Barriers to retention in substance use treatment: Validation of a new, theory-based scale

Abstract

Purpose:

Few studies and no theory-based scales have addressed specific barriers to substance use disorder (SUD) treatment retention. The current study, building on the Theory of Planned Behavior (TPB), sought to (a) identify those barriers that are most strongly associated with treatment retention, and most common, and (b) develop and validate a new scale of retention barriers, focusing on TPB attitude and perceived control components.

Methods:

The study administered surveys to 200 participants initiating SUD treatment at a public, outpatient program in Northern California; the analytic sample (N = 156) included only those not strongly coerced into treatment. Surveys included TPB-based measures of treatment barriers; other motivational readiness measures; treatment coercion and social desirability measures; and clinical severity variables and demographics. Discharge status was collected from program records.

Results:

Item and scale analyses identified three dimensions of attitudinal barriers (i.e., Low Perceived Treatment Need/Value, Social Concerns, and Concerns about Missing Substances) and two dimensions of perceived control barriers (i.e., Personal Limitations and Basic Logistic Barriers). Results informed creation of a 19-item Barriers to Retention Scale (BRS) with 5 subscales and very good internal reliability (alpha = 0.88). While all subscale scores were correlated with treatment completion, only Concerns about Missing Substances and total BRS scores predicted treatment completion in multivariate analyses.

Conclusions:

The present study identified core dimensions of treatment retention barriers and developed a new scale predictive of treatment completion and potentially useful as a screener and in future research. Results suggest that interventions to improve retention should focus strongly on concerns about the negative impacts of abstaining from alcohol and drugs on craving and quality of life.

1. Introduction

Rates of treatment-seeking and treatment completion remain very low in the U.S. (Lappan et al., 2020). National data from 2017 to 2018 suggest that, among those reporting a past-year SUD, only 11% received specialty treatment in the prior year (Substance Abuse and Mental Health Services Administration, 2019a), and among those discharged from publicly-funded programs, just 41% completed treatment (Substance Abuse and Mental Health Services Administration, 2019c). Thus, understanding—and maximizing—treatment retention among the few who do seek treatment for an SUD is imperative. Accordingly, the current study investigates the specific perceived barriers to SUD treatment retention that are most influential and most common, and develops and validates a new, theory-based scale describing those barriers. The study focuses on barriers as perceived by treatment clients, as clients have unique and valuable perspectives on treatment-related processes (Laudet et al., 2009; Orford, 2008).

1.1. Prior research on barriers to SUD treatment

Just a handful of studies have addressed barriers to treatment retention as perceived by treatment clients, and they have produced mixed and inconclusive results. In one mixed-methods study, Palmer et al. (2009) conducted focus groups and quantitative surveys with 21 outpatients who prematurely terminated treatment, assessing problems or concerns related to their decision to leave early. The most common reasons emerging from focus group data concerned low social support for treatment (such as from friends and family), staff limitations/poor staff connection, and low motivation to change/treatment readiness (in that order). Conversely, survey results showed that the most highly ranked reasons for termination among provided options were client substance use, competing family responsibilities/problems, transportation/financial problems, and ambivalence surrounding the need to stop using (in that order). In another, qualitative, study, Yang et al. (2018) individually interviewed 60 clients receiving short-term inpatient substance use treatment just prior to discharge to identify barriers to and facilitators of treatment “engagement,” defined as “treatment participation and positive treatment experience.” Thematic analysis suggested that barriers to engagement included lack of perceived treatment need; low trust and counselor rapport; and dissatisfaction with the amount and type of services offered (or with non-clinical staff); by contrast, positive experiences in groups with peers could support engagement. However, this study was limited by the fact that essentially all participants completed treatment, so retention per se was not the focus.

Two additional studies have collected survey data on reasons for premature termination of SUD treatment. Laudet et al. (2009) surveyed 135 outpatients who prematurely terminated treatment using open-ended survey questions. Investigators asked respondents to identify the most important reason that they dropped out, as well as whether the program could have done anything differently to promote continued attendance. The most common reasons for leaving treatment, grouped thematically, were dislike of program aspects (including staff, other clients, and program rules), competing activities/responsibilities, use of substances/relapse, lack of interest in treatment, personal issues unrelated to the program, finances, and lack of perceived helpfulness (in order). Respondents reported that programs could improve by offering services to better address their needs; more supportive staff; and more flexibility/range of hours. Previously, Ball et al. (2006) administered their 28-item Reasons for Leaving Treatment scale to 24 adults prematurely terminating outpatient treatment. The most commonly endorsed reasons for ending treatment included general disinterest in continued attendance, loss of hope in ability to change, and lack of a reason to stop using alcohol or drugs; conflicts or lack of connection with staff and dissatisfaction with treatment services were also common. Researchers have also conducted more targeted studies, including a study among mothers identifying perceived barriers related to childcare (Seay et al., 2017); a study of rural women identifying barriers such as concerns around treatment helpfulness and being judged by others/self-disclosure (Godlaski et al., 2009); and a study of homeless persons, with results combining provider and client perspectives (Orwin et al., 1999).

Meanwhile, a large literature has examined perceived barriers to SUD treatment initiation. In these quantitative and qualitative studies, populations with SUDs have reported multiple attitudinal, social, and logistic barriers to seeking treatment, with lack of perceived treatment need being identified again as a predominant barrier (Keen II et al., 2014; Motta-Ochoa et al., 2017; Perron et al., 2009; Pinedo et al., 2018; Saunders et al., 2006; Schmidt et al., 2007; Schonbrun et al., 2011; Schuler et al., 2015; Verissimo & Grella, 2017). Nonetheless, studies addressing barriers to treatment initiation may not be very relevant to understanding barriers to treatment retention specifically. This is because (a) individuals must largely (if not entirely) surmount any barriers to treatment initiation in order to begin treatment, and (b) proximal interactions and experiences within treatment, such as interactions with staff (which are not addressed in studies on treatment initiation barriers), are likely to contribute to continued engagement. Additionally, (c) those surveyed in studies on barriers to treatment initiation may differ substantially from the SUD treatment population (e. g., have less severe SUDs overall, since studies of barriers to treatment initiation tend to focus on those who considered but never received treatment).

1.2. The Theory of Planned Behavior in relation to treatment retention

The Theory of Planned Behavior (TPB; Ajzen, 1985, 1991) is a general, social-cognitive theory of human behavior that can help to identify and organize specific perceived barriers that may be impactful for treatment retention. The TPB assumes that the probability of engaging in a given behavior is proximally determined by the intention to engage in that behavior, itself a function of one’s attitude, subjective norm, and perceived control regarding the behavior. The attitude component represents one’s personal evaluation of a given behavior, and is based on the expected outcomes of engaging in the behavior. The subjective norm component represents social influences on a given behavior, and is based on the perceived opinions of important referents regarding the behavior. Perceived control refers to the subjective ease or difficulty of a given behavior, and is based on beliefs about factors that may facilitate or impede it. Because perceived control often reflects actual control, it is typically treated as a direct predictor of behavior as well as a predictor of intention. The TPB has been successfully applied to modeling a very wide range of behaviors in public health, such as alcohol and drug use, smoking, safe sex behavior, speeding, exercise, and medication adherence, and has been used in well over 1000 independent studies (Ajzen, 2020; Armitage & Conner, 2001; Fishbein & Ajzen, 2010; Godin & Kok, 1996; Notani, 1998; Sheeran, 2002; Webb & Sheeran, 2006), including in studies modeling participation in 12-step groups for a substance use problem (Vederhus et al., 2015; Zemore & Kaskutas, 2009) and initiation of mental health treatment among veterans screening positive for an alcohol use disorder and/or other psychiatric disorder (Stecker et al., 2010). However, no known studies (besides that of our team) have used the TPB to model SUD treatment retention.

1.3. The current study

The first author led a study designed to test application of the TPB to modeling SUD treatment retention, and to identify specific factors contributing to greater retention and amenable to intervention. The study developed an initial pool of items based on the TPB and designed to measure treatment incentives and barriers, and administered these items, at treatment entry, to individuals seeking outpatient SUD treatment; program records were used to assess retention (operationalized as treatment completion). Prior, published analyses targeting the direct/global measures of TPB components (e.g., for attitude, agreement that “It would be extremely rewarding for me to complete my program at [program name]”) have supported the predictive power of the TPB, though subjective norm was not predictive: Model tests showed that a more favorable attitude and higher perceived control predicted greater intention to complete treatment (both ps < .001), and that greater intention to complete treatment predicted treatment completion (p < .05) (Zemore & Ajzen, 2014). These analyses also found that TPB components, as measured, correlated with existing treatment motivation and problem severity measures as expected, confirming measure validity (Zemore, 2012).

The current study extends the analyses described above by investigating the specific barriers to retention that were most strongly associated with treatment completion in this sample and most common, analyzing the detailed attitude belief measures (i.e., expected outcomes of treatment completion, such as cravings) and perceived control belief measures (i.e., perceived obstacles to treatment completion, such as difficulty affording treatment). This study does not analyze the subjective norm belief measures because published analyses (described above) showed no association between the direct subjective norm measure and treatment intention or completion, and because preliminary analyses for the current study revealed largely null associations between subjective norm belief measures and treatment completion. The study then uses the results to develop and validate the first known theory-based scale of treatment retention barriers, the Barriers to Retention Scale (BRS). While existing scales assess barriers to treatment initiation and reasons for leaving treatment (e.g., the Reasons for Leaving Treatment Questionnaire, Ball et al., 2006), there are no known scales that (a) systematically address psychological factors theoretically related to treatment retention, (b) can be used at treatment entry to predict retention, and (c) can be used to evaluate the independent contributions of distinct psychological factors to retention (e.g., whether social concerns predict retention when controlling for other attitudinal barriers). Study results should help to identify those at risk of premature treatment termination, and why; inform interventions to improve adherence; and support further research.

The present study conducts preliminary analyses of attitude and control belief measures separately among gender and racial/ethnic groups, so that the results and final scale adequately represent barriers that may be important overall but also only in under-represented groups. This is particularly important because treatment barriers vary across genders (Greenfield et al., 2007; Keen II et al., 2014; Pinedo, Zemore, Beltran-Giron, Gilbert, & Castro, 2020; Verissimo & Grella, 2017; Zemore et al., 2009) and racial/ethnic groups (Pinedo et al., 2018; Pinedo, Zemore, & Mulia, in press; Schmidt et al., 2007; Verissimo & Grella, 2017). Further, studies have found lower treatment retention among Black/African American and Latinx/Hispanic vs. White SUD treatment seekers (Bluthenthal et al., 2007; Lucabeche & Haney, 2018; McCaul et al., 2001; Mennis & Stahler, 2016; Saloner & Cook, 2013; Stahler & Mennis, 2018) due in part to socioeconomic factors (Jacobson et al., 2007; Saloner & Cook, 2013), and some studies have reported lower treatment retention among women vs. men (Greenfield et al., 2007; Lucabeche & Haney, 2018; McCaul et al., 2001). The study excludes participants reporting strong legal coercion to obtain treatment because treatment-related barriers may differ in this population (Klag et al., 2005) and because preliminary analyses revealed much weaker associations between TPB measures and treatment completion among coerced participants.

In view of the somewhat inconsistent prior findings on barriers to treatment retention and lack of clarity regarding how barriers may best be grouped, the current study offers only broad hypotheses. Multiple studies among those reviewed above (i.e., Ball et al., 2006; Godlaski et al., 2009; Laudet et al., 2009; Palmer et al., 2009; Seay et al., 2017; Yang et al., 2018) identified barriers to treatment retention in the general classes of treatment/change motivation (e.g., lack of perceived treatment need, lack of perceived treatment helpfulness); desire to use or continued use of alcohol/drugs; staff relationships; dissatisfaction with services; and logistic factors. Based in part on these findings, the study expected to identify barriers related to low perceived treatment need/value, concerns about missing substances (i.e., concerns about the immediate consequences of stopping or reducing substance use), and logistic factors, and that all such barriers would be associated with lower odds of treatment retention. The study expected that concerns about missing substances would present as a unique barrier distinct from perceived treatment need/value because a given client might feel that treatment is generally needed and valuable, but simultaneously feel concerned about the immediate impacts of making changes in alcohol or drug use. Because participants were surveyed immediately following intake, the study did not expect to observe salient barriers related to staff or services. The study also hypothesized that scores on the full BRS would correlate with existing motivational readiness scales (confirming construct validity) and higher drug and psychiatric severity (consistent with findings that higher severity on these dimensions predicts lower retention) (Syan et al., 2020; Zemore, 2012). For exploratory purposes, the study additionally examined associations between the BRS and measures of coercion and social desirability.

2. Materials and methods

2.1. Study site and recruitment

The current study administered surveys to 200 participants (with a 91% response rate) from August 2009 to August 2010 at a large, public, outpatient SUD treatment program in Contra Costa County, California (see Zemore, 2012; Zemore & Ajzen, 2014). Treatment involved group sessions conducted twice weekly; the program also invited clients to attend an educational group once weekly while waiting for a treatment slot to open. Programming addressed the individual in the context of family and social relationships, and emphasized involvement in 12-step groups. The typical prescribed length of stay was 8–12 weeks, with clients encouraged to attend weekly continuing care groups for a minimum of one year post-discharge.

Eligible participants included adults aged 18+ presenting for intake. English fluency was required; the study excluded those mandated to treatment via California Proposition 36 (i.e., for simple drug possession) and those with incapacitating mental or health problems. All participants had received an SUD/substance abuse diagnosis from the program. The present analyses included only those who did not report strong legal coercion (defined below), N = 156. The analytic sample was 63% White, and the 57 racial/ethnic minorities included 22 Latinx/Hispanic, 21 Black/African American, 8 Asian American/Pacific Islander, 3 American Indian/Alaska Native, and 3 Other participants.

The study solicited participants from their first or second education group, prior to entering formal treatment. Participants completed interviews immediately, onsite, using audio computer-assisted self-interviews. All participants completed consent procedures approved by the ethical review board of the Public Health Institute and received $50 gift cards. Comparisons of eligible clients who refused versus consented to participate showed no differences on gender, race/ethnicity, or age.

2.2. Measures

2.2.1. BRS item pool

The original study developed an initial item pool for the BRS based on a literature review, a focus group of clients initiating treatment, and consultation with an expert panel. Wording closely followed Ajzen’s (2002) recommendations for TPB measure development. Attitude belief measures asked respondents to rate their agreement that “If I continue attending treatment at [program name], rather than dropping out,” they would experience each of 24 positive outcomes (e.g., “be more likely to achieve my goals”) and 21 negative outcomes (e.g., “worry about what the staff are doing with my personal information”). Consistent with the current focus on barriers to retention and with results of preliminary analysis—which indicated generally weak or counter-theoretical associations between perceived positive outcomes and treatment completion—the present analyses target the 21 negative outcome expectancies exclusively. Perceived control belief measures asked respondents to rate their agreement that “Assuming that you wanted to complete your treatment plan at [program name], you could have trouble attending your treatment sessions” due to each of 12 possible barriers (e.g., “you can’t afford treatment”). Participants responded to all questions using 7-point Likert-type scales (e.g., strongly agree to strongly disagree).

2.2.2. Additional measures of motivational readiness and treatment coercion

Surveys included three existing scales addressing motivational readiness and deemed useful for scale validation; due to time constraints, these were balloted. The first ballot received the University of Rhode Island Change Assessment (URICA; McConnaughy et al., 1989), a 32-item scale designed to measure four stages of change readiness using 8 items each. Following Project MATCH (Project MATCH Research Group, 1997), the current study created total scores by reverse-coding the Precontemplation items and averaging all items (current α = 0.93). The second ballot received Freyer’s Treatment Readiness Tool (TREAT; Freyer et al., 2004), a 15-item scale designed to assess five stages of readiness to seek help for an alcohol and/or drug problem using 3 items per stage. Similar to coding for the URICA, the study reverse-coded the Precontemplation and Contemplation items and summed with the Preparation items. The study excluded the Action and Maintenance subscales because these subscales assess actual help-seeking (vs. motivation to seek help) and because their inclusion reduced the scale’s internal reliability (current α = 0.81). Participants receiving the TREAT also received the Treatment Motivation Questionnaire (TMQ; Ryan et al., 1995). The TMQ Internal subscale includes 11 items measuring personal, positive and negative reinforcement value for seeking and completing treatment, while the TMQ External subscale includes 5 items assessing legal and other coercion. The study averaged scores within each subscale.

Additionally, all surveys included both the Legal and Employment subscales of the Perceived Coercion Questionnaire (PCQ; Klag et al., 2006), with 5 items each. Prior research has supported the PCQ’s dimensional structure, internal reliability, test-retest reliability, and validity as a measure of coercion (Klag et al., 2006). The study again created total scores by averaging within subscales (Legal α = 0.86, Employment α = 0.85). Additionally, the study used one item from the Legal subscale to identify and exclude those experiencing strong legal coercion: Those who strongly agreed that “I felt pressured to enter this alcohol/drug treatment program because I was legally required” (constituting 22% of cases) were categorized as strongly coerced (vs. not). The study selected this item and threshold due to the item’s excellent face validity for assessing coercion, and because strong agreement with the Legal subscale differentiates those involuntarily admitted vs. voluntarily admitted to substance use treatment (Opsal et al., 2016).

2.2.3. Social desirability

Surveys included Ballard’s 11-item short form (Ballard, 1992) of the Marlowe-Crowne Social Desirability Scale (MC-SDS) (Crowne & Marlowe, 1960), identified as the best of the short forms (Loo & Loewen, 2004; Reynolds, 1982) (current α = 0.74). Used as a covariate here, the MC-SDS measures a response bias reflecting the need to “obtain approval by responding in a culturally appropriate and acceptable manner” (Crowne & Marlowe, 1960, p. 350).

2.2.4. Clinical and demographics variables

The study used items from the Alcohol, Drug, and Psychiatric Severity subscales of the Addiction Severity Index (ASI; McLellan et al., 1985; McLellan et al., 1980) to measure problem severity. The ASI is a standard, structured clinical interview showing high reliability and validity (McLellan et al., 1980). Other variables included self-identified gender, primary racial/ethnic group identification, age, education, employment status, annual household income, marital status, number of children under 18, and number of prior alcohol or drug treatment episodes.

2.2.5. Treatment discharge status

The current study’s main outcome was treatment discharge status (“complete” vs. “incomplete”), collected from program records. Program staff assigned discharge status codes, coding all those who left the program for any reason before completion as “incomplete.”

2.3. Analysis

To describe the sample, the study first examined baseline socio-demographics, clinical characteristics, and treatment completion rates, using t-tests and chi square tests to establish significant differences across genders (males vs. females) and racial/ethnic groups (Whites vs. people of color). (The study included only those identifying as male or female in gender-focused analyses, thus excluding one person identifying as transgender. Also, due to limited power, the study combined all racial/ethnic minorities and compared this group to White respondents.)

To identify the most influential and most common barriers to treatment completion, the study first conducted Pearson point-biserial correlations testing associations between all TPB belief items and treatment completion separately by gender and race/ethnicity, retaining any item showing at least a marginally significant association with treatment completion (p < .10) for any subgroup. The study specified a liberal item selection criterion to ensure that the BRS included all barriers potentially relevant to any major subgroup (for further study) and a liberal statistical significance threshold due to the small sample and exploratory stage of research.¹ Using only retained items, the study then conducted two exploratory factor analyses (using varimax rotation) in the sample overall, one each for the attitude and control belief items. The study flagged items that did not load highly on a given factor (i.e., lambda<0.40), loaded highly on multiple factors, and/or had poor face validity as indicators of a given factor for potential exclusion. Finally, the study replicated the point-biserial correlations with treatment completion and described item means for the sample overall. The study did not perform all scale analyses by gender and race/ethnicity because results would be difficult to address fully in a single paper; future analyses will address gender and race/ethnicity in more detail.

To further develop and validate the BRS, the study computed scale alphas for each subscale, as defined by results of the factor analyses above, and for the total scale. The study also created summary scores for each subscale and the total scale by averaging across relevant items. The study then examined correlations among each subscale, the total scale, and treatment completion. Further examining validity, the study also tested correlations between both subscale and total scale scores and clinical severity indicators, other motivational readiness measures, treatment coercion, and social desirability. Last, the study performed logistic regressions predicting odds of treatment completion from both the BRS subscale scores and total scores (considered separately) entered alone and adjusting for covariates. To identify covariates, analyses tested associations between treatment completion and all demographic variables, all clinical variables, and social desirability; variables that were associated with treatment completion at p < .10 were included in the model.

3. Results

3.1. Sample characteristics and treatment completion rates

Table 1 presents baseline sample characteristics and discharge status by gender and racial/ethnic minority status. Findings suggest a sample low on socioeconomic status, with, for example, 66% of participants reporting no employment and 42% reporting annual household incomes less than $10,000 at baseline. Overall, the sample was also high on SUD severity: For example, 63% reported some prior treatment, and less than 40% were abstinent from alcohol and drugs (asked separately) for the 30 days prior to baseline. Drug use was common, with 72% reporting a primary drug or poly-substance use problem at baseline. A minority (41%) completed treatment, consistent with national statistics.

Table 1.

Baseline sample characteristics and discharge status, overall and by gender and racial/ethnic minority status.

	Total (N=156)	Women (N=47)	Men (N=108)	People of Color (N=57)	Whites (N=99)
Age
Mean (SD) age	36.9 (11.01)	36.0 (11.44)	37.4 (10.82)	35.1 (10.45)	37.9 (11.24)
Education
% Less than high school	21.8	21.3	22.2	28.1	18.2†
% High school diploma	36.5	29.8	38.9	42.1	33.3
% Some college or more	41.7	48.9	38.9	29.8	48.5
Employment status
% Employed full-time	14.1	8.5	15.7	12.3	15.2
% Employed part-time	19.9	12.8	23.1	17.5	21.2
% Not employed1	66.0	78.7	61.1	70.2	63.6
Annual household income
% Under $10,000	41.6	56.5	35.5†	43.9	40.2
% $10,001–30,000	33.1	26.1	35.5	35.1	32.0
% $30,001 or more	25.3	17.4	29.0	21.1	27.8
Marital status
% Single	66.0	78.7	60.2*	61.4	68.7
% Married or partnered	34.0	21.3	39.8	38.6	31.3
Children
Mean (SD) number children	1.10 (1.25)	1.28 (1.21)	1.04 (1.27)	1.46 (1.33)	0.90 (1.17)**
% any children living at home	36.5	46.8	32.4†	42.1	33.3
Prior treatment
% None	36.5	38.3	36.1	47.4	30.3*
% 1–2 episodes	39.7	42.6	38.9	38.6	40.4
% 3 or more episodes	23.7	19.1	25.0	14.0	29.3
Drug of choice
% Alcohol only	27.6	27.7	27.8	35.1	23.2
% Other drugs or poly-drug	72.4	72.3	72.2	64.9	76.8
Addiction Severity Index scores
Mean (SD) ASI Alc. Severity	0.25 (0.23)	0.28 (0.25)	0.23 (0.22)	0.25 (0.25)	0.24 (0.22)
Mean (SD) ASI Drug Severity	0.09 (0.09)	0.09 (0.10)	0.09 (0.09)	0.07 (0.08)	0.10 (0.10)*
Mean (SD) ASI Psych. Severity	0.26 (0.24)	0.34 (0.24)	0.23 (0.23)**	0.19 (0.21)	0.30 (0.24)**
30-day abstinence
% abstinent from alcohol	39.7	40.4	39.8	45.6	36.4
% abstinent from drugs	37.2	40.4	36.1	38.6	36.4
Discharge status
% completed treatment	41.0	36.2	43.5	49.1	36.4

^***p<.001

^**p<.01

^*p<.05

^†p<.10;

males compared to females, Whites compared to people of color.

¹Includes both unemployed and out of the labor force.

Several differences emerged across gender and racial/ethnic groups. Compared to women, men were marginally more likely to report incomes in the higher brackets, significantly less likely to be single, and marginally less likely to report having children living at home. They also reported significantly lower psychiatric severity, as measured by the Addiction Severity Index (ASI). Compared to people of color, Whites were marginally more likely to report high levels of education, and they reported significantly fewer children. Whites were also significantly more likely than people of color to report prior treatment episodes (especially multiple episodes), and they reported significantly higher drug and psychiatric severity, again as measured by the ASI. Nonetheless, gender and racial/ethnic groups did not differ on treatment completion.

3.2. BRS item results

Among the 33 attitude and perceived control belief items, 20 were associated with treatment completion at p < .10 for at least one gender or racial/ethnic group (people of color or Whites); Table 2 shows factor loadings, correlations with treatment completion, and descriptive statistics for those 20 items in the sample overall, grouping items into subsets based on factor scores. As this table shows, factor analysis suggested three attitudinal dimensions, labeled Low Perceived Treatment Need/Value, Social Concerns, and Concerns about Missing Substances. Notably, the analysis discriminated between Low Perceived Treatment Need/Value and Concerns about Missing Substances, showing little cross-loading for items within these dimensions. Further, all items assessing Concerns about Missing Substances were associated with treatment completion in the sample overall (whereas only two other attitude items, both loading on Social Concerns, were correlated with treatment completion); items assessing Concerns about Missing Substances were also among the most strongly endorsed, as shown by item means (particularly for the craving item). Item loadings are reasonably concordant with theory/prior research, and all items show reasonable face validity as measures of the underlying factor, with the exception of the fifth Social Concerns item (regarding feeling frustrated with program rules). This item also showed moderately high loadings on both Factors 1 and 2 (lambdas = 0.45 and 0.49 respectively). Therefore, the study dropped the item from further analysis.

Table 2.

Barriers to Retention Scale (BRS) items: Factor loadings (with factor analyses performed separately for attitude and perceived control items) and item descriptives.

	Factor 1 Loading	Factor 2 Loading	Factor 3 Loading	Pearson r for Tx. Compl.	Item Mean (SD)1
ATTITUDE ITEMS: How much do you agree with the following statements? If I continue attending treatment at (PROGRAM) I will…
1. Low Perceived Treatment Need/Value
Feel like I am wasting my time because my alcohol/drug use is not my main problem.	.86	.15	.14	−.06	2.05 (1.44)
Feel like I am wasting my time because I’m not sure I need to quit alcohol and drugs completely.	.76	.12	.23	−.06	1.88 (1.37)
Feel isolated because I am different from people here.	.70	.33	.11	−.07	2.35 (1.49)
Feel like I am wasting my time because treatment doesn’t usually work.	.63	.29	.03	−.01	2.05 (1.34)
2. Social Concerns
Lose important relationships.	.18	.74	.08	−.03	2.40 (1.87)
Feel disconnected from people who are important to me.	.21	.66	−.05	−.03	2.62 (1.73)
Worry that people I know are going to find out about my problem.	.16	.61	.10	−.10	2.38 (1.62)
Feel uncomfortable because I will be asked to share private things.	.20	.58	.23	−.21 **	2.78 (1.79)
Feel frustrated by all the rules.2	.45	.49	.33	−.15†	2.91 (1.77)
3. Concerns about Missing Substances
Miss getting drunk or high.	.07	.08	.86	−.19 *	3.34 (2.05)
Miss my old lifestyle.	.35	.00	.80	−.20 *	3.15 (1.99)
Have a lot of cravings.	.10	.12	.66	−.24 **	3.50 (2.06)
Have a hard time experiencing intimacy/sex because I will not be using alcohol/drugs.	.03	.46	.54	−.16†	2.21 (1.56)
PERCEIVED CONTROL ITEMS: Assuming you wanted to complete treatment at (PROGRAM), how much do you agree that you could have trouble attending sessions because…
4. Personal Limitations
You are sometimes tempted to break rules.	.80	.14	--	−.17 *	2.78 (1.80)
You might be drunk, high, or hungover.	.78	.06	--	−.07	1.89 (1.54)
You tend to forget things or are not very organized.	.62	.32	--	−.14†	3.00 (1.91)
You don’t like being told what to do.	.58	.28	--	−.10	2.82 (1.82)
5. Basic Logistic Barriers
It might be difficult to find transportation to or from sessions.	.05	.85	--	−.21 *	3.03 (2.08)
It might be difficult to fit treatment into your schedule.	.22	.73	--	−.02	3.12 (1.91)
You often feel ill, tired, or down.	.40	.59	--	−.23 **	2.90 (1.80)

^***p<.001

^**p<.01

^*p<.05

^†p<.10;

tx. compl. is treatment completion.

¹Scale responses range from 1 (disagree very strongly) to 7 (agree very strongly).

²Item dropped from scale due to low face validity and high loadings on multiple factors.

Factor analysis suggested two dimensions of perceived control barriers, labeled Personal Limitations and Basic Logistic Barriers. Generally, items loading on Personal Limitations seem to address personality, preferences, and habits, whereas items loading on Basic Logistic Barriers address concrete, practical barriers. Two items from each dimension were correlated with treatment completion, though more strongly for those assessing Basic Logistic Barriers. Endorsement of the Basic Logistic Barriers items was also somewhat higher than for the Personal Limitations items. Items show reasonable face validity as measures of the underlying factors, and all were retained.

3.3. BRS scale results

Table 3 displays scale alphas for each BRS subscale, as defined by the factor analysis results (dropping one Social Concern item), and for the 19-item total scale. This table also shows correlations among each subscale, the total scale, and treatment completion. Internal reliability of each subscale was fair to good ( αs = 0.65–0.80) and very good (α = 0.88) for the total scale. Subscale scores were only moderately correlated with each other. Scores on both the total score and all but one subscale (i.e., Low Perceived Treatment Need/Value) were associated with treatment completion, though the correlation was only marginally significant for Social Concerns; the strongest correlations emerged for Concerns about Missing Substances and Basic Logistic Barriers.

Table 3.

Barriers to Retention Scale (BRS) subscales and total scale: Alphas, inter-correlations, and correlations with treatment completion.

	1. Low Perceived Treatment Need/Value (4-item α=.80)	2. Social Concerns (4-item α=.65)	3. Concerns about Missing Substances (4-item α=.75)	4. Personal Limitations (4-item α=.70)	5. Basic Logistic Barriers (3-item α=.65)	6. Total Scale (19-item α=.88)	7. Treatment Completion
1. Subscale 1	1
2. Subscale 2	.52***	1
3. Subscale 3	.39***	.35***	1
4. Subscale 4	.44***	.57***	.56***	1
5. Subscale 5	.45***	.44***	.38***	.51***	1
6. Total Scale	.72***	.76***	.74***	.83***	.72***	1
7. Tx. Compl.	−.07	−.14†	−.26***	−.17*	−.22**	−.23**	1

^***p<.001

^**p<.01

^*p<.05

^†p<.10;

tx. compl. is treatment completion.

Table 4 shows correlations between BRS subscale scores and total scores and the following: clinical severity indicators, other motivational readiness measures, treatment coercion, and social desirability. Total scale scores were significantly (or marginally significantly) associated with every indicator tested: that is with higher alcohol, drug, and psychiatric severity (as defined by the ASI), lower motivational readiness (as defined by the URICA, TREAT, and TMQ Internal), higher treatment coercion (as defined by the TMQ External and PCQ), and lower social desirability (as defined by the MC-SDS). While all subscales tended to be positively associated with the coercion indicators (especially the PCQ), different patterns emerged for the other indicators. For example, the Concerns about Missing Substances subscale and (to some extent) both perceived control subscales were positively associated with the severity indicators, but Low Perceived Treatment Need/Value and Social Concerns were almost entirely uncorrelated with these indicators. By contrast, the Concerns about Missing Substances subscale was almost entirely uncorrelated with the motivational readiness scales, whereas the Low Perceived Treatment Need/Value and Social Concerns subscales were very strongly associated with those scales overall. Among all subscales, the Concerns about Missing Substances subscale and perceived control subscales were most strongly associated with social desirability.

Table 4.

Barriers to Retention Scale (BRS) subscales and total scale: Correlations with clinical severity, other motivational readiness measures, treatment coercion, and social desirability.

	1. Low Perceived Treatment Need/Value	2. Social Concerns	3. Concerns about Missing Substances	4. Personal Limitations	5. Basic Logistic Barriers	6. Total Scale
ASI Alcohol	.01	.03	.31***	.11	.05	.15†
ASI Drug	−.06	.06	.26***	.25**	.13	.18*
ASI Psychiatric	.03	.14†	.37***	.36***	.32***	.34***
URICA1	−.56***	−.32**	−.04	−.24*	−.19	−.35**
TREAT1	−.60***	−.48***	−.22†	−.46***	−.32**	−.53***
TMQ Internal1	−.42***	−.22†	−.07	−.14	−.14	−.24*
TMQ External1	.21†	.15	.21†	.28*	.12	.26*
PCQ Legal	.33***	.34***	.27***	.22**	.31***	.39***
PCQ Employ.	.30***	.38***	.29***	.28***	.18*	.38***
MC-SDS	−.02	−.01	−.33***	−.30***	−.23**	−.25**

^***p<.001

^**p<.01

^*p<.05

^†p<.10;

ASI is Addiction Severity Index; URICA is University of Rhode Island Change Assessment Scale; TREAT is Treatment Readiness Tool; TMQ is Treatment Motivation Questionnaire; PCQ is Perceived Coercion Questionnaire; Employ. is Employment; MC-SDS is Marlowe-Crowne Social Desirability Scale.

¹Measure balloted (N=78).

Table 5 shows the unadjusted and adjusted logistic regressions predicting odds of treatment completion from BRS subscale scores and total scores (considered separately). Results of the unadjusted models reveal significant effects for both the Concerns about Missing Substances subscale (even accounting for other subscales) and the total scale on treatment completion. These effects were robust when adjusting for psychiatric severity and social desirability, both of which were nonsignificant in adjusted models. None of the alternative motivational readiness scales (i.e., the URICA, TREAT, and TMQ Internal) or coercion measures (i.e., TMQ External and PCQ scales) included in the present study were significantly associated with treatment completion (not shown).

Table 5.

Barriers to Retention Scale (BRS) subscales and total scale: Logistic regressions of treatment completion.

Unadjusted Models
Model 1: Subscales (R2=.13)	OR	(CI)
Low Perceived Treatment Need/Value	1.26	(0.86, 1.87)
Social Concerns	0.88	(0.61, 1.27)
Concerns about Missing Substances	0.68*	(0.51, 0.92)
Personal Limitations	1.03	(0.71, 1.50)
Basic Logistic Barriers	0.83	(0.63, 1.09)
Model 2: Total Scale (R2=.08)	OR	(CI)
TPB Retention Barriers Scale	0.61**	(0.42, 0.87)
Adjusted Models 1
Model 1: Subscales (R2=.14)	AOR	(CI)
Low Perceived Treatment Need/Value	1.19	(0.79, 1.78)
Social Concerns	0.87	(0.60, 1.26)
Concerns about Missing Substances	0.72*	(0.53, 0.98)
Personal Limitations	1.09	(0.74, 1.61)
Basic Logistic Barriers	0.86	(0.65, 1.14)
Model 2: Total Scale (R2=.10)	AOR	(CI)
TPB Retention Barriers Scale	0.68*	(0.47, 0.99)

^***p<.001

^**p<.01

^*p<.05

^†p<.10.

¹Adjusted models control for ASI Psychiatric Severity and Marlowe-Crowne Social Desirability.

4. Discussion

4.1. Results and implications

The current study identified five distinct domains of attitudinal and perceived control barriers, and successfully developed and validated the 19-item Barriers to Retention Scale (BRS), with subscales corresponding to the five barrier dimensions. The study used rigorous methods, drawing on theory, prior research, and item and scale analysis to develop the BRS. Both BRS subscales and the total scale showed acceptable internal reliability, and the total scores were associated with other treatment readiness and clinical severity measures as predicted. Although total BRS scores were also associated with social desirability, the scale nonetheless predicted treatment completion when controlling for social desirability (and psychiatric severity).

The identified barrier dimensions largely corresponded to those hypothesized and included Low Perceived Treatment Need/Value, Social Concerns, Concerns about Missing Substances, Personal Limitations, and Basic Logistic Barriers. Associations for individual barrier items were modest overall, with some notable exceptions (e.g., concerns about craving; concerns about having to share private things; concerns about transportation; and expectations of feeling ill, tired, or down). Nonetheless, subscale scores for all primary dimensions excepting Low Perceived Treatment Need/Value were associated with treatment completion. Both researchers and providers often emphasize Low Perceived Treatment Need/Value as a barrier to treatment utilization, but this barrier may be more relevant to treatment initiation than retention.

Concerns about Missing Substances emerged as a surprisingly robust predictor of treatment completion. Among all subscales, only Concerns about Missing Substances independently predicted treatment completion in multivariate models including other subscales and covariates. Also, endorsement of items assessing Concerns about Missing Substances was high, with the four-item subscale including those three items most highly endorsed among all scale items: concerns about having a lot of cravings, missing my old lifestyle, and missing getting drunk. This suggests that concerns about the immediate consequences of stopping or reducing alcohol or drug use—including physical impacts as well as more psychological effects—may be salient among those attending treatment, and can interfere with treatment retention even if perceived treatment need/value is high.

Effect sizes for Concerns about Missing Substances and the total BRS score were modest; models without covariates explained 8–13% of the variance in treatment completion. Nonetheless, these results must be viewed in the context of the lack of predictive power for all other variables studied. None of the other motivational readiness scales included here was associated with treatment completion, suggesting that the BRS makes an important contribution. Further, while psychiatric severity (negatively) and social desirability (positively) predicted treatment completion in bivariate models, these covariates were nonsignificant in the multivariate models, which may suggest that effects for both variables on treatment completion are at least partially mediated by Concerns about Missing Substances. Supporting this supposition, both psychiatric severity and social desirability were strongly and significantly associated with Concerns about Missing Substances (Table 4). Findings for psychiatric severity may suggest that substance use can play an instrumental role in helping those with psychiatric problems to cope with negative feelings and experience positive feelings—and that this functional reliance on substance use can heighten risk for premature treatment termination.

The present findings—and findings for Concerns about Missing Substances in particular—suggest that programs enhance their focus on client concerns about the immediate negative impacts of stopping or reducing substance use as an issue separate from low problem recognition or perceived treatment need. For example, programs might emphasize strategies to cope with withdrawal, cravings, and the loss of a favored lifestyle that can come with change early and often in treatment, particularly for those with recent, severe alcohol and/or drug use. Programs might also highlight how they will address the negative impacts of stopping or reducing substance use in their marketing, which could increase optimism about the treatment process and enhance both treatment initiation and retention.

Future research could support efforts to address concerns about missing substances by investigating several approaches to doing so. As one example, future research might examine whether psychosocial interventions addressing how to cope with cravings and enjoy life (including relationships and sex) without substances can improve retention. Future research might also study whether a recovery plan including moderation goals, at least for some substances, can improve retention. Prior research has suggested that some people who receive treatment may continue to engage in heavy drinking, yet report good health outcomes up to 9 years after treatment (Witkiewitz, Pearson, Wilson, Stein, and Votaw, 2020); further, not all those who have recovered from an alcohol or drug problem consider abstinence to be fundamental to their definition of recovery (Kaskutas et al., 2014). Program acceptance of moderation goals (at least for some people and some substances) may help to head off potential negative consequences of abstinence and retain treatment contact; as appropriate, programs might renegotiate substance use goals at a later stage.

Third, research might address whether enhancing access to medications addressing craving and psychiatric symptoms can improve treatment retention. SUD medications exist for several drug classes and can reduce cravings and improve treatment outcomes (Douaihy et al., 2013). Additionally, studies show that psychiatric medications can be effective for addressing both psychiatric and substance use problems among those with co-occurring disorders (though yielding mixed results for substance use; Iqbal et al., 2019; Murthy & Chand, 2012). Increasing access to SUD and psychiatric medications, which remains limited in public treatment (Substance Abuse and Mental Health Services Administration, 2019b), may thus improve treatment retention, psychiatric outcomes, and SUD outcomes.

A complication related to the incorporation of both moderation goals and psychiatric medications at treatment outset, however, is that proper diagnosis of psychiatric conditions may require a period of abstinence lasting up to several months (Iqbal et al., 2019). Thus, clients presenting with psychiatric symptoms may be best served by a period of abstinence combined with psychosocial intervention addressing concerns about missing alcohol and drugs, followed by a re-evaluation of psychiatric symptoms, and potentially substance use goals, subsequently.

Last, future research would be helpful to enrich our understanding of concerns around missing substances, and to identify common sources of these concerns. Concerns about missing substances may include concerns not well-addressed in the current scale, such as concerns about the loss of identity that may come with change; future research could help to elaborate these. Research addressing how prior experiences (such as prior change attempts and treatment experiences) may have contributed to current fears could help to address those fears.

4.2. Potential limitations

First, the study’s small sample size prevented analysts from assessing differential scale performance in relevant subgroups (e.g., racial/ethnic minorities and women). Related, the study administered surveys in English only. Results may thus fail to generalize to key racial/ethnic minority and gender subgroups (as well as recent immigrants), who could show barriers to treatment retention differing from those of the total population. The study team encourages future studies that validate the BRS in additional subgroups, as a better understanding of differences in barriers to treatment retention may help ameliorate disparities in retention. Future research might pay particular heed to women with children. Women with children constituted only a very small minority (N = 22) of the present sample, yet among this group, findings suggested a trend toward an association between lower odds of treatment completion and higher agreement that one might have trouble completing treatment because “it might be difficult to find and pay for childcare” (r = −0.33, p = .135). The current study did not include the childcare item in the BRS as the item did not meet inclusion criteria, but this finding does suggest that women with children may have unique concerns (see also Seay et al., 2017).

A second limitation is that the current study excluded participants reporting strong legal coercion to attend treatment (i.e., 22% of the sample). The present findings thus provide little insight about barriers to treatment completion among those strongly coerced into treatment, which is a significant limitation given the high value of understanding treatment and recovery processes in this population. The study team encourages future research to model factors that might contribute to treatment completion among those strongly coerced into treatment.

Finally, the study used liberal inclusion criteria to select scale items. Thus, the BRS may include some barriers that are not very relevant to treatment completion overall, despite the fact that scale reliability and validity tests were supportive.

5. Conclusions

Analyzing a diverse sample, the current study produced the first theoretically derived scale of treatment retention barriers, and identified key dimensions of treatment retention barriers. The study also identified concerns about missing substances as a particularly important barrier. As treatment retention has received considerably less attention than treatment initiation, these are advances that may ultimately lead to better recovery outcomes. Study results pointed to a number of avenues for improving treatment retention, and suggest that the BRS may be useful for screening in clinical practice (i.e., to identify and intervene with those at risk for premature termination) and in supporting further research on treatment retention barriers. The study team looks forward to future research to further refine the BRS, including validation of the scale in under-represented subgroups and application to understanding disparities in treatment retention.