Medication-Assisted Treatment and Violent Outcomes in Community-Based Offenders with Alcohol and Drug Use Problems

Abstract

Objective

This study explored medication-assisted treatment (MAT), the combined use of medication and psychosocial treatment, as a strategy for reducing violent outcomes in community-based offenders. The primary aims were to: 1) examine associations between participant characteristics and treatment adherence; 2) examine associations between treatment adherence and substance use; 3) examine associations between treatment adherence and violent outcomes; and 4) determine whether associations between treatment adherence and violent outcomes may be attributable to reductions in substance use.

Method

Baseline interviews were completed with 129 male offenders in community-based treatment prior to their first MAT appointment. Follow-up interviews (n = 91) were conducted approximately 90 days later.

Results

Participant age was associated with medication adherence. Medication nonadherence was associated with at least occasional alcohol use, but not drug use. Conversely, missing several counseling sessions was associated with at least occasional drug use, but not alcohol use. Results of multivariable analyses suggested MAT may be effective in reducing violent outcomes, and victimization specifically, through reductions in alcohol use.

Conclusion

Findings provide evidence supporting MAT as an intervention for victimization. Continued efforts are needed to explore strategies to promote treatment adherence and reduce violent outcomes in community-based offenders with alcohol and drug use problems.

Medication-assisted treatment (MAT) involves the use of medication combined with psychosocial treatment, such as cognitive behavioral therapy or other counseling services, to improve outcomes among individuals with substance abuse problems (Pecoraro, Ma, & Woody, 2012). Despite evidence regarding treatment effectiveness, MAT is not available to many offenders with alcohol and drug use problems in the United States (Friedmann et al., 2012; Perry et al., 2014). Prior research suggests that MAT has the potential to be an effective approach for reducing violence in this population (Gerra et al., 2011; Havnes et al., 2012), but further research is needed. To that end, this paper reports on the findings of a prospective, naturalistic study examining MAT as a strategy for reducing violent outcomes in community-based offenders, including probationers and parolees, with alcohol and drug use problems.

Background

Alcohol and drug use problems are widely recognized as risk factors for violent behavior (Grann & Fazel, 2004). Not surprisingly, then, we find very high rates of substance use problems in correctional populations. In the United States, for example, estimates suggest that approximately eight out of 10 offenders in state prisons have a history of illicit drug use, with two-thirds having a history of regular illicit drug use, and almost half of offenders incarcerated for violent crimes in state and federal prisons meet criteria for a drug use disorder (Mumola & Karberg, 2006). Unfortunately, even after successful completion of treatment programs while incarcerated, many offenders with alcohol and drug use problems relapse following their release into the community (Binswanger et al., 2012), and relapse increases risk of recidivism and violence (Dowden & Brown, 2002; Haggård-Grann, Hallqvist, Lånsgtröm, & Möller, 2006).

Beyond the risk to public safety, offenders with alcohol and drug use problems also are at increased risk for victimization after their release into the community (Zweig, Yahner, & Rossman, 2012). Alcohol use at the time of the offense is frequently reported by victims, as well as offenders (Lipsey, Wilson, Cohen, & Derzon, 1997), and offenders with alcohol and drug use problems may be more likely to find themselves in situations in which they could be victimized (e.g., while drug seeking) (Pizarro, Zgoba, & Jennings, 2011). In fact, research suggests that the overlap between violent victimization and perpetration is greatest among adults with substance use disorders compared to those with other psychiatric disorders (Johnson, Desmarais, Van Dorn, & Grimm, 2015). Thus, substance abuse treatment represents a potential strategy for reducing risk of violent outcomes, and ultimately, for promoting community reintegration among offenders with alcohol and drug use problems (Casey & Day, 2014; Resor & Blume, 2008).

MAT is a state-of-the-art approach to treating substance use problems (Substance Abuse and Mental Health Services Administration, 2015) and has been used to treat opioid dependence for more than 40 years (Pecoraro et al., 2012). There are many different medications available for use in MAT that operate through varied mechanisms, such as methadone, buprenorphine, and naltrexone. A preponderance of evidence demonstrates the effectiveness of MAT—and naltrexone, in particular—in treating opioid and alcohol dependence (e.g., Krupitsky & Blokhina, 2010; Rösner, Leught, Lehert, & Soyka, 2008). However, the implementation of MAT remains relatively limited in correctional settings in the United States (Fiscella, Moore, Engerman, & Meldrum, 2004). This is especially true for the treatment of community-based offenders (Friedmann et al., 2012; Patapis & Nordstrom, 2006). Many policymakers, providers, and members of the public are of the view that MAT allows offenders to simply substitute one drug for another, without taking responsibility for their recovery (Perry et al., 2014: Friedman et al., 2012). Others cite concerns regarding treatment coercion in this population (O’Brien & Cornish, 2006). Others, still, argue that the research evidence from studies conducted in correctional settings is insufficient (Patapis & Nordstrom, 2006).

Extant research provides some support for the effectiveness of MAT in reducing criminal behavior, presumably through its effects on alcohol and drug use. A review of MAT for opioid dependence among prison and post-release populations, for example, identified some positive, but mixed findings across studies regarding the impact of MAT on criminal activity (Hedrich et al., 2011). However, some research has found MAT to be associated with increased likelihood of recidivism (Mitchell, Wilson, & MacKenzie, 2007). Recent meta-analytic results suggest that these inconsistencies may reflect the medication prescribed as part of the MAT under investigation; reductions in criminal activity are associated with some medications, such as naltrexone, but not others, such as methadone (Perry et al., 2014).

Only a handful of studies have examined the effectiveness of MAT in reducing violent behavior (Gerra et al., 2011; Havnes et al., 2012). Though limited, the evidence is promising. For example, among 3,221 offenders who started opioid maintenance treatment, those who completed the treatment demonstrated lower rates of violent crime in the three years after compared to prior to treatment (Havnes et al., 2012). There is even less empirical evidence regarding the effectiveness of MAT in reducing victimization. To our knowledge, no studies have examined the effects of MAT on rates of victimization; instead, studies have examined victimization histories as predictors of MAT outcomes (Branstetter, Bower, Kamien, & Amass, 2008; Oviedo-Joekes et al., 2011; Pirard, Sharon, Kang, Angarita, & Gastfriend, 2005).

Treatment adherence may be a key variable in understanding the effects of MAT on violent outcomes. Clinical guidelines caution against the use of some MATs for patients who are not sufficiently motivated and, thus, likely to be nonadherent (Center for Substance Abuse Treatment, 2004). Treatment adherence may be of heightened concern within the criminal justice context (Resor & Blume, 2008), especially among community-based offenders (Zanis, Coviello, Lloyd, & Nazar, 2009) and with respect to MAT (Coviello et al., 2012; Friedmann et al., 2012). Though adherence to the prescribed MAT regimen per se has not been tested vis-à-vis violent outcomes, research on related constructs suggests that treatment nonadherence may be associated with higher rates of violence perpetration. For example, in the study by Havnes and colleagues (2012) mentioned earlier, rates of violent crime for men who failed to complete MAT were higher after compared to before they began treatment. Findings from another study of 300 heroin-dependent outpatients referred for long-term methadone treatment similarly found that those who were supervised or received incentives to participate in MAT, and ostensibly were more likely to follow the prescribed regimen, were less likely to perpetrate violence compared to those who were not supervised or who did not receive incentives (Gerra et al., 2011).

The Present Study

The present study explores MAT as a strategy for reducing violent outcomes in a sample of male community-based offenders with alcohol and drug use problems, with a focus on treatment adherence. The primary aims were to: 1) examine associations between participant characteristics and treatment adherence, including attendance at counseling sessions and medication adherence; 2) examine associations between treatment adherence and alcohol and drug use; 3) examine associations between treatment adherence and violent outcomes, including violence and victimization; and 4) determine whether associations between treatment adherence and violent outcomes may be attributable to reductions in alcohol and/or drug use.

Method

Participants

The study recruited probationers and parolees enrolled in an outpatient substance abuse treatment program in a large Midwest region of the United States. Participants were referred for MAT services at their request or after recommendation by their substance abuse counselor or their supervising officer from the state Department of Corrections. The baseline sample included 129 male offenders with a recent history of alcohol or drug use problems who were currently engaged in substance abuse treatment at a community-based program. The follow-up sample included 91 participants (70.5% of the baseline sample) who continued in treatment through the study period. Reasons for treatment discontinuation included unsuccessful discharge (e.g., non-attendance, or program non-compliance; 50.0%, n = 19), incarceration (44.7%, n = 17), and inappropriate placement (e.g., higher level of care was needed, resulting in transfer to a detox or residential program; 5.3%; n = 2).

Treatment Procedures

The pharmacotherapy agency that provided MAT services to the participants specialized in treating chemical dependence and was accredited by The Joint Commission (TJC). The initial pharmacotherapy visit included a medical assessment, toxicology testing for substances of abuse, and MAT recommendations. The MAT prescribed for each participant depended on the toxicology report and the medical assessment completed at the time of program intake. If toxicology screening was positive for opiates, then medications that act to partially inhibit brain receptor receptivity to opiates (e.g., a combination of buprenorphine and naloxone) were prescribed. The purpose of the mixed agonist/partial antagonist medication was to reduce craving and promote abstinence. No more than a 7-day supply was dispensed. Clients returned to the agency for weekly toxicology screening until abstinence from opiates was achieved.¹

The target MAT was naltrexone, a full antagonist medication that was only prescribed after client abstinence was attained. Originally developed to treat heroin dependence, naltrexone is also approved and has demonstrated effectiveness is treating alcohol dependence (Chandler et al., 2009). The 30-day extended release Vivitrol^® depot naltrexone was used to promote adherence to the medication regimen and improve treatment retention (Gray, Desmarais, Cohn, Doherty, & Knight, 2015). Typically, initial prescriptions of naltrexone in pill format were stipulated to rule out side effects from the medication, and then a 1-month extended release injection was applied. A different course of MAT treatment was followed for alcohol abuse. If toxicology screening was positive for alcohol, naltrexone was prescribed from the beginning because the human body metabolizes alcohol differently and more quickly than opiates. Applications of naltrexone could be initiated several hours after a positive toxicology screen for alcohol, after it was metabolized.

In addition to medication, study participants were offered psychosocial services customized to their individual needs. Treatment consisted of one-on-one sessions with a primary counselor and as many as four group sessions per week. The counselors in the program used evidence-based treatment approaches, including cognitive behavioral therapy and motivational interviewing. In addition, this outpatient program was linked with a MAT provider, to facilitate MAT referrals for those who identified as opiate-dependent or those who requested MAT. (Further details on the treatment provider and services are available in Gray et al., 2015.)

Study Procedures

Potential participants were contacted at the time of their referral for substance abuse treatment services. The treatment agency provided a list of referred probation and parolees to the research team, and graduate-level research staff contacted the clients outside of group or individual counseling meetings. Following written informed consent, in-person baseline interviews were completed prior to the scheduled appointment with the MAT service provider (typically within seven days). The baseline interviews queried personal and family background, criminal activity and legal involvement, alcohol and drug use, and treatment history using questions drawn from the TCU Criminal Justice Comprehensive Intake (TCU CI), which has been used in four national treatment effectiveness studies (see Joe, Simpson, Greener, & Rowan-Szal, 2004). Violent outcomes in the three months prior to treatment (or prior to incarceration for participants incarcerated immediately preceding the study period) were assessed using the MacArthur Community Violence Screening Inventory (MCVSI; Steadman et al., 1998), an accepted standard for the measurement of violence (Harris, Oakley, & Picchioni, 2013).

In-person follow-up interviews with participants were scheduled approximately 90 days after the baseline interview ± 14 days. During these follow-up interviews, participants again responded to questions regarding their personal and family situation, criminal activity and legal involvement, alcohol and drug use, and violent outcomes in the 90-day period following baseline. They additionally reported on how often they took the medications they received during treatment. To promote recall accuracy, researchers helped participants anchor their responses to the 90-day period of treatment that followed the baseline interviews using milestone events (e.g., holidays and birthdays) (Sobell & Sobell, 1992). Participants received gift cards as compensation for completion of each research interview, to a maximum of $60.

There was considerable variability in the follow-up period (M = 120.64 days, SD = 109.31, Range = 41 – 853). The vast majority of participants available at follow-up completed their interviews within 120 days of their baseline interviews (84.6%, n = 77). Three participants (3.3%) were incarcerated or transferred to a higher level of care during the follow-up period, but subsequently returned to treatment while the study was ongoing. For a handful of participants (4.4%, n = 4), researchers knew ahead of time that supervision was ending early (e.g., the participant was being transferred to a different level of care, probation or parole was being revoked) and follow-up interviews were completed more than 14 days prior to the target interview date. In these circumstances, the follow-up interviews queried behaviors occurring since the baseline interview. Thus, events would not be double counted if they happened less than 90 days after the baseline interview for these “early follow-up” participants. All participants available at follow-up were included in the present analyses, adjusting for length of follow-up.²

Variables

Treatment adherence

Participants reported separately on their adherence to counseling and medication in the 90-day follow-up period on 5-point, ordinal scales (1 = never missed a session / dose; 2 = missed only one or two sessions / times; 3 = missed several times, but attended / took at least half; 4 = attended / took less than half; 5 = never attended a session / never took the medication). For analyses, two dichotomous variables were created: (1) counseling attendance (0 = attended all or most sessions, 1 = missed several or more sessions), and (2) medication adherence (0 = took all or most doses, 1 = missed several or more doses).

Substance use

Alcohol use and drug use were assessed at baseline and follow-up. At baseline, participants self-reported frequency of use in the past 30 days on a 4-point, ordinal scale (1 = none / not at all, 2 = occasional, 3 = weekly, or 4 = daily) for alcohol and each of 19 drugs, such as heroin, heroin and cocaine combined, other street opiates, and prescription medications not prescribed by a doctor, excluding tobacco. At follow-up, frequency of alcohol and drug use in the past 90 days was measured as described above. Responses then were coded to capture the highest reported frequency use during the referent period to create the alcohol use and drug use variables, respectively. For analyses, responses were collapsed to reflect a dichotomous coding of 0 = none and 1 = at least occasional use. Participants additionally were asked to identify the substance that caused them the most problems during the referent period.

Violent outcomes

Prevalence of violence perpetration and victimization during 90-day referent periods were assessed at baseline and follow-up using the MCVSI. The MCVSI includes eight behaviorally-based self-report items. Items assess: 1) pushing, grabbing, or shoving; 2) kicking, biting, or choking; 3) slapping; 4) throwing an object; 5) hitting with a fist or object; 6) sexual assault; 7) threatening with a weapon in hand; and 8) using a weapon. For each item, participants were first asked if someone did this to them, then were asked if they did this to someone else. When an affirmative answer was given, a follow-up question asked for the number of times that behavior or victimization occurred in the referent period. Consistent with previous psychometric analyses (Michie & Cooke, 2006), we found that the measure’s eight items mapped onto unidimensional factors of victimization and violence, respectively, in the current sample. Each scale demonstrated strong internal consistency (Cronbach’s αs ≥ .75).

Following the approach used in prior research (e.g., Desmarais et al., 2014; Johnson et al., 2015), violence was operationalized as a positive response to at least 1 item querying if the participant did this behavior to someone else (0 = no, 1 = yes), and victimization, a positive response to at least 1 item querying if someone else did this behavior to the participant (0 = no, 1 = yes). We additionally created an any violent outcome variable that reflected a positive response to any of MCVSI behaviors (0 = no, 1 = yes).

Participant characteristics

At baseline, participants reported on their legal and treatment history using the TCU CI, including their age at first arrest (in years), number of lifetime arrests¸ and prior substance abuse treatment (0 = no, 1 = yes). The TCU CI additionally queried sociodemographic characteristics at baseline, such as age (in years), race/ethnicity (0 = white, 1 = black, or 2 = other), employment status (0 = unemployed, 1 = employed), relationship status (0 = not living with a spouse or partner, 1 = living with a spouse or partner), and education level (0 = did not complete high school/GED, 1 = completed high school/GED).

Analytic Strategy

Prior to conducting analyses addressing our research aims, we first computed descriptive statistics for all variables. We also conducted bivariate analyses to compare baseline characteristics between participants who completed follow-up and those who were not available at follow-up, to test for attrition bias. To address our first three research aims, we conducted bivariate logistic regression analyses to examine associations between participant characteristics and treatment adherence (Research Aim 1), treatment adherence and substance use (Research Aim 2), and treatment adherence and violent outcomes (Research Aim 3), adjusting for length of follow-up. Finally, we conducted multivariable logistic regression analyses to examine the effects of treatment adherence on each of the violent outcome measures, above and beyond substance use, to determine whether associations between treatment adherence and violent outcomes may be attributable to reductions in substance use (Research Aim 4). Participant characteristics that differed between participants who completed follow-up and those who were not available at follow-up and variables that showed bivariate associations with violent outcomes at p < .10 were included as covariates in these multivariable models. All analyses were conducted with SPSS, v.22. Statistical significance was set at p < .05, unless otherwise specified.

Results

Participant Characteristics

Table 1 describes baseline characteristics for the sample overall and for participants who did and did not complete follow-up. Approximately three-quarters of participants were black and average age at baseline was 35 years. More than two-thirds reported being on probation, having completed their high school education or equivalent, being unemployed, and not living with a spouse or partner. More than half reported at least occasional alcohol use in the 30 days prior to the baseline interview and more than three-quarters, at least occasional drug use. The vast majority identified heroin or another opiate as the substance that caused them the most serious problems, and most indicated prior substance abuse treatment. Almost half reported some form of violence in the 90 days prior to their index incarceration. Just under one-third indicated that they perpetrated violence and slightly more than one-third indicated that they were the victims of violence. Thirty-six (27.9%) reported both perpetration and victimization, with 10.9% (n = 14) reporting victimization only, and 4.7% (n = 6) reporting perpetration only.

Table 1.

Participant Characteristics at Baseline

	Total		Completed Follow-up		Not Available at Follow-up		Comparison
Categorical Variables	N 129	% 100	N 91	% 70.5	N 38	% 29.2	χ2
Race/ethnicity
White	28	21.7	20	22.0	8	21.1	1.78
Black	97	75.2	67	73.6	30	78.9
Other	4	3.1	4	4.4	0	0.0
Employment status							0.00
Unemployed	95	73.6	67	73.6	28	73.7
Employed	34	26.4	24	26.4	10	26.3
Relationship status							0.05
Not living with spouse/partner	90	69.8	64	70.3	26	68.4
Living with spouse/partner	39	30.2	27	29.7	12	31.6
Education level							0.02
< High school/GED	42	32.6	30	33.0	12	31.6
≥ High school/GED	87	67.4	61	67.0	26	68.4
Legal status							0.02
Probation	96	74.4	68	74.7	28	73.7
Parole	33	25.6	23	25.3	10	26.3
Prior substance abuse treatment							2.12
No	15	11.6	13	14.3	2	5.3
Yes	114	88.4	78	85.7	36	94.7
Alcohol use							3.41
Never/not used	61	47.3	40	44.0	21	55.3
Occasional	20	15.5	17	18.7	3	7.9
Weekly	30	23.3	20	22.0	10	26.3
Daily	18	14.0	14	15.4	4	10.5
Drug use							4.27
Never/not used	27	20.9	23	25.3	4	10.5
Occasional	14	10.9	9	9.9	5	13.2
Weekly	23	17.8	17	18.7	6	15.8
Daily	65	50.4	42	46.2	23	60.5
Substance caused most problems							1.12
Alcohol	11	8.5	9	9.9	2	5.3
Heroin or other opiate	105	81.4	72	79.1	33	86.8
Other	13	10.1	10	11.0	3	7.9
Recent history of any violent outcomea							0.04
No	73	56.6	52	57.1	21	55.3
Yes	56	43.4	39	42.9	17	44.7
Recent history of violence							0.32
No	87	67.4	60	65.9	27	71.1
Yes	42	32.6	31	34.1	11	28.9
Recent history of victimization							0.81
No	79	61.2	58	63.7	21	55.3
Yes	50	38.8	33	36.3	17	44.7
Continuous Variables	M (SD)		M (SD)		M (SD)		F
Age (in years) at baseline	35.65 (9.41)		35.67 (9.86)		35.61 (8.34)		0.00
Age at first arrest (in years)	16.86 (3.53)		17.24 (3.47)		15.95 (3.56)		3.67‡
Number of prior arrests	18.29 (19.83)		15.42 (12.89)		25.18 (29.79)		6.80*

Notes.

^aIncludes violence perpetration and victimization. Recent history refers to the 90 days prior to the index incarceration.

^‡p < .10,

^*p < .05,

^**p < .01,

^***p < .001.

Bivariate comparisons of participants who completed follow-up interviews compared to those who were not available at follow-up revealed two possible sources of attrition bias: 1) age at first arrest, and 2) number of prior arrests. Specifically, participants who were not available at follow-up were younger at the time of their first arrest (p = .058) and reported significantly more prior arrests than participants who completed follow-up (p = .010). Thus, these variables were tested as potential covariates for inclusion in the multivariable analyses.

Behaviors during Follow-up

Treatment adherence during the 90-day follow-up period was relatively high, particularly for counseling session attendance. Approximately three-quarters of participants (75.8%, n = 69) were classified as adherent to their counseling sessions, with more than one-third reporting they never missed a counseling session (37.4%, n = 34) or missed two sessions at most (38.5%, n = 35). Only five participants (5.5%) indicated they attended less than half of the sessions. Approximately two-thirds of participants (69.2%, n = 63) were classified as medication adherent and almost half (47.3%, n = 43) reported never missing their medication. That said, more than one in five participants (22.0%, n = 43) reported taking less than half of their medication.

Consistent with the harm reduction approach of the treatment services, some substance use was reported during the 90-day follow-up period. More than half of participants (60.4%, n = 55) reported at least occasional drug use, but fewer than half (42.9%, n = 39) reported at least occasional alcohol use. We observed greater decreases in drug use compared to alcohol use, overall. Specifically, almost half of participants (44.0%, n = 40) reported a decrease in the frequency of their drug use from baseline to follow-up, compared to one-third of participants (33.0%, n = 30) who reported a decrease in frequency of alcohol use. However, almost twice as many participants reported drug use compared to alcohol use at baseline (see Table 1) and, as such, fewer participants were able to decrease their alcohol use over follow-up.

Prevalence of violence and victimization were notably lower in the 90-day follow-up period than during the 90-day period prior to participants’ index incarceration. Few participants – less than one in five – reported experiencing some form of violent outcome in the 90-day follow-up period (18.7%, n = 17). Rates of violence and victimization were identical (15.4%, n = 14), but do not represent perfect overlap in participants.

Research Aim 1

To address our first research aim, we conducted binary logistic regression analyses examining associations between participant characteristics and treatment adherence, adjusting for length of follow-up. Results failed to identify significant associations (ps ≥ .072). We did, however, find an association between attendance at counseling sessions and medication adherence: Participants who missed several or more counseling sessions were almost three times as likely to miss several or more doses of medication than participants who missed no more than two counseling sessions (Odds Ratio_adj = 2.90, 95% CI = 0.99–8.50, p = .052).

Research Aim 2

Bivariate logistic regression analyses provided some evidence of associations between treatment adherence and substance use during follow-up. Results showed that, after adjusting for length of follow-up, participants who missed several or more counseling sessions were more than five times as likely to report at least occasional drug use compared to participants who attended most or all sessions (Odds Ratio_adj = 5.03 95% CI = 1.34 – 18.85, p = .013). Frequency of attendance at counseling sessions, in contrast, was not associated with alcohol use (p = .591). Conversely, medication nonadherence was associated with alcohol use, but not with drug use. Medication nonadherent participants were more than twice as likely to report at least occasional alcohol use (Odds Ratio_adj = 2.74, 95% CI = 1.00 – 7.51, p = .050). The pattern of results was similar for drug use, but the association was not significant (p = .134).

Research Aim 3

To address our third research aim, we conducted binary logistic regression analyses examining associations between treatment adherence and each of the outcome variables, adjusting for length of follow-up. Results, presented in Table 2, provided some evidence for associations between treatment adherence and violent outcomes. Participants who missed several or more counseling sessions and who missed several or more doses of medication were about three and a half times more likely to experience at least one violent outcome, violence and/or victimization, during follow-up compared to their treatment adherent counterparts. Bivariate associations between treatment adherence and victimization were even stronger. Participants who missed several or more counseling sessions were more than five times as likely and those who missed several or more doses of medication were almost four times as likely to be victimized compared to their treatment adherent counterparts. In contrast, we did not observe any associations between treatment adherence and violence perpetration.

Table 2.

Bivariate Predictors of Any Violent Outcomes, Any Violence, and Any Victimization during the 90-day Follow-up Period

	Any Violent Outcome			Any Violence			Any Victimization
Categorical Variables	None	1 or More	AOR (95% CI)	None	1 or More	AOR (95% CI)	None	1 or More	AOR (95% CI)
Race/ethnicity
White (ref)	14 (18.9)	6 (35.3)	2.19 (0.65–7.34)	15 (19.5)	5 (35.7)	2.14 (0.57–8.01)	15 (19.5)	5 (35.7)	2.14 (0.58–7.96)
Black	57 (77.0)	10 (58.8)	1.57 (0.14–17.64)	59 (76.6)	8 (57.1)	1.95 (0.17–23.0)	59 (76.6)	8 (57.1)	2.00 (0.17–23.18)
Other	3 (4.1)	1 (5.9)		3 (3.9)	1 (7.1)		3 (3.9)	1 (7.1)
Employment status
Unemployed (ref)	54 (73.0)	13 (76.5)	0.90	55 (71.4)	12 (85.7)	0.45	56 (72.7)	11 (78.6)	0.80
Employed	20 (27.0)	4 (23.5)	(0.26–3.13)	22 (28.6)	2 (14.3)	(0.09–2.24)	21 (27.3)	3 (21.4)	(0.20–3.23)
Relationship status
Not living with spouse/partner (ref)	53 (71.6)	11 (64.7)	1.28	55 (71.4)	9 (64.3)	1.26	56 (72.7)	8 (57.1)	1.88
Living with spouse/partner	21 (28.4)	6 (35.3)	(0.40–4.04)	22 (28.6)	5 (35.7)	(0.36–4.42)	21 (27.3)	6 (42.9)	(0.56–6.33)
Educational level
< HS/GED (ref)	24 (32.4)	6 (35.3)	1.19	24 (31.2)	6 (42.9)	0.84	25 (32.5)	5 (35.7)	1.26
≥ HS/GED	50 (67.6)	11 (64.7)	(0.33–3.97)	53 (68.8)	8 (57.1)	(0.24–2.98)	52 (67.5)	9 (64.3)	(0.33–4.76)
Legal status
Probation (ref)	52 (70.3)	16 (94.1)	0.14	55 (71.4)	13 (92.9)	0.18	55 (71.4)	13 (92.9)	0.19
Parole	22 (29.7)	1 (5.9)	(0.02–1.21)‡	22 (28.6)	1 (7.1)	(0.02–1.62)	22 (28.6)	1 (7.1)	(0.02–1.62)
Recent history of violent outcome
No (ref)	46 (62.2)	6 (35.3)	3.08	55 (71.4)	5 (35.7)	5.43	53 (68.8)	5 (35.7)	4.28
Yes	28 (37.8)	11 (64.7)	(0.99–9.57)‡	22 (28.6)	9 (64.3)	(1.49–9.73)*	24 (31.2)	9 (64.3)	(1.22–14.93)*
Prior substance abuse treatment
No (ref)	7 (9.5)	6 (35.3)	0.19	9 (11.7)	4 (28.6)	0.35	7 (9.9)	6 (42.9)	0.13
Yes	67 (90.5)	11 (64.7)	(0.05–0.73)*	68 (88.3)	10 (71.4)	(0.08–1.53)	70 (90.9)	8 (57.1)	(0.03–0.52)**
Alcohol use
None (ref)	47 (63.5)	5 (29.4)	4.47	47 (61.0)	5 (35.7)	3.02	48 (62.3)	4 (28.6)	4.51
At least occasional use	27 (36.5)	12 (70.6)	(1.35–14.82)*	30 (39.0)	9 (64.3)	(0.86–10.6)‡	29 (37.7)	10 (71.4)	(1.20–16.91)*
Drug use
None (ref)	32 (43.2)	4 (23.5)	2.18	32 (41.6)	4 (28.6)	1.47	33 (42.9)	3 (21.4)	2.36
At least occasional use	42 (56.8)	13 (76.5)	(0.64–7.46)	45 (58.4)	10 (71.4)	(0.41–5.39)	44 (57.1)	11 (78.6)	(0.59–9.36)
Counseling attendance
Attended all/most (ref)	60 (81.1)	9 (52.9)	3.56	61 (79.2)	8 (57.1)	2.48	63 (81.8)	6 (42.9)	5.61
Missed several or more	14 (19.0)	8 (47.1)	(1.12–11.33)*	16 (20.8)	6 (42.9)	(0.70–8.74)	14 (18.2)	8 (57.1)	(1.61–19.5)**
Medication adherence
Took all/most (ref)	55 (74.3)	8 (47.1)	3.49	56 (72.7)	7 (50.0)	2.56	57 (74.0)	6 (42.9)	3.91
Missed several or more	19 (25.7)	9 (52.9)	(1.11–11.03)*	21 (27.3)	7 (50.0)	(0.75–8.88)	20 (26.0)	8 (57.1)	(1.13–13.49)*
Continuous Variables	M (SD)	M (SD)	AOR (95% CI)	M (SD)	M (SD)	AOR (95% CI)	M (SD)	M (SD)	AOR (95% CI)
Age (in years) at baseline	37.30 (9.60)	28.59 (7.84)	0.89 (0.81–0.97)**	36.91 (9.63)	28.86 (8.51)	0.91 (0.93–0.99)*	37.26 (9.60)	26.93 (6.10)	0.84 (0.76–0.94)**
Age at first arrest (in years)	17.20 (3.18)	17.41 (4.64)	1.05 (0.90–1.23)	17.47 (3.40)	16.00 (3.70)	0.89 (0.72–1.09)	17.19 (3.35)	17.50 (4.20)	1.07 (0.90–1.26)
Number of prior arrests	16.03 (13.45)	12.76 (9.99)	0.97 (0.91–1.02)	15.52 (13.43)	14.86 (9.77)	0.99 (0.94–1.04)	15.94 (13.25)	12.57 (10.60)	0.96 (0.89–1.03)

Note. N = 91. No violent outcome, no violence, and no victimization served as the references for the outcome variables, respectively.

^aIncludes violence perpetration and victimization. Recent history refers to the 90 days prior to the index incarceration. AOR = Odds ratios adjusted for length of follow-up.

^‡p < .10,

^*p < .05,

^**p < .01,

^***p < .001.

We additionally examined associations between participant characteristics and substance use with the three outcome variables using binary logistic regression analyses to identify covariates for inclusion in the multivariable models. Results provided support for substance use and recent history of violent outcomes as predictors of violent outcomes. At least occasional alcohol use during follow-up was associated with increased risk for all three outcomes, increasing the risk of violence more than twofold, and the risk of victimization or any violent outcome more than threefold. In contrast, drug use during follow-up was not associated with any of the outcome measures (ps ≥ .215). All three recent history variables emerged as robust predictors of the violent outcome in question. Probationers and participants with no history of substance abuse treatment were more likely than parolees and those with a history of substance abuse treatment, respectively, to report any violent outcome and victimization. No such effects of legal status (p = .127) or prior substance abuse treatment (p = .164) were observed for violence. Older age was associated with decreased risk for violent outcomes (see Table 2).

Research Aim 4

We conducted two sets of direct entry hierarchical logistic regression analyses to determine whether associations between treatment adherence and violent outcomes may be attributable to reductions in substance use. The first regressed the treatment adherence variables on any violent outcome, and the second regressed the treatment adherence variables on victimization. Because the bivariate associations between treatment adherence and violence were not significant, we did not run multivariable analyses for this outcome. Prior to conducting these analyses, we tested for multicollinearity between the predictors and covariates. All tolerance values were > .10 and variance inflation factor values were < 10. In the sections that follow, we present the results of the multivariable victimization regression analyses. The patterns of results for any violent outcome were identical and are available upon request.

Length of follow-up was added in Step 1. In Step 2, age, legal status, and prior substance abuse treatment were entered. The overall model was significant, χ²(4, N = 91) = 22.54, p < .001, −2 LL = 50.51, Nagelkerke R² = .40. Age (p = .016) and prior substance abuse treatment (p = .009), but not legal status (p = .345), provided unique contributions to the prediction of victimization. In Step 3, recent history of victimization improved the model (Δχ²[1, N = 91] = 7.63, p = .006) and demonstrated unique contributions to the prediction of victimization (p = .026). The addition of alcohol use in Step 4 improved the model further still (Δχ²[1, N = 91] = 5.92, p = .015). Alcohol use predicted victimization (p = .022), even after accounting for length of follow-up, participant characteristics, and recent history of victimization.

The final model is presented in Table 3, with counseling session attendance and medication adherence added in the fifth step. After accounting for length of follow-up, participant characteristics, recent history of victimization, and alcohol use in the 90-day follow-up period, the two treatment adherence variables no longer demonstrated significant associations with victimization and their inclusion did not improve the overall model (p = .174).

Table 3.

Final Logistic Regression Model with Treatment Adherence Predicting ] Victimization in the 90-Day Follow-Up Period

Predictors		1 or More Incidents
	β	SE	Wald	OR	95% CI	p
Length of follow-up (in days)	0.00	0.01	0.39	1.00	0.99–1.01	.533
Age (in years) at baseline	−0.10	0.07	2.07	0.91	0.79–1.04	.150
Legal status
Probation (ref)
Parole	−1.47	1.25	1.37	0.23	0.02–2.69	.241
Prior substance abuse treatment
No (ref)
Yes	−3.86	1.44	7.17	0.01	0.00–0.36	.007
Recent history of victimization
No (ref)
Yes	1.94	0.94	4.27	6.96	1.10–43.84	.052
Alcohol use in past 90 days
None (ref)
At least occasional use	1.59	0.78	4.15	4.91	1.06–22.73	.042
Counseling attendance
Attend all/most sessions (ref)
Missed several or more	1.37	0.96	2.02	3.92	0.60–25.79	.155
Medication adherence
Took all/most medication (ref)
Missed several or more	1.06	1.07	0.99	2.88	0.36–23.21	.321

Notes. N = 91. No victimization served as the reference for the dichotomous outcome variable. Recent history refers to the 90 days prior to the index incarceration. Model statistics: χ²(8, N = 91) = 37.12, p < .001, −2 LL = 35.93, Nagelkerke R² = .61.

Discussion

Though empirical evidence regarding the effectiveness of MAT in reducing alcohol and drug use is well-documented, less is known regarding the effects of MAT on violent outcomes, including violence and victimization. We conducted a prospective, naturalistic study exploring MAT as a strategy for reducing violent outcomes in a sample of male community-based offenders with alcohol and drug use problems, with a focus treatment adherence. Findings provide some evidence supporting MAT for treating alcohol use and preventing victimization in community-based offenders, but less support for its role in treating drug use and preventing violence. Findings also underscore the importance of examining treatment adherence in studies of MAT effectiveness. In the sections that follow, we discuss the study findings in further detail.

Summary of Findings

Our first aim was to examine differences in treatment adherence, including medication adherence and counseling session attendance, as a function of participant characteristics. There were no characteristics that differed across participants who attended all or most counseling sessions and those who missed several or more sessions nor between those who took all or most of their medication and those who missed several or more doses. These findings, or lack thereof, add to the mixed results of prior research regarding differences in MAT nonadherence as a function of severity of substance use at baseline, criminal history, or other clinical-legal factors. It is possible that factors not measured in the current study, such as a psychiatric comorbidity, may have been associated with MAT nonadherence (Tzack, Severt, Cacciola, & Ruetsch, 2012). Additionally, the individualized nature of the MAT treatment, including the prescribed dosage of medication, frequency of administration, or frequency of psychosocial services, may have affected adherence rates. Likely, treatment adherence increased as the required frequency of treatment contacts decreased (e.g., monthly depot naltrexone vs. daily oral medication). These differential regimens are very much in keeping with real world practices and represent important avenues of future research on MAT treatment adherence in community-based offenders.

Our second aim was to examine whether adherence to the prescribed MAT regimen was associated with alcohol use and drug use during the 90-day follow-up period. Findings were mixed. Whereas medication nonadherence was associated with increased likelihood of at least occasional alcohol use, it was not associated with drug use during follow-up. Conversely, missing several or more counseling sessions was associated with increased likelihood of at least occasional drug use, but not with alcohol use. As such, findings of the present study support the effectiveness of MAT in reducing alcohol use among community-based offenders, at least in terms of the medication itself. There was less support for its utility in treating drug use. These findings are in contrast with prior research that has found compelling evidence of the effectiveness of MAT, and naltrexone specifically, in treating both alcohol use and drug use (Krupitsky & Blokhina, 2010; Rösner et al., 2008). Instead, findings suggest that the counseling session component of MAT, rather than the medication, contributed to reductions in drug use.

Our third aim was to examine associations between adherence to the prescribed MAT regimen and violent outcomes. The results of bivariate analyses showed strong negative associations of counseling session attendance and medication adherence with violent outcomes, and with victimization in particular. However, these associations were no longer significant in multivariable analyses conducted to test our fourth and final aim. This pattern of results suggests potential mediating effects. Specifically, the lack of significant association between medication adherence and violent outcomes in the multivariable models may be attributable to the impact of medication adherence on alcohol use. One prior study similarly has found that reductions in crime rates among heroin users who received methadone treatment are mediated by reductions in heroin use following treatment (Gossop, Marsden, Stewart, & Rolfe, 2000); yet, this potential mediating mechanism is rarely tested empirically.

Neither treatment adherence variable was associated with risk of violence. These findings are contrary to those of Havnes et al. (2012) and Gerra et al. (2011), which showed reductions in rates of violence associated with MAT. However, as noted earlier, inconsistencies between the findings of the current and prior studies may reflect the specific medication under investigation; specifically, Gerra and colleagues examined the effects of methadone maintenance, and the opioid maintenance medication was not specified in the Havnes et al. study. Inconsistences also may be attributable to differences in study designs; for example, Havnes and colleagues compared official rates of violent crime between treatment completers and noncompleters over a 3-year follow-up period, whereas the present study compared self-reported rates of violence over a 90-day period between adherent and nonadherent offenders.

Limitations

This study has several limitations, including our reliance on self-report data. Though self-report may be susceptible to recall bias and errors, as well as social desirability effects, it is a valid and reliable method for collecting data on violence, victimization, and substance use (Crisanti, Laygo, & Junginger, 2003; Darke, 1998). The study also is limited by the variable lengths of follow-up. That said, during the follow-up interviews, participants were prompted to report on the same 90-day period of treatment that followed the baseline interview, and we adjusted for follow-up length in all analyses. Furthermore, the naturalistic design of the study did not allow for randomized assignment to treatment conditions; as such, we can only speak to associations observed between study variables. Finally, approximately one-third (29.5%, n = 38) of participants were lost to follow-up. We identified two potential sources of attrition bias: age at first arrest and number of prior arrests. These variables are known predictors of recidivism, suggesting that the participants lost to follow-up were higher risk and, consequently, those who should be receiving more intensive treatment services (Andrews & Bonta, 2010).

Research Implications

The present study represents an initial step in establishing MAT as a treatment for reducing violent outcomes in community-based offenders. This is one of a handful of studies examining the use of naltrexone in correctional settings, and with community-based offenders in particular (Coviello et al., 2012; Crits-Christoph, Lundy, Stringer, Gallop, & Gastfriend, 2015; Patapis & Nordstrom, 2006). To that end, further empirical evidence is needed to establish MAT as a strategy for reducing alcohol and drug use, as well as research examining its impact on violent outcomes, in this population. Specifically, there is a need for controlled research with randomization to treatment conditions that examines the impact of MAT over longer follow-up periods and with diverse samples, including female offenders. Additionally, conclusions would be strengthened with the use of multiple sources of data on variables of interests, including alcohol use and drug use, treatment adherence, and violent outcomes, such as data drawn from official records, collateral reports, and biological testing.

Findings also highlight the value of measuring alcohol use and drug use separately, as these two variables showed differential associations with both treatment adherence and violent outcomes in the current study and other recent work (e.g., Johnson et al., in press). Finally, there is very little research examining the independent associations of the medication and adjunctive psychosocial treatment that comprise MAT with alcohol and drug use after treatment. Given the varied associations of medication adherence and counseling session attendance with alcohol and drug use, such efforts are needed to better understand the “active ingredients” in MAT and their contributions to reductions in alcohol and drug use.

Clinical and Policy Implications

Much has been written about the importance of treating substance use as a risk management strategy in correctional populations, such that reducing alcohol and drug use will, in turn, reduce criminal and violent behavior (Casey & Day, 2014; Grann & Fazel, 2004; Resor & Blume, 2008; Mitchell et al., 2007). Findings suggest that naltrexone may assist in reducing alcohol use in community-based offenders and provide the first empirical evidence supporting MAT as a strategy for reducing risk of victimization. However, violence risk is typically of greater concern and our findings failed to show associations between adherence to MAT and reductions in either drug use or risk of violence, attributable to substance use or otherwise. Participant age and recent history of violence were the only predictors of violence in this study (though alcohol use was associated with violence at p = .078). The predictive capacity of these two static risk factors is well-established (Piquero, Jennings, Diamond, & Reingle, 2015; Witt, Van Dorn, & Fazel, 2013) and they represent useful considerations in decisions regarding levels of intervention, community access, and risk management strategies (Andrews & Bonta, 2010).

Our focus on treatment adherence, including attendance at counseling sessions and medication adherence, merits some discussion. In correctional settings, there frequently are concerns regarding offender motivation to participate in substance use treatment programs and, consequently, adherence to prescribed treatment regimens (Resor & Blume, 2008; Zanis et al., 2009). Moreover, nonadherence is one of the problems that has been associated with naltrexone as part of MAT (O’Brien & Cornish, 1996). The extended-release formulation of naltrexone used in the present study removes some (but not all) barriers to medication adherence, such as the need for daily oral dosing (Chandler et al., 2009). However, almost one-quarter of participants in the current study still reported missing several or more doses during the follow-up period. Consequently, further efforts are needed to identify and reduce barriers to MAT adherence in community-based offenders with alcohol and drug use problems. For example, evidence-based practices may be employed during the adjunctive psychosocial therapy to increase medication adherence, such as motivational interviewing (Julius, Novitsky, & Dubin, 2009).

Conclusion

Continued efforts are needed to identify and implement evidence-based treatments to reduce violent outcomes in community-based offenders with alcohol and drug use problems. MAT shows promise as an intervention to reduce alcohol use and victimization, but evidence from the current study is less compelling regarding the use of MAT to reduce drug use and violence. Adherence appears to be key to MAT effectiveness, but examinations of treatment adherence vis-à-vis violent outcomes are scarce. Knowledge of factors associated with treatment adherence may inform policy and practice regarding the use of MAT among offenders with alcohol and drug use problems, including decisions regarding which medication may be most effective, which offenders may be most likely to benefit from MAT, and the conditions required to promote adherence. Identification of offender characteristics associated with treatment nonadherence also may assist clinicians in identifying patients at greater risk of nonadherence and, hopefully, preventing relapse to alcohol and drug use. Doing so, ultimately, should promote successful community reintegration among offenders with alcohol and drug use problems.