Prevalence of Non-Partner Sexual Violence: A Global Review PDF

# CHAPTER 7 ## Prevalence of non-partner sexual violence: a review of global data Naeemah Abrahams, Karen Devries, Charlotte Watts, Christina Pallitto, Max Petzold, Simukai Shamu, and Claudia García-Moreno ## Introduction to the Prevalence of Non-Partner Sexual Violence Sexual violence is recogn...

# CHAPTER 7 ## Prevalence of non-partner sexual violence: a review of global data Naeemah Abrahams, Karen Devries, Charlotte Watts, Christina Pallitto, Max Petzold, Simukai Shamu, and Claudia García-Moreno ## Introduction to the Prevalence of Non-Partner Sexual Violence Sexual violence is recognized as a human rights violation with profound public health impact and huge health and social consequence (Jewkes et al. 2002). The terms 'sexual violence', 'rape', and 'sexual abuse' encompass a broad range of forms of violence which have been defined in the *World Health Organization's World Report on Violence and Health* as: > any sexual act, attempt to obtain a sexual act, unwanted sexual comments or advances, or acts to traffic, or otherwise directed, against a person's sexuality using coercion, by any person regardless of their relationship to the victim, in any setting, including but not limited to home and work (Jewkes et al. 2002, p. 149). It is universally accepted that gender power inequality in society is the most important social force behind the problem of sexual violence (Jewkes 2002; Jewkes et al. 2011). These include prevalent notions of male sexual entitlement and male rights to use rape as punishment as has been shown in studies of sexual violence perpetrators (Jewkes et al. 2011; Wood et al. 2008). In addition, a culture of tolerance to sexual violence against women is reflected in, and reinforced by, the lack of seriousness with which the crime is treated by some members of the community, police, parts of the criminal justice system, and policymakers (Seedat et al. 2009). The study of sexual violence can be approached in a number of different ways, including how it happens over the lifespan (e.g. female genital mutilation, child sexual abuse, rape in marriage); the context in which it happens (e.g. within intimate relations, date rape, sexual abuse at school, sexual harassment within work settings, and incest within family settings); and by perpetrator type (intimate partners and non-intimate partners). Non-partner perpetrators include strangers, acquaintances, friends, colleagues, peers, teachers, neighbours, police, military personnel, and family members such as fathers, brothers, uncles, cousins, and step relations. It might be asked why one should analyse rape by perpetrator since research shows that non-partner sexual violence has much in common with intimate partner sexual violence (Plichta and Falik 2001; Tjaden and Thoennes 2006), sharing some of the same risk factors, and also having similar health and social impact to that described in Chapter 6 (Plichta and Falik 2001; Tjaden 2006). However, there are also important differences between the two phenomena, including different population prevalence levels, and different impacts on health. The differences in population prevalence may suggest different drivers, such as varying levels of gender inequality, masculinities emphasizing heterosexual performance, community tolerance in the form of acceptance of male sexual entitlement for different forms of violence, and general levels of specific types of community violence such as gang violence. In addition, there may be different factors that influence reporting, with stigma of intimate partner rape and non-partner rape being different and therefore influencing reporting and help seeking differently (Abrahams and Mathews 2013). In addition, data suggest stranger rapes are more violent and can more often involve weapons, leading to greater physical injury, while those by known perpetrators may be less violent but the betrayal of trust may have a larger impact on mental health (Culbertson et al. 2001; Jones et al. 2004; Temple et al. 2005). Other differences include the period of abuse and when the sexual violence occurs: within an intimate partnership it often occurs over a long period, accompanied by controlling behaviour by the perpetrator, a pattern which may not be present in non-partner sexual violence. In addition, all forms of child sexual abuse can be considered non-partner sexual violence. It is also recognized that women can perpetrate violence against men within intimate relationships as well as in non-intimate partner relationships. Our focus in this analysis however is on the more common form of violence, that is violence against women by men. This chapter presents the results of a systematic review and meta-analysis of existing data on the population prevalence of non-partner sexual violence reported by women from the age of 15 years. This review was done as part of the work for the Global Burden of Disease Study, to contribute to the quantification of the ## SECTION 2 The descriptive epidemiology of violence burden of disease and injury attributed to interpersonal violence (Murray et al. 2012). In addition, this analysis of global data will contribute to the recent call from United Nations bodies for better data to assist in the development of prevention interventions and for scaling up responses (United Nations 2013). ## Methods A global review of population-based prevalence estimates of non-partner sexual violence against women 15 years and older was undertaken, involving both peer-reviewed and grey literature. A systematic search was made of the following databases: the Cochrane Library, Medline, PubMed, Embase, CINAHL, British Nursing Index, Science Direct, British Medical Journal, Wiley InterScience, Health Management Information Consortium; social sciences databases (International Bibliography of Social Sciences, PsychINFO, Web of Science), and international databases (ADOLEC, Global Health, African Healthline, LILACS, Index Medicus of the Eastern Mediterranean, Southeast Asian, and Western Pacific Regions, Medcarib, Popline). Articles published from 1 January 1998 to 31 December 2010 were included. The search was updated to include studies and reports from 1 January 2011 to end 31 December 2011. We also hand-searched citations and requested experts to suggest other materials. We also made contact with the authors and the data managers of large studies dedicated to violence against women. We requested data disaggregated by age and sex on non-partner sexual violence from the authors of the *International Violence against Women Survey (IVAWS)* (eight countries) (Johnson et al. 2008); the *World Health Organization Multi Country Study (WHO-MCS) on Women's Health and Domestic Violence* (Garcia-Moreno et al. 2005) (ten countries); *Demographic and Health Surveys (DHS)* (eight countries) (DHS 2007); *Gender Alcohol and Culture International Study (GENACIS)* (16 countries) (GENACIS 2007); the *Centres for Disease Control Reproductive Health Surveys (CDC RHS)* (two countries) (CDC), and *Crime Victimization Surveys* across the globe (no relevant estimates) (ICVS). We included representative population-based studies with estimates of non-partner sexual violence. We considered including non-population-based studies in regions where data were limited, but did not find any to include. We included data on women aged 15 years and older on both lifetime and current (past year) exposure to non-partner sexual violence. We recognize that sexual violence between the ages of 15 and 18 years is also considered child sexual abuse, but the lower age range of 15 years is commonly used in intimate partner violence estimates, and we therefore used the lower age limit of 15 years for this analysis. Because we were interested in sexual violence perpetrated by all perpetrators other than intimate partners, that is strangers, acquaintances, friends, family members, colleagues, teachers, police, military personnel, etc., we excluded studies where the analysis combined intimate and non-intimate perpetrators (e.g. *International Crime Victimization Surveys*). We accepted any author definition of sexual violence (i.e. rape and any other form of sexual violence), and excluded studies that combined sexual and non-sexual violence in the analysis (e.g. combined sexual and physical). Unlike the measurement of intimate partner violence, most studies that we reviewed used a single, general question, to ask women about their experiences of non-partner sexual violence. One of the most common questions was: 'Were you ever forced to have sex or to perform a sexual act when you did not want to with someone other than your partner'. Narrow definitions were used by a few individual studies and these measured specific acts such as ... ever touched sexually against your wishes' (Ackard and Neumark-Sztainer 2003). We also looked at whether prevalence estimates were based on a single or multiple non-partner perpetrators, and included the estimate for combined perpetrators (if available) to ensure consistency, as most studies did not ask separate questions for different types of non-partner perpetrators. We screened the abstracts and extracted the data into an EpiData database. For our meta-analysis we required prevalence and uncertainty estimates, including the numerator, denominator, and design effect for studies with clustered sampling. Information on methodological variables that could assist in the identification of potential biases and assist with the assessment of the quality of the studies was extracted. We extracted information on whether the perpetrator data were analysed as a single or a combination of perpetrators; how sexual violence was defined; the exposure period, for example ever or current (last year); study sites (national/regional/urban/rural); whether the study was part of a larger data set, if the questionnaire was derived from the questionnaire used in the WHO multi-country study on Women's Health and Domestic Violence; and whether fieldworkers received special training in how to ask about violence sensitively, and to respond appropriately if respondents became distressed. ## Meta-Regression Models We used random effects meta-regression in Stata 12.1 to produce both an adjusted and unadjusted prevalence model and produced summary prevalence estimates for all Global Burden of Disease regions. We did not calculate estimates by age group because of limited data on specific age groups. The standard deviation of the prevalence was calculated as the value of the upper limit of the 95 per cent confidence interval divided by 1.96. If no standard deviation or no confidence interval was reported, the Wilson method was used to estimate the confidence limits based on the prevalence data. We controlled for covariates in the models and these were selected based on previous knowledge. Prior to model fitting, covariates were checked for correlation to avoid multi-collinearity. The covariates were whether the study was a national study, if the study was dedicated to violence against women, if the study measured lifetime violence or past year violence, if a broad definition of sexual violence was used, and if the study measured single perpetrators or combined different non-partner perpetrators. To obtain a global estimate, the regional estimates were weighted by region population sizes of women aged 15-49 years for the year 2010. ## Results We identified 7,231 abstracts/records for screening. The main reasons for exclusion were incorrect study design (non-population-based studies), studies focused on partner violence, or analysis combined perpetrators or type of violence (Figure 7.1). 189 records/abstracts were identified for full text screening, and after assessment 77 studies covering 56 countries were included, producing 412 estimates from women 15 years and older. Table 7.1 presents key characteristics of the 412 estimates. Data were available from all the Global Burden of Disease study regions with four regions (Asia Pacific, high income; North Africa/Middle-East; Europe, Eastern; sub-Saharan Africa, Central) having less than six estimates each. Only one estimate was found for the sub-Saharan African Central region (Democratic Republic of Congo) (Johnson et al. 2010) with eight regions having estimates from only one country. The regions with the largest proportions of estimates were Europe, Western (58 estimates) followed by sub-Saharan Africa, East (43 estimates). Six conflict setting countries contributed population estimates (Liberia, Timor-Leste (Timor East), Democratic Republic of the Congo, Kosovo, Philippines, and Sri Lanka), with the studies in Africa focussing on capturing conflict related sexual violence. More than half (59.7 per cent) of the estimates were derived from dedicated violence against women studies, and a similar proportion was from nationally representative samples (53.8 per cent). The majority of estimates measured lifetime non-partner sexual violence (81.8 per cent), combined perpetrators (93.7 per cent), and used a single question to capture any forced sexual act (91.5 per cent). Globally, 7.2 per cent (95 per cent CI 5.3-9.1) of women experienced non-partner sexual violence (Table 7.2). There were variations across the regions and the prevalence ranged from 3.3 per cent (95 per cent CI 0-8.3) in Asia, South, to 21 per cent (95 per cent CI 4.5-37.5) in sub-Saharan Africa, Central. Regions with high prevalence were sub-Saharan Africa, Central and Southern as well as Australasia. The wide confidence interval in the sub-Saharan Africa, Central region is most likely due to this being based on a single estimate. Regions with lower estimates were Asia South, Asia Southeast, and North Africa/Middle East. The Asia Pacific high-income region had a considerably higher estimate than the other four Asian regions, while Eastern Europe had a much lower prevalence than the other two European regions. Similarly among the Latin American regions, the southern region had a much lower prevalence, and among the sub-Saharan African regions the western regions also had a considerably lower estimate than the other three African regions. We present a forest plot of both the adjusted and unadjusted estimates (Figure 7.2) and similar results are found with all the confidence intervals of the unadjusted estimates overlapping with those of the adjusted estimates. ## Rates of Non-Partner Sexual Violence Identified The analysis has shown that, globally, 7.2 per cent of all women had experienced non-partner sexual violence. Levels above 15 per cent were found in four regions: Australasia; Latin America Andean; sub-Saharan Africa, Central, and sub-Saharan Africa, South, while the lowest estimates were found for Asia, South, and North Africa/Middle East. This is the first time that global and regional estimates of non-partner sexual violence have been calculated based on a comprehensive systematic review. Generally, higher prevalence estimates were found for regions with more data points, except for sub-Saharan Africa, Central, for which a single conflict setting country (Democratic Republic of the Congo) provided the estimate for this region and this is the highest estimate found in our review. With the exception of this region, lower prevalence was found for regions with fewer data. The findings confirm that non-partner sexual violence is widespread. The regional variations may reflect true variations but may also be linked to the availability of data, levels of disclosure, and the context of sexual violence such as war related sexual violence, although sexual violence does not occur in every armed conflict (Cohen et al. 2013). Sexual violence is highly stigmatized in most settings and this affects disclosure, with fear of being blamed and lack of perceived support from families, friends, and services leading to under-reporting (Abrahams et al. 2013; Kelly et al. 2005), and influencing help seeking and recovery (Moore and Farchi 2011). The estimates found in this review are therefore most likely underestimates, and this is confirmed by more recent population studies on male perpetration. South Africa was the first country to publish population-based perpetration results where non-partner and partner rape was disaggregated; this showed that non-partner rape was more common than intimate partner rape (21.4 per cent raped a woman who was not a partner; 14.3 per cent raped a current or ex-partner) (Jewkes et al. 2012). This was followed by studies conducted in the Asia and the Pacific regions with the UN Multi-country Cross-sectional Study on Men and Violence: the results on non-partner rape from interviews with 10,178 men showed that single perpetrator rape prevalence varied between 4 and 41 per cent across the nine sites (Jewkes et al. 2013). Men's exaggeration of perpetration in these studies has been considered as a possible explanation for the discrepancy between men and women's reported rates, but comparing the prevalence reported by women in the same populations shows that women invariably report lower prevalence of victimization across the different sites in South Africa and the Asia-Pacific, implying male rates of perpetration are not exaggerated but that women under-report victimization (Machisa et al. 2011; Jewkes et al. 2013). ## Gaps in the Evidence The review had limitations. The study was largely constrained by the limited availability of quality population-based data. Eight regions had data from only one country, and many countries did not have any population-based data at all. This is reflected in the wide uncertainty bounds in the regional estimates. The highest prevalence in this review was found for the sub-Saharan, Central region (21 per cent), but this is based on a single estimate from the eastern region of the Democratic Republic of the Congo, which explains the wide uncertainty (Johnson et al. 2010). This estimate was derived from a study that focused on conflict-related sexual violence, perpetrated by combatants during the past 16 years of conflict (Johnson et al. 2010), with great detail given to identifying types of perpetrators. It is worth noting that conflict-affected settings provide major challenges in the execution of population-based surveys and obtaining a representative sample may be difficult due to logistical and security issues. In addition, many conflicts are localized to specific parts of a country and data from national studies may not reflect the situation of specific, conflict-affected sites fully. A recent report on wartime sexual violence from Cohen et al. (2013) warns of the many misconceptions related to sexual violence in conflict and how much remains unknown because of the lack of data. Although more than 50 per cent of the 412 estimates were derived from dedicated violence against women studies, the focus in most of these studies was on intimate partner violence, and the measure of non-partner sexual violence was most often based on a single question. This single question is not in-line with current recommendations on violence-related surveys, which are to ask about experience of behaviourally specific acts. In addition, the lack of information on the range of perpetrators of sexual violence is an important limitation. Other study limitations include the inability to conduct age-disaggregated analysis, which made it impossible to calculate age-specific estimates. In addition, we included in our review studies that interviewed women aged 15 years and older, who reported their experiences of sexual violence since age 15. We did this in order to make a distinction with child sexual abuse. However, young women in the age group 15-18 years, who report having experienced non-partner sexual violence could also be considered, by some legal definitions, to have experienced child sexual abuse, so these two categories are not mutually exclusive. We also did not include men as victims, although this is increasingly recognized as an issue needing attention. Population level data are also limited for men. In spite of the limitations, this review demonstrates that sexual violence is a common experience in the lives of many women. Sexual violence, irrespective of perpetrator, violates the human rights of victims and has a profound and enduring impact on their lives. Studies have shown that sexual violence can lead to short- and long-term health consequences. There is increasing evidence of the health effects following exposure to non-partner sexual violence, in particular mental health problems including depression, anxiety, and alcohol use disorders (Hankin et al. 1999; Kimerling et al. 2007; Plichta and Falik 2001), which is similar to that found for intimate partner violence. Furthermore, any one form of sexual violence increases the risk for other forms of violence, with many child sexual abuse studies showing an increased risk for later victimization (Dunkle et al. 2004; Fang and Corso 2007; Fergusson et al. 1997), and more recently the link between harsh childhoods and adult perpetration (Jewkes et al. 2010; Mathews et al. 2011). ## Conclusion This first global review of the prevalence of non-partner sexual violence has shown that one in 14 women globally has been sexually assaulted by someone other than a partner, since the age of 15. For reasons described here, including the stigma and blame attached to sexual violence, this is likely to be an underestimate. The psychological effects of sexual violence and the high prevalence confirm that it is a pressing health and human rights concern, requiring serious attention. There are important data gaps, and the further standardization of research tools and methods would improve future measurement and monitoring. The review highlights the need for countries to have their own population-level data on the levels of sexual violence by different perpetrators to understand the magnitude of the problem, to better understand the main risk factors, and to develop appropriate policies and responses, including both comprehensive sexual assault services and primary prevention interventions. Reducing gender-based violence and building gender equality is a global challenge and an important development goal for governments across the world. Addressing and preventing non-partner sexual violence is a critical aspect of achieving this goal. ## Acknowledgement Data in the chapter are from The Lancet, Abrahams, N. et al, Global prevalence of non-partner sexual violence: a systematic review, 12 February 2014, DOI: 10.1016/S0140-6736(13)62243-6, Copyright © 2014 World Health Organization and World Health Organization, Global and regional estimates of violence against women: prevalence and health effects of intimate partner and non-partner sexual violence, World Health Organization, Geneva, Switzerland, Copyright© 2013, available from http://www.who.int/reproductivehealth/publications/violence/9789241564625/en/index.htm