Findings: reducing and preventing violence against women

Evidence on the impact of prevention interventions

This section presents findings relating to the first guiding question: What does recent research and practice tell us about the effectiveness of prevention interventions, in terms of impact on rates of perpetration and/or victimisation of violence against women?

Well crafted prevention interventions can reduce violence against women in program timeframes

There is solid evidence that well-crafted prevention interventions can reduce violence against women perpetration and/or victimisation, and in relatively short intervention timeframes. The rigorous review undertaken in 2020 by the What Works program (reference 1) looked at peer-reviewed studies published over nearly two decades from 1 January 2000. It included studies undertaken as part of the What Works program itself, alongside others evaluated using randomised controlled trial or a quasi-experimental methodologies. ‘Impact’ was defined as ‘whether the intervention prevented physical or sexual intimate partner violence; or non-partner sexual violence experienced by women or perpetrated by men or child and youth peer violence.’

The What Works review found that globally, ninety-six randomised controlled trials or quasi-experimental studies assessing interventions’ impact on men’s perpetration or women’s victimisation for physical or sexual intimate partner violence, and non-partner sexual violence had been conducted as of 2020. Thirty-one percent of these were in high-income countries, and 69 per cent in low-to-middle income countries.

• Forty-one (including 14 in high-income countries) were rated ‘positive’ (statistically significant (footnote 1) at the end of the intervention,  
• Eighteen (including 3 in high-income countries) were rated ‘promising’ (footnote 2) (e.g. significant for a sub-group, or a non-significant trend), and  
• Thirty-seven (including 13 in high-income countries) showed ‘no impact’ (footnote 3) on the forms of violence described above. However, they may have produced statistically significant reductions on other forms of violence, such as emotional or economic violence against women. 

Under the definitions used in this review (see section: Used specific definitions related to type of intervention), the vast majority of the interventions encompassed in the What Works review were conceptualised as primary prevention, along with a smaller number that would be understood as secondary prevention or early intervention. Both types of intervention resulted in reduced reports of past 12-month victimisation/perpetration by participants.

Among the primary prevention interventions in high-income countries (the context closest to Victoria’s), those that were evaluated as ‘positive’ or ‘promising’ in the What Works review included: three that were schools-based, one based in a college sports setting, one couple/parenting intervention, and two that were ante- or post-natal interventions. However, the evidence on ante- and post-natal interventions as a category was conflicting, with an equal number of evaluations in high-income countries showing no impact.

In more recent studies, a meta-analysis of randomised controlled trials examining adolescent relationship and dating programs, confirmed that such interventions can result in small but statistically significant decreases in overall dating violence, specifically in relation to physical and psychological violence (the authors suggested sexual violence was more difficult to address (reference 39)). The interventions included in the meta-analysis were almost all conceptualised as primary prevention, with a focus on addressing gendered drivers. 

For programs categorised as early intervention in this study, those shown in the What Works review to have a ‘positive’ or ‘promising’ impact in high-income countries all targeted young women. While one worked with ‘young, high-risk pregnant women’, the remaining seven were either empowerment-based self-defence interventions (footnote 4), or aimed at tackling behaviours around alcohol and substance abuse among young women who were self-reported ‘episodic drinkers’ (some of whom reported previous experience of sexual assault). The What Works review found no evaluations of interventions working with men or boys at higher risk of perpetration, whether due to alcohol and substance abuse or otherwise, or with children and young people who may have experienced violence. The forthcoming ANROWS review found a more recent evaluation of a relationship and parenting program delivered to men with substance abuse issues, which did not reduce intimate partner violence (reference 7).  More promisingly, an evaluation of the impact of providing supported housing to young people leaving out-of-home care in the US found participants were less likely to experience partner violence than a control group (reference 8). 

Evaluations analysed in the above reviews rarely made clear whether measured reductions were due to preventing statistically-expected new incidents of violence from emerging, or because violence that was already occurring ceased. According to a recent study (reference 9) (discussed below), this may be due to coding practices that confound the two categories, rather than a lack of raw data per se that would otherwise enable this analysis (an important consideration for future evaluations).

Neither the evidence reviews analysed here, nor the broader literature search undertaken as part of this scoping review found any prevention of violence against women interventions in Victoria (or Australia) that had been rigorously evaluated for impact on perpetration and/or victimisation of violence against women, though the ANROWS review described two policy changes that correlated with positive impact (footnote 5). However, a number of relatively long-standing Victorian interventions, such as the Respectful Relationships Education in Schools program and the Baby Makes Three program are similar to some of the impactful examples in similar contexts elsewhere.

Limited evidence suggests impact can result from preventing new incidents of violence, or reducing recurring violence, or both

Very recent research (reference 9) looked at how evaluations in this field defined outcomes, and how different conceptualisations and coding of variables influenced interpretations about impact. The researchers re-analysed data from six existing randomised controlled trial (all in low- and middle-income contexts) interventions that had demonstrated impact to understand if the impact reported was the result of an intervention preventing new cases of violence (i.e. among those who did not report ongoing violence at baseline, and compared to the new cases we would expect to emerge in the intervention period based on population data), and/or of it reducing the frequency of violence (i.e. among women already reporting violence at baseline). They found a range of results:

• In one case, the intervention prevented the onset of physical intimate partner violence (i.e. in relationships where violence had not existed previously, and compared to statistical expectations) but did not reduce or stop ongoing physical violence (i.e. in relationships where violence already existed at baseline) (reference 9, reference 40).
• Another prevented the onset of physical intimate partner violence (where none previously existed), and reduced ongoing physical violence (where it did already exist) but did not stop this latter ongoing violence completely (reference 9, reference 41).
• One case had a contrary result to the first: it worked by reducing and/or stopping ongoing physical and sexual intimate partner violence but was ineffective at preventing the onset of new violence that had not previously existed in the relationship (reference 9, reference 42).
• A final case (SASA! – discussed later in this report) was effective both in reducing ongoing sexual and physical intimate partner violence, and at preventing the onset of both types, but slightly more effective at the former than the latter (reference 9). 

These interventions would be conceptualised as primary prevention under the definitions used here (See section: Used specific definitions related to type of intervention), and included activities aimed at shifting gender norms, building skills, engaging leaders, and other relevant factors. They did not specifically target ‘at risk groups’ or adopt early intervention approaches, though some included counselling for existing violence as part of their imperative to ensure or establish appropriate responses to victim survivors.  

There is not yet enough research to indicate which factors determine whether an intervention will have more impact on existing or ‘new’ violence. The study did find differential impacts between interventions depending on the inclusion criteria and age group of their target populations. The authors speculated that couples-based interventions, for example, ‘may be better suited to older, cohabiting couples who may be more invested in transforming their relationships as opposed to younger populations who may not yet be in long-term committed relationships, and therefore less invested in working with their partners to resolve relationship issues,’ (reference 9) but noted further research is needed. The authors recommended that, in order to identify intervention strategies that may be more or less effective for preventing new or reducing recurring violence, interventions and their evaluations ‘need to specify a clear theory of change and pathways of impact for their outcomes of interest.’ (reference 9) 

Existing evidence is too limited and skewed to enable definitive conclusions about effectiveness by type of intervention

The number of rigorous evaluations of prevention of violence against women interventions has not yet reached a level that enables definitive conclusions about the types of intervention that are likely to be the most effective in reducing violence. Attempts have been made to classify interventions by strength of available evidence, but as the What Works review notes: a ‘shortcoming of systematic reviews is that they may not capture locally developed solutions or community programming that hold promise but lack rigorous evidence.’ (reference 43)

Categorisations of effectiveness by intervention type therefore show what interventions are getting evaluated for violence against women impact, and how they compare against the indicators those evaluations measured, but do not provide an objective assessment of everything that works, or could work, among all the prevention interventions that exist. Even less is known about how different intervention types might enhance the impact of others when implemented as part of a coordinated, multi-component approach. Those that may be categorised, based on available evidence, as having limited or no effectiveness as standalone interventions, may nevertheless contribute to the impact of other interventions as part of a phased and comprehensive approach (reference 43) (see section: Combining interventions to maximise or expand impact).

Importantly, most existing evaluations are of interventions directly targeting/engaging individuals to promote changes in attitudes, beliefs and behaviours. The interventions generally seek, whether through workshops, social marketing, peer-to-peer influencing or other strategies, to directly impact those engaged to prevent or reduce future experiences of violence perpetration or victimisation. As such, it makes sense to evaluate the impact of the intervention on these individuals accordingly, and it is possible to attribute changes in individual behaviour to the intervention.

Such interventions are a crucial part of the spectrum of prevention work. However, it should be noted that not all prevention interventions seek to engage individuals this way: some, for example, build processes and capacity that enable direct engagement work, such as curriculum development or teacher training, others might seek to create an environment that enables gender equality and addresses the gendered drivers of violence, such as by promoting organisational policy or system changes (sometimes called ‘environmental interventions’). It is difficult, and arguably undesirable, to directly attribute changes in victimisation/perpetration rates to such environmental interventions through evaluative processes (due, for example, to time lag, overlapping interventions, and the complexity of interactions between people and their environments). But these types of intervention are no less important when it comes to creating the enabling environment for change in the drivers, reinforcing factors, and ultimately rates of violence against women.

Finally, variations in the quality of design and implementation, along with contextual factors such as baseline prevalence, norms, and the strength of the enabling environment, mean that an intervention such as a parenting program may be impactful at reducing levels of violence against women in one case, and not in others. The relative effectiveness of any initiative rests on how well it has been conceptualised, how accurately it addresses context-specific drivers and reinforcing factors, how well and with what ‘intensity’ it engages with its target audience, what supports are in place, how long it runs for, and the skills of the practitioners involved. These factors likely matter more than the ‘type of intervention’ per se (reference 10) (see section: Conditions and factors that influence impact). 

Return to the top of this page

Evidence on impact at the population level

This section further addresses the research question: What does recent research and practice tell us about the effectiveness of prevention interventions, in terms of impact on violence against women perpetration or victimisation?  However, it explores in more detail, the evidence for impact at a population-level – that is, beyond those individuals or groups directly engaged in an intervention.

Few evaluations seek to measure population-level impact, but there is some emerging evidence

Very few evaluations have sought to measure population-wide impact on rates of violence against women. Of the 96 robust evaluations conducted in the past two decades and identified in the What Works review, only eight appeared to evaluate impact at the population-level.

There are good reasons for this: in any context, most existing prevention interventions that are impact evaluated have a distinct target population such as college students or first-time parents. An intervention may well aspire to contribute to broader population-level change in concert with other interventions, but the intervention’s scope, and therefore its evaluation, is generally limited to measuring impact on the target population group, and usually only those directly involved in the intervention (footnote 6).

Some single-component or stand-alone prevention interventions may target a larger-scale population group, for example social marketing campaigns, edutainment, or digital technologies such as apps. However, these intervention types captured in the What Works review either did not measure impact at the population-level or did so but found they had no effect (reference 1) (footnote 7).

There is emerging evidence showing that well-conceptualised prevention interventions can have population-level impact. These fall into two broad categories: i) policy, legislative and environmental interventions, and ii) place-based, multi-component community mobilisation interventions.

Certain policy, legislative and environmental interventions correlate with reductions in violence against women

A small number of studies showed that certain policy, legislative or environmental reforms were associated with impact at the population-level. Two of these involved a suite of initiatives (in the US and Nicaragua), which are notable for demonstrating the potential contribution of a multiple intervention approach to a population-wide decrease in violence against women. The other studies involved specific welfare or other policy or environmental changes that correlated with subsequent reductions in victimisation, as demonstrated by comparing administrative data in implementation versus non-implementation municipalities (such as demographic surveys or homicide rates), or with surveyed population groups.

In the first category, the US Violence Against Women Act (VAWA) – first authorised in 1994 – funded a number of prevention, early intervention and response programs nationwide (footnote 8). Decreases in rates of, and homicides due to intimate partner violence between 1993 and 2008 have been attributed to the Act.  A study of more than 10,000 jurisdictions between 1996 and 2002 showed that ‘jurisdictions that received VAWA grants had significant reductions in the numbers of sexual and aggravated assaults compared with jurisdictions that did not receive VAWA grants.’ (reference 3)

In Nicaragua, a 20-year longitudinal survey-based study (reference 44) undertaken at the municipal-population-level, attributed reductions in women’s reported experience of physical intimate partner violence (footnote 9) to several initiatives implemented over the same period. These included a range of legal and policy reforms, improved police, justice and service responses, and a series of awareness-raising and ‘edutainment’ initiatives aimed at transforming gender norms. The research noted the significant influence of strong feminist activism in policy and legal reforms and their subsequent success.  

The forthcoming ANROWS umbrella review shows further promising results relating to the influence of large-scale policy reforms but notes that evidence on this remains limited. Promising results were shown for reforms such as firearms control, off-premises alcohol outlet density and school-based restraining orders (reference 6). In the What Works review, two separate economic analyses conducted in Peru and Columbia of the deployment of social protection programs, involving conditional cash transfers to women, showed statistically significant reductions in women’s reported experiences of physical intimate partner violence (reference 12). These were based on demographic survey data comparing municipalities where the intervention had been implemented to those where it had not.

The above policy and legislative examples demonstrate the crucial role of the state in creating the systemic conditions to enable change, and, in the Nicaraguan case, the role of women’s movements in advocating for change, and then in supporting and diffusing messaging for policy and legislative reform at the local level. These examples also suggest that legal and policy-driven interventions can themselves measurably contribute to reductions in violence against women, especially by supporting multiple or multi-component initiatives.

Three multi-component, place-based community mobilisation interventions have demonstrated population-level impact

Place-based community mobilisation interventions, the second broad category demonstrating population-level impact, is particularly pertinent to the objectives of this review. Three such interventions found by the What Works review showed population-level impact, all of which were implemented in low- and middle-income countries in Africa (two in Uganda and one in Ghana). Population size was limited by the anticipated reach of the interventions, and in all three cases involved several thousand people across one or more administrative areas of a local municipality. Impact in all cases was measured through randomised controlled trials using surveys in control and ‘treatment’ communities, including people without any direct involvement in the intervention. These interventions and their results are described in the following section, and their defining features further examined.

As these interventions were classified as community mobilisation interventions, it is worth clarifying what this prevention strategy or technique entails. As the What Works researchers note, community activism/mobilisation interventions differ from others they evaluated in that they are not so much a single intervention as ‘a set of interventions in which multiple components are deployed (reference 1).’ Change the Story defines ‘community mobilisation and strengthening’ as supporting communities to address violence and shift norms, increasing their access to resources, and addressing broader community-level factors that may be contributing to violence against women (reference 4). That is, the work may be ‘powered by individuals or groups’ but it cannot be limited to a simple peer-to-peer engagement strategy – it requires that other prevention techniques and strategies are incorporated. As such, it has much in common with truly participatory and multi-component place-based community development approaches.

Of the three interventions, the SASA! program is the best known and the earliest evaluated. SASA! began in Uganda and has since been implemented in several other countries and contexts (reference 45). The program’s initial aim was to prevent a range of forms of violence against women and concurrently reduce HIV risk factors (this focus has since been narrowed (see section: Features of impactful interventions). It used multiple strategies at different levels including: advocacy and institutional strengthening; engagement; local government, cultural leader, police officer and health care provider capacity building; and, above all, community activism and diffusion – based on the hypothesis that ‘sustained activism from and within communities drives change rather than external people or projects (reference 45).’

The original evaluation focused on intimate partner violence and HIV outcomes. Over 2007–2012, the program demonstrated a 52 per cent reduction in women’s past-year experience of physical intimate partner violence in the treatment communities, with smaller shifts in past year sexual intimate partner violence, and other ‘results consistent with positive intervention impacts on all of the primary outcomes assessed’ (e.g. help-seeking behaviours, knowledge, skills) (reference 13). The evaluators noted that ‘in contrast to most current evidence, these intervention effects were demonstrated at the community level, and not limited to those with high reported levels of intervention exposure’ (reference 13). This suggested the result was likely due to the success of the community diffusion model. The SASA! evaluation was the first randomised controlled trial applied to such a model for prevention of violence against women. However, a key limitation was the unexpected degree of variation in results among control communities, which meant the treatment communities results did not reach statistical significance (footnote 10). Therefore, subsequent evaluations of similar interventions discussed below powered their randomised controlled trials accordingly to ensure statistical significance.

The Safe Homes and Respect for Everyone (SHARE) project – also conducted in Uganda – aimed to reduce physical and sexual intimate partner violence and HIV incidence. It used a multi-strategy approach centred on the role of trained and supported community activists to engage peers in conversations to change attitudes and social norms. It also included policy advocacy, capacity building police, social welfare officers, teachers and other officials, a men and boys’ program, as well as screening and brief interventions to reduce violence and sexual risk for women seeking HIV counselling and testing. The evaluation showed a significant reduction in women’s reports of past year physical and sexual intimate partner violence three years after baseline: women’s experience of physical violence decreased from 17 to 12 per cent, and sexual violence from 13 to 10 per cent for the treatment population—approximately a 23 per cent reduction for both, compared to steady rates in the control population (reference 14). While women’s reported experience of emotional intimate partner violence reduced for the treatment group over the course of the study (from 25 to 18%), reported experiences also reduced for the control group (from 25 to 20%), therefore results could not be attributed to the SHARE project. Men’s reported perpetration across all types of intimate partner violence also reduced, but the reductions were not statistically significant. As the study was nested within a larger, previously existing community survey, the effective sample size was established by that cohort. For intimate partner violence outcomes, there were approximately 3,500 people in the intervention group and 11,0000 people in the control group, making it the most robust of the three studies presented here.

Finally, the Rural Response System (RRS) intervention – conducted in Ghana, focused on addressing intimate partner violence. It also centred the role of trained and supported community activists and included a range of other strategies, including training for the police, health, social welfare staff, institutional/policy advocacy, and engaging other community-based organisations, traditional and religious leaders. It also included a violence response component, providing counselling, support and assistance to victim survivors to seek redress, and developing a referral system between community-based response systems and state agencies, and to strengthen appropriate traditional systems of resolution of violence against women (reference 15). The evaluation showed a statistically significant reduction in women’s past-year experiences of sexual intimate partner violence (from 17.1 to 7.7%, versus 9.3 to 8.0% in the control communities). There was also a non-significant trend towards reduction in women’s past-year experience of physical intimate partner violence (from 16.5 to 8.3% versus 14.6 to 10.9% in the controls). That is, physical intimate partner violence also decreased in the control arm over the course of the study, but by less than the treatment group, and this was considered in the assessment of the significance of the treatment group reductions.

The researchers noted that, while food insecurity was not a direct correlate with intimate partner violence, it was identified as a potential risk factor at baseline and reduced over the course of the study in both the control and treatment arms (reference 15). Men reported less physical and sexual intimate partner violence perpetration over the same period, but the changes did not reach statistical significance. Unlike the results of SHARE and SASA!, reports of emotional intimate partner violence perpetration by men in the RRS intervention were significantly lower at endline, compared to baseline (and the control group). Reports of male partner controlling behaviour also significantly reduced in the intervention compared to the control arm, as did women’s report of depression (reference 15). This latter finding is particularly interesting as it demonstrates that interventions aimed at preventing violence against women can have population-level co-benefits that extend beyond the intended violence prevention outcomes. The researchers noted that ‘the improvement in women’s depression may have resulted from actual support received from [the intervention] or the perceived societal support stemming from the visibility of [violence against women] as a social issue created by the intervention. Women’s exposure to information on how to handle cases of [violence against women] and the perceived affirmation of a woman’s power to seek help or redress could also have contributed to a reduction in depression (reference 15).’

Features of impactful place-based interventions demonstrating population level impact

The three prevention interventions presented above demonstrated impact on the levels of violence against women at a population level – the only randomised controlled trial-evaluated interventions found by the What Works review to do so in the past two decades (footnote 11). These interventions were all characterised by conditions and quality principles that are essential for impact, which are discussed later in this report. They also had several other factors in common:

• the demonstrated impact was limited to intimate partner violence against women
• they were multi-component and included activities at multiple levels of the social ecological model
• they adopted a ‘people powered’ and benefits-based approach
• they were designed and phased around theories of behaviour change.

Demonstrated impact limited to intimate partner violence against women

In all three interventions, violence against women impact was limited to women’s past-year experience of intimate partner violence (footnote 12).  

• SHARE aimed to reduce physical and sexual intimate partner violence and succeeded in significantly reducing both (footnote 13).
• RRS aimed to prevent and improve responses to all forms of violence against women and girls, but the evaluation only quantitatively measured outcomes for intimate partner violence. It demonstrated a statistically significant reduction in sexual and emotional intimate partner violence and a non-significant reduction in physical violence (footnote 14).  
• SASA! (in its original evaluated iteration) aimed to prevent all forms of violence against women, but the evaluation only quantitatively measured outcomes for intimate partner violence. It demonstrated reductions in physical intimate partner violence, and to a lesser extent, sexual intimate partner violence, but not emotional intimate partner violence (reference 13).

That is, none of these interventions demonstrated impact across a range of forms of violence against women. This could be because the evaluations were not constructed or powered to measure changes beyond intimate partner violence. The SASA! evaluation did collect and review qualitative and programmatic data related to other forms, such as non-partner sexual violence, and found no indication that the intervention was effective in these areas (reference 13). The SASA! programmers concluded that prevention of this form of violence against women requires a specialized approach (reference 16).

Indeed, lessons from the SASA! program, distilled through a recent review process, indicated that ‘including content and messaging to communities highlighting diverse forms of violence can be confusing or overwhelming, and, depending on the specific context, could potentially alienate community members (reference 16).’ This, combined with findings showing its limited effectiveness on non-partner sexual violence, led to a revised version of the program (SASA! Together), which has a narrower focus on male intimate partner violence against women.

Multi-component and multi-level

All three interventions used multiple techniques or strategies in their program models, such as institutional and policy advocacy, organisational development, direct participation programs, communications and social marketing. They worked at multiple levels of the social ecological model, using a phased approach and ensuring high-quality design and implementation. This recognised that institutional and systemic support are essential enablers to the normative and behavioural shifts sought in the community activism components of the interventions (see section: Enabling conditions for effective primary prevention interventions).  

However, these interventions were implemented in contexts with limited state support or resourcing for sectoral/institutional prevention activity, and in some cases limited response mechanisms. This meant that work at the organisational or institutional level of the social ecology was largely limited to capacity-building and engaging leaders in locally-based sectors such as education, police and health, combined with advocacy towards state-level policy makers for whole-of-setting approaches and institutional support.

‘People powered’ and benefits-based

While working at several levels of the social ecology and employing a range of strategies and techniques, a distinct feature of these interventions was their community-level, place-based focus, their use of training and their approach to supporting and empowering individuals and groups. This approach is based on social norm and diffusion of innovation theories (references 46-47) positing that once a certain ‘critical mass’ of people adopts new attitudes and behaviours, others in their circles of influence begin making similar shifts (reference 16).

The focus in all three cases was on diffusion of messaging through grassroots activism and political organising strategies. Local people were engaged as de-facto practitioners: trained and supported to further engage their peers, who then further engage their peers, the messages passed through conversations about gender, power and violence. The aim of these conversations was not to convey simple messaging, but to stimulate critical thinking, develop skills and help inspire and guide the actions and behaviours that positively shift drivers of violence across the different settings where people live, work and play (reference 14-15, reference 17).

The deep engagement of individuals and informal networks (as opposed to a uniquely settings-driven approach, for example) was seen as crucial for impacting levels of violence. The evaluation of SHARE, for instance, considered the program’s impacts linked to the consistently high exposure of people to its activities and messages as a result of community activist dissemination (reference 14). 

The interventions also largely employed a benefits-based approach – that is they emphasised what would be gained from change over what was negative about the status quo. A recent review of the longest-running of the three interventions, SASA!, claimed such an approach increases and sustains engagement, whether at the individual, community or organisational/institutional level (reference 16). However, ‘selling a vision’ of what a world without violence against women would look like was not considered enough: a benefits-based approach also required providing the supports for, and showing pathways towards, change. For example, SASA! emphasised the ‘benefits of non-violence rather than the negative emphasis on the consequences of violence, (reference 18)’ and, among other strategies, trained activists within communities to have conversations in a problem-solving and supportive way. Their practice evidence, developed over several decades of work showed that change ‘requires a feeling of hope and practical alternatives, as well as specific skills and opportunities to try new things while being supported by family, friends and neighbours (reference 18).’ SASA! practitioners also found that in relationship-level work, an emphasis on ‘the quality of relationships – intimacy, communication, mutual care and respect – resonates more strongly than an emphasis on equal household roles (reference 18).’

The contention that a benefits-based approach is an enabler of successful programming is supported by research on other types of prevention intervention. For example, research reviewing 25 years of gender-based violence prevention with adolescents and young people noted that interventions ‘conceptualized as holistic community wellness initiatives (vs. narrower, deficit-based programming)’ seemed to garner more engagement, and ‘work with LGBTQ+ youth has shown a similar pattern of youths wanting identity-affirming, strengths-based programming (that also addresses, but is not limited to, [gender-based violence] prevention).’ (reference 48)

Designed and phased based on theories of behaviour change

Based on the documentation available, the interventions that demonstrated population level impact all appeared to be informed by behaviour change theories and approaches (reference 15, reference 47, reference 49). That is, they were specific about the individual behaviours they were trying to change, not simply in terms of the behaviour of violence perpetration, but also the behaviours associated with the drivers of that violence at different levels of the social ecology, with individuals in their different roles. They also drew on theories of behaviour change to plan and stage their activities.

Both SASA! and SHARE drew on the transtheoretical model of behaviour change (reference 50), which ‘proposes that change occurs in sequential stages, although there may be relapse, and that at any time people are positioned in these different stages and need to be enabled to move to the next stage to effect behaviour change (reference 10).’ The stages are: pre-contemplation (not thinking of it yet), contemplation (thinking of it), preparation (taking steps), action (attempting to practice the new behaviour), and maintenance (or relapse). These stages are considered as relevant to institutional or organisational behaviour/practice change as they are to individuals and groups.

Both interventions therefore planned activities across multiple levels of the social ecology in a way that aligned with this understanding of the behaviour change process (footnote 15). Each was implemented in four phases, set different objectives and implemented different interventions at every phase (reference 47, reference 49). Practitioners observed that not only did this provide a framework for programming, but crucially assisted and enabled ‘time for reflection, internalisation and experimentation with new behaviours or practices (reference 18)’ (which evidence reviews have identified as a variable that maximises impact (see section: Factors that make a difference to scale and sustainability of violence against women impact).

Return to the top of this page

Conditions and factors that influence impact

The research presented above confirms that violence against women impact can indeed be achieved with well conceptualised prevention interventions. Further, several interventions have demonstrated that their impact can extend to the population level, including several place-based interventions.  

This section examines the findings relating to the second guiding question: What do we know about the foundational conditions that determine effective prevention practice, and which variables or criteria affect the extent of its impact?

Enabling conditions for effective primary prevention interventions

Recent reviews (reference 1, reference 3, reference 10, reference 19) of the research and evidence-base all provided a summary of the conditions common to effective and impactful prevention work in some from, whether at the level of individual interventions, or across a multi-component program. While these conditions were expressed differently in different reports and reviews, they can be summarised under three broad categories:

1. A prevention infrastructure: the essential core components, facilities, services and mechanisms for enabling and sustaining effective prevention of violence against women;
2. Adherence to quality design principles: address the identified drivers, use an explicit gender analysis, follow a program logic, tailor to context, setting and population (footnote 16); and
3. Effective partnerships with, and connections to, response services including a high level of support to victim survivors.

The factors that constitute good practice in each of these three categories are well-documented elsewhere, across international, national and Victorian literature (footnote 17). It is outside of scope of this report to reproduce that evidence here, however the major features of thee enabling conditions are worth noting.

A prevention infrastructure

A prevention infrastructure refers to the core components, facilities, services and mechanisms for enabling and sustaining effective prevention of violence against women activity. The extent to which such an infrastructure is present and well-functioning varies from context to context, but reviews of interventions across contexts have noted the importance of, for example, coordination mechanisms, a trained workforce, high-quality materials, a supportive legal and policy environment and adequate resources (reference 51). While these features are rarely studied in intervention-level evaluations, and there are conceptual difficulties in evaluating their direct contribution to impact, it is clear that they provide an environment that enables quality prevention practice. Interventions shown to have had an impact have generally had elements of this infrastructure in place to varying degrees, and indeed in contexts where certain elements were lacking (such as adequate legislative or policy support), advocating for the development of that element was usually a work area of the intervention itself (see examples in section: Three multi-component, place-based community mobilisation interventions have demonstrated population-level impact).

Adherence to quality principles

Evidence reviews and meta-evaluations conducted globally in the past decade have examined distinctions in the design, implementation and evaluation of impactful versus non-impactful prevention interventions. These have distilled several principles that appear to make a difference, that is, are common to impactful interventions, and frequently missing or incomplete in non-impactful ones. Ellsberg et al’s review in The Lancet, for instance, notes that impactful programs and interventions engage multiple stakeholders with multiple approaches (often across multiple sectors) and ‘not only challenge the acceptability of violence, but also address the underlying risk factors for violence including norms for gender dynamics, the acceptability of violence, and women’s economic dependence on men. They also support the development of new skills, including those required for communication and conflict resolution (reference 3).’ Similarly, Jewkes and colleagues (reference 10) identified ten key elements for effective design and implementation, which are reproduced in Table 2 below, and further discussed in the next section. 

Return to the top of this page

Return to the top of this page

Table 2: Ten design and implementation elements of effective interventions to prevent violence against women and girls

Source: Jewkes et al (2020). p.33

Return to the top of this page

Return to the top of this page

A high level of support to victim- survivors

It is well-established that prevention work can increase disclosures of existing violence (reference 2, reference 4, reference 35, reference 36). It is basic ethical practice to liaise with response services prior to beginning a comprehensive program of prevention work, and to ensure that pathways are in place for victim survivors to access appropriate support.  

What is less well-established is the relationship between this practice and violence prevention outcomes. However, recent research confirmed that if linkages and pathways between prevention and response services has not been established, or if response systems are not in place and readily accessible, it contradicts prevention messages about the acceptability of violence, and undermines the effectiveness of primary prevention efforts. For example, the What Works evidence review showed that interventions that failed to provide or ensure direct assistance for victim-survivors of violence had little to no impact on experience/perpetration of violence compared to those that ensured such support (reference 10). Providing this support was understood to reduce stigma and shift norms and was a key part of the theories of change for those programs that did show impact on violence against women.

Factors that make a difference to scale and sustainability of violence against women impact

While it has been observed that impactful interventions meet most if not all the quality principles outlined above, it is difficult to distinguish which elements are the most important to their success. However, recent evaluations and reviews looked at the factors that seem to influence the level of impact achieved and the sustainability of that impact (i.e. above the ‘baseline’ of the conditions and quality principles above).

Some initial (cautious) conclusions were drawn in the What Works distillation of effectiveness in prevention programming (see Table 2 above), the clearest of which was that across a number of different types of interventions, the ‘intensity’ of engagement with participants was a determining factor for level of impact (reference 10). The What Works review concluded that there appeared to be an intensity threshold in order for even well-conceptualised interventions to achieve impact, noting that ‘none of the interventions [reviewed] that had sub-optimal intensity were effective (reference 20).’ They also noted that, above this ‘optimal’ threshold, and among the impact ranges of interventions reviewed, intensity seemed to have a roughly linear relationship with the level of impact, whereby ‘more is more (reference 20).’

This relationship between intensity and impact has also been observed in other evidence reviews in specific settings or with specific population groups (reference 38), and the following factors identified as indicators of intensity:

• an adequate number of skilled and supported practitioners
• appropriate program length  
• the number, duration and frequency of sessions allowing time for reflection and experiential learning.

More detail on each of these categories is provided below.

The review also identified several factors that increased the effectiveness of different types of intervention. For example, direct participation interventions were more impactful when they included participatory, group-based methods with empowerment as their core goal (reference 10). It is beyond the scope of this review to distil or detail the elements of effective/impactful practice in each intervention area, but it will be important that any interventions implemented as part of the saturation project model are designed with such research in mind, and meet any standards or criteria that emerged from the literature on practice in that area. What Works’ Effective design and implementation elements in interventions to prevent violence against women and girls (reference 10) outlines success factors by type of intervention (e.g. couples’ interventions, community activism, gender transformative and economic empowerment approaches, etc) and provides some guidance for quantitative minimums or benchmarks with regard to intensity (e.g. duration, dosage) in these different intervention types.

An adequate number of skilled and supported practitioners

A capability development system to recruit and retain a skilled, valued and appropriately remunerated workforce is a key element of prevention infrastructure. The more successful interventions in the What Works review had a large workforce on the ground (reference 20), employed practitioners selected for their skills and experience (or with long training times), and ensured ongoing support was available to such personnel (reference 10). That is, greater effectiveness and/or impact was associated not only with an adequate number of practitioners, but with their skill and the level of support provided to them to do their job well.

The most impactful interventions trained staff for longer (three or more weeks), took staff and volunteers through the entire intervention as participants and built in time for practice prior to implementation (footnote 24). For example, for schools-based work: ‘The key to the effectiveness appears to be sufficient time for selecting and training personnel and providing an accompanying manual to help them (reference 54).’

In community mobilisation interventions where people from the local population, as opposed to professional practitioners. were employed to deliver aspects of the program (e.g. as peer-to-peer influencers), successful interventions ensured they were chosen for their gender equitable and non-violent attitudes and behaviours prior to training, as it ‘is not possible within typical training periods to change attitudes on gender from very conservative to sufficiently equitable (reference 10).’ Interventions that did not make this a recruitment requirement were ultimately evaluated as ineffective (reference 20). Ensuring new community activists/practitioners were in turn well-supported by an adequate number of experienced practitioners with proven skills was also found necessary ‘to appropriately facilitate gender transformative programming (reference 10).’

Appropriate program length

Observing the length of impactful programs analysed in evidence reviews, a recent Prevention Collaborative brief, on ‘investing wisely’ in prevention, suggests that successful interventions require at least three years for the implementation phase (i.e. excluding design), and need to be implemented at high intensity to have a measurable impact on violence against women prevalence. This is ‘in addition to the nine to 12 months of preparatory work needed to map local resources and stakeholders, adapt a program to a new setting, train staff, and most importantly, build trust and partnerships among implementing organisations and the communities they serve (reference 55).’

When formal evaluations or processes that capture practice-based learning are undertaken (as is recommended), the authors concluded that implementing an impactful multi-component intervention ‘requires four to five years of dedicated and flexible funding.’ Importantly, they concluded that ‘attempting to do more than money and timing allow is not value neutral […] and can provoke backlash from male partners and community members (reference 43).’ Adaptation and scaling to new contexts while maintaining fidelity to intervention design obviously requires even longer timeframes (reference 43).

Number, duration and frequency of sessions, and allowing time for reflection

The What Works evidence review similarly noted that for interventions directly engaging individuals and groups, multiple sessions of sufficient duration, and activities spanning a significant duration were needed for impact. The most impactful of the workshop- or group-based interventions were ‘sufficiently intense’ – 40 to 50 hours long in total. Holding ‘weekly meetings for two to three hours at a time, once or twice a week, enabling in-depth discussions, recall of the previous session and a period for reflection and experiential learning (reference 10)’ also contributed to greater impact.

In other research reviewing schools-based interventions, impact seemed to rely on ‘delivery over several sessions, with the most effective programs […] being delivered over several years with 20-150 sessions (reference 56).’  A 2015 review of schools-based interventions found that those demonstrating violence against women impact and/or driver outcomes all ‘used a comprehensive methodology over a sustained period of time (reference 57).’ 

Return to the top of this page

Combining interventions to maximise or expand impact

The findings presented so far have shown that prevention interventions that have demonstrated violence against women impact are those that meet the quality design and implementation principles identified in international and national prevention frameworks. Ensuring an appropriate infrastructure (such as prevention workforce and coordination mechanisms) alongside links to well-resourced response services, are also key conditions for impact. The review also found that several other criteria appear to influence the extent of impact, especially with regard to the ‘intensity’ of activity. Finally, a common feature of the interventions that demonstrated population-level impact was the use of multiple-component interventions.  

This section presents findings from prevention of violence against women and other fields relating to the third guiding question: What do we know about how and whether outcomes from individual interventions are strengthened because of the ways they interact when coordinated with other interventions, and what design, implementation or contextual conditions contribute to any ‘mutually-reinforcing effect’?

Rationale for combining interventions: the mutually-reinforcing effect

The socio-ecological models that underpin health promotion theory and practice suggest combining interventions in a coordinated way across settings and levels of the social ecology to achieve a ‘mutually-reinforcing effect (reference 21).’ The literature review undertaken to inform the first edition of Change the Story found this practice had been documented extensively in the US as part of efforts to reduce tobacco-use and other health priorities (reference 23). It was supported by the theory of multi-directional causation, suggesting, in the example of prevention of violence against women, ‘a bi- or multi-directional relationship between gender norms and gender equity at macro-level (that is, norms shape social structure, and social structure shapes norms (reference 58))’. This implies the need for multi-directional prevention activities to influence how structures, norms and practices are formed and challenged across levels of the social ecology.  

The pathways through which multiple, coordinated interventions might support sustainable change in the field of prevention of violence against women were summarised in the final framework as follows:

Single techniques employed in single settings may well have positive effects, but these will likely be limited to those participating, and – if a ‘one off’ project – may not be sustained. […] The effects of prevention initiatives are strengthened […] when their messages are reinforced by simultaneous complementary initiatives, such as when [a] schools program is accompanied by a social media campaign, a local community initiative, and a sports or recreation-based program.[…] In order to achieve this mutually-reinforcing effect, different techniques need to be employed simultaneously across multiple settings, in a coherent and sustained way (reference 4). 

Many studies across other areas of prevention or health promotion refer to the need for ‘mutually-reinforcing’ components. However, while some refer to the advantage of using multiple techniques, others speak of interventions taking place in multiple settings/sites, and still others to interventions aimed at multiple levels of the social ecology. ‘Mutual reinforcement’ might therefore occur through a range of component combinations, with distinct rationales for each:

• Using multiple strategies or techniques (i.e. within a single site or setting), which allows, for example, individual skills-building to be supported by organisational policy development (reference 21);
• In multiple settings/sites (whether in a specific geographic location or among the places most accessed by a particular community), so that complementary messaging is delivered across a range of environments and spaces (footnote 25);
• At multiple levels of the social ecology (individual, relationship, community, organisational, system, institutional, societal), to target the various structures, norms and practices that operate across levels (reference 23); or
• A combination of the above. 

In practice, it is likely that any multi-component prevention intervention will include at least two of the above three categories. Different techniques tend to be more applicable to certain levels of the social ecology (organisational development, for instance, obviously works best in organisations with formal structures, while direct participation necessarily involves working at the individual/group level), so using multiple techniques will simultaneously mean working at multiple levels of the social ecology. Similarly, choosing to work across multiple settings will likely require adapting techniques or choosing different ones that are best suited to the setting. Given ‘whole-of-setting’ approaches are best practice for settings-based work, this also means working at different levels of the social ecology. There is therefore an inter-relatedness to these different types of components and a circularity to their influence on each other, which perhaps itself provides pathways for mutual reinforcement. 

While it is unclear which of the above ‘components’, and in which combination, offer the most potential for maximum impact, this review has surfaced some recent and longer-standing analyses that can guide such an approach (see section: Initial principles for combining interventions to maximise impact). 

Mutual reinforcement in prevention of violence against women

This review has revealed that a mutually-reinforcing effect has been demonstrated in many different areas of health promotion (reference 23). In the field of prevention of violence against women, several evidence reviews noted that multi-component interventions tend to have more impact that single-component ones (reference 2-3, reference 10, reference 19).

Only one study identified through this review attempted to quantify the extent to which having multiple components increased or magnified impact. This was undertaken by the Dalberg consultancy group for the EU/UN Spotlight Initiative (reference 24), and involved a meta-analysis of violence against women interventions impact evaluations. While the analysis drew on the evaluations of prevention interventions cited elsewhere in this report, it also included evaluations of response-end interventions (footnote 27). 

Of the thirty-four interventions that had violence against women impact, fourteen were classified as ‘multi-component’, meaning they included at least one prevention component (footnote 28), along with either (an)other prevention component(s), or (a) component(s) addressing another pillar (usually around supporting women’s movements, legal reform, or service and justice responses). The remaining 20 were classified as ‘single-component’ – in all cases, a single prevention component. Of the 14 multi-component interventions, the average time-adjusted violence against women impact was 24 per cent, and of the twenty single-component prevention interventions, the average time-adjusted violence against women impact was 13.75 per cent (footnote 29). As a result, researchers concluded that combination or multi-component interventions were on average 60 per cent more effective than single-component ones (footnote 30).  

There are many caveats to consider here, with regard to comparing evaluations that are not ‘like for like’ and span a range of contexts and baseline prevalence rates. In addition, because the analysis included interventions across the spectrum from primary prevention to response, we cannot consider it measures the mutually-reinforcing effect of a multiple-component prevention program (footnote 31). But the conclusion does suggest that a multi-component approach to violence against women will engender a mutually-reinforcing effect that increases impact compared to stand-alone interventions.

Challenges in understanding and evaluating the mutually-reinforcing effect

While theoretical understandings of prevention and health promotion support the idea that combining activities will lead to a mutual reinforcement of impact, and while a magnification of impact has been observed for complex interventions compared to stand-alone ones, exactly how the ‘mutually-reinforcing effect’ operates is poorly understood.  

Limited evaluative data exists that examines how interventions have been combined to achieve ‘more impact.’ A framework developed in the US for assessing the value of multi-component interventions across health promotion spheres noted that evaluations ‘tend to focus on individual rather than comprehensive interventions, to attribute changes in health behaviours and health outcomes to specific interventions instead of multiple or synergistic efforts (reference 59).’ This, as noted in section: Evidence on the impact of prevention interventions and section: Evidence on impact at the population level, is still the case for the vast majority of evaluated prevention of violence against women interventions.  

There are also challenges to collecting meaningful data about complex multi-component programs, precisely because of their complexity:

Evaluating non-standardized, constantly changing, community-directed, slow-moving changes at all the levels in ecological models from programs to policies presents methodological, logistical, and economic feasibility challenges. […] Deconstructing complex interventions may not even be advisable, given […] the reciprocal dependency of many of the interventions (reference 59).  

The authors do not suggest abandoning the quest to better understand mutual reinforcement, given the important impacts attributed to complex interventions, and the fact that a better evidence base is required to design such work. Rather, they promote evaluations that include ‘qualitative methods to support the generation of systems science maps or diagrams that capture the underlying theories of change and causal structures in the system (reference 59).’ 

Initial principles for combining interventions to maximise impact

Despite the above limitations to the evidence base, one important finding from the review of evidence undertaken here is that a direct relationship between ‘more interventions’ and ‘more impact’ should not be assumed (reference 21), even if the interventions meet the quality standards and other conditions for impact covered in previous sections. However, a number of guiding principles or features were identified in the literature that can support mutual reinforcement. The combination of interventions should be:

• based on the principle of addressing multiple drivers and reinforcing factors (in prevention of violence against women interventions)
• sequenced in such a way as to support the phases of behaviour change
• synchronised across settings and levels.

Address multiple drivers and reinforcing factors

A long-standing practice of socio-ecological approaches to health promotion is to choose interventions that ‘focus on “high-leverage” factors – that is, those personal and environmental factors that research indicates have a disproportionate influence on the specific health issue in question (reference 21).’ However, although ‘useful for narrowing options [this does not] indicate which interventions are likely to work together in mutually-reinforcing ways, and which are not (reference 21).’

The What Works review of effective prevention programs and interventions over two decades is informative here. While researchers confirmed the importance of addressing high-leverage factors, and that there was increased impact through multi-component over single-component prevention interventions (footnote 32), they also found that ‘what is critically important is addressing multiple drivers of violence (reference 10).’  

The idea that preventing violence against women requires addressing multiple drivers and contributing factors is not new (footnote 33). What is new in this conclusion, is the implication for mutual reinforcement. The authors go on to suggest that ‘multiple components’ are not, in and of themselves, the key to mutual reinforcement. The reason multi-component interventions have greater impact is more likely due to the fact that ‘more components’ means greater reach and range of influence to address multiple drivers (reference 10). 

Sequence implementation

This review identified some guidance on the sequencing of individual interventions or components as part of multi-component health promotion programs to maximise impact. For example, there is some support for implementing ‘environmental’ interventions before ‘educational’ interventions to avoid promoting unrealistic behaviour change (reference 21). Examples given include the promotion of walking in an area with insufficient pedestrian infrastructure or high levels of street crime. This finding would imply, for place-based prevention of violence against women, that it may be more effective to work on creating enabling structures, norms and practices across the physical or organisational/institutional environment of the place, before undertaking work aimed at the individual or group level. To some degree the GEAR program in Victoria took this approach, focusing on building leadership, organisational capacity, community readiness and infrastructure for prevention, prior to direct participation interventions with community members (reference 60).

However other recommendations in the health promotion literature include being guided by an in-depth needs assessment of the target population and testing any proposed sequencing with community members in the design phase. An example given is ‘creating walking trails in a community populated predominantly with older adults [that] might not be effective until the joint pain that many residents experience is overcome (reference 21).’ Learnings from GEAR are helpful here too: in reflecting on the program, implementation stakeholders cautioned against ‘over-engineering’ and suggested continuous assessment of community (and partner) needs, and the flexibility to adapt sequencing and type of interventions, as the program progressed (footnote 34).

In terms of multi-component interventions that have successfully reduced violence against women, section: Features of impactful interventions looked at how two interventions had sequenced interventions based on behaviour change models. Intervention components were planned in four phases, setting different objectives and implementing different techniques in each one (reference 47, reference 49). Flexibility was built into the process with the specifics of activities evolving ‘in direct response to community priorities, needs and characteristics (reference 49).’ However, ‘environmental’ activities were not sequenced before educational ones. In SASA!, for example, activities were undertaken at each level of the social ecology in each phase (using techniques appropriate to that level). Organisational readiness was built concurrently with individual and group engagement, consistent with the grassroots model (and different to GEAR), but ‘each phase’s content sets up the next in critical ways because it introduces activities and ideas gradually, which builds support and reduces backlash (reference 18).’

In summary, while some guidance was found in the literature on how to sequence multi-component work, there is no universal method. The important finding is the need to sequence activity in a way that is intentional and logical regarding the changes being sought (and their dependencies), while being responsive to community advice and needs.

Synchronise components across settings and levels

A final feature of impactful multi-component programs that combined interventions or activities was the attention to synergies across settings and levels. For example, one evidence review (reference 25) compared multi-level, multi-component interventions around healthy eating and highlighted the challenge of creating consistency between activities across levels. Another review of multi-level/component interventions in childhood obesity noted that it was essential to ‘create linkages between intervention components based on complementarity, mutual promotion and mutual reinforcement (reference 61).’  

This means ensuring ‘the intervention components and activities that are taking place at the different levels are synchronized in terms of optimal timing’ and with ‘a sense of coherence in the themed intervention activities (reference 25).’ It also requires a certain intensity of activity, not just within a single intervention but between them, ‘creating repetition of program activities and messages’ through ‘a multiple-exposure approach (reference 25).’

Synchronising interventions was also a challenge for GEAR. An underlying assumption in GEAR was that that the programs would reinforce each other, but the ways in which the different components supported each other in practice and in real time were not fully defined. As one key informant put it: ‘Jim works at Bosch, he takes his nephew to story time at the local library [where healthy relationships and gender equitable norms are promoted], he and his wife attend Baby Makes Three for their first child, etc. But there wasn't overt recognition of how the program's activities interacted outside of this (footnote 35).’

SASA! and other multi-component programs to prevent violence against women have used actor-mapping tools to identify synergies across levels and settings, and to plan interventions accordingly. This identified key individuals, groups and organisations important to involve, and subsequently strategies to ‘reach within and stretch throughout these layers (footnote 36).’ 

Return to the top of this page