by residents	0.916	0.547	1.535	0.739	0.746	0.342	1.629	0.462

Notes:

^aMulti-level logistic regression; All models adjusted for the same resident-level characteristics at intake (e.g., demographics, socioeconomic characteristics including employment and housing status, SUD severity, age of first use, SUD treatment history, receipt of other state services); Treatment duration was included as a covariate when not used as an outcome measure; Cells with “—” indicate that variable was not included in a model; Each model only controlled for program-variables in the same conceptual category (i.e., “Organizational Characteristics”, “House Processes”, “Social Model

4. Discussion

The goal of this study was to expand upon prior research that examined the outcomes of individuals in residential treatment by also including facility-level characteristics. Overall, few organizational characteristics were associated with resident outcomes. However, we did observe an association between the number of non–clinical services offered and lower odds of completing treatment. Facilities offering more services may do so because they serve residents with greater biopsychosocial acuity, but we are not able to determine the direction of this association.

House processes, including house meeting frequency and eating meals together, were also associated with outcomes at discharge. Facilities with requirements that are too stringent (meetings more frequently than once per week, eating all meals together) may cause some residents to choose between remaining in treatment and pursuing employment opportunities in the community. To accommodate residents with external employment opportunities, facilities could consider relaxing some requirements particularly related to participation to balance resident treatment engagement and other important recovery outcomes, such as meaningful employment in the community.

Almost none of the programs met the minimum threshold for a “true” social model, and programs’ overall SMPS scale score was significantly associated with treatment completion only in bivariate analyses. In multivariate models, treatment duration was significantly longer in houses where residents made and enforced the rules. As Polcin, et al. (2014) have suggested, this type of community building likely imparts a sense of ownership on residents that may translate to longer stays.

Our limited findings regarding social model features could be an artifact of too few Social Model Recovery Home programs in our sample, which on average scored higher on the SMPS than either RH or TC types in our post hoc analyses (results not shown). The predominance of RHs in our sample may overshadow potential effects that we might observe with a larger sample of facilities that more closely adhere to a social model philosophy. However, prior research on Sober Living Homes in California (an even lower intensity settings than those in our sample) found that these programs on average also did not meet the minimum threshold to be considered a “true” social model (Mericle, Mahoney, Korcha, Delucchi, & Polcin, 2019). Future research should help to disentangle possible confounding due to governance structure and authority base to better understand the effect of social model treatment philosophy on individual outcomes. Additionally, future research should re-examine the validity of the SMPS for measuring social model adherence in different types of residential settings than were used in the scale’s original development.

While none of the facilities in our sample were either detox or other high intensity residential treatment, we conducted post hoc exploratory bivariate and regression analyses to examine potential confounding by facility type between the three types of long-term residential treatment in our sample. This study observed some differences in resident milieu, such as a greater proportion of female residents in TCs compared to RHs and SMRHs (81% female vs. 26% and 29%, respectively; p< 0.001) as well as more TC residents with a history of mental health service use (91% compared to 77% and 71%, respectively; p< 0.001). The study observed few statistically significant differences in facility characteristics, which could be partly due to the smaller sample sizes for the SMRH and TC facility types, but we also observed less variation in most facility-level characteristics across facility type than might be expected. We observed significant differences in bivariate analyses (results not shown) by facility type on the following features: frequency of house meetings, overall SMPS score, and whether there were rules that were made and enforced by residents.

This led us to run additional regression analyses with indicators for TC and SMRH facility type, but for the most part these results did not differ from our main analyses with the exception of the employment outcome models. In models adjusting for individual-level characteristics and only TC and SMRH facility-level covariates added, residents had significantly lower odds of being employed in both the TC (aOR= 0.171, 95% CI= 0.067, 0.436) and SMRH (aOR= 0.357, 95% CI= 0.162, 0.786). We also observed the effect for whether “residents can leave without permission” increase and approach significance (aOR= 1.968, p= 0.057), while the effect for “% of staff in recovery” was greatly diminished (aOR= 1.672, p= 0.451). The confidence intervals for these variables were overlapping with those of the main analyses so we cannot say with certainty that these are empirically important differences. Still, future research should further explore potential differences in resident characteristics and subsequent treatment outcomes for each distinct type of residential treatment.

In addition to facility type, organizational structure (i.e., private versus public, for-profit vs. non-profit) may relate differentially to resident outcomes. Our sample comprised exclusively state licensed, publicly funded facilities, all of which adhered to minimum licensure requirements and were required to conduct thorough intake exams to determine the most appropriate placement for care. However, future research should incorporate a combination of both public and private residential settings given recent findings that private residential programs’ organizational practices differ significantly from publicly funded settings, with implications for resident outcomes. For example, private settings were more likely to deny enrollment to individuals with greater clinical severity—the very individuals for whom residential treatment is most beneficial (Beetham et al., 2021).

Another important area for future study is the effectiveness of residential treatment for individuals with OUD. The majority of residents in our sample reported an opioid as either their primary or secondary substance at intake (results not shown), and almost all program survey respondents reported that they accept residents who use medication treatment. However, 15% or less survey respondents reported that MOUD was available on-site, and among residents with opioid use reported at intake, only 16% received either methadone or buprenorphine during their stay (results not shown). This finding is consistent with results from national surveys that show very low acceptance and uptake of MOUD acceptance in both short-term or long-term residential treatment (Huhn et al., 2020; Stahler & Mennis, 2020), even in states with additional provider incentives for doing so (Maclean, Wen, Simon, & Saloner, 2021). Widespread adoption of 12-step principles in residential settings likely contributes to anti-medication stigma by operators and residents alike (Majer et al., 2018; Miles, Howell, Sheridan, Braucht, & Mericle, 2020). Given the extensive evidence of MOUD efficacy (National Academies of Sciences & Medicine, 2019), additional research should examine strategies to increase MOUD uptake in residential treatment, patterns of MOUD utilization during and after residential treatment, and its association with residents’ outcomes.

Several limitations are important to note. Measuring program characteristics of residential treatment programs is challenging. Currently, no single cohesive measure exists for examining program characteristics, and the measures used in this study are a combination of those that have been used in previous studies of recovery residences and of other formal treatment programs. Furthermore, the sample in this study was part of a licensed system of residential treatment programs. While licensing serves to standardize treatment, licensing may make it difficult to empirically identify which characteristics matter more for supporting residents’ outcomes. Our sample primarily comprised ASAM 3.3 licensed residential treatment settings, potentially masking the importance of certain facility characteristics and limiting the generalizability of our findings to that specific facility type.

Additionally, while 63% of program operators did respond to the survey, a significant proportion of programs did not, and having only limited information on these programs (i.e., average length of stay, average program size) makes it challenging to know the extent to which these programs differ from those included in the sample. The smaller sample size at the program level also meant that we were more limited in the number of program variables that we could include in multivariate models. Each model, therefore, has confounding due to other facility-level features that could not be included in each model. As with many treatment studies, observed improvements could simply be due to regression to the mean given that by definition residents in our sample began their treatment episode at a high severity level.

Despite these limitations, the size of the sample of residents and programs represents a significant improvement on previous research on recovery residences that typically relied on smaller samples of each. Incorporating both resident and program factors also represents a significant improvement on previous research studies that until recently did not describe or account for these critical differences.

5. Conclusion

This study contributes evidence that program characteristics may be important factors for residential treatment outcomes. However, the mixed findings from this study on how program characteristics affected resident outcomes could indicate that resident, rather than program, characteristics are more important drivers for outcomes. Additional research should examine outcomes for specific patient populations, including individuals with opioid use disorder who may benefit from medication treatment during a stay in a residential setting (Miles, Howell, Sheridan, Braucht, & Mericle, 2020; Walley, Lodi, Li, Bernson, Babakhanlou-Chase, Land, & Larochelle, 2020). As Medicaid and other payers continue to expand access to residential treatment, expanding the evidence base for both resident and program-level features that affect residents’ outcomes is critical to inform both policy-makers and providers (Miles, 2019).

Highlights.

• Administrative data linked with a facility-level survey enabled multilevel analysis

• Residents with greater involvement and fewer contingencies saw improvements

• Future research should study nuances of different types of residential treatment