Searching for the mechanismsof change: a protocol for a realist reviewof batterer treatment programmes

Alisa J Velonis,1,2 Rebecca Cheff,1 Debbie Finn,1 Whitney Davloor,1

Patricia O’Campo1,3

To cite: Velonis AJ, Cheff R,Finn D, et al. Searching forthe mechanisms of change: aprotocol for a realist reviewof batterer treatmentprogrammes. BMJ Open2016;6:e010173.doi:10.1136/bmjopen-2015-010173

▸ Prepublication history forthis paper is available online.To view these files pleasevisit the journal online(http://dx.doi.org/10.1136/bmjopen-2015-010173).

Received 2 October 2015Revised 15 January 2016Accepted 1 March 2016

1Centre for Research on InnerCity Health, St Michael’sHospital, Toronto, Ontario,Canada2Division of CommunityHealth Sciences, University ofIllinois at Chicago, School ofPublic Health, Chicago,Illinois, USA3Dalla Lana School of PublicHealth, University of Toronto,Toronto, Ontario, Canada

Correspondence toDr Alisa J Velonis; avelonis@uic.edu

ABSTRACTIntroduction: Conflicting results reported byevaluations of typical batterer intervention programmesleave many judicial officials and policymakers uncertainabout the best way to respond to domestic violence,and whether to recommend and fund theseprogrammes. Traditional evaluations and systematicreviews tend to focus predominantly on whether theprogrammes ‘worked’ (eg, reduced recidivism) often atthe exclusion of understanding for whom they may ormay not have worked, under what circumstances, andwhy.Methods and analysis: We are undertaking a realistreview of the batterer treatment programme literaturewith the aim of addressing this gap. Keeping with thegoals of realist review, our primary aims are to identifythe theory that underlies these programmes, highlightthe mechanisms that trigger changes in participantbehaviour and finally explain why these programmeshelp some individuals reduce their use of violence andunder what conditions they are effective or noteffective. We begin by describing the process ofperpetrator treatment, and by proposing an initialtheoretical model of behaviour change that will betested by our review. We then describe the criteria forinclusion of an evaluation into the review, the searchstrategy we will use to identify the studies, and theplan for data extraction and analysis.Ethics and dissemination: The results of this reviewwill be written up using the RAMESES Guidelines forRealist Synthesis, and disseminated through peer-reviewed publications aimed at the practitionercommunity as well as presented at community forums,and at violence against women conferences. Ethicsapproval was not needed.

INTRODUCTIONSince the emergence of early batterer inter-vention or treatment programmes (BIPs) inthe 1970s and 1980s, discussions about theirefficacy have proliferated within research,professional and policy circles.1–3 In attempt-ing to answer the general question, ‘Do theseprogrammes work?’ a number of subsurface

debates have emerged, highlighting pointsof contention about the nature of intimatepartner violence (IPV), about the multiplelevels and types of influences that may contrib-ute to abusive behaviour, about what ‘success’means in terms of batterer treatment andabout the ‘right’ approach to both achievingand evaluating this success.1 4 These differ-ences in perspective lead to conflicting con-clusions about the effectiveness of theseprogrammes, making the job of navigating theliterature that surrounds batterer interventionor treatment programmes challenging.Yet, as part of the official response to

domestic violence across North America,guidelines on sentencing, or other codifiedjudicial requirements frequently require indi-viduals who are convicted of crimes againstintimate partners to attend treatment or edu-cational programmes (the content and formatof which can vary widely) as a condition toreceiving a deferred sentence, probation orparole.5–7 This, among other factors, has

Strengths and limitations of this study

▪ Realist syntheses are based on the developmentof a solid initial theory describing how, why, forwhom and under what conditions programmestrategies generate key outcomes.

▪ We present our initial theory for a realist synthe-sis of batterer intervention programme evalua-tions, and our explanation for the importance ofconducting such a review.

▪ Our initial theory draws from existing theory aboutbatterer intervention programmes, and presents ahypothesis about how the primary strategies ofeducation, skills building and group processmight generate immediate outcomes includingparticipants’ desire to develop alternatives to vio-lence, use of non-violent communication and vio-lence avoidance skills, development of empathyfor partners and shifts among others.

▪ We present a search strategy and approach toanalysis that are consistent with realist principles.

Velonis AJ, et al. BMJ Open 2016;6:e010173. doi:10.1136/bmjopen-2015-010173 1

Open Access Protocol

on 29 May 2018 by guest. P

rotected by copyright.http://bm

jopen.bmj.com

/B

MJ O

pen: first published as 10.1136/bmjopen-2015-010173 on 6 A

pril 2016. Dow

nloaded from

spurred a proliferation of programme evaluations and sys-tematic reviews, most sharing a defined goal of determin-ing whether or not the programmes that currently existcan be proven to directly reduce subsequent violence andcriminal behaviours.3 8 While there has been significantdebate over what theoretical approach(es) should be usedto guide these programmes (eg, feminist theory, familysystems theory, cognitive behavioural theory), the natureof these discussions tends to be as political (eg, profemi-nist or antifeminist in rhetoric) as it is scientific.9–12 Fewsystematic reviews have attempted to examine the under-lying programmatic theory and understand how, whyand in what contexts these programmes work, or do notwork.

Focusing on theory rather than on programmesDrawing on the work of Sayer13 and other realist philoso-phers, Pawson14 describes interventions as ‘complexprocess that are inserted into complex structures’(p.79). Budgets get cut, referrals increase or decrease,participants resent attending mandated programmes,staff feel overworked and underappreciated, and the listgoes on. Even under the best of conditions, behaviourchange is difficult, and programmes often use multipletheoretical strategies to help clients move along the pathto improvement.Intervention programmes for batterers are particularly

tricky. While multiple models for batterer interventionexist, most function within the framework of a largercommunity and criminal justice-oriented response todomestic violence.1 Participants are primarily—althoughnot exclusively—required to attend BIPs as part of pro-bationary or deferred sentencing agreements,6 andwhile the programmatic details vary across jurisdictions,most BIPs are designed as a series of educational andskill-building group sessions that run from 12 to52 weeks.7 Regardless of the specific therapeutic, philo-sophical or political framework used, these programmesare impacted by a variety of internal and external factors,including the characteristics and experiences of partici-pants and staff, the mission of the lead organisation, thelevels of communication between the programmes, thelocal courts or probationary departments, victim-centreddomestic violence services, and the social and politicalclimate of the larger community. In turn, these factorsinfluence how programmes run, how closely they adhereto the programme design, how the strategies are receivedby participants and more.Yet, a great deal of the research describing the ‘effect-

iveness’ of batterer intervention programmes has beendesigned to minimise the influence of these real-worldcontextual factors, generally by controlling for many ofthe very forces that could explain programme success orfailure (eg, cultural backgrounds, income, substanceuse). Often, experimental and quasiexperimental eva-luations are held up as the gold standard; these modelsendeavour to link the intervention—and ONLY theintervention—to a narrowly defined outcome, usually

recidivism or reoffense.3 8 15 16 A result has been the pro-liferation of a vast body of literature that shows mixed evi-dence of programme success with little explanation ofwhy. It was our frustration with this lack of explanation(and the frustration that programme administrators anddomestic violence advocates expressed to us about notknowing what to do to improve treatment) that led us toconduct this realist synthesis of the literature.What follows is a description of a protocol we have

developed for undertaking this review. As with otherrealist reviews, the purpose of this synthesis is explana-tory: to articulate underlying programme theories anduse evidence to determine their usefulness and rele-vance for batterer treatment.14 In writing and sharingthis protocol, we set forth three goals. First, based onour initial understanding of BIPs, to propose andexplain a ‘preliminary rough theory’ that captures theframework used by the majority of BIPs with the aim ofidentifying both strengths and gaps. Second, we wish toillustrate how a realist perspective can add value to ourunderstanding and interpretation of BIP evaluation byshifting focus from whether programmes work to howand why they (should) work. Finally, we desire to providetransparency for the forthcoming review, enablingreaders to know the specific steps that will be followedthroughout the review process and to highlight some ofthe specific challenges we face as we move forward.Part of what makes realist synthesis unique is the

emphasis it places on proposing, testing and ultimatelyrefining theory. Initially, reviewers conceptualise arough, preliminary theory that explains ‘what is sup-posed to happen?’, and ‘why is that supposed towork?’17 for the programme in question (a programmetheory). In the language of realist synthesis, the key ele-ments that need to be identified and understood in rela-tion to one another include the programme strategies(the activities that comprise the programme), contextualfactors that influence how and for whom the pro-grammes operate (such as the individual-level character-istics mentioned above as well as the structural contextin which programmes operate, such as funding andstatutory requirements), participant outcomes (eg, parti-cipants reduce their use of violence or recidivism) andthe hidden mechanisms or ‘generative processes’ thatoften take place within participants’ minds and triggerthe outcome (or, as Wong et al (ref. 17, p.6) put it, ‘whatit is about a programme that generates change’).17–19

For each intervention strategy that a programmeincludes, there may be multiple pathways that lead tomultiple outcomes; each pathway has its own set of con-textual factors and mechanisms at play, and the out-comes can be visible or hidden, intermediate or final,and intended or unintended.20 At the end of a realistsynthesis, the primary goal is the generation of a refinedtheory that takes the shape of a set of context–mechan-ism–outcome (CMOs) configurations, a heuristic usedto illustrate these relationships and pathways.20 It is atthis point that our theoretical lens becomes less focused

2 Velonis AJ, et al. BMJ Open 2016;6:e010173. doi:10.1136/bmjopen-2015-010173

Open Access

on 29 May 2018 by guest. P

rotected by copyright.http://bm

jopen.bmj.com

/B

MJ O

pen: first published as 10.1136/bmjopen-2015-010173 on 6 A

pril 2016. Dow

nloaded from

on the programmatic elements of BIPs and morefocused on the mechanisms of change.This protocol paper is organised into several sections

following the steps of realist synthesis laid out by Pawsonet al19 (figure 1). The core of any realist synthesis orevaluation is the description of the programmatic theorythat underpins the activity being evaluated (Step 1). Webegan this process by formulating our preliminary(rough) theory describing the strategies most BIPsemploy, the circumstances in which they operate, whatthey aim to accomplish and how it appears to us thatthese strategies will lead to those outcomes; essentially,this is a programme theory that describes, in generalterms, what is supposed to change and why as a result ofthe programme. As we develop this protocol, we are inthe initial stages of the review process, and what isreflected here reflects our initial thinking about BIPs.Furthermore, as Jagosh and colleagues remind us, realistreviews are abductive in nature, meaning that we infer‘to the best explanation’, iteratively ‘examining evidenceand developing hunches or ideas about the causalfactors linked to that evidence’ (ref. 20, p135). Afterlaying out our preliminary programme theory, we go onto describe the remaining steps (2–5) that will be takento conduct this review. Finally, we conclude with a discus-sion of why we believe this approach will contribute toour understanding of batterer treatment programmes,and how it can inform not only programme implementa-tion but also larger policy.

METHODS/DESIGNThe impetus for this review was a series of conversationswith programme facilitators and judicial personnel whohad grown frustrated with the lack of conclusive infor-mation about what their systems could do to reduce theperpetration of partner violence. As they saw it, the evi-dence was insufficient to declare batterer interventionsuseless, yet these programmes clearly did not work in allcircumstances. Using this to inform our approach, weframed our research question as for whom, and underwhat conditions, will batterer intervention programmeshelp men who have been identified as perpetrators ofpartner violence reduce their violent behaviour and why?

Step 1a: establishing the scope of our workBefore refining our scope and articulating the processeswe believe to be at work, we needed to understand what

the bigger picture of batterer response looked like inNorth America. After selecting a handful of evaluations,systematic reviews and reports from both the scientificand grey literature,2 3 5 8 15 21–27 we learned that themajority of individuals who attend batterer interventionin the USA and Canada undergo a two-part process:first, they enter the criminal justice system and are adju-dicated for an offense against a partner; and second,they are mandated to attend an educational/therapeutic‘treatment’ programme—what we refer to as a BIP—aspart of their sentence or agreement with the court.Since the vast majority of participants in BIPs first havecontact with the criminal justice system (and becauseour interest is in explaining how and why these pro-grammes work), we consider this to be part of the largerenvironment in which these programmes exist, ratherthat as one of the strategies or interventions that makeup BIP programmes. We recognise that this decision maybecome a limitation for this study, and that to gain a trueunderstanding of the mechanisms that underlie BIPs, wemay need to look more critically at the system-level issuesat work (eg, whether or not the level of communicationbetween BIP programme staff and court officials has animpact on BIP outcomes); if this appears to be the case aswe proceed, we will revisit this decision. However, theongoing debate over the efficacy of the programmes them-selves (and the lack of information about community-levelsystems contained within most programme evaluations)leads us to begin this review by focusing on BIPs.Another issue we encountered while considering the

scope of the review was the enormity of what we werecalling ‘individual perpetrator characteristics’, and howthey may interact with programme strategies to impactwhether participants respond to the programme (posi-tively or negatively). A substantial amount of literaturesupports the contention that men who engage in vio-lence against intimate partners and family members arenot all alike.28 Over the past 30 years, researchers haveattempted to categorise batterers according to psycho-logical, behavioural, attitudinal and/or motivationalcharacteristics using descriptors such as family only ortypical batterers, dysphoric/borderline or passive-aggressive-dependent batterers, and sociopathic or generally violent/antisocial batterers.29–32 Likewise, while Johnson doesnot provide a batterer typology, per se, he differentiatesbetween types of IPV (coercive-controlling violence, situ-ational partner violence and violent resistance), whichsuggests that the motivation behind these categories ofviolence would be different for different instigators.33

Finally, substantial evidence points to an overlapbetween substance abuse, a history of trauma, neglectand/or victimisation, and other psychological condi-tions, some of which are captured in the batterer typolo-gies described above, but which may also emerge asindependent issues that programmes may or may not beprepared to address.34 35

In no way are we claiming that individual-levelcharacteristics cause someone to engage in abusiveFigure 1 Key steps in realist review.

Velonis AJ, et al. BMJ Open 2016;6:e010173. doi:10.1136/bmjopen-2015-010173 3

Open Access

on 29 May 2018 by guest. P

rotected by copyright.http://bm

jopen.bmj.com

/B

MJ O

pen: first published as 10.1136/bmjopen-2015-010173 on 6 A

pril 2016. Dow

nloaded from

behaviours; rather, those factors may interfere with theeffectiveness of programme strategies.36 37 Unfortunately,individual programmes are often unable to assess for andaddress the myriad of issues that participants may bringto the intervention, and even in jurisdictions where com-pleting an intake assessment for substance abuse andmental health problems is mandatory, the availability ofand coordination between various treatment modalitiescan vary widely. Understandably, much of this detail is leftout of evaluation write-ups and formal reports, yet theseare influential factors that need to be acknowledged andaddressed. For the purposes of this review, we havedecided that we will consider them to be among the con-textual factors that can influence programme effective-ness, and where they are mentioned, we will note themaccordingly. However, because of the wide breadth of pos-sible influences, and the relative dearth of informationcollected about them in most evaluations, we have chosennot to limit our review to only those evaluations thataddress these issues.As we began developing our rough programme theory

for BIPs, we first drafted a flow chart illustrating the ‘bigpicture’ processes at work in most perpetrator interven-tions (figure 2). After reviewing various programmedescriptions and evaluations, we differentiated two setsof outcomes that were generally discussed: ‘proximal’outcomes, or changes that happen within the partici-pants as a result of participation in the programme activ-ities (such as changes in attitudes, skills and intentions);and ‘final’ outcomes, which would include recidivismand reassault, and which are generally measured as long-term consequences of programme participation.The majority of batterer intervention evaluations are

concerned primarily or exclusively with recidivism and/or reassault, which we found problematic for severalreasons. First, we contend that for BIPs to impact recidiv-ism or reassault, they first must achieve these moreimmediate changes in attitudes, motivations and skills.The final outcomes are at least one step removed fromthe BIP itself in that any changes in recidivism or violentbehaviour that are not preceded by changes in attitudes,motivations and skills may not be the result of the BIPprogramme. Furthermore, even after the proximal out-comes are met (if and when they are met), numerousfactors unrelated to the BIP programme itself may influ-ence whether participants reoffend. For example, if aperpetrator with co-occurring substance abuse or clinicaldepression receives little or no additional treatment forthose problems (whether because he refuses treatmentor because treatment is not available, affordable oraccessible), the progress he makes towards realising anon-abusive relationship (that may be achieved throughBIP participation) may be offset by the lack of assistancefor these other problems.Another problem with using recidivism or reoffense

measurements as the sole indicators of success or failureof BIPs is that definitions of these outcomes often differacross studies and may not be limited to assaults against

a partner, but can include a conviction for any violentcrime, an arrest for a violent crime (regardless of convic-tion) or even any subsequent arrest.1 Finally, becausemany—if not most—acts of abuse do not result in lawenforcement intervention, using this as an outcomelikely underestimates the recurrence of these beha-viours.1 3 38 Unfortunately, other measures of reassaultare likely to be equally unreliable, in that unless evalua-tors contact the perpetrators’ current partners, thesedata rely primarily on participants’ self-reported descrip-tions of their behaviours. Even when victim reports areincluded, the numbers are often small and subject toself-report bias. For these reasons, we chose to limit ourreview to the relationship between BIPs and these prox-imal outcomes, which are addressed in greater detailbelow.

Step 1b: forming an initial theory of change: looking forthe mechanismsOwing to the lack of agreement in the scientific commu-nity about whether partner violence is primarily a psy-chological, cognitive, developmental or social problem,the literature is replete with a variety of interventionapproaches. One of the challenges we faced was inidentifying a universal programme model that reflectsall or most BIPs; not only do different state and provin-cial jurisdictions outline different standards for pro-gramme length and content,6 7 but even at a local level,programmes can vary tremendously in how they are im-plemented and who they serve. In spite of this hetero-geneity, the vast majority of programmes across theUSA and Canada employ what could be called afeminist-informed cognitive-behavioural therapy (CBT)i

model.10 28 39 While the specific content, philosophicalemphases and details differ across jurisdictions and pro-grammes, our initial review of programme descriptionsindicate that most BIPs include a common set of ele-ments, including: (1) the use of educational strategiesthat challenge beliefs about gender equity, relationshipsand the impact of abuse; (2) skill-building activitiesintended to provide alternatives to abuse and violence;and (3) a facilitated group process that offers partici-pants both support and accountability.40

Based on this understanding of ‘generic’ BIPs, thereview team constructed our initial (rough) program-matic theory. After identifying what we felt were themost essential programme-level (proximal) outcomes—those that would be necessary to lead to further changein longer term (final) outcomes like recidivism—weworked backwards, asking ourselves what strategies werelikely to be linked to each outcome, and how that

iAlthough this approach is often referred to as ‘the Duluth Model’because of the influential community-based programme that emergedfrom that city in the early 1990s, we are referring to it as afeminist-informed CBT model because it is likely that very few currentprogrammes are true replications of the Pence and Paymar’s Duluthprogramme.16

4 Velonis AJ, et al. BMJ Open 2016;6:e010173. doi:10.1136/bmjopen-2015-010173

Open Access

on 29 May 2018 by guest. P

rotected by copyright.http://bm

jopen.bmj.com

/B

MJ O

pen: first published as 10.1136/bmjopen-2015-010173 on 6 A

pril 2016. Dow

nloaded from

strategy triggers the outcome and what contexts are themost relevant in allowing that to happen? We recognisedthat each strategy may be linked to multiple outcomes,

that multiple mechanisms are often at work within eachstrategy, and that sometimes change has to happen in aparticular order. For example, one ‘education strategy’

Figure 2 Batterer intervention process, outcomes and influencing factors.

Figure 3 Sample context–mechanism–outcome configurations from the preliminary rough theory of batterer intervention

treatment programmes.

Velonis AJ, et al. BMJ Open 2016;6:e010173. doi:10.1136/bmjopen-2015-010173 5

Open Access

on 29 May 2018 by guest. P

rotected by copyright.http://bm

jopen.bmj.com

/B

MJ O

pen: first published as 10.1136/bmjopen-2015-010173 on 6 A

pril 2016. Dow

nloaded from

that appears common to most BIPs is to discuss thenegative impact that violence and abuse has on one’spartner and children. For participants who are capableof feeling remorse and empathy (ie, who do not havesociopathic tendencies or an antisocial personalitydisorder) (context), learning about these damagingimpacts triggers both shame and guilt for past beha-viours (mechanism), and a desire to stop hurting peoplewhom he loves (mechanism). Both these mechanismscan lead to the perpetrator feeling motivated to stopusing abusive behaviours (outcome) and to desire alter-natives to violence (outcome). Another strategy com-monly employed is to teach perpetrators skills they canuse to avoid becoming violent (such as recognisingemotional triggers and calmly walking away). Learning,and then practising, these behaviours triggers a level ofself-confidence in participants (mechanism) that leadsto the eventual adoption of these skills (outcome).However, before a participant is likely to truly benefitfrom these skill-building sessions, he most likely needs toalready feel motivated to stop using abusive behaviourand to desire alternatives to violence. Figure 3 illustratesseveral pathways in which programme strategies may linkto our proximal outcomes; this is a partial model in-tended to exemplify our process, and is by no means acomplete outline of our preliminary, rough theory.At this stage, we wish to acknowledge that we fully

anticipate revising and refining this as we proceed withthe review. In their reflection on realist review, Jagoshand colleagues describe their struggle with identifying asingular theoretical construct that guided the subject oftheir review, and observed the ways in which context,mechanisms and outcomes often overlap, with anoutcome in one chain of evidence serving as a contextin a subsequent one.20 We also recognise that usingproximal outcomes may pose a challenge, as the natureof these outcomes—especially motivations and attitudes—can be subjective in nature and difficult to capture,and some may not be captured at all. As we moveforward with our review, we will assess how well theseconstructs are assessed, and where we believe criticalgaps may exist.As we constructed our protocol, we identified several

existing theories of behaviour change (eg, social cogni-tive theory, stages of change, etc) that appear useful forunderstanding how BIPs are supposed to work,25 41 yetour attachment to these is preliminary. Throughongoing discussions within the team and continuouslyasking what causes a particular strategy to lead to a par-ticular outcome during the synthesis phase of the review,we will continue to refine our models and flesh out themechanisms at work.

Step 2: search for evidenceSearch strategyIn partnership with a medical reference librarian (whois conducting the searches but is not part of the reviewteam), we selected search terms that have been identified

in previous literature searches (examples include varia-tions on words such as ‘batterer’, ‘perpetrator’, ‘inter-vention’ ‘evaluation’, etc). The disciplinary andinterdisciplinary databases in our scope include (but arenot limited to): MEDLINE, EBM Reviews (includingCochrane Database of Systematic Reviews), EMBASE,PsycINFO, CINAHL, Criminal Justice Abstracts, SocialSciences Abstracts, International Bibliography of the SocialSciences (IBSS), Applied Social Sciences Index andAbstracts (ASSIA), ProQuest Criminal Justice, ProQuestDissertations & Theses Full Text, ProQuest Social ServicesAbstracts and Sociological Abstracts. Additionally, databasesand electronic resources such as the Minnesota CenterAgainst Violence and Abuse (MINCAVA) ElectronicClearinghouse and Google will be used to search the ‘greyliterature’ for unpublished and informal evaluations.Based on prior experience with realist reviews as well asthe literature describing them, we expect that as weproceed with the review, we will return to this step severaltimes as we expand and/or refine our scope as necessary.All searches will be limited to English language articlespublished from 1995 to present.

Inclusion/exclusion criteriaUnlike more traditional syntheses, we are less concernedwith whether or not an evaluation meets certain meth-odological standards (eg, is a randomised trial orincludes a control/treatment group design) and morewith the type of information it can provide about how,why and for whom BIPs work; this is reflected in ourinitial inclusion and exclusion criteria. Quantitative,qualitative and mixed-methods evaluations, regardless ofstudy design, are to be included if the programmes theyassess:A. Are offered in community-based rather than institutionalised

settings, such as the military or prison. Even if the pro-gramme activities and format resemble those foundin community-based settings, BIPs conducted inthese institutions likely involve a unique set of con-textual factors and mechanisms that impact how andfor whom the programmes work (eg, a soldier’s com-manding officer often ensures compliance with pro-gramme requirements, unlike in civilian settings)3;

B. Include some form of facilitated group treatment/educationcomponent, the most common format required acrossNorth America. Most importantly, for purposes ofthis review, we are intentionally excluding researchinvolving couples counselling or individual psycho-therapy, as these are sufficiently different from BIPprogrammes, are considered controversial by manypractitioners, and are often specifically prohibited byjudicial statute7;

C. Run at least 8 weeks or 16 h in duration. Most judicialstatutes require at least 12 h,7 24 and we do not believethat programmes that are shorter could be compar-able in scope or content;

D. Involve primarily male, court-mandated partner violenceperpetrators. Both research and observation suggest

6 Velonis AJ, et al. BMJ Open 2016;6:e010173. doi:10.1136/bmjopen-2015-010173

Open Access

on 29 May 2018 by guest. P

rotected by copyright.http://bm

jopen.bmj.com

/B

MJ O

pen: first published as 10.1136/bmjopen-2015-010173 on 6 A

pril 2016. Dow

nloaded from

that the use of violence may be different for womenthan for men, necessitating different approaches.42

Likewise, men who voluntarily choose to participatein perpetrator treatment likely have differentcharacteristics, and programmes designed to caterprimarily to this population may be addressing differ-ent causes of violence;

E. Measure at least one proximal outcome, such as skills, atti-tudes, intentions. This criterion emerged as we devel-oped our preliminary programme theory (describedabove), and reviewed articles identified during anearly search. At the start of the project, we plannedto include evaluations that measured at least one ofthe final outcomes, but the iterative process oftheory identification led us to shift to a more limitedscope.

Articles that are programme descriptions or evaluativereviews will be set aside for use as background, but willnot be included in the formal synthesis process. Articlesthat focus solely on reducing programme attrition,increasing completion rates, or are limited to identifyingthe impact that participant’s stage of change has on pro-gramme success will also be excluded.

Article screeningUsing the search strategies outlined above, our librarianwill generate a list of articles and (when available)abstracts, which we will divide among members of thereview team, who will review the titles and abstracts todetermine if the paper is (a) focused on domestic vio-lence perpetrator programmes at all, and if it appears to(b) fit within the aforementioned inclusion/exclusioncriteria. Screeners will be asked to categorise eacharticle as ’include’, ‘exclude’, and ‘maybe’. When anabstract is not available, titles will be used to determinewhether the article is appropriate for the review (eg,does the title mention BIPs?); if a title is insufficient tomake this determination, the article will remain in thelist of potential evaluations to be included until the com-plete text can be reviewed.To ensure inner rater reliability, we will randomly

select a handful of titles and abstracts that all screenerswill review; as a group, we will discuss each screener’scategorisation and, as necessary, come to consensusabout articles in which screeners disagree. Once we aresatisfied that all screeners share an understanding of thecriteria and screening objectives, each team memberwill complete her/his assignments.After this initial screening phase, all articles labelled

as ‘include’ and ‘maybe’ will be redistributed among thereview team members, who will complete a secondscreen of the remaining titles and abstracts. Once allmembers completed this task, the review team will againdiscuss this process. Articles that the first screener labelsas ‘include’ but the second screener decides to ‘exclude’will be discussed, and consensus reached.Finally, the complete article or paper will be obtained

for all remaining titles. Once again, inner rater reliability

will be assessed by having all reviewers read the same setof five articles, individually make recommendations onexclusion or inclusion, and then meet as a group todiscuss the process. Each remaining article will be dis-tributed among the review team members, and skimmedin order to make a final determination of whether toinclude or exclude it based on our screening criteria. Ifscreeners are uncertain about whether or not to includea particular article at this stage, the article will be sharedamong other team members, and consensus will bereached.

Step 3: study appraisal and data extractionFor the appraisal process, each article will be read care-fully by two reviewers, each assessing the relevance ofthe document to our inquiry (ie, how much informationcan it contribute to our development of programmetheory?) and the rigour (ie, whether that informationwas generated using credible and trustworthy methods).43

Reviewers will use a tool designed to identify and recordthe following information: (1) the programme strategiesthat are described, (2) what proximal outcomes are mea-sured, (3) how the proximal outcome(s) is/are mea-sured, (4) the contextual factors that are mentioned inthe article (eg, if participants with addictions need to bein treatment or recovery prior to joining the pro-gramme), (5) whether the authors describe possiblemechanisms that could lead to the outcome(s), and ifso, what those mechanism are; and (6) the studydesign/fit for purpose making a clear note if it seems tobias the results (eg, if only successful programme partici-pants were included in the evaluation). Based on thesefindings, the reviews will give an overall impression ofthe richness of the data available from this article andhow much it can contribute to our understanding ofprogramme theory.We expect additional articles will be excluded after

this in-depth review process if it is decided that theycannot contribute to our understanding of BIPs. If thetwo individuals who review a single article come to dif-ferent conclusions, the larger team will discuss the issuesand, if necessary, others will be asked to review thearticle(s) as well. Finally, we will comb through the cita-tions of our articles as well as through our initial searchresults for additional articles or reports describing thesame programme, and will review these sets or ‘families’of articles as a single unit.As reviewers read and re-read these papers, particu-

larly pertinent passages will be directly extracted andincluded in the spreadsheet, and other data summarisdand annotated as necessary. The reviewers will meet ona regular basis to discuss their findings.

Step 4: Analysis and synthesisHaving started with a hypothesised theoretical model forBIPs, this stage will involve examining the evidence gath-ered and determining whether it supports or contradictsour proposed programmatic theory. Using our proximal

Velonis AJ, et al. BMJ Open 2016;6:e010173. doi:10.1136/bmjopen-2015-010173 7

Open Access

on 29 May 2018 by guest. P

rotected by copyright.http://bm

jopen.bmj.com

/B

MJ O

pen: first published as 10.1136/bmjopen-2015-010173 on 6 A

pril 2016. Dow

nloaded from

outcomes as our organising framework, we will lookcarefully at each evaluation that pertains to a particularoutcome (or, if multiple publications describe the sameprogramme, we will look at these as a ‘family’ of articles)and will assess how the data that were extracted from thestudies inform our understanding of how batterer in-tervention works. Specifically, we will use the data toconstruct CMOs for each programme. Owing to theemphasis that many evaluators have placed on looking atfinal, rather than proximal, outcomes, and at whetherBIPs lead to reductions in recidivism/reoffense (insteadof how they lead to them), we anticipate that the data-describing mechanisms that underlay BIPs may be thin;thus, we will apply abductive reasoning as necessary toformulate our series of CMOs.20 We will be particularlycognisant of the ways in which different contextualfactors—when addressed—appear to influence themechanisms that lead to these proximal outcomes.Relying on the interdisciplinary perspectives and

expertise of our research team, we will look at the infor-mation that arises from the construction of each CMO,as well as across programmes, to identify similarities anddifferences. Ultimately, we anticipate using the synthesisprocess to refine our original theoretical model in lightof our review findings.

Step 5: Presentation and disseminationThe findings from this process will be presented in atleast two formats: through at least one peer-reviewedarticle that conforms to the RAMESES publication stan-dards put forth by Wong et al43 that is intended toinform implementation scientists and others in aca-demic settings; and through targeted outreach and con-versations intended to reach decision-makers andpractitioners: programme directors, policymakers, andcommunity coalitions charged with overseeing coordi-nated responses to partner violence. This will includepresentations at domestic violence and batterer treat-ment coalitions and conferences, and the preparation ofplain-language reports and briefs for disseminationthrough national and international networks.

DISCUSSIONThis review was conceived after informal conversationswith batterer intervention treatment providers, domesticviolence advocates and judicial system personnel re-vealed frustration over the current understanding of bat-terer intervention programmes. Results from evaluationsand systematic reviews have been found to be conflict-ing, inconclusive and without substantial guidance abouthow approaches to batterer treatment could be im-proved.2 44 In response, our team decided to completethis realist synthesis of perpetrator treatment pro-gramme evaluations with the aim of clarifying how andwhy BIPs work for some men, and the role that certaincontextual factors may play in that success (or lack ofsuccess).

This is the first realist review of the perpetrator treat-ment evaluation literature that we know of, and webelieve it will offer key insights into the debate over howcommunities can respond to partner violence. By focus-ing on the mechanisms that lead participants to change(rather than only looking at whether or not change wasachieved), we believe we will provide much neededinsight into promising (and not-so-promising) theoretic-ally informed strategies that can lead to the types ofchanges that will allow men to reduce their use ofviolent and/or controlling behaviours.One of the key contributions of this review, in relation

to the majority of BIP programme evaluations and sys-tematic reviews that have been done, is that we will focuson the impact that programmes have on proximal out-comes, rather than on reoffending or recidivism. Webelieve this is important for several reasons. To trulyunderstand why programmes are or are not successful(and for whom and under what conditions), we need togain a clear picture of the processes that lead to theseoutcomes. In the case of batterer treatment, it is unlikelythat participants are magically transformed into non-violent partners simply because they attended a pro-gramme; rather, the programme promotes certain out-comes within participants that then lead to these moredistal behavioural changes. Identifying what these prox-imal outcomes are and how programmes can achievethem is a key part of understanding this process.This is not to say that we think the elimination of

violent and controlling behaviours on the part of perpe-trators should not be the ultimate goal of communityresponses; it most definitely should. We anticipate thatone of the conclusions that may be drawn by this reviewis that significantly more work needs to be done to showhow the achievement of these proximal outcomes ultim-ately can lead to the cessation of violence. However,as we look more closely at the literature surroundingBIPs, it has become apparent that factors unrelated tothe programmes themselves also contribute to the likeli-ness that men will cease to engage in violence againstpartners and family members. Influences at both theindividual and interpersonal levels, as well as the com-munity and social levels, are at play, and these need tobe identified and accounted for within the larger coordi-nated response.

Acknowledgements The authors extend their thanks to Pamela Dittman, staffto the Washington State Supreme Court Gender and Justice Commission,whose initial questions to the lead author sparked this work and whocontinues to provide ongoing review and input to ensure relevance to judicialsystems across the USA. Their appreciation also goes to Dr Maritt Kirst whoprovided methodological advice during the early stages of protocoldevelopment, and to St Michael’s Hospital Information Specialist CarolynZiegler, who designed and conducted the search strategy.

Contributors AJV conceptualised the study, reviewed the overview literatureon batterer intervention, and completed the initial programme process andtheoretical models. AJV and POC led the design and drafting of the reviewprotocol, assisted by RC, DF and WD. AJV and RC wrote the first draft of thismanuscript, which was critically reviewed by POC, WD and DF. All authorshave approved submission of the manuscript to BMJ Open.

8 Velonis AJ, et al. BMJ Open 2016;6:e010173. doi:10.1136/bmjopen-2015-010173

Open Access

on 29 May 2018 by guest. P

rotected by copyright.http://bm

jopen.bmj.com

/B

MJ O

pen: first published as 10.1136/bmjopen-2015-010173 on 6 A

pril 2016. Dow

nloaded from

Funding This research received no specific grant from any funding agency inthe public, commercial, or not-for-profit sectors.

Competing interests None declared.

Provenance and peer review Not commissioned; externally peer reviewed.

Open Access This is an Open Access article distributed in accordance withthe Creative Commons Attribution Non Commercial (CC BY-NC 4.0) license,which permits others to distribute, remix, adapt, build upon this work non-commercially, and license their derivative works on different terms, providedthe original work is properly cited and the use is non-commercial. See: http://creativecommons.org/licenses/by-nc/4.0/

REFERENCES1. Gondolf EW. Batterer intervention systems: issues, outcomes, and

recommendations. Thousand Oaks, CA: Sage, 2002.2. Carter LS. Batterer intervention: doing the work and measuring the

progress. A Report on the December 2009 Experts Roundtable.San Francisco: Family Violence Prevention Fund & National Instituteof Justice, 2010.

3. Babcock JC, Green CE, Robie C. Does batterers’ treatment work? Ameta-analytic review of domestic violence treatment. Clin PsycholRev 2004;23:1023–53.

4. Bowen E, Gilchrist E. Comprehensive evaluation: a holistic approachto evaluating domestic violence offender programmes. Int J OffenderTher Comp Criminol 2004;48:215–34.

5. MacLeod G, Pi R, Smith D, et al. Batterer intervention systems inCalifornia. California, Judicial Council of California/AdministrativeOffice of the Courts, 2009.

6. Pepin D, Hoss L. Menu of State batterer intervention program laws.Atlanta, GA: Center for Disease Control and Prevention Office forState, Tribal, Local and Territoral Support, Public Health LawProgram, 2015.

7. Maiuro RD, Eberle JA. State standards for domestic violenceperpetrator treatment: current status, trends, and recommendations.Violence Vict 2008;23:133–55.

8. Miller M, Drake E, Nafziger M. What works to reduce recividism bydomestic violence offenders. Olympia, WA: Washington StateInstitute for Public Policy, 2013.

9. Dutton DG, Corvo K. Transforming a flawed policy: a call to revivepsychology and science in domestic violence research and practice.Aggression Violent Behav 2006;11:457–83.

10. Gondolf EW. Theoretical and research support for the Duluth model:a reply to Dutton and Corvo. Aggression Violent Behav2007;12:644–57.

11. Dixon L, Archer J, Graham-Kevan N. Perpetrator programmes forpartner violence: are they based on ideology or evidence? LegalCriminol Psychol 2012;17:196–215.

12. Gelles RJ. The politics of research: the use, abuse, and misuse ofsocial science data?The cases of intimate partner violence. FamCourt Rev 2007;45:42–51.

13. Sayer A. Critical realism and the limits to critical social science.J Theory Soc Behav 1997;27:473–88.

14. Pawson R. Evidence-based policy. London: SAGE Publications Ltd,2006.

15. Jackson S, Feder L, Forde DR, et al. Batterer intervention programs:where do we go from here. Washington DC: U.S. Department ofJustice, Office of Justice Programs, 2003.

16. Stover CS, Meadows AL, Kaufman J. Interventions for intimatepartner violence: review and implications for evidence-basedpractice. Prof Psychol Res Pract 2009;40:223–33.

17. Wong G, Westhorp G, Pawson R, et al. Realist synthesis:RAMESES training materials: RAMESES Project. 2013.

18. Connelly JB. Evaluating complex public health interventions: theory,methods and scope of realist enquiry. J Eval Clin Pract2007;13:935–41.

19. Pawson R, Greenhalgh T, Harvey G, et al. Realist review—a newmethod of systematic review designed for complex policyinterventions. J Health Serv Res Policy 2005;10(Suppl 1):21–34.

20. Jagosh J, Pluye P, Wong G, et al. Critical reflections on realistreview: insights from customizing the methodology to the needs ofparticipatory research assessment. Res Synth Methods2014;5:131–41.

21. Sartin R, Hansen D, Huss M. Domestic violence treatment responseand recidivism: a review and implications for the study of familyviolence. Aggression Violent Behav 2006;11:425–40.

22. Saunders D. Group interventions for men who batter: a summary ofprogram descriptions and research. Violence Vict 2008;23:156–72.

23. Paymar M, Garnes G. Countering confusion about the Duluth model.Duluth, MN: Domestic Abuse Intervention Programs, 2007.

24. Price BJ, Rosenbaum A. Batterer intervention programs: a reportfrom the field. Violence Vict 2009;24:757–70.

25. Mankowski ES, Haaken J, Silvergleid CS. Collateral damage: ananalysis of the achievements and unintended consequences ofbatterer intervention programs and discourse. J Fam Violence2002;17:167–84.

26. Gondolf EW. How batterer program participants avoid reassault.Violence Against Women [HWWilson—SSA] 2000;6:1204–222.

27. Rosenbaum A, Leisring PA. Group intervention programs forbatterers. J Aggression Maltreat Trauma 2001;5:57–71.

28. Gilchrist E. The cognition of domestic abusers: explanations,evidence and treatment. In: Gannon T, Ward T, Beech A, et al, eds.Aggressive offenders’ cognition: theory, research, and practice. JohnWiley & Sons, 2007:247–66.

29. Holtzworth-Munroe A, Stuart GL. Typologies of Male batterers: threesubtypes and the differences among them. Psychol Bull1994;116:476–497.

30. Carlson RG, Jones KD. Continuum of conflict and control: aconceptualization of intimate partner violence typologies. The FamilyJournal 2010;18:248–54.

31. Gondolf EW. Who are those guys? Toward a behavioral typology ofbatterers. Violence Vict 1988;3:187–203.

32. Hamberger LK, Hastings JE. Personality correlates of men whoabuse their partners: a cross-validation study. J Fam Violence1986;1:323–41.

33. Johnson MP. A typology of domestic violence: intimate terrorism,violent resistance, and situational couple violence. Boston, MA:Northeastern University Press, 2008.

34. Buzawa ES, Buzawa CG, Stark E. Responding to domesticviolence: the integration of criminal justice and human services.Thousand Oaks, CA: Sage, 2017.

35. Stuart GL, Temple JR, Moore TM. Improving batterer interventionprograms through theory-based research. JAMA2007;298:560–2.

36. Day A, Chung D, O’Leary P, et al. Programs for men who perpetratedomestic violence: an examination of the issues underlying theeffectiveness of intervention programs. J Fam Violence2009;24:203–12.

37. Saunders DG. Feminist-cognitive-behavioral andprocess-psychodynamic treatments for men who batter: interactionof abuser traits and treatment models. Violence Vict1996;11:393–414.

38. Bennett L, Violence PCAD. Controversies and recent studies ofbatterer intervention program effectiveness. National ResourceCenter on Domestic Violence VAWnet, 2001.

39. Diefenbeck C. Group therapy for male batterers: comparison ofcognitive-behavioral & object relations approaches. J PsychosocNurs Ment Health Serv 2003;41:18–25.

40. MacLeod D, Pi R, Smith D, et al. Batterer intervention systems inCalifornia: an evaluation. San Francisco, CA: Judicial Council ofCalifornia, 2009.

41. Scott KL, Wolfe DA. Readiness to change as a predictor ofoutcome in batterer treatment. J Consult Clin Psychol2003;71:879–89.

42. Melton H, Belknap J. He hits, she hits: assessing gender differencesand similarities in officially reported intimate partner violence. CrimJustice Behav 2003;30:328–48.

43. Wong G, Greenhalgh T, Westhorp G, et al. RAMESES publicationstandards: realist syntheses. BMC Med 2013;11:21.

44. Gondolf EW. Evaluating batterer counseling programs: a difficult taskshowing some effects and implications. Aggression Violent Behav2004;9:605–31.

Velonis AJ, et al. BMJ Open 2016;6:e010173. doi:10.1136/bmjopen-2015-010173 9

Open Access

on 29 May 2018 by guest. P

rotected by copyright.http://bm

jopen.bmj.com

/B

MJ O

pen: first published as 10.1136/bmjopen-2015-010173 on 6 A

pril 2016. Dow

nloaded from