Effectiveness of a Prison-Based Treatment Program for Male Perpetrators of Intimate Partner Violence: A Quasi-Experimental Study of Criminal Recidivism

Abstract

Despite increasing interest in programming for perpetrators of intimate partner violence (IPV), the literature provides weak support for the effectiveness of these interventions. However, there are few studies that evaluate programs offered to felony IPV offenders who are serving prison sentences. This study uses a quasi-experimental design to evaluate the effectiveness of a prison-based implementation of a popular IPV intervention in reducing general and offense-specific recidivism among 169 men released from state prison in 2017. Because recidivism data were collected through early March 2019, the average follow-up period was approximately 20 months for both the control group and experimental group. Observable selection bias was minimized by using propensity score matching to create a comparison group of 169 nonparticipants released in 2017 who were not significantly different from the program participants. Cox regression models were used to predict general rearrest, reconviction, reincarceration, and supervised release revocation, as well as rearrest and reconviction for any violent offenses and rearrest and reconviction for domestic violence offenses in particular. No significant differences in any type of recidivism were found between the comparison group and those who participated in treatment, regardless of whether the participant completed or failed to complete the program. The findings suggest that the intervention is no more successful when offered within prisons than when offered in the community. The study concludes by making recommendations for increasing the effectiveness of prison-based domestic violence programming.

Keywords

domestic violence domestic violence treatment batterers

Introduction

In the United States, 22% to 25% of women and 8% to 14% of men experience assault by an intimate partner at some point in their lifetime (Breiding et al., 2014; Tjaden & Thoennes, 2000). Intimate partner violence (IPV) has serious medical, psychological, and financial effects for victims and for other members of the family, including children (Tjaden & Thoennes, 2000; Weithorn, Carter, & Behrman, 1999). In the 1980s, policies advocating mandatory arrest for domestic violence became common (Sherman & Berk, 1984), resulting in an increasing number of prosecutions and incarcerations of men who assault intimate partners. Due to this increase, intervention programs aimed at IPV perpetrators—especially those that are facilitated in group settings—have increasingly been implemented within the criminal justice system.

One of the most common of these interventions is the Duluth Model, which is rooted in feminist arguments that partner abuse against women is caused by patriarchal attitudes. Approximately, half of programs in the United States (47%-53%) use it as the primary or secondary treatment approach (Cannon, Hamel, Buttell, & Ferreira, 2016; Price & Rosenbaum, 2009). Research suggests that programs based on the Duluth Model have only limited success in reducing recidivism (see meta-analyses by Arias, Arce, and Vilariño (2013); Babcock, Green, and Robie (2004); Feder and Wilson (2005); Miller, Drake, and Nafziger (2013)). However, most studies have evaluated interventions for IPV offenders serving sentences in the community, often for misdemeanor offenses. Therefore, the effectiveness of the program in a prison setting is still not well known. The current study addresses these gaps in the literature by conducting an outcome evaluation of a Duluth Model–based intervention for felony offenders incarcerated in Minnesota state prisons.

The Duluth Model Intervention for Domestic Abuse

The Duluth Model is based on the assumption that IPV stems from patriarchal attitudes that allow men to dominate relationships, leading to violence against their partners. Programs based on this model attempt to target and change those attitudes so that participants develop skills for maintaining healthier intimate relationships, thereby reducing future IPV. The group treatment program employs group facilitators who challenge men’s perceptions that they have a right to control their partners, using the “Power and Control Wheel” to help participants understand their patterns of abusive behavior (Pence & Paymar, 1993). The facilitators then emphasize features of egalitarian relationships, demonstrated by the “Equality Wheel,” which men who participate in the program are encouraged to adopt (Pence & Paymar, 1993). The Duluth Model also requires a coordinated community response (CCR) in which different criminal justice agencies make efforts to protect survivors and hold perpetrators accountable (Shepard, Falk, & Elliott, 2002).

The literature on the effectiveness of this treatment is mixed. Some evaluations found that men who complete programs based on the Duluth Model were significantly less likely to reoffend (Babcock & Steiner, 1999; Gondolf, 2004, 2002; Hasisi, Shoham, Weisburd, Haviv, & Zelig, 2016; Murphy, Musser, & Maton, 1998; Palmer, Brown, & Barrera, 1992). For example, Davis, Taylor, and Maxwell’s (2000) experimental evaluation found lower rates of recidivism among treatment participants; however, their examination of the effects of treatment length led them to conclude that the intervention may have only reduced violence while offenders were still under court control rather than actually changing their behavior (see also Maxwell, Davis, & Taylor, 2010). Other evaluations also show lower recidivism rates among completers than dropouts (Bennett, Stoops, Call, & Flett, 2007; Coulter & VandeWeerd, 2009; Dobash, Dobash, Cavanagh, & Lewis, 1996). In addition, the program may also influence factors other than reassault; for example, Taylor and Maxwell (2009) found that program participants self-reported less controlling behavior and less alcohol and marijuana use.

However, many experimental (Dunford, 2000; Feder & Dugan, 2002; Feder & Forde, 2000; Ford & Regoli, 1993; Labriola, Rempel, & Davis, 2008; Lin et al., 2009; Saunders, 1996; Taylor & Maxwell, 2009; Tolleffson & Gross, 2006) and quasi-experimental studies (Bowen, Gilchrist, & Beech, 2005; Gordon & Moriarty, 2003; Haggård, Freij, Danielsson, Wenander, & Långström, 2017) showed no significant reduction in violence among participants of programs based on the Duluth Model. Indeed, some studies found that men in these programs had significantly higher rates of violence than nonparticipants (e.g., Harrell, 1991; Shepard, 1992). In addition, Feder and Dugan’s (2002) evaluation of a Florida program showed that the program did not affect attitudes toward women or beliefs about wife beating.

Furthermore, several meta-analyses on domestic violence interventions provide little support for the Duluth Model. In 2004, a meta-analysis by Babcock and colleagues examining 19 effect sizes found a 5% reduction in recidivism among participants of Duluth programs. Importantly, however, they noted that more support was found in quasi-experimental studies than in true experiments. Feder and Wilson’s (2005) meta-analysis of 10 studies showed a 13% to 20% reduction in recidivism according to official statistics, but no effect according to victim reports. Similarly, meta-analyses conducted by Miller et al. (2013), which included six effect sizes for Duluth-like programs, and by Arias et al. (2013), which included 29 effect sizes, found that programs based on the Duluth Model had no statistically significant effect on recidivism. More recently, a meta-analysis conducted by Cheng, Davis, Jonson-Reid, and Yaeger (2021) echoed previous results: Studies showed less impact of batterer intervention programs when they had greater methodological rigor and when they measured recidivism through victim reports.

Overall, the literature provides only limited support for the effectiveness of domestic violence treatment based on the Duluth Model. However, many of the programs that have been evaluated have targeted offenders who committed misdemeanors or low-level felonies. Past studies’ focus on offenders who commit less serious offenses might explain the limited support found to date. Men who commit more serious acts of IPV may respond differently to the treatment, but it is not well known whether the program is effective with this population. In addition, most of the evaluated programs discussed above were provided to offenders in the community; therefore, it is still not clear whether similar programs that target incarcerated offenders are effective. To address these gaps, the current study evaluates general and offense-specific recidivism among participants of a prison-based IPV intervention based on the Duluth Model. To that end, the study employs a quasi-experimental design to compare offenders who completed the domestic violence program with eligible offenders who did not participate in the program.

Minnesota Domestic Violence Intervention Pilot Program (DVIPP)

Due to a rising population of incarcerated offenders serving sentences for domestic violence offenses in Minnesota, the Minnesota Department of Corrections (MnDOC) has identified domestic violence treatment as an important programming need. However, with the exception of an optional track that has been available as part of some chemical dependency (CD) programs since 2014, offenders in Minnesota state prisons have not had opportunities to participate in programming to address their behavior related to IPV. Therefore, in November 2016, the MnDOC Victim Assistance and Restorative Justice unit obtained an Office of Justice Programs (OJP) Community Crime Prevention Grant to implement the DVIPP, a program for male offenders serving sentences for IPV offenses.

The DVIPP is a 16-week group treatment program that utilizes the Duluth Model’s curriculum, Creating a Process of Change for Men Who Batter. The program, in which groups met for two sessions per week, used interactive classes, video vignettes, and control and equality logs to help men identify their beliefs that support using violence against women and learn strategies to change their behavior. The program was administered by one full-time and one part-time grant-funded corrections transitions coordinators (CTCs) who cofacilitated the group treatment sessions. In accordance with the Duluth Model design, group sessions were cofacilitated by one male and one female. The full-time CTC also worked with supervision agents to incorporate community reentry support that is specific to IPV and conducted follow-up conversations with supervision agents at 30 and 60 days after release and with offenders at 90 and 120 days after release.

To be eligible for the program, male offenders had to meet the following criteria: (a) have active or inactive sentences for felony domestic violence offenses against an intimate partner or a history of misdemeanor domestic violence offenses against an intimate partner, (b) be scheduled for release during the calendar year 2017,¹ (c) have at least 16 weeks of confinement time remaining on their sentence,² (d) be housed in general population in a participating facility,³ and (e) not have participated in the optional domestic violence track of CD programming.

Data and Method

Data and Sample

The study used a quasi-experimental design comparing recidivism between program participants and a matched comparison group of nonparticipants scheduled for release from Minnesota prisons in the calendar year 2017. A total of 169 offenders who participated in the program were scheduled for release in 2017. Of them, 109 successfully completed the program.⁴ The program participants were 50.9% Black, 33.1% White, 9.5% Native American, 5.9% Hispanic, and 0.6% Asian. The participants ranged in age from 22 to 64 years, with an average age of 35.75 years. The majority of participants (91%) were incarcerated for a new sentence, whereas 9% were incarcerated due to revocation of supervised release. The participants were incarcerated for a period between 4 and 142 months, with an average length of stay of 17.62 months.

During the same period, 853 offenders, with active or inactive sentences for domestic violence against an intimate partner, who did not participate were released. Offenders who were offered a chance to enter the program but refused to participate were not included in the comparison group. Offenders who were not located at a participating facility but were otherwise eligible for the program were included in the comparison group. The comparison group was reduced to 552 after 301 offenders were removed from the sample because they participated in CD treatment after a similar curriculum had been introduced within some programming groups. Therefore, the overall sample for this study consisted of 721 offenders, of whom 23% participated in the program and 15% completed the program.

Dependent Variables

Recidivism was measured in four ways: rearrest for a new offense, reconviction for a new offense, reincarceration for a new felony offense, and return to prison due to revocation of supervised release. In addition to general recidivism for any new offending, the study examines offense-specific recidivism. The dependent variables include measures of rearrest and reconviction for any violent (or “person”) offenses (i.e., homicide, assault, robbery, criminal sexual conduct, kidnapping, false imprisonment, child abuse, harassment, stalking, threats of violence, and violation of orders of protection) as well as rearrest and reconviction for domestic violence (i.e., domestic assault and domestic abuse).⁵ These outcomes include both “status” variables, indicating whether an offender recidivated, and “time” variables, measuring the number of months between release and the first recidivism event. Recidivism data were collected through March 10, 2019, resulting in a follow-up period between 9 and 26 months, with an average of 19.7 months.

Arrest and conviction data were obtained electronically from the Minnesota Bureau of Criminal Apprehension (BCA). Reincarceration and revocation data were obtained from the Correctional Operations Management System (COMS) database maintained by MnDOC. It is important to note that the dependent variables only measure recidivism that took place in Minnesota. Because they did not include reoffending that occurred in other states or that went undetected by the criminal justice system, these variables may underestimate the true rates of reoffending.

Independent Variables

Three independent variables relating to program participation were examined. The first treatment variable compares offenders who entered DVIPP (given a score of 1) with a comparison group of offenders who did not (given a score of 0). The second variable, program completion, compares offenders who completed treatment or successfully participated until release (1) with untreated offenders (0). The third variable measures program dropouts by comparing offenders who quit the program or were terminated from treatment (1) with untreated offenders (0).

Control Variables

Several correlates of recidivism identified by previous research were included as controls. Complete data for each of the measures below were obtained for all offenders. Below is a list of these variables⁶ along with a description of how they were created:

Minnesota Screening Tool Assessing Recidivism Risk (MnSTARR) 2.0: The probability that an offender will be reconvicted within 3 years of release (see Duwe & Rocque, 2017). Scores are then classified into one of four risk levels: “low risk” contains the bottom 40% of scores, “medium risk” contains the next 20%, “high risk” contains the next 20%, and “very high risk” contains the top 20% of scores.

Nonsexual violent MnSTARR: The probability that an offender will be reconvicted for a violent offense within 3 years of release.

Prior convictions: Total number of convictions, excluding the conviction(s) that resulted in the offender’s incarceration.

Prior felony convictions: Number of convictions for felonies.

Felony specialization: Proportion of convictions that were felonies.

Prior violent convictions: Number of convictions for violent offenses.

Violent specialization: Proportion of convictions that were violent offenses.

Prior drug convictions: Number of convictions for drug offenses.

Drug specialization: Proportion of convictions that were drug offenses.

Prior property convictions: Number of convictions for property offenses.

Property specialization: Proportion of convictions that were for property offenses.

Metro area: A rough proxy of urban and rural Minnesota, this variable dichotomizes the offender’s county of commitment as within the Twin Cities metropolitan area (1) or Greater Minnesota (0). The seven counties in the Minneapolis/St. Paul metropolitan area are Anoka, Carver, Dakota, Hennepin, Ramsey, Scott, and Washington.

Length of stay: The number of months between prison admission and release.

Institutional discipline: The number of discipline convictions received during the term of imprisonment prior to release.

Suicidal history: Dichotomized as history of suicidal tendencies (1) or no such history.

Gang involvement: Number of criteria for security threat group involvement met by the offender.

Prison admission type: Dichotomized as new commitment (1) or release return (0).

Offender race: Dichotomized as minority (1) or White (0).

Offender age: The age of the offender in years at the time of release.

Offender education: Dichotomized as at least high school diploma or equivalent (1) or less than high school diploma (0).

Prison visitation: Dichotomized as visited during last confinement (1) or not visited (0).

Postrelease supervision type: Dichotomized as intensive supervised release (ISR = 1) and standard supervision or no supervision (0).⁷

Propensity Score Matching (PSM)

PSM is a popular method used in program evaluations to help achieve balance between treatment and comparison groups. In doing so, PSM can be used to control for observable selection bias. PSM provides estimates of the conditional probability of selection to a particular treatment based on several observed covariates (Rosenbaum & Rubin, 1985). These estimates (i.e., propensity scores) are generated by estimating a logistic regression model in which the dependent variable is program participation. Then, the propensity scores are used to match individuals who entered programming with similar individuals who did not.

The results of the logistic regression model predicting participation in DVIPP are presented in Table 1. The variables included in a propensity score estimation model should consist of those related to the outcome—even if weakly—that affect treatment selection and are not caused by the treatment (Shadish, Cook, & Campbell, 2002). Therefore, the covariates included in the logistic regression model are those that temporally preceded the beginning of DVIPP participation. As shown in Table 1, several covariates significantly predicted whether offenders participated in DVIPP. Offenders with more prison discipline convictions were less likely to participate in the program. The total number of felony convictions was positively related to program participation, whereas the number of property convictions was negatively related to program participation. Offenders who were committed to prison for a new commitment were more likely to participate in the program than those who were committed to prison for a revocation of supervised release. Finally, two variables were marginally related to participation in DVIPP: offenders with longer prison stays were more likely to participate, and offenders who received visits while incarcerated were also more likely to participate.

Table 1.

Logistic Regression Model for Program Selection.

Predictor	Predictor Description	Coefficient	SE
Minority	Minority = 1, White = 0	0.17	0.13
Age	Offender age in years at time of release from prison	−0.002	0.01
MnSTARR violent probability	Likelihood of reconviction for a violent offense within 3 years of release	−0.002	0.01
Prison discipline	Number of discipline convictions during last confinement	−0.01*	0.01
Supervision revocations	Number of prior revocations while under correctional supervision	−0.06	0.05
Total convictions	Total number of convictions, including index conviction(s)	0.01	0.01
Felony convictions	Total number of felonies, including index conviction(s)	0.12**	0.04
Felony specialization/diversity	Specialization/diversity in felony offending	0.37	0.42
Violent convictions	Total number of violent offenses, including index conviction(s)	−0.001	0.04
Violent specialization/diversity	Specialization/diversity in violent offending	−0.10	0.50
Drug convictions	Total number of drug offenses, including index conviction(s)	0.03	0.07
Drug specialization/diversity	Specialization/diversity in drug offending	1.15	2.82
Property convictions	Total number of property offenses, including index conviction(s)	−0.10*	0.04
Property specialization/diversity	Specialization/diversity in property offending	−0.52	1.13
Married	Married = 1, unmarried = 0	−0.01	0.20
Metro commitment	Prison commit from Twin Cities metropolitan area = 1, Greater Minnesota = 0	−0.05	0.12
New commitment	New court commitment = 1, probation or release violator = 0	0.66***	0.17
Length of stay	Number of months between prison admission and release dates	0.01^†	0.002
Suicidal history	History of suicidal tendencies = 1, no history of suicidal tendencies = 0	0.15	0.13
Security threat group	Security threat group (STG) or gang affiliation criteria, 0-10	0.02	0.04
Secondary degree	Secondary degree = 1, less than secondary degree = 0	0.19	0.13
Visited during last confinement	Received any visits during last confinement = 1, no visits = 0	0.21^†	0.12
Constant		−2.83	3.11
N		721
Log-likelihood		–352.79
Pseudo R²		.10

Note. MnSTARR = Minnesota Screening Tool Assessing Recidivism Risk.

†

p < .10. *p < .05. **p < .01. ***p < .001.

After obtaining propensity scores on the 721 offenders included in the propensity score model, the study utilized a “greedy” matching procedure using a without-replacement method in which DVIPP participants and nonparticipants were matched within a caliper (i.e., range of propensity scores) of 0.3. Matches were obtained for all 169 offenders who participated in the DVIPP, resulting in a final sample of 338.

Table 2 presents statistics that measure the degree to which PSM was effective in reducing observable selection bias. The bias measure presented in the table quantifies the amount of bias (i.e., the standardized mean difference) between the treatment and comparison groups. According to Rosenbaum and Rubin (1985), bias values over 20 are considered unbalanced. Before matching, there were seven unbalanced covariates—supervision revocations, total convictions, felony convictions, violent convictions, commitment type, length of stay, and prison visitation. Table 2 shows that, after matching, balance between the DVIPP group and the comparison group was achieved as all of the covariates and propensity score had bias values below 20.

Table 2.

Propensity Score Matching and Covariate Balance for Program Selection.

Variable	Sample	Treatment Mean	Control Mean	Bias	Bias Reduction	t Test p Value
Propensity score	Total	0.32	0.21	79.5		.00*
Propensity score	Matched	0.32	0.32	0.7	−99.1	.95
Minority	Total	0.67	0.59	16.9		.06^†
Minority	Matched	0.67	0.68	2.5	−85.5	.82
Age	Total	35.75	34.76	10.8		.23
Age	Matched	35.75	35.86	1.2	−88.7	.91
MnSTARR violent probability	Total	27.58	26.04	11.0		.23
MnSTARR violent probability	Matched	27.58	26.91	4.8	−56.1	.66
Prison discipline	Total	4.53	5.86	10.1		.34
Prison discipline	Matched	4.53	3.76	5.8	−42.2	.27
Supervision revocations	Total	1.15	1.54	30.4		.00*
Supervision revocations	Matched	1.15	1.20	3.7	−87.8	.72
Total convictions	Total	14.65	12.64	28.0		.00*
Total convictions	Matched	14.65	14.62	0.5	−98.2	.96
Felony convictions	Total	5.41	4.33	40.1		.00*
Felony convictions	Matched	5.41	5.44	1.1	−97.3	.93
Felony specialization/diversity	Total	0.79	0.79	1.0		.92
Felony specialization/diversity	Matched	0.79	0.80	1.7	78.1	.86
Violent convictions	Total	3.54	3.03	24.2		.01*
Violent convictions	Matched	3.54	3.34	9.3	−61.5	.40
Violent specialization/diversity	Total	0.89	0.88	9.4		.31
Violent specialization/diversity	Matched	0.89	0.90	5.8	−38.2	.53
Drug convictions	Total	0.77	0.57	16.7		.04*
Drug convictions	Matched	0.77	0.86	7.9	−52.8	.52
Drug specialization/diversity	Total	0.99	0.99	2.4		.79
Drug specialization/diversity	Matched	0.99	0.99	9.7	305.4	.43
Property convictions	Total	1.87	1.91	1.5		.87
Property convictions	Matched	1.87	1.97	3.8	150.6	.73
Property specialization/diversity	Total	0.97	0.96	8.8		.34
Property specialization/diversity	Matched	0.97	0.96	8.1	−8.4	.51
Married	Total	0.07	0.09	5.5		.54
Married	Matched	0.07	0.09	4.2	−23.5	.69
Metro commitment	Total	0.53	0.50	5.7		.52
Metro commitment	Matched	0.53	0.51	3.5	−37.6	.75
New commitment	Total	0.91	0.74	44.5		.00*
New commitment	Matched	0.91	0.90	1.6	−96.4	.86
Length of stay	Total	17.62	13.25	21.9		.02*
Length of stay	Matched	17.62	17.38	1.2	−94.5	.93
Suicidal history	Total	0.25	0.23	3.5		.69
Suicidal history	Matched	0.25	0.24	2.8	−20.2	.80
Security threat group	Total	0.83	0.64	12.8		.14
Security threat group	Matched	0.83	0.92	6.2	−51.6	.60
Secondary degree	Total	0.79	0.73	14.8		.10
Secondary degree	Matched	0.79	0.76	6.9	−52.9	.51
Visited during last confinement	Total	0.41	0.32	20.2		.02*
Visited during last confinement	Matched	0.41	0.41	1.2	−93.9	.91

Note. MnSTARR = Minnesota Screening Tool Assessing Recidivism Risk.

†

p < .10. *p < .05.

Analytical Procedure

Because information on the timing of recidivism events was available, this study used survival analysis to examine recidivism. Survival analyses are preferable over logistic regression because they allow for an examination of not only whether offenders recidivate, but also when they do so. In particular, this study used Cox regression models, which incorporate both a “status” variable (a binary variable with a value of 1 if the event occurred) and a “time” variable. The “time” variable measured the amount of time (in months) between the release date and the date of the first recidivism event (or March 10, 2019, for those who did not recidivate).

Scholars argue that there should be a minimum of five to 10 events—or cases with a score of 1 on the dependent variable—for each independent variable in a Cox regression model (Penduzzi, Concato, Kemper, Holford, & Feinstein, 1996; Vittinghoff & McCulloch, 2007). The sample size for this study is small and, as a result, there are few cases with recidivism events; the number of events for the analyses predicting general recidivism ranged from 20 (reincarceration) to 201 (rearrest), the number of events for the analyses predicting violent recidivism ranged from 28 (reconviction) to 122 (rearrest), and the number of events for the analyses predicting domestic violence recidivism ranged from 13 (reconviction) to 82 (rearrest). Many of the control variables described above were included in the calculation of the MnSTARR; therefore, they were not included in the recidivism analyses to maximize the number of events per variable (EPV).

In addition to participation in DVIPP, the independent variables included in the recidivism analyses were postrelease supervision type, offender race, and either the MnSTARR overall risk level (to predict general recidivism) or the MnSTARR nonsexual violent assessment (to predict violent and domestic violence recidivism). Reducing the number of independent variables resulted in EPVs above five for all analyses except the one predicting domestic violence reconviction among program dropouts (EPV = 3.25). No problems with collinearity were found; the correlations between predictors were all lower than 0.5 and the tolerance values for all predictors were above 0.4.

Results

Recidivism Rates

Table 3 presents descriptive statistics comparing DVIPP participants and offenders in the comparison group for the eight outcome measures. The results show that general recidivism rates were not substantially different between the comparison group and those who entered, completed, or dropped out of DVIPP. When looking at offense-specific recidivism rates, rearrest for violent offenses and for domestic violence offenses appear to be higher among DVIPP completers (41.3% and 32.1%, respectively) than among the comparison group (33.1% and 21.9%, respectively). However, it is possible that these observed differences may be due to other factors. To statistically control for the impact of other factors on reoffending, Cox regression models were estimated for each of the 10 recidivism variables across the three treatment measures (entered treatment, completed treatment, and failed to complete treatment).

Table 3.

Recidivism Rates by Treatment Participation and Outcome.

	Comparison Group	Entered Program	Completed Program	Program Dropout
General recidivism
Rearrest	56.8%	62.1%	63.3%	60.0%
Reconviction	27.8%	29.0%	28.4%	30.0%
Reincarceration	8.9%	7.1%	6.4%	8.3%
Revocation	34.3%	37.9%	33.0%	46.7%
Violent recidivism
Violent rearrest	33.1%	39.1%	41.3%	35.0%
Violent reconviction	13.0%	11.2%	11.9%	10%
Domestic violence recidivism
Domestic violence rearrest	21.9%	26.6%	32.1%	16.7%
Domestic violence reconviction	7.1%	7.1%	10.1%	1.7%
N	169	169	109	60

Multivariate Results

The results of the Cox regression models predicting recidivism are presented in Table 4. The four columns on the left show the results for the four measures of general recidivism. As shown in the table, participation in DVIPP was not related to any of the measures of general recidivism, regardless of whether the participant completed or failed to complete the program. However, two of the control variables were related to recidivism. First, offenders on ISR were 62% to 81% more likely to return to prison due to a violation of supervised release; in addition, although the results only approached significance, offenders on ISR were 29% to 31% less likely to be rearrested, 39% less likely to be reconvicted, and 35% less likely to be reincarcerated for a new felony. Second, the MnSTARR risk assessment was significantly or marginally related to all four types of recidivism; for each one-level increase of the MnSTARR score, the risk of rearrest increased by 34% to 41%, the risk of reconviction increased by 37% to 46%, the risk of reincarceration increased by 89% to 390%, and the risk of supervised release revocation increased by 28% to 46%.

Table 4.

Results of Cox Regression Models Predicting Recidivism.

	General Recidivism				Violent Recidivism		Domestic Violence Recidivism
	Rearrest	Reconviction	Reincarceration	Revocation	Rearrest	Reconviction	Rearrest	Reconviction
Program participant	1.02 (0.14)	0.98 (0.21)	0.70 (0.39)	1.05 (0.18)	1.10 (0.18)	0.80 (0.31)	1.15 (0.22)	0.96 (0.41)
ISR	0.77 (0.17)	0.61 (0.26)^†	0.42 (0.54)	1.64 (0.19)*	1.28 (0.20)	0.54 (0.42)	1.37 (0.24)	0.39 (0.62)
MnSTARR	1.34 (0.09)**	1.37 (0.13)*	2.24 (0.35)*	1.38 (0.12)**	1.02 (0.01)***	1.01 (0.01)	1.03 (0.01)***	1.00 (0.02)
Minority	1.01 (0.15)	1.14 (0.23)	1.24 (0.44)	1.40 (0.21)	0.90 (0.20)	1.15 (0.35)	0.90 (0.24)	0.99 (0.44)
N = 338
Program completer	1.07 (0.16)	0.98 (0.23)	0.63 (0.46)	0.88 (0.21)	1.24 (0.20)	0.87 (0.35)	1.51 (0.24)^†	1.38 (0.42)
ISR	0.69 (0.19)^†	0.63 (0.29)	0.41 (0.63)	1.62 (0.22)*	1.12 (0.22)	0.55 (0.45)	1.19 (0.26)	0.43 (0.62)
MnSTARR	1.41 (0.10)***	1.46 (0.15)*	1.89 (0.35)^†	1.46 (0.14)**	1.03 (0.01)***	1.03 (0.01)*	1.04 (0.01)***	1.01 (0.02)
Minority	1.02 (0.17)	1.25 (0.26)	1.98 (0.56)	1.37 (0.24)	0.83 (0.21)	1.29 (0.39)	0.89 (0.26)	1.06 (0.46)
N = 278
Program dropout	0.93 (0.20)	0.99 (0.28)	0.80 (0.52)	1.38 (0.23)	0.91 (0.26)	0.73 (0.46)	0.64 (0.36)	0.24 (1.04)
ISR	0.71 (0.21)^†	0.63 (0.31)	0.35 (0.63)^†	1.81 (0.23)**	1.35 (0.24)	0.55 (0.50)	1.65 (0.31)	0.23 (1.04)
MnSTARR	1.36 (0.11)**	1.39 (0.17)*	4.90 (0.68)*	1.28 (0.14)^†	1.02 (0.01)**	1.00 (0.01)	1.03 (0.01)**	0.98 (0.02)
Minority	0.96 (0.19	0.99 (0.27)	1.68 (0.56)	1.40 (0.25)	1.10 (0.26)	1.11 (0.41)	1.01 (0.33)	0.91 (0.58)
N = 229

Note. Hazard ratios are presented with standard errors in parentheses. ISR = intensive supervised release.

†

p < .10. *p < .05. **p < .01. ***p < .001.

Table 4 also displays the results of the Cox regression models predicting offense-specific measures of recidivism. There were no significant differences in violent recidivism between DVIPP participants and the comparison group, regardless of whether the participant completed the program or failed to complete the program. One control variable was related to violent recidivism: The MnSTARR violent recidivism assessment was significantly related to violent recidivism in four of the six models; as the probability of violent recidivism calculated by the MnSTARR increased, the risk of rearrest increased by 2% to 3% and the risk of reconviction increased by 3%.

The right-most columns of Table 4 show the results of the models predicting domestic violence recidivism. The results show one marginally significant relationship between DVIPP status and domestic violence recidivism: Offenders who completed DVIPP or participated until their release from prison were 51% more likely to be rearrested than were offenders in the comparison group (p < .10). In addition, the MnSTARR violent assessment successfully predicted domestic violence–specific rearrest; for each percent increase in the probability of violent recidivism predicted by the MnSTARR, the risk of rearrest for a domestic violence offense increased by 3% to 4%. None of the independent variables were related to the risk of reconviction for a domestic violence offense.

Supplemental Analyses

Supplemental analyses (available on request) were conducted in which offenders were only included in the comparison group if they were housed in one of the two participating facilities during the approximate study period (November 2016-December 2017). About 40% of the comparison group (221 offenders) were housed in a participating facility during this time. Using the same PSM procedure as the main analyses, 35 offenders in the treatment group could not be matched, resulting in a smaller sample of 268 offenders. As with the main analyses, balance was achieved on all variables after matching. The multivariate results were substantially similar to the main analyses, with two exceptions: When examining domestic violence rearrest, (a) there was no difference between program completers and the comparison group, and (b) program dropouts were less likely to be rearrested than the comparison group although this difference was only marginally significant.

Discussion

While the results presented above show that the DVIPP did not reduce general, violent, or domestic violence recidivism, there are limitations of the study that must be acknowledged when considering the implications of the findings. First, this study was only able to examine official measures of reoffending. The Bureau of Justice Statistics estimates that 46% of IPV victimizations between 2006 and 2010 were not reported to the police (Langton & Berzofsky, 2012); therefore, this study likely does not capture all behavior related to IPV. Third, this study focuses on reoffending rather than other outcomes that could be affected by the program, such as attitudes toward women and other behavior within romantic relationships Therefore, the program could have had positive impacts on offenders’ and victims’ lives that were not measured in this study (see Lee & Stohr, 2012).

Still, as the main goal of a correctional program is to reduce reoffending, it is important to understand the possible reasons that the program did not meet this goal. There were some obstacles to implementation that may explain the DVIPP’s lack of success. First, Creating a Process of Change for Men Who Batter was designed as a 26-week intervention. Therefore, the DVIPP did not implement the curriculum as intended, as the curriculum was covered in only 16 weeks. Second, the Duluth Model was designed as a community-based intervention for misdemeanor domestic violence offenders. Therefore, the program may not have ever been intended to address men with significant histories of severe domestic violence or those with histories of general violence, both of which made up the population targeted by the DVIPP.

Third, there was turnover in the CTC role, who served as the group facilitators; there were three staff changes during the time period covered by the study. This turnover can affect trust between participants and the facilitator and reduce program effectiveness (Farabee et al., 1999; Gendreau, Goggin, & Smith, 1999). A growing body of literature suggests that strategies such as motivational interviewing may increase therapeutic alliance and confidence between participants and facilitators (Alexander, Morris, Tracy, & Frye, 2010; Babcock et al., 2016; Crane & Eckhardt, 2013; Eckhardt et al., 2013; Lila, Gracia, & Catalá-Miñana, 2018; Musser, Semiatin, Taft, & Murphy, 2008). Therefore, programming implemented within prisons may benefit from incorporating these strategies as they may negate the negative impacts of turnover. Fourth, because the program was implemented by Victim Assistance staff, offenders may have had trouble establishing a rapport with the facilitators, thereby affecting their participation in the program (see Simpson, 2001).

Fifth, some may not consider the DVIPP a true implementation of the Duluth Model because it was implemented as a stand-alone correctional program rather than in coordination with presentencing initiatives to compile information from different sources and establish standard responses to IPV across multiple agencies (see Shepard et al., 2002). Sixth, any offender incarcerated in a participating facility who had a history of IPV was eligible for the program; according to the risk–needs–responsivity (RNR) model, programming should be offered to high-risk offenders whose criminogenic needs will be addressed by the curriculum (Andrews et al., 1990; Lowenkamp, Latessa, & Holsinger, 2006).

In addition, some aspects of the prison setting affected the program’s implementation and success. First, it was sometimes necessary to cancel group meetings due to lockdowns and other security reasons.⁸ In those instances, homework was assigned and participants had one-on-one sessions with the facilitator; therefore, while the entire curriculum was provided, some offenders were not able to work through all the material in a group setting. Second, unlike those serving sentences in the community, incarcerated offenders typically do not have continuing face-to-face contact with the victims of their offenses.⁹ Therefore, offenders are not able to put the skills they learn in the program into practice until well after they have completed the program. Similarly, the program was not able to incorporate victims or relationship partners into the treatment, which is an important aspect of effective IPV programs (Pitts, Givens, & McNeeley, 2009). Programs for incarcerated offenders should carefully consider adjustments that account for these issues.

The null results found here may be the result of these problems with implementation and the other limitations described above and therefore should not necessarily be taken as evidence of the ineffectiveness of the Duluth Model. Still, considering the wealth of previous research showing that these programs have only limited effects on recidivism (e.g., Arias et al., 2013; Babcock et al., 2004; Cheng et al., 2021; Feder & Wilson, 2005; Miller et al., 2013), it is also possible that the DVIPP did not reduce recidivism because the Duluth Model—on which it was based—is itself flawed. As explained by Latessa, Cullen, and Gendreau (2002), a program that does not address major criminogenic needs—factors known to contribute to offending, such as antisocial thinking styles, antisocial personality traits, and antisocial peers—is unlikely to reduce recidivism. While the curriculum focuses on particular attitudes that may be criminogenic (i.e., attitudes about women and gender roles in relationships), not all IPV can be attributed to the influence of patriarchal attitudes over offenders’ relationships. In fact, Ellen Pence, one of the creators of the program, later noted that, in contrast to the philosophy that the program is based on, “many of the men [she] interviewed did not seem to articulate a desire for power over their partner” (Pence, 1999, p. 29). Other research supports this notion: Most domestic violence offenders do not specialize in domestic violence (Piquero, Brame, Fagan, & Moffitt, 2006; Richards, Jennings, Tomsich, & Gover, 2012) and domestic violence offending is related to the same risk factors as other types of offending (Sellers, 1999).

Importantly, the Duluth Model is only designed to address violence that is perpetrated by a male partner within the context of a heterosexual relationship. Furthermore, research suggests that issues related to race, ethnicity, immigration, socioeconomic status, religion, mental health, and disability may contribute to IPV (Bograd, 1999; Ellison, Trinitapoli, Anderson, & Johnson, 2007; Elvey & McNeeley, 2018; Kessler, Molnar, Feurer, & Appelbaum, 2001; Mays, 2006; Menjívar & Salcido, 2002). However, the curriculum is not designed to address these issues. This is an important shortcoming given that approximately two thirds (67%) of the program participants were members of minority groups.

The results point to several recommendations for IPV programming. First, programs should follow the RNR model by, first, targeting high-risk offenders and, second, using a needs assessment to select the most appropriate participants. Johnson (2006) identified four types of violence that occurs within relationships—situational couple violence, violent resistance, intimate terrorism, and mutual violent control—that are often grouped together and labeled “domestic violence.” Programs may be more effective if they target a specific type of domestic violence offender (see Langlands, Ward, & Gilchrist, 2009). Accordingly, programs based on the Duluth Model may be more effective if a screening tool is developed to identify offenders whose attitudes that are favorable toward IPV contribute to their offending.

Second, correctional programs for incarcerated offenders are most effective when they begin in the facility and continue through reentry (Ndrecka, 2014). Programs for incarcerated IPV offenders should create a continuum of care in which the program continues in the community after the offender is released. This can be done by beginning treatment later in the offender’s sentence so that he completes the program shortly before release, which has been shown to aid in reducing recidivism (Duwe, 2018).

Third, as cognitive-behavioral therapy (CBT) has been shown to be an effective correctional treatment (e.g., Landenberger & Lipsey, 2005), it may be useful for IPV programs to incorporate more CBT components. Although many IPV programs do incorporate some aspects of CBT techniques, there is insufficient research to determine its effectiveness (see Babcock et al., 2004; Smedslund, Dalsbø, Steiro, Winsvold, & Clench-Aas, 2007). A recent review (Nesset et al., 2019) found only six controlled studies published since 2007; the results suggested that, while some studies show reduced violence among IPV offenders who participate in CBT, there was insufficient evidence to determine its success. Given the effectiveness of CBT programming among violent offenders more generally (Lee & DiGiuseppe, 2018), it is assumed that the effectiveness of IPV programming may increase with additional focus on helping offenders think creatively, develop problem-solving skills, and manage their emotions, as well as assignments such as thinking reports. However, it is imperative that future CBT programs developed for IPV offenders include an evaluation component. Furthermore, as offenders may find it difficult to work through programming when they live in general population, it may be helpful to create therapeutic communities in which offenders are surrounded by staff who are trained in CBT programming and other members of the therapeutic group, allowing them to spend more time engaged in the cognitive-behavioral aspects of the program.

Prior research suggests that programs based on the Duluth Model have only limited success in reducing reoffending. However, most of the existing research examines programs that were offered in the community. This study adds to the literature by evaluating an intervention offered to offenders incarcerated for IPV. The results suggest that programs based on this model are not more successful at reducing recidivism when offered in a prison setting than when offered in the community. Programs for incarcerated IPV offenders should utilize risk and needs assessments to select participants, include community-based aftercare after release, incorporate elements from successful correctional programs such as CBT, and make adaptations to account for cultural differences (see Primm, Osher, & Gomez, 2005).

Footnotes

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) received no financial support for the research, authorship, and/or publication of this article.

ORCID iD

Susan McNeeley

Notes

Author Biography

Susan McNeeley is a research analyst with the Minnesota Department of Corrections. In addition to corrections, her research interests include criminological theory and victimology. Her recent work is published in journals such as Justice Quarterly, Journal of Criminal Justice, Journal of Experimental Criminology, Criminal Justice and Behavior, and Crime and Delinquency.

References

Alexander

P. C.

Morris

Tracy

Frye

(2010). Stages of change and the group treatment of batterers: A randomized clinical trial. Violence and Victims, 25, 571-587.

Andrews

D. A.

Zinger

Hoge

R. D.

Bonta

Gendreau

Cullen

F. T.

(1990). Does correctional treatment work? A clinically relevant and psychologically informed meta-analysis. Criminology, 28, 369-404.

Arias

Arce

Vilariño

(2013). Batterer intervention programmes: A meta-analytic review of effectiveness. Psychosocial Intervention, 22, 153-160.

Babcock

Armenti

Cannon

Lauve-Moon

Buttell

Ferreira

. . . Lehmann

(2016). Domestic violence perpetrator programs: A proposal for evidence-based standards in the United States. Partner Abuse, 7, 355-460.

Babcock

J. C.

Green

C. E.

Robie

(2004). Does batterers’ treatment work? A meta-analytic review of domestic violence treatment. Clinical Psychology Review, 23, 1023-1053.

Babcock

J. C.

Steiner

(1999). The relationship between treatment, incarceration and recidivism of battering: A program evaluation of Seattle’s coordinated community response to domestic violence. Journal of Family Psychology, 13, 46-59.

Bennett

Stoops

Call

Flett

(2007). Program completion and re-arrest in a batterer intervention system. Research on Social Work Practice, 17, 42-54.

Bograd

(1999). Strengthening domestic violence theories: Intersections of race, class, sexual orientation, and gender. Journal of Marital and Family Therapy, 25, 275-289.

Bowen

Gilchrist

E. A.

Beech

A. R.

(2005). An examination of the impact of community-based rehabilitation on the offending behavior of male domestic violence offenders and the characteristics associated with recidivism. Legal and Criminological Psychology, 10, 189-209.

10.

Breiding

M. J.

Smith

S. G.

Basile

K. C.

Walters

M. L.

Chen

Merrick

M. T.

(2014). Prevalence and characteristics of sexual violence, stalking, and intimate partner violence victimization—National Intimate Partner and Sexual Violence Survey, United States, 2011. MMWR Surveillance Summaries, 63, 1-18.

11.

Cannon

Hamel

Buttell

Ferreira

R. J.

(2016). A survey of domestic violence perpetrator programs in the United States and Canada: Findings and implications for policy intervention. Partner Abuse, 7, 226-276.

12.

Cheng

Davis

Jonson-Reid

Yaeger

(2021). Compared to what? A meta-analysis of batterer intervention studies using nontreated controls or comparisons. Trauma, Violence, & Abuse, 23(3), 496-511. doi:10.1177/1524838019865927

13.

Coulter

VandeWeerd

(2009). Reducing domestic violence and other criminal recidivism: Effectiveness of a multilevel batterers intervention program. Violence and Victims, 24, 139-152.

14.

Crane

C. A.

Eckhardt

C. I.

(2013). Evaluation of a single-session brief motivational enhancement intervention for partner abusive men. Journal of Counseling Psychology, 60, 180-187.

15.

Davis

R. C.

Taylor

B. G.

Maxwell

C. D.

(2000). Does batterer treatment reduce violence? A randomized experiment in Brooklyn (Document No. NCJ 180772). New York, NY: Victims Services Research. Retrieved from https://www.ncjrs.gov/pdffiles1/nij/grants/180772.pdf

16.

Dobash

R. E.

Cavanagh

Lewis

(1996). Reeducation programs for violent men: An evaluation. Research Findings, 46, 309-322.

17.

Dunford

(2000). The San Diego Navy experiment: An assessment of interventions for men who assault their wives. Journal of Consulting and Clinical Psychology, 68, 468-476.

18.

Duwe

(2018). The effects of timing and dosage of correctional programming on recidivism. Journal of Offender Rehabilitation, 57, 256-271.

19.

Duwe

Rocque

(2017). Effects of automating recidivism risk assessment on reliability, predictive validity, and return on investment (ROI). Criminology & Public Policy, 16, 235-269.

20.

Eckhardt

C. I.

Murphy

C. M.

Whitaker

D. J.

Sprunger

Dykstra

Woodard

(2013). The effectiveness of intervention programs for perpetrators and victims of intimate partner violence. Partner Abuse, 4, 196-231.

21.

Ellison

C. G.

Trinitapoli

J. A.

Anderson

K. L.

Johnson

B. R.

(2007). Race/ethnicity, religious involvement, and domestic violence. Violence Against Women, 13, 1094-1112.

22.

Elvey

McNeeley

(2018). Target congruence as a means of understanding risk of intimate partner violence: A comparison of male and female college students in the United States. Crime & Delinquency, 65, 1823-1849.

23.

Farabee

Prendergast

Carter

Wexler

Knight

Anglin

M. D

. (1999). Barriers to implementing effective correctional drug treatment programs. The Prison Journal, 79, 150-162.

24.

Feder

Dugan

(2002). A test of the efficacy of court-mandated counseling for domestic violence offenders: The Broward experiment. Justice Quarterly, 19, 343-375.

25.

Feder

Forde

D. R.

(2000). A test of the efficacy of court-mandated counseling for domestic violence offenders: The Broward experiment (Final report, Document No. NCJ 184752). Memphis, TN: University of Memphis. Retrieved from https://www.ncjrs.gov/pdffiles1/nij/199729.pdf

26.

Feder

Wilson

D. B.

(2005). A meta-analytic review of court-mandated batterer intervention programs: Can courts affect abusers’ behavior? Journal of Experimental Criminology, 1, 239-262.

27.

Ford

Regoli

M. J.

(1993). The criminal prosecution of wife assaulters. In Hilton

(Ed.), Legal responses to wife assault: Current trends and evaluation (pp. 127-164). Newbury Park, CA: SAGE.

28.

Gendreau

Goggin

Smith

(1999). The forgotten issue in effective correctional treatment: Program implementation. International Journal of Offender Therapy and Comparative Criminology, 43, 180-187.

29.

Gondolf

E. W.

(2002). Batterer intervention systems. Thousand Oaks, CA: SAGE.

30.

Gondolf

E. W.

(2004). Evaluating batterer counseling programs: A difficult task showing some effects. Aggression and Violent Behavior, 9, 605-631.

31.

Gordon

J. A.

Moriarty

L. J

. (2003). The effects of domestic violence batterer treatment on domestic violence recidivism: The Chesterfield County experience. Criminal Justice and Behavior, 30, 118-134.

32.

Haggård

Freij

Danielsson

Wenander

Långström

(2017). Effectiveness of the IDAP treatment program for male perpetrators of intimate partner violence: A controlled study of criminal recidivism. Journal of Interpersonal Violence, 32, 1027-1043.

33.

Harrell

A. V.

(1991). Evaluation of court-ordered treatment for domestic violence offenders (Final report). Washington, DC: Urban Institute.

34.

Hasisi

Shoham

Weisburd

Haviv

Zelig

(2016). The “care package,” prison domestic violence programs and recidivism: A quasi-experimental study. Journal of Experimental Criminology, 12, 563-586.

35.

Johnson

M. P.

(2006). Conflict and control: Gender symmetry and asymmetry in domestic violence. Violence Against Women, 12, 1003-1018.

36.

Kessler

R. C.

Molnar

B. E.

Feurer

I. D.

Appelbaum

(2001). Patterns and mental health predictors of domestic violence in the United States: Results from the National Comorbidity Survey. International Journal of Law and Psychiatry, 24, 487-508.

37.

Labriola

Rempel

Davis

R. C.

(2008). Do batterer programs reduce recidivism? Results from a randomized trial in the Bronx. Justice Quarterly, 25, 252-282.

38.

Landenberger

N. A.

Lipsey

M. W.

(2005). The positive effects of cognitive-behavioral programs for offenders: A meta-analysis of factors associated with effective treatment. Journal of Experimental Criminology, 1, 451-476.

39.

Langlands

R. L.

Ward

Gilchrist

(2009). Applying the good lives model to male perpetrators of domestic violence. Behaviour Change, 26, 113-129.

40.

Langton

Berzofsky

(2012). Victimizations not reported to the police, 2006-2010. Washington, DC: U.S. Department of Justice, Bureau of Justice Statistics.

41.

Latessa

E. J.

Cullen

F. T.

Gendreau

(2002). Beyond correctional quackery—Professionalism and the possibility of effective treatment. Federal Probation, 66, 43-49.

42.

Lee

A. H.

DiGiuseppe

(2018). Anger and aggression treatments: A review of meta-analyses. Current Opinion in Psychology, 19, 65-74.

43.

Lee

L. C.

Stohr

M. K.

(2012). A critique and qualified defense of “correctional quackery.” Journal of Contemporary Criminal Justice, 28, 96-112.

44.

Lila

Gracia

Catalá-Miñana

(2018). Individualized motivational plans in batterer intervention programs: A randomized clinical trial. Journal of Consulting and Clinical Psychology, 86, 309-320.

45.

Lin

Chou

F. H.

Chen

Huang

G. T.

Chen

(2009). Domestic violence recidivism in high-risk Taiwanese offenders after the completion of violent treatment programs. Journal of Forensic Psychiatry & Psychology, 20, 458-472.

46.

Lowenkamp

C. T.

Latessa

E. J.

Holsinger

A. M.

(2006). The risk principle in action: What have we learned from 13,676 offenders and 97 correctional programs? Crime & Delinquency, 52, 77-93.

47.

Maxwell

C. D.

Davis

R. C.

Taylor

B. G.

(2010). The impact of length of domestic violence treatment on the patterns on subsequent intimate partner violence. Journal of Experimental Criminology, 6, 475-497.

48.

Mays

J. M.

(2006). Feminist disability theory: Domestic violence against women with a disability. Disability & Society, 21, 147-158.

49.

Menjívar

Salcido

(2002). Immigrant women and domestic violence: Common experiences in different countries. Gender & Society, 16, 898-920.

50.

Miller

Drake

Nafziger

(2013). What works to reduce recidivism by domestic violence offenders? (Document No. 13-01-1201). Olympia: Washington State Institute for Public Policy. Retrieved from https://www.wsipp.wa.gov/ReportFile/1119/Wsipp_What-Works-to-Reduce-Recidivism-by-Domestic-Violence-Offenders_Full-Report.pdf

51.

Murphy

C. M.

Musser

P. H.

Maton

K. I.

(1998). Coordinated community intervention for domestic abusers: Intervention system involvement and criminal recidivism. Journal of Family Violence, 13, 263-284.

52.

Musser

P. H.

Semiatin

J. N.

Taft

C. T.

Murphy

C. M.

(2008). Motivational interviewing as a pregroup intervention for partner-violent men. Violence and Victims, 23, 539-557.

53.

Ndrecka

(2014). The impact of reentry programs on recidivism: A meta-analysis (Unpublished doctoral dissertation, School of Criminal Justice, University of Cincinnati, OH).

54.

Nesset

M. B.

Lara-Cabrera

M. L.

Dalsbø

T. K.

Pedersen

S. A.

Bjørngaard

J. H.

Palmstierna

(2019). Cognitive behavioural group therapy for male perpetrators of intimate partner violence: A systematic review. BMC Psychiatry, 19, Article 11.

55.

Palmer

Brown

Barrera

(1992). Group treatment program for abusive husbands: Long-term evaluation. American Journal of Orthopsychiatry, 62, 276-283.

56.

Pence

(1999). Some thoughts on philosophy. In Shepard

Pence

(Eds.), Coordinating community responses to domestic violence: Lessons from Duluth and beyond (pp. 25-40). Thousand Oaks, CA: SAGE.

57.

Pence

Paymar

(1993). Education groups for men who batter: The Duluth Model. New York, NY: Springer.

58.

Penduzzi

Concato

Kemper

Holford

T. R.

Feinstein

A. R.

(1996). A simulation study of the number of events per variable in logistic regression analysis. Journal of Clinical Epidemiology, 49, 1373-1379.

59.

Piquero

A. R.

Brame

Fagan

Moffitt

T. E.

(2006). Assessing the offending activity of criminal domestic violence suspects: Offense specialization, escalation, and de-escalation evidence from the Spouse Assault Replication Program. Public Health Reports, 121, 409-418.

60.

Pitts

W. J.

Givens

McNeeley

(2009). The need for a holistic approach to specialized domestic violence court programming: Evaluating offender rehabilitation needs and recidivism. Juvenile & Family Court Journal, 60, 1-21.

61.

Price

B. J.

Rosenbaum

(2009). Batterer intervention programs: A report from the field. Violence and Victims, 24, 757-770.

62.

Primm

A. B.

Osher

F. C.

Gomez

M. B.

(2005). Race and ethnicity, mental health services and cultural competence in the criminal justice system: Are we ready to change? Community Mental Health Journal, 41, 557-569.

63.

Richards

T. N.

Jennings

W. G.

Tomsich

E. A.

Gover

A. R.

(2012). A longitudinal examination of offending and specialization among a sample of Massachusetts domestic violence offenders. Journal of Interpersonal Violence, 28, 643-663.

64.

Rosenbaum

P. R.

Rubin

D. B.

(1985). Constructing a control group using multivariate matched sampling methods that incorporate the propensity score. American Statistician, 39, 33-38.

65.

Saunders

D. G.

(1996). Feminist-cognitive-behavioral and process-psychodynamic treatments for men who batter: Interaction of abuser traits and treatment models. Violence and Victims, 11, 393-414.

66.

Sellers

C. S.

(1999). Self-control and intimate violence: An examination of the scope and specification of the general theory of crime. Criminology, 37, 375-404.

67.

Shadish

W. R.

Cook

T. D.

Campbell

D. T.

(2002). Experimental and quasi-experimental designs for generalized causal inference. Boston, MA: Houghton Mifflin.

68.

Shepard

(1992). Predicting batterer recidivism five years after intervention. Journal of Family Violence, 7, 167-178.

69.

Shepard

M. F.

Falk

D. R.

Elliott

B. A.

(2002). Enhancing coordinated community responses to reduce recidivism in cases of domestic violence. Journal of Interpersonal Violence, 28, 551-569.

70.

Sherman

L. W.

Berk

R. A.

(1984). The specific deterrent effects of arrest for domestic assault. American Sociological Review, 49, 261-272.

71.

Simpson

D. D.

(2001). Modeling treatment process and outcomes. Addiction, 96, 207-211.

72.

Smedslund

Dalsbø

T. K.

Steiro

A. K.

Winsvold

Clench-Aas

(2007). Cognitive behavioural therapy for men who physically abuse their female partner. Cochrane Database of Systematic Reviews, 3, CD006048. doi: 10.1002/14651858.CD006048.pub2

73.

Taylor

B. G.

Maxwell

C. D.

(2009). The effects of a short-term batterer treatment program for detained arrestees: A randomized experiment in the Sacramento County, California jail. Sacramento, CA: Department of Justice, Crime and Violence Prevention Center.

74.

Tjaden

Thoennes

(2000). Extent, nature, and consequences of intimate partner violence: Findings from the National Violence Against Women Survey. Washington, DC: U.S. Department of Justice.

75.

Tolleffson

D. G.

Gross

(2006). Predicting recidivism following participation in a batterer treatment program. Journal of Social Service Research, 32, 39-62.

76.

Vittinghoff

McCulloch

C. E.

(2007). Relaxing the rule of ten events per variable in logistic and Cox regression. American Journal of Epidemiology, 165, 710-718.

77.

Weithorn

L. A.

Carter

L. S.

Behrman

R. E.

(1999). Domestic violence and children: Analysis and recommendations. Future of Children, 9, 4-20.