Article Text
Abstract
Objectives To test associations between men’s past year alcohol use and patterns of drinking, and their perpetration of intimate partner violence (IPV) and non-partner sexual violence (NPSV). To test the associations between women’s reports of partner alcohol use and their experience of IPV, in three countries in Africa.
Design Pooled analysis of cross-sectional baseline data from men and women participating in four IPV prevention studies across Africa and Asia.
Setting Data from five data sets generated by four violence against women and girls prevention studies in three countries in sub-Saharan Africa, South Africa, Ghana and Rwanda.
Participants 8104 men 18+ years old and 5613 women 18+ years old from a mix of volunteer and randomly selected samples.
Main outcome measures Studies employed comparable measures of past year alcohol use, harmful alcohol use (Alcohol Use Disorder Identification Test scale) and items from modified WHO Women’s Health and Domestic Violence to measure physical IPV and NPSV perpetration among men and IPV experience among women.
Findings Overall harmful alcohol use among men was associated with a substantially increased odds of perpetrating physical IPV (adjusted OR (aOR)=3.45 (95% CI 2.56 to 4.64)) and NPSV (aOR=2.64 (95% CI 1.85 to 3.76)) compared with non-drinkers. Women who had seen their partner occasionally drunk (aOR=2.68 (95% CI 2.13 to 3.36)) or frequently drunk (aOR=5.94 (95% CI 4.19 to 8.41)) in the past 12 months had an increased odds of experiencing physical IPV.
Conclusions Alcohol use is associated with increased IPV and NPSV perpetration for men and (physical) IPV experience for women. Reported frequency of IPV and NPSV increase with increasing levels and frequency of alcohol use. Interventions aimed at reducing alcohol may also lead to reductions in IPV and NPSV perpetration and experience.
- public health
- statistics & research methods
- substance misuse
- sexual and gender disorders
Data availability statement
Data are available in a public, open access repository. De-identified individual participant data for Stepping Stones and Creating Futures (South Africa), Sonke CHANGE trial (South Africa) and Evaluation of the COMBAT intervention (Ghana), are available to anyone who wishes to access the data for any purpose at https://medat.samrc.ac.za/index.php/catalog/WW . De-identified individual participant data from the Indashyikirwa couples cohort and community surveys (Rwanda) are available from the Principal Investigator of the study, Dr Kristin Dunkle: kristin.dunkle@mrc.ac.za, but may require permission from the Rwandan Ministry of Gender and Family Promotion (MIGEPROF) before transfer.
This is an open access article distributed in accordance with the Creative Commons Attribution Non Commercial (CC BY-NC 4.0) license, which permits others to distribute, remix, adapt, build upon this work non-commercially, and license their derivative works on different terms, provided the original work is properly cited, appropriate credit is given, any changes made indicated, and the use is non-commercial. See: http://creativecommons.org/licenses/by-nc/4.0/.
Statistics from Altmetric.com
Strengths and limitations of this study
Provides a synthesis of comparable regional data (Africa) in the global South, thereby addressing the limited geographical scope of studies examining alcohol’s association with violence against women and girls (VAWG).
Extends the evidence base on the nexus between alcohol and violence in sub-Saharan Africa, where both alcohol use and VAWG are prevalent.
Extends previous research on the association between alcohol and gender-based violence by examining past year alcohol use and violence as opposed to lifetime.
Highlights the need to address alcohol use in prevention of intimate partner violence, and the need to address the experiences of violence in programmes addressing harmful alcohol use.
The study does not adjust for women’s own drinking, which has been found to increase their subsequent experiences of violence, as it was not included in all data sets.
Only one of the four studies is population-based, limiting generalisability.
Introduction
Harmful alcohol use is a recognised driver of intimate partner violence (IPV) and non-partner sexual violence (NPSV), but its importance is still contested, and it is infrequently directly addressed in violence against women and girls (VAWG) prevention programmes. Although the associations between alcohol use and VAWG have been recognised and described for some time, most of the research evidence is from North America, and there has been considerable reluctance in some quarters to frame the link as causal due to a perception that this would lead to alcohol being blamed rather than the perpetrator.1–3 There is also debate about whether the alcohol-VAWG associations described are driven by alcohol abuse/dependence or whether they hold at much lower levels of alcohol drinking. In North America, the links with alcohol abuse/dependence have been shown in a 30-year longitudinal study where having alcohol abuse/dependence symptoms was associated with a twofold increased risk of perpetration of IPV by men, and a similar increase in risk of IPV victimisation among women.4
Research in low-and-middle income countries has not always found consistent associations between alcohol use and VAWG. For example, in the WHO Multi-Country Study on Women’s Health and Domestic Violence, the association between women’s past year IPV experience and their male partner’s alcohol use were inconsistent.5 Similarly, in nine urban and non-urban settings included in a six-country study on men’s violence in Asia and the Pacific, harmful alcohol abuse was associated with lifetime perpetration of IPV in only three of the settings.6 However, in four settings in the pooled sample, women’s reports of their partner’s alcohol use were associated with her having a higher risk of past year IPV experience.7 Community-based research in South Africa with men on perpetration found that past year drug use, but not harmful alcohol used, was associated with lifetime rape perpetration.8
Recent research in sub-Saharan Africa (SSA), drawing on the Demographic and Health Survey (DHS) has shown positive associations between women’s lifetime experiences of IPV and their current partner’s use of alcohol, but the analysis excluded many countries where heavy episodic drinking is relatively common.9 Patterns of alcohol use are critical to understand because prior research suggests that they may differ both in terms of IPV risk,10 as well as cultural and contextual norms around alcohol use and gender.11 For example, heavy drinking is more strongly associated than moderate drinking with partner aggression and more severe IPV resulting in injuries and even death.12 13 Furthermore, as with much of the literature on the association between alcohol and IPV, analysis drawing on the DHS is limited as the DHS does not have measures of all the key known risk factors for IPV that need to be adjusted for.
To strengthen the evidence base from SSA and address some of the limitations of studies in the current literature, we undertook a pooled analysis of baseline data using consistent measurement methods from four IPV prevention studies conducted in three countries in Africa (South Africa, Ghana and Rwanda). We sought to answer the following questions relating to alcohol and IPV and NPSV perpetration and experience in the past 12 months1: What proportion of men who currently use alcohol, and who used alcohol at harmful levels, perpetrated IPV against female partners?2 What proportion of men with current, and harmful alcohol use perpetrated NPSV?3 What proportion of women who reported seeing their partner drunk or frequently drunk experienced IPV?4 Is there an increasing odds with increasing levels of alcohol use or frequency of drunkenness and IPV perpetration and experience?5 Is there an increasing odds with increasing levels of alcohol use and NPSV perpetration?
Methods
The studies were conducted under the UK-Aid funded What Works to Prevent Violence Against Women and Girls? Global Programme (What Works). The primary goal of What Works was to advance knowledge on the prevalence and drivers of VAWG; test efficacy of interventions to prevent VAWG, estimate the cost-effectiveness of violence prevention interventions; and the costs of violence to households and businesses in 13 countries of the Global South. The current study used the baseline data of 8104 men and 5613 women from four VAWG prevention studies in three countries in SSA to assess the association between alcohol use and IPV and NPSV perpetration and experience. These studies include the Stepping Stones and Creating Futures intervention, Sonke CHANGE trial, Community Based Action Teams (COMBAT) community intervention, Indashyikirwa couples intervention and Indashyikirwa community diffusion. Further information on the studies is available in the study sources referenced in table 1.
Patient and public involvement
Patients and the public were not involved in its design, implementation or analysis presented here. However, the individual projects had different levels of participant and public involvement as part of the wider What Works research uptake strategy.
Measures
All study participants were asked socio-demographic questions including age, current marital status, relationship residence status and education. All studies except Indashyikirwa in Rwanda asked whether male participants had worked in the past 3 months or not; and two studies asked whether female participants had worked in the past 3 months or not. Income assessments in Indashyikirwa were not included here, as they were tailored to the local context in which most participants engaged in subsistence agriculture.
The alcohol use and violence measures are presented in table 2. In all the studies, except in the Indashyikirwa couples and community samples, harmful alcohol use among men was measured using the 10-item Alcohol Use Disorder Identification Test (AUDIT) scale.14 A shortened version of the AUDIT scale (3-item AUDIT-C) was used in the Indashyikirwa couples and community samples. An alcohol use score was derived for each study and relevant cut-offs were used to assess the prevalence of problem drinking in each sample (≥8 for the 10-item scale and ≥4 for the shortened version). We created a three-level variable, where men were coded either: 0=non-drinkers, 1=moderate drinkers and 2=harmful drinkers. We asked women in the studies about the frequency of seeing their partner drunk in the past year. Responses were ‘never’, ‘less than once a month’, ‘once a month’, ‘weekly’ and ‘everyday or nearly every day’. We created a three-level variable coded as follows: 0=never seen drunk (if the response was ‘never’), 1=occasionally drunk (if response was either ‘less than a month’ or ‘once a month’) and 2=frequently drunk (if response was either ‘weekly’ or ‘everyday/nearly everyday’). Violence perpetration (IPV and NPSV) and experience (IPV) were measured using items from the WHO Women’s Health and Domestic Violence survey,15 modified where needed to assess men’s perpetration.6 The NPSV scale was first developed in South Africa16 and subsequently refined and used extensively in Asia-Pacific region.7 There was no question on NPSV perpetration in the Indashyikirwa couples and community questionnaires because it was not a target of the intervention.
Data analysis
Descriptive statistics (frequencies and percentages) were used to summarise participants’ socio-demographic characteristics within each study and in the pooled analysis. Within-study and pooled estimates considered any clustering with each study’s sampling procedures. All pooled estimates were weighted according to study sample size. We used forest plots, I2 and Cochran’s Q statistics to assess consistency of outcomes across the studies. The I2 values showed moderate heterogeneity in both physical IPV (66%, p value=0.018) and NPSV outcomes (40%, p value=0.189). To account for any heterogeneity across the studies due to methodological diversity, we opted for mixed effects models to estimate overall effects. We thus performed a one-stage individual patient data meta-analysis using mixed effects logistic regression models to account for within-study and between-study variances (heterogeneity) across studies.17 The model considered the study hierarchy (participants with study). Study specific estimates and forest plots were derived as part of the post-estimation model to the mixed effects logistic regression model. Both the main and post-estimation models included participant’s age as fixed effects. All data were analysed in Stata V.16 and all models adjusted for participants’ ages.
Results
Men
The mean age of men across all studies was 32 years (table 3). Two-thirds (67%) of the men were living with their partner or spouse, 46% were married, 66% were cohabiting and a quarter (26%) were in a relationship, but not living with their partner. Most of the men had secondary school education or above (58%) and 56% were employed in the past 3 months.
Among the 8104 men, the prevalence of any (past year) alcohol use was 51% with study specific prevalence ranging from 37% to 69%. The overall prevalence of past year harmful alcohol use was 24% (range: 11%–44%). Almost one-third (29%) of the men across the four studies had perpetrated physical IPV in the past 12 months (range: 12%–50%). In the three studies where NPSV perpetration was measured, just over a quarter (26%) reported perpetrating NPSV in the past 12 months (range: 10%–39%).
Association between alcohol and violence perpetration
The association between moderate alcohol use and physical IPV perpetration was mixed. In the two South African studies (Stepping Stones and Creating Futures and Sonke), there were no significant associations between moderate alcohol use and past year physical IPV perpetration, compared with those who did not drink alcohol in the past year, in individual studies. However, there were significant associations in the Ghanaian and Rwandan studies (COMBAT, Indashyikirwa-couples and Indashyikirwa-community) between moderate alcohol use and IPV perpetration, with adjusted associations ranging from (adjusted OR (aOR)=1.91 (95% CI 1.47 to 2.48)) in the Indashyikirwa-couples sample to (aOR=3.18 (95% CI 2.28 to 4.45)) in COMBAT. In the pooled analysis of all five studies, there was a significant association between moderate drinking and physical IPV perpetration (aOR=1.57 (95% CI 1.04 to 2.38)).
There were consistent associations between harmful alcohol use in the past year and past year physical IPV perpetration across all the individual studies, and in the pooled analysis. Across individual studies the adjusted association for harmful alcohol use and past year IPV perpetration ranged from (aOR=2.32 (95% CI 1.61 to 3.34)) in Stepping Stones and Creating Futures to (aOR=5.68 (95% CI 3.96 to 8.15)) in COMBAT (table 4). Among men who used alcohol at harmful levels, in the pooled analysis the odds of perpetration of physical IPV were increased almost three and a half times (aOR=3.45 (95% CI 2.56 to 4.64)), compared with those who did not report alcohol use in the past year.
There were no associations between moderate alcohol use and perpetration of NPSV in the two South African studies (Stepping Stones and Creating Futures and Sonke CHANGE), but there was a significant association between moderate alcohol use and perpetration of NPSV in the Ghanaian study (COMBAT), compared with those who reported no alcohol use in the past year. In COMBAT men with moderate alcohol use were more likely to perpetrate NPSV, with an adjusted association of (aOR=1.98 (95% CI 1.38 to 2.83)).
Harmful alcohol use was significantly associated with NPSV perpetration in the past year across the three studies, compared with those who did not use alcohol in the past year, with adjusted associations ranging from (aOR=2.02 (95% CI 1.40 to 2.92)) in Stepping Stones and Creating Futures to (aOR=4.45 (95% CI 3.06 to 6.47)) in COMBAT and a pooled estimate of (aOR=2.64 (95% CI 1.85 to 3.76)) (table 4).
Women
Overall, women’s mean age was 31 years, just over half (54%) of women were married, 41% were in a relationship, 80% were living with their partner or spouse and 15% were not living with their partner (table 5). Among the 5163 women, just over one-third (37%) reported having seen their partner drunk in the past 12 months with study specific prevalence ranging from 14% to 53%; 16% reported seeing their partner frequently drunk in the past 12 months, (range: 8%–23%).
Association between alcohol and violence experience
Among women there were consistent associations between occasionally having seen their partner drunk in the past 12 months, compared with never, and their experience of physical IPV, with adjusted associations ranging from (aOR=1.63 (95% CI 1.12 to 2.37)) to (aOR=3.42 (95% CI 2.41 to 4.85)). Similarly, among women who had seen their partner frequently drunk in the past year, the adjusted associations ranged from (aOR=2.93 (95% CI 1.92 to 4.46)) in Stepping Stones and Creating Futures to (aOR=8.27 (95% CI 5.42 to 12.61)) in Indashyikirwa-Couples (table 6).
In the pooled analysis women who occasionally saw their partner drunk in the past 12 months, compared with never, had a more than 2.5-fold increased odds of experiencing IPV in the past 12 months (aOR=2.68 (95% CI 2.13 to 3.36)). While women reporting seeing their partner frequently drunk in the past 12 months had an almost sixfold increased odds of experiencing physical IPV in the past year (aOR=5.94 (95% CI 4.19 to 8.41)), compared with those who had never seen them drunk.
Discussion
In a pooled sample of data from five data sets arising from four studies in three countries across SSA, we found strong evidence of a clear association between men’s harmful alcohol use, and their perpetration of violence, with a mixed association between men’s moderate alcohol use and violence perpetration. Among women, we found a consistent association between witnessing their partner drinking and experience of IPV.
The association between men’s moderate consumption of alcohol and perpetration of physical IPV varied by country. In three of the data sets (COMBAT, Indashyikirwa-couples and Indashyikirwa-community), moderate alcohol use was associated with a greater likelihood of IPV perpetration, while in the two South African studies (Stepping Stones and Creating Futures and Sonke), there was no association. One explanation for these findings may be related to the difference in the background level of drinking and drinking patterns between the countries. In South Africa, men’s drinking is normative; 69% and 64% of men in the South African studies were current drinkers, compared with 44% in Ghana and 45% and 37% in Rwanda couples and community samples, respectively. Also, in the South African studies, 44% of men in Stepping Stones and Creating Futures and 40% of men in Sonke CHANGE trial drank harmfully (8+ on AUDIT scale). Therefore, in places like South Africa where alcohol consumption is more prevalent and socially normative, it may be that moderate alcohol use is both unremarkable and too common to serve as an indicator of risky or anti-social behaviour, whereas in settings where alcohol use is less common, any use may be indicative of being somewhat socially rebellious and associated with this is violence perpetration. A second explanation may lie in the generalised very high rates of violence in South African informal settlements. In both the Stepping Stones and Creating Futures and the Sonke CHANGE samples, 50% and 40% (respectively) of men reported perpetrating past year physical IPV, suggesting that even in the absence of alcohol, IPV perpetration may be driven by other salient factors such as poverty, patriarchal privilege and the normative use of violence in interpersonal relationships.3 Future mixed method research into the drivers of violence in informal settlements is merited.
Across all studies we found significant associations between harmful alcohol use, and IPV perpetration among men, and IPV experience among women. This is consistent with previously published literature.5 6 We also found correlated increases, whereby men who used more alcohol were also more likely to report recent perpetration of violence. Addressing harmful alcohol use should remain a focus globally in alcohol and IPV prevention interventions. These findings are especially important in the Global South, considering the high levels of harmful and heavy episodic drinking in the region.18 This points to the need for interventions that seek to reduce violence against women and girls in the Global South to take account of the significant contribution of men’s drinking and harmful patterns of alcohol use in violence perpetration, alongside other currently recognised drivers of violence.3
We also found that harmful alcohol use was strongly correlated with perpetration of NPSV. In SSA, there has been little data assessing the association between alcohol use and NPSV. A previous population-based study in South Africa covering three districts across two provinces showed an association between NPSV and drug use, but not alcohol use,8 while population-based samples across four countries in the Asia and Pacific did show an association between alcohol use and NPSV.7 Our findings show a clear association between harmful alcohol use and NPSV perpetration, and again, that increasing alcohol consumption was correlated to increased likelihood of reporting recent perpetration of violence. This makes the current analysis an important contribution to understanding the alcohol NPSV nexus in the Global South and underscores the potential importance of addressing alcohol in VAWG prevention.
Among women, we similarly found that occasionally seeing their partner drunk in the past year, and seeing their partner drunk frequently, were both associated with an increased likelihood of reporting past year IPV. As with men’s reports in the three studies, the findings indicate that any drunkenness around an intimate partner, and not just a pattern of heavy drinking, or a culture of drinking, is a risk factor for alcohol-related violence. It also confirms that the risk for violence increases with men’s increasing levels and frequency of alcohol use. Two plausible explanations may account for this association. One, it is possible that alcohol use directly increases aggression in at least some people through disinhibition or other physiological impacts, and therefore men’s risk of use of violence increases with increasing levels and frequency of alcohol use. Second, and particularly in contexts where many men drink, drinking alcohol to harmful levels may concurrently embody an image of ‘what it means to be a man’ that includes demonstrations of toughness and strength and dominance and control over women. However further research is required to unpack these issues and better understand which, if either, may predominate in any given setting. These reflections have important implications for planning future violence prevention interventions, and our findings indicate that in settings where harmful drinking is common, IPV prevention interventions should also address alcohol consumption.
In the settings where we have comparable data for women’s reports of their partners drunkenness and men’s reports of their harmful drinking (Indashyikirwa couples, Indashyikirwa community and COMBAT), we found that in Indashyikirwa (couples and community samples), men reported less harmful drinking than women reported their partners drunkenness. However, in Ghana, the opposite is true, with men reporting more harmful drinking than women do their partner’s drunkenness. It is unclear whether this is due to men’s under-reporting of harmful drinking or differences in understanding of what constitutes drunkenness in the different settings.
Addressing the intersections of alcohol-related VAWG is a critical intervention challenge. A variety of interventions have been shown to be effective in reducing alcohol-related VAWG. Structural interventions focusing on legal and policy reforms to make alcohol more expensive (price increases, minimum unit pricing, increases on excise tax) and less available (limiting hours and days of week sale, limit on quantities purchased and alcohol outlet density) have been shown to reduce drinking and alcohol-related IPV,19 20 but this has largely been in high-income settings. Community-based interventions in SSA have primarily focused on working directly with couples and groups. For example, psychotherapeutic interventions with couples with exceedingly high levels of alcohol use and IPV in their relationships show promise in reducing both of these.21 Intervening with couples to shift beliefs and social norms that drive IPV, such as poor conflict resolution, key triggers of IPV including raised awareness of alcohol (ab)use as a key trigger of IPV is an important approach for impacting alcohol use, and reducing alcohol-related IPV.22 Another set of interventions implemented in SSA have provided a group context in which to offer gender-transformative, social empowerment interventions and these have shown promise with men and reduced alcohol use and IPV perpetration.23 24 Gender transformative programmes provide men with opportunities to reflect on, and reframe their lives as men, change their relationships, and reduce their use of violence and other harmful practices, including heavy drinking.24 They have great potential to incorporate an assessment of the use and patterns of alcohol use in their design to comprehensively address the link between alcohol use and VAWG. It is notable that some of the latter interventions, implemented in SSA, notably Stepping Stones, although not specifically designed to reduce alcohol use have showed significant declines in alcohol use.23 24 Therefore, addressing the intersecting issues of gender, masculinity and alcohol use on IPV, may be an important target in IPV prevention planning, both as part of a multicomponent approach and by exploring the huge potential of gender transformative programme within which to address alcohol-related violence.
Strengths and limitations
One of the study’s key strengths is the pooled analysis of data from multiple settings and interventions across the SSA, using comparable measures of IPV and NPSV perpetration, experience of IPV and alcohol use. This overcomes the limitation of many previous studies both globally and in the Global South examining the associations between alcohol use and IPV and NPSV. It also examines associations between alcohol use and IPV/NPSV over the same recall period, that is, over the past 12 months, thereby reducing recall bias and problems with the temporal nature of the association. This shorter recall period is particularly important for alcohol-related research, where accuracy and volume of recollections may already be compromised. The study is also strengthened by the synthesis of comparable regional data (Africa) in the Global South, thereby addressing the limited geographical scope of studies examining alcohol’s association with VAWG. It extends the evidence base on the alcohol violence nexus, as we have increased the number of settings in SSA and included (South Africa), where alcohol use is highly prevalent. Furthermore, it examines both violence perpetration and experience, from the perspectives of both men and women, thereby strengthening our understanding of the relationship between IPV/NPSV and alcohol use. However, we also acknowledge limitations in the research. Previous research suggests that there is likely a bidirectional relationship between alcohol and IPV, particularly for women.25 Future longitudinal studies focused on examining the temporal relationships are needed to examine this. We acknowledge that both men and boys may also experience alcohol-related violence from women and other men. However, there was no data on this from these studies. Only one of the four studies (Ghana) was population-based and other studies were limited in generalisability, as they were based on populations recruited for the purpose of impact evaluation. We did not adjust for women’s own drinking, which has been found to increase their subsequent experiences of violence, as it was not included in all data sets. Despite the current limitations, the study confirms alcohol as a robust correlate of IPV and NPSV perpetration and of women’s experiences of IPV in SSA.
Conclusion
In five studies in SSA, harmful alcohol use was associated with increased odds of IPV and NPSV, while moderate alcohol use showed varied associations with IPV and NPSV, potentially shaped by background levels of alcohol use and violence perpetration. Public health programming needs to address alcohol use in prevention and treatment of IPV, and experiences of violence needs to be addressed in programmes addressing harmful alcohol use.
Data availability statement
Data are available in a public, open access repository. De-identified individual participant data for Stepping Stones and Creating Futures (South Africa), Sonke CHANGE trial (South Africa) and Evaluation of the COMBAT intervention (Ghana), are available to anyone who wishes to access the data for any purpose at https://medat.samrc.ac.za/index.php/catalog/WW . De-identified individual participant data from the Indashyikirwa couples cohort and community surveys (Rwanda) are available from the Principal Investigator of the study, Dr Kristin Dunkle: kristin.dunkle@mrc.ac.za, but may require permission from the Rwandan Ministry of Gender and Family Promotion (MIGEPROF) before transfer.
Ethics statements
Ethics approval
Ethical clearance for all studies was obtained prior to the studies commencing. For the Stepping Stones and Creating Futures intervention, clearance was obtained from the South African Medical Research Council’s Ethics Committee (EC006-2/2015) and the University of KwaZulu-Natal’s Biomedical Research Ethics Committee (BFC043/15). For the Sonke CHANGE intervention, clearance was obtained from the University of Witwatersrand’s Ethics Committee (M150443). For Indashyikirwa in Rwanda, ethical approval was obtained from the Rwandan National Ethics Committee (340/RNEC/2015) and the South Africa Medical Research Council Ethics Committee (EC033-10/2015). A required research permit was obtained from the National Institute of Statistics Rwanda (0738/2015/10/NISR). For the Community Based Action Teams (COMBAT) intervention in Ghana, the Noguchi Memorial Institute for Medical Research, University of Ghana ((# 006/15–16) and the South African Medical Research Council’s Ethics Committee (EC031-9/2015) granted ethical clearance for the study. All studies followed ethical and safety guidelines for research on violence against women. All participants provided written informed consent before participation. Further information on the studies ethics is available in the study sources referenced in Table 1.
Acknowledgments
The authors wish to thank the research teams, programme staff and participants from all studies included in this pooled analysis.
References
Footnotes
Contributors LR conceptualised the analysis with AG and RJ. EDC led the statistical analysis. AG was a principal investigator (PI) on the Stepping Stones and Creating Futures study in South Africa, EDC was a PI on the COMBAT study in Ghana, KD was a PI on the Indashyikirwa study in Rwanda and RJ was the Director of the overall What Works to Prevent Violence Against Women and Girls Global programme. LR led the drafting of the manuscript; all authors contributed to comments and revisions.
Funding All studies presented and this pooled analysis of the data were funded through the What Works to Prevent Violence? A Global Programme on Violence Against Women and Girls (VAWG) funded by the UK Government’s Department for International Development (DFID), Grant number: PO 6254. However, the views expressed do not necessarily reflect the department’s official policies and the funders had no role in study design; collection, management, analysis and interpretation of data; writing of the report; and the decision to submit the paper for publication. Funding was managed by the South African Medical Research Council. Time drafting this manuscript was funded by the South African Medical Research Council.
Competing interests None declared.
Patient and public involvement Patients and/or the public were not involved in the design, or conduct, or reporting, or dissemination plans of this research.
Provenance and peer review Not commissioned; externally peer reviewed.