Introduction

Domestic violence is considered a major global public health problem (Hart 1995; Pallitto et al. 2013). Battered women tend to suffer from psychological and physiological problems such as chronic pain, irritable bowel syndrome, gynecological problems, infertility and a variety of reproductive disorders, eating disorders, suicidal thoughts, clinical depression, psychosis, anxiety, emotional and behavioral disorders (Haggård et al. 2015; Howard et al. 2013; Sartin et al. 2006; Whitaker et al. 2006). Implications of the phenomenon are also seen in children, who often tend towards promiscuity, dropping out of school, substance abuse, depression, eating disorders, suicidal thoughts, aggression, early pregnancy, as well as deviant and criminal behaviors (Moylan et al. 2010; Sousa et al. 2011).

In light of domestic violence’s serious consequences for the victim, family, and society, domestic violence has become a significant social issue that requires intervention and treatment (Mankowski et al. 2002). Many Western countries have begun to focus efforts and resources towards the development of intervention programs, both private and public, seeking to reduce the risk that individuals will assault their partners. A variety of treatment methods may be found among those domestic violence programs, such as psychodynamic therapy (Montgomery 2002; Robertson 1999; Shedler 2010), educational therapy (Coulter and Vandeweerd 2009), psycho-educational therapy (Lee et al. 2014), Duluth domestic abuse intervention (Taylor et al. 2001) and cognitive behavioral treatment (CBT) (Coulter and Vandeweerd 2009; Enosh et al. 2013). These programs seek to reduce the probability that already violent partners will assault their spouses again in the future (Coulter and Vandeweerd 2009; Day et al. 2009; Raeder 2011).

There have been a number of studies that have examined whether or not domestic violence programs (voluntary or court mandated) are effective in reducing future violence/recidivism (Arias et al. 2013; Babcock et al. 2004; Babcock and Steiner 1999; Bennett 2001; Coulter and Vandeweerd 2009; Enosh et al. 2014; Feder and Wilson 2005; Feder et al. 2008; Gordon and Moriarty 2003; Hamm and Kite 1991; Sartin et al. 2006; Shepard 1992). However, the findings with respect to the effectiveness of these programs in actually reducing future violence are still mixed and inconclusive (Akoensi et al. 2013; Bennett 2001; Feder et al. 2008; Feder and Wilson 2005; Gordon and Moriarty 2003; Sartin et al. 2006). Moreover, only a few studies examined domestic violence programs that are operated in prison settings (Enosh et al. 2013; Ley 2005; Pascual-Leone et al. 2011).

More generally, in the context of prison-based rehabilitation, the literature suggests that a supportive climate (Liebling and Arnold 2004) and environment (Hunter 1993), together with a therapeutic milieu that exists in a prison setting have a positive influence on prisoners’ rehabilitation (Gideon et al. 2010). Some argue that the combination and integration of multiple, evidently successful approaches would lead to a greater reduction in recidivism among domestic violence offenders (Day et al. 2009). Thus far, there remains a lack of empirical evidence in the area of domestic violence programs in prisons regarding the impact of integrative treatment, where different methods and approaches toward domestic violence prisoners are employed within the same rehabilitation-program framework (Gondolf 2007).

The present study focuses on the evaluation of a domestic violence program in the Israeli prison system called “The House of Hope” (Beit HaTikva). This program combines and integrates three different basic therapies in its rehabilitation program: cognitive behavioral therapy, which has been shown to be effective in several studies (Landenberger and Lipsey 2005), psychodynamic therapy, and psycho-educational therapy. The program has an operational duration of approximately 1 year and also incorporates components of therapeutic community intervention, which encourages a supportive social climate.Footnote 1 Moreover, the treated prisoners in The House of Hope are subject to a well-structured routine based on work and educational training. Our main argument is that in cases of domestic violence, this kind of integrative and intensive therapeutic “package” is the most promising approach to the prisoners’ rehabilitation, especially when the treatment is extended over a sufficient duration of time and takes place in a supportive environment, as we shall see below.

The aim of our study is to examine the effects on recidivism of an integrative domestic violence program that takes place inside the Israeli prison system. For this purpose, we have drawn our comparison sample to treatment from all convicted prisoners who were released from prison between 2004 and 2012. In order to match the prisoners who participated in the domestic violence program to an appropriate control group, a propensity score matching method was used. The findings indicate that the rates of reincarceration and rearrests of inmates who participated in the domestic violence program were significantly lower during a period of up to 4 years after release. In the discussion, we attribute these positive results to the integrative characteristics of the prison-based domestic violence program.

Domestic violence programs in prison

Today, there are many different types of domestic violence programs in the Western world, constituting an important and integral part of the rehabilitation system operating within the detention facilities (Babcock and Steiner 1999; Coulter and Vandeweerd 2009; Day et al. 2009; Hamm and Kite 1991; Shepard 1992). Among these programs, a wide range of treatment methods may be found, such as Psychodynamic Therapy, Psycho-Educational Therapy and therapy based on the Cognitive–Behavioral Model (Coulter and Vandeweerd 2009). Additionally, most of the current domestic violence programs involve some version of group therapy (Enosh et al. 2013). The programs vary according to their stated goals and objectives, their disciplinary emphases and with regard to their understandings and assumptions with respect to the causes of violence (Day et al. 2009). Below we begin by elaborating briefly on the rationale behind these programs and then present empirical evidence with regard to their effectiveness.

Psycho-educational therapy

Many domestic violence offenders suffer from poor cognitive-behavioral skills and hold distorted beliefs and expectations regarding the role of their partners as part of the relationship. Therefore, many psycho-educational (also known as the Duluth model) programs adopt an approach that seeks to challenge perceptions such as that a man has the right to control or to dominate his spouse (Feder and Wilson 2007; Lee et al. 2014). Consequently, it is not surprising that many prisons use a psycho-educational therapy approach in domestic violence rehabilitation programs. This approach seeks to develop in the offenders’ minds an awareness of their violent behavior and associated consequences. Moreover, the psycho-educational therapy approach helps the offenders to take responsibility for their violent behavior and makes them aware of the power, force, and dominance which they wield against their partners. This type of treatment provides the offenders with new anger control and management tools as well as non-violent approaches to dealing with problems between themselves and their partners (Lee et al. 2014).

Psychodynamic therapy

Psychodynamic therapy relies on the principles of a psycho-analytical approach that views mental disorders as unconscious and often related to distress, repressed impulses and conflicts rooted in the individual’s past (Robertson 1999). In the context of domestic violence, the psychodynamic therapy method assumes that the offender’s use of violence stems from internal conflicts or unresolved problems from childhood such as relations with parents (i.e., abuse, rejection, neglect). Additional factors may be found in personality disorders, problems in communicating with parents, fear of intimacy, low self-esteem, and obsessive-compulsive disorder behavior (Carden 1994; Montgomery 2002). As part of psychodynamic therapy treatments the therapists help the offender to express their harassing or threatening unconscious feelings, especially those that create stress or resistance (Shedler 2010). This treatment aims to increase the offender’s awareness of the source of their aggressive behavior toward their partner.

Cognitive behavioral therapy

Another important intervention program for domestic violence is based on cognitive behavioral therapy (CBT). This program emphasizes the offender’s changing positions concerning their violent behavior and abuse of power against their partners (Coulter and Vandeweerd 2009). The goal of CBT is to change the offender’s cognitive perceptions, their feelings and finally the way in which they respond to situations of conflict. CBT focuses on the identification, recognition, and modification of the offender’s negative thoughts (anger, envy, hostility, etc.). These unconscious thoughts arise in the offender’s mind during conversations, conflicts or any other interactions with their partners. CBT tries to replace the negative feelings with rational thoughts and a constructive way of thinking. This treatment also aims to identify early indicators of violent behavior in order to provide the offender with techniques and strategies (communication skills and assertiveness training) to control their violent behavior (Enosh et al. 2013).

Group therapy

Most contemporary domestic violence programs involve some version of group therapy (Enosh et al. 2013). Initially group therapy sessions tend to focus on developing confidence, trust and achieving consensus between the therapists and the participants with regard to the purposes and objectives of the program. In these sessions the participants share their traumatic personal stories with the group; this initial stage is a critical first step on the way to an offender eventually changing their behavior (Mankowski et al. 2002). One of the major benefits of group therapy is its ability to demand that participants accept responsibility for their behavior and the damage they caused to their victims (Enosh et al. 2013). Group therapy also encourages its participants to express themselves by communicating freely and openly, skills that may be essential for the development of empathy and a nonviolent attitude towards the victims. Finally, the group setting encourages the participants to challenge destructive “masculine” beliefs and to adopt male roles that do not include harming of partners (Mankowski et al. 2002). All these elements help in dealing with issues such as violence, power and sex, communication skills, coping strategies and anger management (Coulter and Vandeweerd 2009; Day et al. 2009).

Prison-based therapeutic community

The setting of a therapeutic community is usually discussed in the context of drug abuser rehabilitation. However, the idea, techniques, and the implementation of a therapeutic community can be applied to other fields and types of rehabilitation (De-Leon 1997: 19), including domestic violence offenders, as in the case of the House of Hope operating in Israel. Therapeutic communities are guided by four interrelated and integrated views, which include “view of disorder”, “view of the person”, “view of recovery” and “view of right living”. The view of disorder refers to personal problems such as the cognitive, behavioral, emotional, medical, and social. The view of the person refers to the psychological dysfunction and/or social deficits of the offender such as self-esteem, difficulty coping with feelings, guilt, dishonesty, unrealistic expectations, problems with responsibility, problems with authority, poor impulse control, poor tolerance for frustration and other deficiencies including literacy, communication and attention. Concerning the view of recovery, De Leon suggests that recovery is a developmental learning and that both self-help and mutual self-help are crucial elements for successful recovery. Lastly, the view of right living places an emphasis on values, perceptions, and beliefs that are crucial for the process of recovery, such as social learning, personal growth and healthy living (De Leon 1997).

A therapeutic community’s purpose is to use the nature of a community or constructed community to teach and enable individuals who make up the community (the participants) to use the community to their advantage in effecting personal change. There are four basic elements that are unique to the therapeutic community, which lend to such capacity: Context refers to the relationship between participants, participants and staff and the daily routine; expectations relate to the expectation of individual participation as a prerequisite for the existence of a community; assessment relates to the constant assessing and reassessing of a participant’s meeting of the expectations; and responses relate to actions taken or not taken in response to the assessment of expectations being met or not (De Leon 1997).

Does it work? Domestic violence programs and recidivism

Most evaluation studies in the field of domestic violence programs have examined the impact on the behavior of offenders after participating in a treatment program (Akoensi et al. 2013; Babcock et al. 2004; Buttell and Carney 2004; Buttell and Pike 2003; Dutton and Corvo 2006; Feder and Wilson 2005; Haggård et al. 2015). One measure of success is the rate of recidivism (arrest or incarceration), either for general offenses, violent offenses or violent offenses against a close family member (Arias et al. 2013; Coulter and Vandeweerd 2009; Haggård et al. 2015; Ley 2005; Pascual-Leone et al. 2011). Many evaluations failed to demonstrate positive outcomes in terms of reductions in violence (Akoensi et al. 2013; Arias et al. 2013; Bennett 2001; Day et al. 2009; Feder and Wilson 2005; Gordon and Moriarty 2003; Haggård et al. 2015; Morrel et al. 2003).

Some of the studies which did find positive effects only identified very moderate improvements (Akoensi et al. 2013; Babcock et al. 2004; Buttell and Pike 2003; Dutton and Corvo 2006). Generally such findings were associated with programs that combine cognitive-behavioral treatment (Babcock et al. 2004; Haggård et al. 2015) with psycho-educational interventions (Babcock et al. 2004; Haggard et al. 2015).

For example, a meta-analysis conducted by Feder and her colleagues examined the impact of court-mandated batterer interventions, which included psycho-educational treatment, cognitive-behavioral treatment or both. The results were analyzed by separating between official reports and victim reports. When analyzing the official reports, they found a small positive reduction in repeat victimization, with such effects reflective of a 13 to 20 % reduction in recidivism rates. Conversely, when analyzing victim reports, no statistically significant differences were found between the treatment and the control groups (Feder et al. 2008). Babcock and his colleagues in their meta-analysis examined the findings of 22 studies evaluating domestic violence programs. They found that the overall effect size based on police and partner reports was d = 0.18. Moreover, experimental designs had an overall effect of d = 0.12, while quasi-experimental designs had an overall effect of d = 0.23. Babcock summarized that in both experimental and quasi-experimental designs, the impact of treatment on the reduction and cessation of repeated domestic assaults was only small, yet still significant (Babcock et al. 2004: 1043).

One issue with these findings, as argued by Feder and her colleagues (Feder et al. 2008; see also Akoensi et al. 2013), is that the positive results obtained in some of these studies were due to weak methodologies, such as creating control groups composed of offenders “who were rejected from the treatment or who rejected treatment” (Feder et al. 2008: 15).

As previously mentioned, studies evaluating the impact of in-prison domestic violence programs are in short (Enosh et al. 2013; Ley 2005; Pascual-Leone et al. 2011). Ley’s study examined a domestic violence program based on the cognitive-behavioral treatment approach. In this program, domestic violence offender prisoners participated in a CBT group over a period of 26 weeks. Ley found a significant reduction in the rate of recidivism at 1 year after release. The treatment group had an 18 % recidivism rate in comparison to 30 % in the control group (Ley 2005). Another important quasi-experimental study evaluated an emotion-focused group psychotherapy program for domestic violence offenders in prison. The study found a significant reduction in recidivism rates between the treated and untreated groups after 7 and 8 months post-release; however, these differences were no longer significant when examining at the 9-month mark or 1–3 years post-release (Pascual-Leone et al. 2011).

What works in domestic violence

Bowen and Gilchrist (2004) suggest that in order to understand what works, for whom and under what circumstances when it comes to domestic violence offenders, it is necessary to incorporate and investigate the psychological characteristics of the offenders as well as the characteristics of the specific treatment in a more holistic approach. Moreover, research indicates that treatment responsivity is influenced by the program theory including the implementation and the integrity of the program,Footnote 2 as well as the characteristics of treated prisoners (Bowen and Gilchrist 2004).

In this context, studies have shown that there is great heterogeneity between treated offenders in domestic violence programs (Bowen and Gilchrist 2004). In other words, existing literature about domestic violence therapy programs points out the difficulty in providing a cohesive psychological profile for all domestic violence offenders (Holtzworth-Munroe and Stuart 1994). Identification of the offender sub-category is of great importance when it comes to responsivity of the program due to the fact that different offenders benefit from different rehabilitative approaches (Bowen and Gilchrist 2004).

Prison climate and integrative treatment

Prison conditions are far from providing an ideal environment for treatment, and treatment providers in prisons face great difficulties (Toch 1977, Zellerer 2003). Prison climate refers to the social, emotional, organizational, and physical characteristics of a correctional institution and may have a major effect on recidivism itself as well as on rehabilitation measures (Ross et al. 2008). Liebling and Arnold (2004) added to these indicators the quality of the relationship between the staff and the prisoners, the level of trust, degree of support, humanity of the regime, perceived fairness, decency and respect shown to prisoners, the prisoners’ social life, the extent of family contact, level of well-being, opportunities for personal development, the degree of order and the meaning attached to the penal experience (Gideon et al. 2010; Liebling and Arnold 2004). Gideon, Shoham, and Weisburd suggest that a therapeutic milieu may improve the treatment process in prison and may reduce recidivism after release. This environment allows more mobility of prisoners within the prison, an open-door policy, which means that the cells remain open throughout the day and where better relations exist between the staff and the inmates (Gideon et al. 2010).

With respect to examining the effectiveness of such programs, the Correctional Program Assessment Inventory (CPAI) “is a standardized tool that assesses how well intervention programs in a given correctional setting adhere to known principles of effective treatment” (Smith and Schweitzer 2012: 8). Smith and Schweitzer subdivided the domains of the CPAI tool into two main categories: capacity and content. The capacity refers to whether the correctional facility is capable of delivering an evidence-based treatment program, which includes the organizational culture, program implementation and maintenance, management and staff characteristics, interagency communication and the evaluation and quality assurance of the intervention. The content, however, refers to the domains concerning the treatment and assessment including the risk-need practices, the program features and core correctional practices (Smith and Schweitzer 2012).

One explanation for the only modest results of existing programs could be the mono-dimensional feature of the most common treatment programs, whereas the common offender actually needs a “package” of treatments within the program. Day and his colleagues argue that integrating and combining several evidently successful approaches increases the chances of program success (Day et al. 2009).Footnote 3 Zellerer argues that education as well as treatment should be integral parts of the efforts in reducing violence of family perpetrators (Zellerer 2003).

In a similar vein, we suggest that integrating promising components of different treatments (such as CBT, educational and occupational trainings or psychodynamic therapy) will lead to the greatest chances for success. The combining and integration of different approaches increases the likelihood that the process will be suited to a wider spectrum of offenders with different needs. When combined with a supportive prison climate and which takes into account the capacity and the content of the intervention discussed above, a “care package” is created that promotes such increased chances for success. Hence, the aim of this paper is to test this assumption. In the foregoing sections we will describe an integrative program for domestic violence offenders in the Israeli prison system and evaluate its outcomes.

An integrative domestic violence program in Israel: the House of Hope

The House of Hope is part of a special division in the Hermon prison that is dedicated to operating a special program whose aims are to treat and rehabilitate domestic violent offenders, especially those offenders who committed crimes against their spouses. The program was inaugurated in the year 2000 and is run for a year at a time. The program’s primary goals are to reduce violent behaviors of offenders vis-à-vis their partners following release, experience meaningful relationships with their partners and children following release and to reduce the risk of recidivism. Other objectives of the program as mentioned by the therapists were reducing inter-generational transfer of violence, improving functioning of the released offender as a partner as well as a parent, and improving the quality of life for the family.

The program is operated as a separate division in the prison that is designed for 40 participants. The participants are divided into four groups, each living in one of the four parts of the building. Each group of ten participants lives in a residential area called a “sleeve.” Each sleeve contains five double bedrooms, designed to allow participants to experience partner-like relationships. By sharing a sleeve, each group of ten participants is made jointly responsible for maintenance of their sleeve, thus creating a “family” type function. The building’s surrounds include gardens that are maintained by the participants themselves. In addition, the department has many public places and rooms, including a large lobby and a laundry room. The therapy groups are generally convened in a large lobby, which has high ceilings, many windows, and natural light. As part of the open-door policy, all cells in the wing of the House of Hope are open during day and night, except for the main door, which is locked only during roll calls.

In contrast to the rigid characteristics of existing prison facilities in Israel, the organizational concepts of the House of Hope are characterized by an approach of adaptability. Therefore, the program’s acceptance criteria, rules and regulations as well as the nature of the therapy program is written in a manual that is flexible and one that is constantly being changed and updated based on the therapists and the program administrators’ knowledge, experience, and seniority. Prisoners who have entered the prison system are usually directed to the House of Hope only if the “sorting committee” of the Israeli Prison Service decides that they fit the treatment and program criteria. Another way of gaining admission into the House of Hope is through a recommendation of a social worker in other prisons or departments who believes that the prisoner fits the program’s framework. In any case, the treatment program is voluntary and a show of willingness on the part of the prospective participant is a requirement.

Program acceptance criteria

  • Prisoners who become participants in the House of Hope are those who have been convicted of domestic violent crime or prisoners with a “domestic violence profile”, which means that in previous convictions during the last 3 years they have been convicted of a domestic violence offense. In addition, in some cases when there is information of involvement or risk for future involvement in domestic violence regarding a specific prisoner, than he may also be able to enter the treatment program.

  • Prisoners who enter the House of Hope must have at least 1 year remaining in their sentence but no more than 3 years. The purpose is that participants should be able to complete the entire year of the program but also that they be released as close as possible to the completion of the program.

  • In order to be admitted to the treatment in the House of Hope, a prisoner must show willingness to participate in the program. In Israel, a general policy prohibits the forced participation in treatment programs in the prison system.

  • Prisoners who want to enter the treatment program must not use alcohol or drugs during the treatment. The intention of the operators is that most of the participants will be free of drug or alcohol abuse problems, and the program is designed with the inclusion of such assumptions.

  • Prisoners with a substantial psychiatric history or mental illness are not accepted to the House of Hope. However, it is possible for those with a mild psychiatric background to participate if they are mentally balanced with medications.

  • It is important that participants in the House of Hope are able to understand and communicate in the Hebrew language, even in a limited fashion. This is because all the therapeutic group activities are delivered in Hebrew.

A prisoner at high risk for escape or a prisoner that is known as dangerous by the IPS intelligence is not able to apply as a participant for the treatment program.

Entering process

Every prisoner entering into the Hermon prison, which is a rehabilitation-oriented prison, passes through a sorting committee a week following their arrival. In this committee, a decision is made about where the prisoner should be assigned to, depending on his skills, abilities, and needs. Some of the prisoners are assigned to formal education while some go to work and others attend workshops. It should be noted that these activities take place only in the morning hours, when there are no therapeutic groups or sessions. The classification and placement of the prisoners into a suitable activity during the morning is the first step in the process of implementing the Risk-Need-Responsivity in the House of Hope.

The treatment in the House of Hope

The House of Hope department (Wing) operates as a semi-structured therapeutic community, which runs a variety of treatment types and group therapies (Goethals et al. 2011). Depending on the therapeutic community perceptions, the offenders, participants in the program (who staff refer to as “patients”) will ideally remain in the facility and program for an extended period (1 year) and participate in the maintenance of the department in almost every aspect of day-to-day life. As is the case in other therapeutic communities there is a widespread perception that the cessation of violent behavior is dependent upon a number of factors, such as resocialization and global changes in identity, attitudes, behavior and lifestyle of the participant. Therefore, an integrated participant-defined agenda that includes a focus on maintenance of the department, participation in daily life, as well as a focus on employment, education, and treatment, are essential.

As mentioned, the department is characterized by intense and continuous daily routines commencing in the early morning. Following wakeup call the participants immediately begin their day with cleaning their rooms and completing other morning related chores. They are subsequently gathered for a short group session called “morning feeling”. In these sessions, the participants share with one another their current feelings and mood. After the session, all participants proceed to their usual, assigned activities, such as employment or education training and such activities operated until lunchtime. Following lunch, the participants return to the department where the routine becomes more therapy oriented and focused, with a variety of therapeutic groups as well as individual therapy. Each participant has a daily one-on-one session or meeting with a program therapist.

Maintaining a steady routine and adhering to the full agenda of the House of Hope is understood to be a primary component of the resocialization process of the participants, in which each of the different activities and therapies accords its own unique contribution (e.g., different care methods, education, and employment). The routine that exists in the House of Hope promotes mutual assistance and self-awareness. All aspects of life in the community are organized around a strict set of rules that are governed by quick responses to any violation.Footnote 4 There are two primary violations for which the result will be the removal of the participant from the program. These “red lines” relate to the use of violence or drugs. With respect to the latter, the program includes routine drug testing.Footnote 5

The drop-out rate in the House of Hope currently stands at approximately 40 %. The program’s facilitators believe that the primary reasons for drop-out are:

  • The participant’s lack of progress in the treatment

  • The participant negatively influences other participants in the therapeutic community.

  • It has been found that the participant used drugs or alcohol during the course of the program.

  • The participant does not follow the rules of the community or the orders given by staff members.

The therapeutic groups

Several therapeutic groups operate in the House of Hope, each of which has a different goal and which are sometimes guided by differing approaches, such as CBT, psychodynamic or psycho-educational. Some of the groups are conducted as “train groups” or continuous groups, which means that the group has no end-point, with new participants entering the group and seniors or those who have completed the program leaving the group without a fixed date, rather being based on each participant’s point of progress. The “morning feeling” group discussed below is a prime example of such a group. Another example for continuous groups is “The large group” —a dynamic group that operates weekly. The aim of this group is to deal with the dynamics and the conflicts surrounding the daily life in the therapeutic community and to raise awareness of past and present behavior patterns of the participants.

Moreover, there are closed groups that start every once in a while and usually last for 12 weeks (12 sessions) at a time. The rationale is that all the participants in the program will be part of all groups during their stay. Among these groups, one can find anger management psycho-educational groups, another type of psycho-educational group, which focuses on nonviolent communication skills; a self-awareness psychodynamic group; and also another psycho-educational group that is oriented to deal with parenting issues. There is also a psychodrama therapy group and a dynamic group, which places emphasis on the childhood of the participant. These groups are managed with the understanding that sometimes the source of the violent behavior is hinged on events that took place in childhood, such as abuse. The variety of the treatments given allows program facilitators to make adjustment for each participant, tailoring the treatments to take individual characteristics and mental states into account.

As described, the House of Hope is characterized by intense and continuous daily routines, which include CBT, psychodynamic, and psycho-educational therapy groups. All of these treatments, as previously noted, have individually demonstrated positive results in prior studies. The routine of the program provides the prisoners with significant opportunities for interpersonal interactions that serve as a platform conducive to ongoing treatment. The program also contains components of therapeutic community by taking place in a supportive environment.

Since the House of Hope contains a multi-dimensional and integrated treatment (CBT, psychodynamic, and psycho-educational therapy groups) together with a positive environment for rehabilitation, our research hypothesis is that this unique therapeutic “package” will significantly decrease recidivism among those who participated in the program.

Method: a quasi-experimental design

Data and sample

In our analysis, we use a large database from the Israel Prison Service (IPS) that contains n = 61,689 prisoners who were released between the years 2004 and 2012 to identify a comparison sample.Footnote 6 The database includes information on the socio-demographic characteristics of the prisoners, their previous arrests, and current imprisonment. In addition, the database contains rich data on the conduct of the inmates during imprisonment, including visits, vacations, and disciplinary hearings as well as participation in rehabilitation programs (education, employment, etc.).

In order to assess the program in the House of Hope, we selected only those prisoners with a domestic violence profile (n = 20,737). We then filtered all prisoners who did not meet the IPS requirements for admission to the program. Due to the IPS requirements, we filtered from the file all non-Israeli prisoners (n = 5098 prisoners), prisoners sentenced to a prison term of less than 1 year (n = 9480 prisoners), prisoners that did not pass a hearing in front of a domestic violence committee in prison (n = 1302 prisoners), and prisoners whose victim’s identity (wife and/or children) is not mentioned in the committee record (n = 2263 prisoners). After the filtering, we were left with a file of n = 2594 prisoners. Out of these prisoners, only 279 inmates participated in the House of Hope program, and had a hearing by a domestic violence committee. The remaining prisoners (n = 2315) had a domestic violence profile but had not participated in the House of Hope program and therefore were the appropriate sample from which we created our control group.

Measures

Dependent variables

Reincarceration – measures whether the prisoner returned to prison each year in a 5-year period (yes = 2, no = 0)

Rearrests – measures whether the prisoner was arrested each year in a 5-year period (yes = 2, no = 0)

Independent variables

Prisoners’ variables included whether a prisoner was an immigrant (yes = 1, no = 0), whether he was married (yes = 1, no = 0), whether he was Jewish (yes = 1, no = 0), whether he was Arab (yes = 1, no = 0), education (measured continuously), number of children (measured continuously), socio-economic indexFootnote 7 (measured continuously), number of prior incarcerations (measured continuously), age (measured continuously), length of sentence (measured continuously), domestic violent crime (yes = 1, no = 0), sexually-based crime (yes = 1, no = 0), violent crime (yes = 1, no = 0), property crime (yes = 1, no = 0), drug-related crime (yes = 1, no = 0), year of release (measured continuously), rated number of vocation (measured continuously), alcohol abuse (yes = 1, no = 0), dangerousnessFootnote 8 (yes = 1, no = 0).

Analytic technique

In order to construct a comparison control group we used a Propensity Score Matching (PSM) method. The first step in calculating the propensity score of each prisoner was to find which variables are important and increase the chance of getting into the treatment group. These variables included socio-demographic characteristics of the prisoner, alcohol consumption, level of dangerousness, imprisonment characteristics, criminal background, and prisoner profile. These propensity scores were calculated using a statistical model that predicts the feasibility by using the subject’s background factors. Each prisoner receives a value between 0 and 1, where 0 represents a zero probability score of participating in the program, and 1 indicates that the probability of participating in the program is 100 %.

After calculating the propensity score of each prisoner, which represents the probability of participating in the program, we matched every prisoner who took part in the House of Hope program to a prisoner that had the most similar score (“his twin”) from the group of prisoners who did not take part in the program. Since the treatment group was large enough, we decided to use the “nearest neighbor” method, which matches every prisoner in the treatment group with only one prisoner from the control group (single match approach). For almost each inmate participating in the House of Hope program we matched a prisoner from the control group with the closest propensity score. We used a “bandwidth” or caliper of 0.01 for selection. This means that the adjustment was carried out only if a prisoner’s score in the control group was far from the score of the prisoner from the treatment group by a maximum of 0.01. After running the propensity score matching and having two matched groups, treatment and control, we examined whether the propensity score analysis indeed provided balanced samples that eliminated the differences between the groups.

In our analysis, prisoners who participated in the House of Hope program were compared to inmates who did not participate in the program, in terms of general recidivism and specific recidivism of violent crimes. Before the matching process the treatment group included 279 prisoners who had been participants in the House of Hope program and who demonstrated an average propensity score of 0.172 (SD = 0.084, min = 0.144, max = 0.472). The control group included 2315 prisoners who did not take part in the intervention in the House of Hope and who had an average score of 0.099 (SD = 0.080, min = 0.0008, max = 0.446). After the matching process the treatment group consisted of 277 inmates with an average score of 0.169 (SD = 0.080, min = 0.014, max = 0.373) and the control group consisted of 277 inmates with an average score of 0.170 (SD = 0.080, min = 0.014, max = 0.370).

Table 1 shows the comparison between the two groups before and after the matching (using χ2 test for nominal variables and t-test for interval variables).

Table 1 Comparison between the treatment and the control groups before and after matching by PSM
Full size table

Table 1 indicates that prior to the matching significant differences existed between the treatment and control groups on eleven variables: Israeli born/Immigrant, nationality (Jewish/Arab), years of education, number of prior incarcerations, age at the time of imprisonment, sexually-based crimes, violent crimes, drug-related crimes, year of release, and alcohol abuse. Following the matching processes no significant differences remained between the groups.

Findings

Among released prisoners, reincarceration or rearrests are the most common measures used in research on recidivism. In our study, recidivism rates were examined by using four measures: reincarceration and rearrests in general; and reincarceration and rearrests on charges of violent offenses specifically. Chi-square tests were used to compare recidivism rates between prisoners who took part in the House of Hope program and the control group. Additionally, Cohen’s d represents the effect size.

General reincarceration

The chi-square tests showed significant differences between the treatment and control groups after a 4-year follow-up period following their release from prison (see Table 2). Prisoners who took part in the program were 61 % less likely to be reincarcerated in the first year after their release than those of the control group.Footnote 9 After 2 years, their chances of reincarceration were 47.5 % lower; after 3 years, 48.5 % lower; and after 4 years, 39.7 % lower lower (see illustration in Fig. 1).

Fig. 1
figure 1

General reincarceration rates for the treatment and control groups

Full size image
Table 2 General reincarceration rates for the treatment and control groups
Full size table

General rearrests

Examination of rearrests revealed a similar pattern where prisoners who took part in the House of Hope program were less likely to be rearrested than those who did not (see Table 3). These differences were significant throughout the 4-year follow-up period after release from prison. Prisoners who took part in the program were 47.9 % less likely to be rearrested in the first year after their release than those of the control group. After 2 years, program participants were 44.8 % less likely to be rearrested than nonparticipants; after 3 years they were 48.3 % less likely; and after 4 years they were 38.7 % less likely to be re-arrested (see illustration Fig. 2).

Table 3 General rearrests rates for the treatment and control groups
Full size table
Fig. 2
figure 2

General rearrest rates for the treatment and control groups

Full size image

Reincarceration on charges of violent offenses

We also examined reincarceration on charges of violent offenses. The chi-square tests showed significant differences between the treatment and control groups over the 3-year follow-up period from the time of release from prison (see Table 4). Prisoners who took part in the program were 72 % less likely to be reincarcerated on charges of violent offenses in the first year after their release than those of the control group. After 2 years, their chances of reincarceration on charges of violent offenses were 49 % lower; and after 3 years, 45 % lower than those in the control group. These differences are illustrated in Fig. 3.

Table 4 Reincarceration rates on charges of violent offenses for the treatment and control groups
Full size table
Fig. 3
figure 3

Reincarceration rates on charges of violent offenses for the treatment and control groups

Full size image

Rearrests on charges of violent offenses

An examination of rearrests revealed a similar pattern; prisoners who took part in the House of Hope program were less likely to be rearrested than those who did not (see Table 5). These differences were significant also throughout a follow-up period of 3 years after release from prison. Prisoners who took part in the program were 64 % less likely to be rearrested on charges of violent offenses in the first year after their release than those who did not participate in the program. After 2 years, program participants were 60 % less likely to be rearrested on charges of violent offenses than nonparticipants were; and after 3 years, they were 55 % less likely to be rearrested on such charges (see illustration in Fig. 4).

Table 5 Rearrests rates on charges of violent offenses for the treatment and control groups
Full size table
Fig. 4
figure 4

Rearrest on charges of violent offenses for the treatment and control groups

Full size image

Sensitivity analysis for average treatment effects

In the PSM method, we would ideally observe all of the variables that have an influence on the odds of entering the treatment and that can influence the outcome variables. However, if the study does not include all these variables in the analysis then a hidden bias might arise as a result of these unobserved variables. The Rosenbaum approach tries to find out to what degree a study is sensitive to this kind of bias (Rosenbaum 2002). In a case where the outcome is binary, such as in the present study, the MHbounds (Mantel and Haenszel bounds) should be used (Aakvik 2001). In this test, when Gamma is equal to 1, it means that there is no hidden bias, whereas when the Gamma rises above 1 this is an indication that there might be unobserved variables that can influence our results. This test only refers to the study’s sensitivity to unobserved variables but the researcher does not know if such an unobserved bias really exists or if it indeed influences the results. If the Gamma becomes significant in an early stage it means that the study becomes less tolerant to unobserved factors.

In our study, the test becomes insignificant: Gamma = 1.8 in the first year, Gamma = 1.5 in the second year, Gamma = 1.6 in the third year and Gamma = 1.2 in the fourth year after release. This suggests that the problem of bias might arise only if there is an unobserved variable that increases the odds of entering the treatment by about 50–80 % or decreases the odds of being recidivist in the treatment group by about 50–80 % in the first 3 years and by about 20 % in the fourth year.

There is no clear definition of what thresholds are required for a model to meet a strong disregardable assumption. However, our results here are consistent with other published studies in criminology and criminal justice and suggest that the results are robust and tolerant of modest violations of PSM approach assumptions (e.g. see Kirk and Hardy 2014; Kirk and Sampson 2013; Lee and Thompson 2008; Wermink et al. 2010).

Limitations

Our analysis and presentation of findings is based on cumulative percentages across time. However, there are certain limitations to such an examination, especially relating to follow-up findings. In our study, we examined follow-up marks at the 1-, 2-, 3-, 4-, and 5-year marks. As stated at the outset of this article our database consisted of prisoners who were released between the years of 2004 and 2012. Additionally, follow-up data was available only until the year 2013. This resulted in a reduction of the N of remaining cases for each year from 2008 onwards. For example, a prisoner released in 2010 will only have 3 years of follow-up data available, whereas a prisoner released prior to 2008 provides full follow-up data. As per Table 2, the number of participants released from the House of Hope decreases from 247 to 53 between the first and fifth years of follow-up. On account of this we can assume that there is no significant difference in recidivism rates between treatment and control group at the 5-year mark due to a reduction of statistical power at this juncture. Additionally, as is the case with many other treatments in a variety of areas, treatment effects may fade over time.

A second limitation exists with respect to the second analysis, which refers to specific type and form of recidivism, namely charges of violent offenses. In our analysis, in order to see whether the treatment influenced both general recidivism and also violent behavior, on account of the unique characteristics of domestic violent offenders, we checked both categories of recidivism. However, due to the limitations of the dataset we only had recidivism information pertaining to violent offenses and not specifically domestic violence offenses. Whilst some prisoners may be listed in the database as having a domestic violence profile, this is not necessarily an indication that the violent offense for which they were rearrested or reincarcerated was related to domestic violence. We note that most research on domestic violence offenders in prison tends to examine only general recidivism. We therefore have attempted to decrease the effects of these limitations by including the examination of specifically violent related recidivism.

Conclusions

Experts in the area of domestic violence treatment programs in prison argue that the combination of multiple, evidently successful approaches would lead to a greater reduction in recidivism among domestic violence offenders (Day et al. 2009). Most studies on the effectiveness of domestic violence programs have analyzed these components separately while the overall evaluation of integrative rehabilitation programs has not yet been convincingly or sufficiently assessed. The purpose of the current study was to assess the effectiveness of integrative domestic violence program in the Israeli prison system, The House of Hope. This program is designed to include different treatments such as cognitive behavior treatment, and psychodynamic and psycho-educational therapy groups. In addition, the program contains components of therapeutic community and takes place in a supportive environment and in a prison with an open-door policy. The integrative characteristics of this program give us an opportunity to evaluate its effectiveness on the recidivism rates of domestic violence offenders.

Our analysis is based on a comparison of recidivism, measured by rearrest and reimprisonment for the years following release, between prisoners who participated in the domestic violence program, and the matched control group. We have examined recidivism rates both for general offenses and for violent offenses in particular. As for general recidivism, the findings indicate that the rates of reincarceration and rearrests of those who participated in the program were significantly lower during a period of up to 4 years after release. After 4 years from release, the risk of reincarceration within the treatment group was 39.7 % lower in comparison to the control group and the risk of rearrests was 38.7 %. Similar findings were observed with regard to recidivism for violent offenses. We found that the rates of reincarceration and rearrests on charges of specifically violent crimes of those who participated in the domestic violence program were significantly lower during a period of up to 3 years after release, in comparison to those who did not participate in this program. After 3 years from release, the risk of reincarceration and rearrest (on charges of specific violence crimes) among the participants in the program was significantly lower than the control group (49.0 and 55 %, respectively).

In developing the explanations for the success of the House of Hope program, we conducted in-depth interviews with past and present ward managers of the House of Hope, ward therapists, the head of rehabilitation and treatment in the Israel Prison Service and the domestic violence and sexual offenses coordinator in the Israel Prison Service. Based on these interviews, we identified two main factors that constitute the treatment package that influence the success of the House of Hope program. First, the integration of several therapeutic paradigms and types of treatment in the ward — Our interviews show that the House of Hope uses a wide range of treatments that include group therapy, CBT, psycho-educational or psychodynamic approaches, along with individual psychodynamic therapy, and answers the psychological needs of a variety of domestic violence offender profiles. Second, the length of treatment and prison climate — It appears that the relatively elongated treatment period at the House of Hope (approximately 1 year) and the on-going therapeutic atmosphere in the ward is conducive to enabling a profound change among participants. The length of treatment is accompanied with a positive rehabilitation environment (open-door policy), and the physical, architectural elements of the building’s structure convey such to the participant prisoners. Additionally, as our findings suggest a potential for the effects of treatment to fade over time, we note that this is an issue that is taken into consideration by the program facilitators in the program’s design and delivery by only taking in participants with between 1–3 years left on their sentences. As noted above and as obtained through interviews with facilitators and managers, the rationale is that treatment closer to release has a more positive and lasting effect. It therefore may be fruitful for professionals constructing treatment programs to include some form of follow-up treatment following release.

The current study shows that integrating a package of evidence-based treatments in a supportive prison environment can lead to significant and strong reductions in general recidivism and recidivism for violent offenses. These are promising findings for addressing an important social problem. While our study is only a single evaluation, it certainly suggests that the approach should be tested more widely. Finally, we had access to a strong body of data from which we developed our propensity score models. Such data allowed us to draw strong conclusions and, as indicated by sensitivity tests, they are robust even under modest violations of model assumptions. Nonetheless, absent a randomized experimental trial it is not possible to completely rule out confounding from our estimates. This suggests the importance of replicating these findings in other jurisdictions using experimental methods.