AbstractAcademic understanding of women's gendered pathways into the criminal justice system has grown significantly over the last 20 years. Allied to this development has been an increasing number of gender responsive practices and interventions designed to address the needs of criminal justice-involved women. This meta-analysis summarizes the trends in 71 interventions extracted from 64 papers involving justice involving women. Subgroup analysis and meta-regression were used, which shows that gender responsive interventions are up to 42% more effective compared to gender neutral, even when controlling for several covariates (b=-.58, p<.01). Other findings in our paper demonstrate features of interventions, such as intervention, format, focus, and length, that appear to be effective in reducing recidivism for criminal justice involved women. Our findings strengthen the case for investment in gender responsive interventions and diversion programs. IntroductionThe question of effective interventions for criminal justice involved (CJI) women rose to prominence in both academic and policy circles from the end of the 1990s. This issue was prompted by rising female prison populations that outstripped the growth in the male prison estate in many advanced industrialized countries and coalesced with the emergence of the 'what works' agenda in criminal justice policy. Whilst some writers attempted to adapt male centric models to explain and respond to female offending, others sought to develop gender specific approaches. Significantly, the work of Bloom et al., (2003) offered a framework of principles that has since influenced the delivery of a growing number of gender responsive services/interventions for CJI women. Running alongside these policy and practice developments is an evolving evidence base. As Van Voorhis (2022) suggests, we now have a greater understanding of:  i) the distinct gendered pathways that women take into the criminal justice system (Brennan et al, 2012; Brennan & Jackson, 2022); ii) the specific gendered needs that if untreated are predictive of criminal justice involvement, as well as the strengths that are protective factors against future offending behavior (Van Voorhis, 2012); iii) the interventions and case management practices that produce optimal results for CJI women (Gobeil et al, 2016). This paper focuses on the later point and seeks to evaluate 'what works' for CJI women.To date, the extant literature suggests that interventions that target either 'criminogenic needs' or 'gendered needs' of CJI women demonstrate a greater impact on recidivism when compared to treatment as normal (Prison and Probationary Supervision). In short, doing something appears to be more effective than the status quo (Gobeil et al., 2016). Perhaps then, a more relevant question that our paper seeks to address, is: 'what might work best' for CJI women. To some extent Gobeil et al's (2016) meta-analysis provides some insights into this question – offering statistically tentative support for the contention that gender responsive approaches are more effective than gender neutral interventions. Beyond these headline findings, important gaps remain in our understanding.   Specifically, a deeper understanding of why and for whom these interventions work is critical to developing our knowledge of what works best. What is a Gender Responsive Intervention?Gender responsive approaches are based on the premise that women respond to interventions in different ways to men. This necessitates a theoretical model that understands the psychological development and social context of women's lives, to ensure effective practices. Bloom et al., (2003) identify three theoretical strands that underpin this approach: i) Relational Theory, which identifies the importance of 'connection' to our understanding of self, and how women, in particular, develop a sense of self through actions that are based on and result in connections with others (Covington, 1998). According to Bloom et al., (2003), a 'healthy connection' with others is seen as both the means and goal of psychological development, and conversely, disconnection can lead to harmful states; ii) Theory of addiction overlaps considerably with the notion of connection, whereby women use substances as a means to make or sustain connections to others or use substances to mask the pain of non mutual and often abusive relationships (Covington & Surrey, 1997); iii) Theory of trauma is critical to understand the needs of women within the criminal justice system, insofar as many women have personal histories shaped by abuse and neglect which the experience of custody often compounds. As Herman (1992) suggests, trauma is a disease of disconnection, and therefore, recovery necessarily involves creating safe spaces whereby women might begin to heal through a relational context, in connection to and not in isolation from others; iv) In addition, we would add a fourth theoretical dimension, a Theory of socio-economic marginalization, which is necessary to understand the material forms of disconnection CJI women experience that significantly frame and determine their life chances (Morash & Kashy, 2022).  Specifically this would acknowledge the gendered patterning of economic resources and opportunities, as well as the disproportionate burden that socially ascribed familial roles places on women.  In summary, these theories are not the end point in themselves but act as a framework to guide the form that gender responsive interventions should take. Drawing on these precepts, we have synthesized the extant literature to identify four key components of a gender responsive practice/intervention. i. Gender responsive interventions must primarily address women's pathways into the criminal justice system. It is widely believed that women offend for different reasons to men and if this is accepted it follows that these factors should inform the design of criminal justice interventions. The literature suggests three gender specific pathways shape criminal justice involvement for women: the Victimisation pathway highlights the impact of childhood abuse or neglect on secondary mental harms (PTSD, depression or anxiety) and substance use that are drivers of criminal justice contacts in adulthood (Daly, 1992); the Relational pathway documents the inter-relationship between dysfunctional and damaging intimate relationships which renders women vulnerable to patterns of co-offending with abusive partners (Covington, 1998);  the Deep marginalization pathway highlights a pattern of poverty, homelessness, and social exclusion which means that some women offend due to a combination of economic necessity and the opportunities provided through their proximity to  criminogenic networks (Richie, 2001; 2018).    Gender responsive interventions should seek to disrupt these pathways with practices and services that simultaneously address constellations of inter-related needs that constitute these routes or build associated strengths that might promote future forms of desistance. ii. A gendered assessment of needs and strengths should inform individualized treatment/service provision. The personal histories of CJI women exhibit key differences to those of CJI males in terms of substance use, trauma, mental illness, parenting stress, employment histories and precarious housing situations. These histories and the needs that arise from them extend beyond the 'Central 8' criminogenic needs (Andrews & Bonta, 2010) and yet traditional assessment tools fail to capture the gendered needs of women.    Gendered assessment tools such as the Women's Risk Needs Assessment are critical to the accurate and comprehensive identification of service users' 'needs' so that appropriate holistic service provision may be determined (Van Voorhis et al, 2010). Alongside needs, gendered assessments should identify the many 'strengths' that CJI women exhibit and the positive relationships they draw on to navigate the challenges in their daily lives – so that these may be built on as known protective factors against the drivers of reoffending (ibid). Moreover, given that the personal histories of CJI women are marked by experiences of abuse and, controlling behavior and social marginalization, for many women levels of self-esteem and self-worth are depleted, meaning that motivational interviewing is a critical feature of assessment/casework approaches in order for women to re-build their own sense of self perception (Miller & Rollnick, 2012).iii. Delivery methods and goals of gender responsive interventions. First, given that many CJI women present with multiple needs, the design of gender responsive interventions should be sensitized to dual or multiple related treatment needs. To meaningfully address one issue may entail addressing several related needs simultaneously, as one need cannot be easily disentangled from another. Second, gender responsive interventions should be multi modal in nature using a range of methods and practices to holistically address needs (Bloom, Owen & Covington, 2003). They should be designed to address the 'whole person', not just isolated cognitive processes, and specifically in the case of women, should be minded towards the importance of relationships and the need to be positively connected to others.   Third, gender responsive interventions should be co-designed drawing on both experiential and expert knowledge. For many CJI women, they have extensive personal histories with system involvement, and are accustomed to not having a say in major decisions that impact their lives. Therefore services should be both 'designed by' and 'designed for' the women who use them to ensure their engagement and to logically enhance the potential for their success (Bloom et al., 2003). Finally, given that the life histories of many CJI women are characterized by limited forms of agency, gender responsive interventions should promote women's journeys towards an ultimate goal of critical autonomy (Bloom et al., 2003). These practices should aim to develop vocational and life skills and provide access to resources and opportunities, so that they might eventually lead healthier and more fulfilling lives of their own choosing (Van Voorhis et al, 2010). iv. An environment designed for a gender responsive intervention.   Gender responsive interventions must take place within a physically and psychologically safe space, where service users are treated with respect and do not feel judged by those supporting them. Importantly, this space should not reflect the conditions or reproduce the emotions associated with control and abuse found within the women's personal histories. Thus, these spaces should be women only and should have staff and mentors with whom the women can relate to in terms of lived experience and background. Critical to creating such an environment is the extent to which interventions are trauma informed, in so far as staff can identify and understand the consequences of trauma and potential triggers that might re-traumatize, alongside the policies/practices that ensure interventions take place in a therapeutic context. There is a critical question posed here, as to whether gender responsive settings can be delivered in custody, given that punishment and control are likely to run counter to the therapeutic alliance between practitioner and service user. In contrast, treatment and healing are more likely to occur within safe, nurturing, and consistent environments. 'What Works' for Criminal Justice Involved WomenIt is well established in the extant literature that an intervention that is designed to address either identifiable 'criminogenic needs' or broader 'gendered needs' is preferable to treatment as usual (prison and/or probationary supervision) (Gobeil et al., 2016). That said, where this consensus unravels is the extent to which interventions that might be categorized, as either 'gender neutral' or 'gender responsive' perform in relation to one another in reducing re-offending for CJI women. This speaks to a broader set of debates over the relevance of gendered pathways and distinct gendered needs/strengths to the design of interventions to address female offending. In particular, these debates have focused on the appropriateness of the Risk Needs Responsivity (gender neutral) model to the experiences and life histories of CJI women. Gender Neutral Interventions – Risk Needs Responsivity ModelTaking each aspect of the Risk Needs Responsivity model in turn. The Risk principle suggests that the form and intensity of interventions should be designed to address a sliding scale of risk. Thus, as the offending risk increases, the treatment dosage should be raised incrementally. There is an intuitive appeal to the risk principle, however, whether this is appropriate for CJI women requires further consideration. The fact that CJI women tend to present with multiple needs whilst demonstrating disproportionately lower levels of risk in terms of criminal harms, somewhat problematizes this logic. Moreover, as Messina and Esparza (2022) note that intervention dosage (length and intensity) may be less significant than is assumed by RNR, rather more critical factors are: the appropriateness of the content of the intervention; and the applicability to the needs of the target population.This leads neatly onto the most debated aspect of RNR in relation to women, which is the Need principle. According to this principle, reductions in recidivism can only be achieved by addressing dynamic risk factors that are probabilistically linked to future offenses. According to Andrews and Bonta (2010), eight central dynamic needs/risk factors drive criminal behavior, which are comprised of the 'Moderate 4' indirect factors (poor/family relationships; lack of education/employment; lack of pro-social leisure time; substance use) and the 'Big 4' (anti-social personality; anti-social cognition; anti-social behavior and anti-social associates) as direct predictors of recidivism. The 'Central 8' needs/risks are purported to be consistent across age, ethnicity, and gender, however, several critiques suggest that the 'Central 8' are inappropriately generalized to women (Blanchette & Brown, 2006). Moreover, as Salisbury et al. (2016) suggest, even where the 'Central 8' might be effective predictors for women of recidivism, validity is significantly improved through the inclusion of gender responsive needs. Factors that do not feature in the 'Central 8', particularly static factors, such as childhood and adulthood physical and sexual abuse, play a key contextual role in offending behaviors. Consequently, the RNR needs principle fails to account for the critical relationship between criminogenic needs and, as Messina and Esparza (2022) put it, 'destabilizing factors' common to the lives of many CJI women. As Brennan et al's (2012) work suggests, for the few CJI women who exhibit characteristics aligned to the 'Central 8', it is likely that gender neutral interventions are effective; however, for the vast majority of system involved women who present with different needs/risk profiles to men, these interventions are unlikely to be optimal.The Responsivity principle asserts that intervention formats should be matched to the learning styles, capacities and characteristics of the individual who is subject to treatment (Andrews & Bonta, 2010). Yet, the influence of gender is often lost here, and in particular the role of trauma and the relational context of substance use and violence in women's lives.  (Messina & Esparza, 2022). Ultimately, these factors for many CJI women will determine the effectiveness of intervention design for them. Given these issues, Belisle et al's (2022:3) summary is apt; the "what works" agenda may have brought improvements to policy in this area, however, there remains 'a concern that when working with justice-involved women, the field frequently settles for "what works" instead of "what works best". We begin to explore this claim in detail in the following section.Gender Responsive InterventionsAs more interventions came to be informed by the gender responsive principles set out in Bloom et al's work, an evidence base has developed through a growing number of evaluation studies, randomized control and match group trials. Van Voorhis (2022:p139) has recently suggested that the evolution of the gender responsive approach and practices has led to 'a large body of research [which] is now offering an impressive empirical picture of "What Works" for CJI women. A number of multi-modal, gender responsive programs combine CBT with different therapies (mindfulness, physical exercise, art therapy) or skills training (budgeting, educational qualifications) to address specific gendered needs and to build related strengths. Positive results have been documented in three areas: i) gender responsive trauma programs (such as Helping Women Recover, Beyond Trauma, Beyond Violence) have reported positive impacts on recidivism and secondary outcome measures (e.g., anxiety, depression, substance use, PTSD) (Messina & Calhoun, 2018; Messina et al., 2010); ii) evaluations of employment skills programs' Moving On' and 'Girls Moving On', report favorable results and in particular decreases in recidivism (Gehring, 2010); and, iii)  parenting programs, such as 'Parenting Inside Out' and 'Emotions' demonstrate these programs successfully support women's relationships with their children, improve emotional regulation and reduce recidivism (Eddy et al., 2013).Gender responsive approaches have also informed the development of case management practices, and specifically, probation models (Belisle and Sailsbury 2022). Advocates suggest that case management practices based on the relational principles of trust, caring and fairness that run counter to the controlling and abusive relationships that have marked the women's personal histories, are more effective in reducing reoffending than traditional forms of supervision that prioritize toughness, threats, and surveillance (Skeem et al., 2009). Moreover, gender responsive case plans that are co-produced and address multiple needs in a staged approach appear to be more effective in reducing recidivism than standard probation approaches (Morash et al., 2016; Millson, et al., (2010).Finally, it is notable up to this point that most gender responsive interventions are delivered in prison or at the point of release into the community. Whether custody or probation supervision is conducive to the delivery of trauma informed services is an important question to pose. In some respects, the development of gender responsive diversion initiatives, either at the point of arrest or the pre-sentence stage, provides a more conducive and appropriate environment for CJI women.   Implicitly, these initiatives recognize that prolonged engagement with the system and specifically imprisonment can be counterproductive in terms of recidivism. In theory, diverting CJI women away from the system into gender responsive services that address the contextual needs that drive female offending should be a more effective and less costly intervention than prison.   The evidence base broadly supports this point (Roe Sepowitz et al., 2011; Myer & Bucholz, 2018), although it should be noted that a couple of these studies are not able to produce statistically significant re-offending outcomes due to low sample size (Messina, Calhoun & Warda, 2012; Brennan et al, 2012).   The growth in this evidence base led to Gobeil et al.'s (2016) important meta-analysis. Drawing on 37 studies (38 effect sizes), the Gobeil et al., (2016) review investigates the effectiveness of interventions for CJI women and in doing so, evaluates the impact of gender responsive vs. gender neutral approaches on recidivism rates, producing important results. First, any intervention (gender neutral or gender responsive) appears to work more effectively than no treatment at all. Second, when gender responsive and gender neutral interventions are compared, only high quality studies detected a larger effect size for gender responsive interventions. Gobeil et al, (2016) were unable to control for participant characteristics and therefore, cannot comment on for whom gender responsive interventions might work. Third, in terms of the type of interventions most likely to reduce recidivism, substance abuse interventions (k=16) produced the strongest effects. Interventions using therapeutic communities similarly produced strong effects. However, as Gobeil et al., (2016) note, 7 out of the 10 interventions also targeted substance use. Therefore, it is difficult to disentangle the influence of one on the other, that means this finding should be treated with some caution.   Finally, the results of Gobeil et al., (2016) suggest that interventions conducted in either institutions or that run between institutions and the community seem more effective than those conducted in the community. As they concede, this finding runs counter to the literature that supports the effectiveness of community over institutional based interventions. As more studies have emerged in the intervening years since the Gobeil et al., (2016) publication, our meta-analysis builds on this work. It extends this analysis to address the following aims: 1) To explore the extent to which gender responsive interventions are effective for reducing recidivism compared to gender neutral interventions; 2) To investigate the specific features of the interventions (e.g. the location of the intervention) and contextual factors that affect their efficacy (e.g. length of follow-up).MethodDatabase searches Databases were chosen for their comprehensive coverage of relevant disciplines and key journals most relevant to the research questions and aligning with the search criteria from Gobeil et al., (2016). Databases included PsychINFO, covering over 2,400 journals in fields like sociology and law; Social Sciences Full Text, with 866 journals in areas such as addiction studies and social work; Sociological Abstracts, focusing on international social and behavioral sciences literature; Web of Science, which catalogs a wide range of scientific and social sciences subjects; and Criminal Justice Abstracts, specializing in criminal justice research. Additionally, grey literature was sourced from OATD.org and the BASE search engine, which includes theses and other unpublished materials. The search terms were aligned with Gobeil et al., (2016), which defined gender ("women," "woman," "female"), population ("offen*," "crim*," "priso*," "incarcerat*," "inmate," "detain*"), and intervention ("program*," "interven*," "treat*," "therap*," "rehab*"). The search syntax was adjusted according to the requirements of each database. Searches were limited to English and dates ranging from 1990-2024.  Gender responsive approaches emerge during the late 1990s and explicitly gender responsive interventions appear thereafter; gender neutral studies existed prior to this point and are generated as a result of the emerging ‘what works’ agenda in the earlier 1990s.Study Screening and CodingFollowing the process outlined in the PRISMA Flowchart (Figure 1): In step 1, the databases were searched in September 2023 by R.S., resulting in 61,173 records. These records were then exported, and subsequently, abstracts were sifted using Abstrakr (Byron et al., 2012). After removing duplicated and completely irrelevant results, this left 14,806 abstracts to be screened (step 2). In the eligibility-assessment stage (step 3), 14,333 studies were excluded because they were not eligible for the meta-analysis (e.g., not a test of an intervention, no recidivism outcome etc.). Clear inclusion/exclusion criteria were defined using the PICO guidelines (Mattos & Ruellas, 2015): Participants (CJS involved women); Intervention/Exposure (Gender responsive interventions); Control (Treatment as Usual/Other Intervention); Outcome (Levels of criminogenic needs/ Recidivism/ Reconviction). This process left k=473 studies for further assessment. R.S. undertook the more in-depth screening, and a 20% sample was also reviewed, and S.P. and J.L. assessed eligibility.  Details of classifications were discussed and agreed in the protocol, which was then used as the criteria by which to classify the papers. There was a high agreement between reviewers (Kappa = .98) regarding the inclusion/exclusion of studies. Following this process, studies identified by Gobeil et al., (2016) were added, leaving a pool of 163 studies. As a final stage (step 4), additional data was requested from 94 study authors, as study results could not be disaggregated for female participants in many instances. As a result of this process, 64 studies were included in the meta-analysis.