Prevalence of sexual violence among refugees: a systematic review

ABSTRACT OBJECTIVE To synthesize data about the prevalence of sexual violence (SV) among refugees around the world. METHODS A systematic review was conducted from the search in seven bibliographic databases. Studies on the prevalence of SV among refugees and asylum seekers of any country, sex or age, whether in English, French, Spanish and Portuguese, were eligible. RESULTS Of the 2,906 titles found, 60 articles were selected. The reported prevalence of SV was largely variable (0% to 99.8%). Reports of SV were collected in all continents, with 42% of the articles mentioning it in refugees from Africa (prevalence from 1.3% to 100%). The rape was the most reported SV in 65% of the studies (prevalence from 0% to 90.9%). The main victims were women in 89% of the studies, all the way, especially when still in the countries of origin. The SV was perpetrated particularly by intimate partners, but also by agents of supposed protection. Few studies have reported SV in men and children; the prevalence reached up to 39.3% and 90.9%, respectively. Approximately one-third of the studies (32%) were carried out in refugee camps and more than half (52%) in health services using mental health assessment tools. No study has addressed the most recent migratory crisis. Meta-analysis was not performed due to the methodological heterogeneity of the studies. CONCLUSIONS SV is a prevalent problem affecting refugees of both sexes, of all ages, throughout the migratory journey, particularly those from Africa. Protection measures are urgently needed, and further studies, with more appropriate tools, may better measure the current magnitude of the problem.


INTRODUCTION
The world is currently experiencing the biggest migratory crisis since World War II, with an increasing number of refugees. According to the United Nations High Commissioner for Refugees (UNHCR) report, 65.6 million people were forced to move because of persecution, conflict, widespread violence or human rights violations in 2016. Of these, 22.5 million were refugees; 2.8 million, asylum seekers; and 40.3 million, internally displaced persons within their own countries 1 .
Sexual violence (SV), defined as a sexual act or attempt to obtain a sexual act without the voluntary consent of the victim or with someone unable to consent or refuse 2 , is considered a present threat during forced displacement and the search for asylum 3,4 . In times of war, women and girls are more vulnerable to rape and are at greater risk for other forms of SV, such as early or forced marriage, intimate partner abuse, child sexual abuse, sexual exploitation and trafficking 4 . SV has also been perpetrated against men and boys as a tactic of war or during detention and interrogation 5 ; they may suffer rape, sexual torture, mutilation, humiliation, enslavement, and forced incest 6 . This risk persists during the escape journey and after the reception in apparently safe destinations 7 .
The consequences can be extremely serious. In women, it can lead to mental disorders, obstetric complications, sexual dysfunctions, unwanted pregnancies, unsafe abortions and sexually transmitted infections 8,9 . Among men, in addition to infections and mental disorders, sexual dysfunction, somatic complaints, sleep disorders, withdrawal from relationships, attempted suicide, alcohol and drug abuse, and violent behavior are common 8,10 . In childhood, sexual abuse may also be accompanied by guilt, shame, eating disorders, cognitive distortions, mental disorders, sexual and relationship problems, and school absenteeism 11 .
Two previous systematic reviews have portrayed SV in refugees and internally displaced persons in emergency humanitarian complexes 12,13 : a meta-analysis aimed at estimating its prevalence in women only 12 , and other aimed at quantifying gender-based violence in three categories: physical violence, by intimate and sexual partner 13 . Neither analyzed the different types, profile of perpetrators and the moment of occurrence of SV in the migratory process. No studies have been conducted on the prevalence of this violence in the total refugee population (children, adults and older adults of both sexes) in different scenarios and moments of their trajectory, for a more comprehensive understanding of the magnitude of the problem.
Thus, we aim to synthesize the literature on the prevalence of SV in refugees around the world through a systematic review, regardless of sex, age and location. With this knowledge, one may better identify the profile of refugees who are victims of SV, contributing to specific prevention, approach, treatment and monitoring strategies in the countries of origin, during migration and in the host countries.

METHODS
The bibliographic search was carried out in January 2018, using the MEDLINE (via Ovid), Embase (via Ovid), PsycINFO (via Ovid), Scopus, Web of Science, Sociological Abstracts (via ProQuest) and LILACS (via VHL) databases. No date limits or language restrictions were applied. Search strategies have involved the following MeSH and free terms: "refugee," "asylum seek," "exiled," "refugee camps," "sexual violence," "sexual harassment," "child abuse," "sexual offense," "sexual abuse," "sexual crime," "rape," "sexual coercion," "sexual assault." Articles addressing any form of SV were included, using the connector "OR." For the calculation by type of SV, we use the definition described in each of the articles. The search strategy is detailed in Appendix A. Articles within the bibliographic reference lists of the review studies and those included in this study were added where applicable.
Studies with data available for calculating the prevalence of SV in refugees or asylum seekers (considered as single population) in any country, sex or age, and published in English, French, Spanish and Portuguese were eligible. Chapters of books, dissertations, annals of congresses, editorials, letters, notes and comments were not included.
The selection of studies was initially conducted through the search of titles and abstracts; then by reading the full texts. Decisions on study eligibility and data extraction were performed by two independent reviewers on electronic forms constructed in EpiData 3.1 (EpiData Association, Odense, Denmark), and the differences were resolved by consensus or by a third reviewer. References were managed in EndNote Web software [Thomson Reuters (SCIENTIFIC), NY, USA].
Information was collected on: (1) study methods and population; (2) prevalence of SV according to sex, age, type of SV, continent/region/country of origin, host country/region, period of occurrence and profile of perpetrators.
In studies that presented additional categories of migrants (e.g. economic migrants), only information on refugees and asylum seekers was used. Likewise, in studies that reported psychological, physical and sexual violence, only SV data were used.
The calculation of global prevalence was estimated from the information on the total cases of the studies. For the calculation of specific prevalence, the following types of SV reported by the articles were considered: rape, attempted rape, unwanted sexual contact, non-contact unwanted sexual experience, sexual harassment, sexual abuse, sexual torture, sexual assault, sexual exploitation, including enforced prostitution and sex for survival, genital mutilation, forced marriage and abortion. When only the prevalence by type were informed and more than one of these forms was inflicted on the same victims, it was not possible to estimate the overall prevalence.

RESULTS
We found 2,906 studies in the databases searched and 10 in the lists of bibliographic references ( Figure 1). After the duplicates were removed (n = 1,111), 1,805 studies were selected for the reading of titles and abstracts. Of these, 1,498 were excluded by the following criteria: language (n = 29), type of publication (comments, letters, books, notes, editorials, abstracts of lectures and dissertations, n = 361), study design (most qualitative or review studies, n = 521), population not composed of refugees or asylum seekers (n = 176), out of scope (did not address SV, n = 131) or both (population and scope, n = 280).
Three hundred and seven studies were selected for the reading of full texts. After the application of the eligibility criteria, 60 studies were included for data extraction. Of the excluded ones, 15 were not original articles, 121 were review studies or with qualitative design and in 27 studies the population was not formed by refugees or asylum seekers.

Characteristics of the Studies and their Populations
The 60 articles selected were all published in English between 1990 and 2017 (45% between 2000 and 2010) and from 31 different countries (14 from the USA). Studies were of crosssectional design (Table 1), except for two cohort studies 48,73 .
The most frequent sites of data collection, according to the 54 articles that contained this information, were health services (n = 28.52%) and refugee camps (n = 17.32%). Most studies (87%) were conducted to evaluate outcomes in mental health, without the main objective of measuring the prevalence of SV cases. Among the 49 studies that informed the instrument used, the Harvard Trauma Questionnaire (HTQ) was the most frequently validated instrument (n = 15, corresponding to 31%), while 29% (n = 14) used questionnaires designed specifically for the research.
Studies involved 28,101 refugees and asylum seekers. The population of each study varied between 15 and 11,458 individuals. In 33% (n = 20) of the studies, the sample included less than 100 people, and in 18% (n = 11), more than 500 people. The mean age of participants ranged from 10.6 to 41.6 years old; 42% (n = 25) of the studies included those younger than 18 years. There was a general predominance of women; in 37% (n = 21) of the studies, the sample was exclusively female. The predominant religion was Muslim, in 12 (63%) of the 19 studies with data about it.

Prevalence of Sexual Violence
The global prevalence variation presented a large amplitude, regardless of the sample size: from 0% to 99.8%, with a total of 2,859 cases of SV. In 15 studies (31%), the prevalence was less than 10% (samples from 80 to 11,458 people), and in 11 (23%), more than 50% (samples from 15 to 919 people), as shown in Table 1. This wide variation occurred independently of the data collection scenario -in refugee camps (n = 12, 0.03% to 99.8%), health units (n = 25, 2.3% to 76.2%) and communities/villages (n = 6, 5.2% to 93.3%)and assessing form -validated instruments (n = 25%; 0.0% to 99.8%) or questionnaires of the own research (n = 14; 0.03% to 93.3%).  Six studies reported SV in children and adolescents, with prevalence varying between 4.6% and 90.9% 16,44,47,52,54,72 . In 32 of the 36 (89%) studies that showed prevalence by sex, the main victims were women. Of these, 12 studies reported SV in both sexes, with a difference of up to 59.2% more of prevalence in women 17 . Two studies reported the opposite, but with disparities less than 2% 16,29 . In men, the prevalence reached 39.3% 14 .
Africa was the most frequent continent of origin in 13 (42%) of the 31 studies with information about it ( Table 2). As to the moment of occurrence, approached by 18 studies, 17 (94%) reported that SV occurred in the country of origin (prevalence between 1% and 92%); in two The most frequent type of SV was rape (65%) (   four by acquaintances 51,53,55,72 , four by relatives 45,54,58,72 , two by unknowns 51,53 , two by rebel soldiers 31,53 , one by police officers 58 , one by armed groups 72 , and one by guards in prison 17 .
In five studies 32,49,55,58,67 , the authors did not report the number of victims, and it was not possible to estimate the overall prevalence. Estimating the sum of prevalence by specific type would overestimate the overall prevalence due to cases that suffered more than one type of SV.

DISCUSSION
Previous studies have shown that SV is a constant threat throughout the refugee migration pathway 3,12,13 , which has been confirmed in the present review. Although most of the studies identified here revealed a higher prevalence among adult women, SV was also a serious problem in men and children. In addition, we observed the SV is perpetrated mainly by intimate partners, but also by military, guards and police. Most cases occur in the country of origin, in the form of rape and in refugees from Africa. In some refugee camps, such as Uganda and Cameroon, the frequency was alarming.
It is possible that prevalence may be underestimated in some studies, since many victimsespecially men -do not report SV because of shame, threats by perpetrators, fear of being found guilty or suffering from stigma and exclusion from family and community 6,74 , with consequent low demand for health care and case records 75 . In addition, the humanitarian crisis caused by armed conflicts in the refugees' countries of origin leads to large displacements of people and demands incompatible with the availability of health services and resources 76 , which may further reduce the chances of case identification. On the other hand, studies focused on the evaluation of mental trauma in health services may overestimate the prevalence.
In the meta-analysis of SV prevalence in women in emergency humanitarian complex scenarios, which also included internally displaced persons and excluded genital mutilation, the mean prevalence was 21.4% and higher in refugees from Africa 12 . In our review, we found several studies with a much higher prevalence. Regardless of the actual prevalence, SV was frequent in the populations studied, and deserves special attention in the health services and the reception of this population already weakened by traumas of war and persecution.
Young women are the main victims of SV, but men, children and adolescents are also victims, a reality little discussed in the literature. Men and unaccompanied minors are also exposed to the risk of sexual exploitation and abuse during migration and arrival in destination countries 3 . Nevertheless, the predominance in women is not surprising. The immigration process is accompanied by difficulties such as economic insecurity, language barriers and acculturation, which lead to the imbalance of power between women and partners, leading to increased tensions 77 . Because of economic, political, and social changes during wars and postwar periods, many men use violence to control women and reestablish their status of power 78 . Such conditions may explain the higher frequency of SV perpetrated by intimate partners.
SV occurs mainly before migration, in the countries of origin of the refugees. This suggests a relation with the conditions generated by the armed conflicts, which potentiate cultural norms of superiority of the masculine power present in these places, even before the condition of search of refuge. High prevalence in Africa supports this view. The Democratic Republic of Congo, where armed conflicts over natural resource reserves have lasted since independence in 1960 79 , is marked by atrocities including group rape, sexual slavery, forced family involvement in rape, genital mutilation, among others 80 . More shocking is the fact that, even when hosted in refugee camps, this already fragile population still faces insecurity and suffers SV perpetrated by those from whom they expect protection, such as officers and police.
Rape was the most mentioned form of this violence. This can be explained by the more concrete definition, by the most remarkable experience, and because most studies have used the HTQ instrument, which has a specific question about rape and sexual abuse, but not about other forms of SV. Rape is considered the cruelest type because it brings serious and severe consequences to the health of the victims. War survivors diagnosed with posttraumatic stress disorder and rape victims report more somatic symptoms than those without a rape experience 81 . Rape also increases the chances of acquiring HIV infection, as reported in sub-Saharan African refugee women in Paris, and is related to social difficulties and lack of fixed residence due to the risk of transactional sex or sexual harassment during lodging by relatives or acquaintances 82 .
Several studies included in this review had many limitations, such as lack of detail on the population, outcome of interest, timing of the occurrence, profile of the perpetrators, gender and age of the victims. In addition, the studies did not include victims of the most recent migratory crisis, which began in 2015.
Our review also has limitations. The literature search did not include the terms "sexual torture" and "genital mutilation," which may have resulted in low sensitivity and explained the number of articles found in reference lists. We did not include the gray literature and no methodological quality evaluation of the selected studies was performed. In addition, we did not restrict the sample size of the articles, which resulted in imprecise estimates in studies with few individuals 38 . Finally, methodological differences between the studies (different data collection sites, such as mental health services and refugee camps; different data collection instruments; studies focusing on mental disorders rather than SV prevalence; and unequal sampling) have contributed to the diversity of the rates found and heterogeneity between the studies, which prevented a meta-analysis to summarize the information.
In summary, results of this review show that SV is a frequent problem among refugees, both women and men, mainly those from Africa, and occurs at all times in the migratory process, including in places of supposed reception and protection. The SV problem among refugees from the most recent migratory crisis must be investigated in unselected scenarios and with more appropriate methods to better guide the necessary protection measures.