Prevalence of intimate partner violence against women in the Arab world: a systematic review

Background Violence against women has particular importance for women’s health and wellbeing in the Arab world, where women face persistent barriers to social, political and economic equality. This review aims to summarize what is known about the prevalence of physical, sexual and emotional/psychological intimate partner violence (IPV) against women in the 22 countries of the Arab League, including geographic coverage, quality and comparability of the evidence. Methods A systematic review of IPV prevalence in Arab countries was carried out among peer-reviewed journal articles and national, population-based survey reports published by international research programmes and/or governments. Following PRISMA guidelines, Medline and the Social Sciences Citation Index were searched with Medical Subject Headings terms and key words related to IPV and the names of Arab countries. Eligible sources were published between January 2000 and January 2016, in any language. United Nations databases and similar sources were searched for national surveys. Study characteristics, operational definitions and prevalence data were extracted into a database using Open Data Kit Software. Risk of bias was assessed with a structured checklist. Results The search identified 74 records with population or facility-based IPV prevalence data from eleven Arab countries, based on 56 individual datasets. These included 46 separate survey datasets from peer-reviewed journals and 11 national surveys published by international research programmes and/or governments. Seven countries had national, population-based IPV estimates. Reported IPV prevalence (ever) ranged from 6% to more than half (59%) (physical); from 3 to 40% (sexual); and from 5 to 91% (emotional/ psychological). Methods and operational definitions of violence varied widely, especially for emotional/psychological IPV, limiting comparability. Conclusions IPV against women in Arab countries represents a public health and human rights problem, with substantial levels of physical, sexual and emotional/psychological IPV documented in many settings. The evidence base is fragmented, however, suggesting a need for more comparable, high quality research on IPV in the region and greater adherence to international scientific and ethical guidelines. There is a particular need for national, population-based data to inform prevention and responses to violence against women, and to help Arab countries monitor progress towards the Sustainable Development Goals.


Background
Despite substantial socio-economic diversity, the 22 countries 1 of the Arab League [1] (also referred to as the Arab region) share commonalities in language (Arabic), culture and religion (majority Islam). Some indicators of women's status across the region have improved in recent years. For example, almost all Arab countries have achieved gender parity in primary and secondary education (with exceptions such as Morocco and Yemen), a dramatic change from parity rates of 40-50% in the 1960s; and women's university enrolment exceeds that of men in Algeria, Bahrain, Jordan, Kuwait, Lebanon, Qatar, Oman and Saudi Arabia [2]. Nonetheless, by many measures, Arab women across the region continue to face barriers to full social, economic and political equality. All 14 Arab countries included in the 2017 World Economic Forum Global Gender Gap index ranked in the bottom 28 of 144 countries, based on gender gaps in health, education, political empowerment and economic participation [3]. A 2013 analysis found that women's labour participation was the lowest of all regions -26% compared with a world average of 51% [2]. While all Arab countries, apart from Somalia and Sudan, have signed the Convention on the Elimination of All forms of Discrimination Against Women, many did so with reservations about gender discrimination; equal rights in marriage, family, employment or nationality; right to choose residence; and/or compatibility with religious laws [4]. While female age at marriage has risen dramatically in some high-income Arab states, rates of child marriage remain substantial in countries such as Mauritania, Somalia, Sudan, Yemen, parts of Egypt, and among Palestinians in Gaza [5].
A recent United Nations (UN) analysis highlighted links between violence against women (VAW) in Arab countries and societal factors such as lack of female political and economic participation, discriminatory legal codes, legal impunity for violence against women and girls, andin some settingsarmed conflict and forced displacement [4]. Studies also suggest that rigid gender norms and notions of masculinity embedded in traditional culture contribute to men's use of VAW [6]. Polarizing debates about women's status in the Arab regionbetween those who argue that traditional interpretations of Islamic doctrine protect women and those who criticize social norms supporting male control over womenmake it all the more important to understand levels of VAW in the region [4]. Intimate partner violence (IPV), the most common form of VAW, is widely recognized as a human rights problem that poses a significant threat to women's health and wellbeingboth globally [7] and within the Arab region [4]. In 2013, the World Health Organization (WHO) estimated that 30% of everpartnered women worldwide had experienced physical or sexual violence by a partner; as had 37% of everpartnered women in the Eastern Mediterranean region, which covers part of the Arab region [7]. Other than South East Asia, this was the highest estimate of any region, though it was based on data from just four Arab countries (Egypt, Iraq, Jordan and Palestine) plus Iran.
All 22 Arab countries agreed to the UN Sustainable Development Goals (SDGs), including Target 5.2: Eliminate all forms of violence against all women and girls in public and private spheres [8]. They also agreed to measure SDG indicators, such as 5.2.1: The proportion of ever-partnered women and girls aged 15 years and older subjected to physical, sexual or psychological violence by a current or former intimate partner. IPV prevalence data from the Arab region are limited, however. As of July 2019, the SDG global databases contained national estimates for Indicator 5.2.1 from only three Arab countries (those with a DHS violence module, Comoros, Egypt and Jordan) [9]. A 2017 UN review identified 14 studies with IPV prevalence from six Arab countries [4], but there has not been a comprehensive review of IPV from the region. To address that gap, this systematic review aims to describe what is known about the prevalence of IPV against women in the 22 countries of the Arab League, including the geographic coverage, quality and comparability of the evidence.

Definitions and forms of IPV
The SDG metadata define IPV as any physical, sexual or emotional/psychological abuse "perpetrated by a current or former partner within the context of marriage, cohabitation or any other formal or informal union" [10]. In practice, however, researchers classify, define and measure IPV in diverse ways. Because prevalence rates in this review came from published sources, we did not try to impose standardized definitions. Instead, operational definitions were extracted from each source (when available) and described as part of the findings. Most sources defined an 'intimate partner' as a husband, but a few included cohabiting partners; this article uses 'intimate partner' and 'partner' interchangeably. For ease of communication, the term 'emotional/psychological' is used to refer to all forms of violence that researchers variously labelled as 'emotional', 'psychological', 'verbal', 'social' or 'threat' abuse, and in some cases controlling behaviour. To limit the scope, data on 'controlling behaviour' was not extracted if researchers considered it a separate domain from IPV.

Search strategy
Eligible data sources were identified, screened and summarized using a two-stage approach: (1) a systematic search for studies reported in peer-reviewed journal articles; and (2) a search for nationally-representative, population-based surveys published by governments or international research programmes such as Demographic and Health Surveys (DHS). The search did not include other grey literature.

Part 1: search for peer-reviewed journal articles
In accordance with PRISMA guidelines, a search for peerreviewed journal articles in Medline and the Social Sciences Citation Index was carried out in duplicate (TE, CA), using Medical Subject Headings (MeSH) terms and key words such as: intimate partner violence, domestic violence, spouse abuse, violence against women, gender-based violence and names of each country in the Arab League, supplemented by hand searching bibliographies. Titles and abstracts were retrieved, downloaded into Endnote and screened independently for eligibility by two authors (TE, CA). Articles selected for full text review were screened by at least two authors (TE, CA, SB). Differences were resolved by consensus.

Inclusion criteria
Journal articles met a priori inclusion criteria if they were: published in a peer reviewed journal between January 2000 -January 2016, in any language (e.g. English, French, Arabic); from a member country of the Arab league (including populations of foreign nationals living in Arab countries, but not Arab populations living outside member countries); presented data gathered among women (not samples composed exclusively of children or adolescents); presented percentages of women who reported any physical, any sexual, any emotional/psychological IPV, or a composite measure, but not just individual acts (e.g. slapping); met basic reporting quality criteria, including a clear description of research design, study population, sample size and respondent selection.
To capture a wider range of studies and to assess both the strengths and weaknesses of the evidence base, we included some articles where the operational definition of IPV was not entirely clear, if they met all other inclusion criteria.
Because the principle aim of the search was to identify surveys that gathered IPV prevalence data, sources were excluded if they were editorials, case reports, correspondence, book reviews, historical or political science analyses; and systematic reviews were retrieved but not included in publication counts. Studies were also excluded if samples were composed entirely of women who reported violence; if they focused on consequences, social determinants, and attitudes but had no prevalence data; or if they focused exclusively on forms of violence other than IPV, such as conflict-related violence, 'honour killings', Female Genital Mutilation (FGM), or forced marriage. Secondary analyses of national surveys were excluded if they duplicated findings reported more completely in reports retrieved through the second search arm (below).

Part 2: search for national, population-based surveys
The second arm searched for national, population-based surveys from Arab countries using databases of UN Women [11], the SDGs [9], the Global Health Data Exchange, DHS, and Google Scholar, and bibliographies of global and regional reviews. Because national survey findings do not always reach journals in a timely way if at all, peer-reviewed journal publication was not required for this arm of the search. National surveys were included if they met all inclusion criteria of the first arm (other than journal publication) and were: population-based, nationally-representative household surveys; part of an international research program (e.g. DHS) or carried out by, or in collaboration with, a government agency such as a national statistical office; published a final report (at least online) with national estimates of IPV against women (not just adolescents) between January 2000-January 2016.

Data extraction from journal articles
Open Data Kit software [12] was used to extract data from eligible articles including: a) article identification (title, authors, year of publication, journal name, country/ies); b) study characteristics (design, methods, setting, years of data collection, and sample size/characteristics; c) instruments and operational definitions of IPV and 'partner'; d) and prevalence rates of physical, sexual and emotional/ psychological IPV against women, controlling behaviours, economic abuse, and composite measures as published.

Risk of bias assessment
Risk of bias was assessed using a checklist adapted from existing tools [13,14], informed by international guidelines for violence research [15,16]. Studies received a point for each quality criteria not met including: 1) national; 2) population-based (not facility-based, except for studies of violence during pregnancy); 3) sample representative of target population, not convenience; 4) adequate sample size (> 300 or based on a precision calculation); 5) breakdown of respondent characteristics; 6) response rate ≥ 70%; 7) clear operational definitions of violence and 'partner'; 8) reasonably valid and reliable measure (e.g. modified conflict tactics scale; partner and behaviourally-specific); 9) clearly defined, sensible numerators and denominators, including timeframe; 10) dedicated violence study, not a module in a health or alcohol study; and 11) clear adherence to WHO ethical guidelines regarding privacy, consent and confidentiality [17]. The possible range of scores was 0-11, with 0 meaning the least risk of bias and 11 meaning the highest. Scoring was performed in duplicate (SB, CA) with discrepancies resolved by consensus.

Presentation of findings
Tables present findings by dataset (i.e. individual survey), not publication. When publications reported different rates based on the same dataset (usually because they used different subsamples to explore a specific correlate), we included rates from the source reporting prevalence among the broadest sample of women. Data from population-based studies are presented separately from (health, religious and educational) facility-based studies because the latter may not be representative of the broader population of women. IPV prevalence rates in pregnancy are presented in a dedicated table.

Eligible sources by type and geographic coverage
The journal database search identified 1104 records ( Fig. 1), of which 926 titles and abstracts were screened (after duplicates removed); 228 were selected for full-text screening, along with 13 records retrieved through manual searches. Full text review excluded articles that did not present IPV prevalence data or meet the other inclusion criteria noted earlier. Full text review also excluded nine articles that duplicated DHS datasets retrieved in the second arm of the searchincluding seven secondary analyses of the 2005 Egypt DHS and two secondary analyses of the 2007 Jordan DHS. Articles reporting on the Palestine 2005/6 national survey (also retrieved in the second search arm) were included, however, because they reported findings more completely than the national report. After full text review was complete, 63 articles that met inclusion criteria remained. The second search arm identified 11 eligible national, population-based surveys, including five DHS surveys and six governmental surveys. National survey reports from Algeria 2005 [18] and Syria 2005 [19] were not included in the group of 11 eligible national surveys due to inadequate description of methods.
Eleven countries had at least one IPV prevalence survey from either search arm; ten had an eligible study from a journal, and seven had an eligible national, population-based survey published by DHS and/or a government (Table 1). Countries with the highest number of individual surveys (not publications) were Egypt (n = 16), Jordan (n = 13), Saudi Arabia (n = 6), Lebanon (n = 4) and Palestine (n = 4).

Characteristics of eligible prevalence sourcesrisk of bias
Characteristics of eligible surveys from both search arms are presented in Table 2 and Table 3, and described below.

Representativeness: study and sample design
Population-based household surveysconsidered the gold standard for IPV prevalence research [16] comprised almost one-third (n = 15) of the 46 individual IPV surveys reported in journals; two-thirds (n = 31) were facility-based. All 11 national surveys from the second search arm were population-based by definition. Distribution by type of survey varied by country. Saudi Arabia had six eligible facility-based studies but no population-based studies. Palestine had four population-based but no facility-based studies. All population-based surveys used multi-stage, cluster sampling, though some older surveys reported an inadequate sampling frame [93]. A majority of facilitybased studies used convenience samples, but some used systematic or random designs or interviewed all women attending health services during a given period.
Except for five facility-based and one population-based study, most surveys met the sample size quality criteria of ≥ 300 or provided reasonable justification for the smaller sample size [58,86]. Samples sizes varied widely, from 80 to 7105 in journals and from 2012 to 18100 in national DHS and governmental surveys. Almost all surveys limited IPV samples to ever-or currently-partnered women, except for studies from Jordan [34] and Syria [91] that included all women, regardless of partnership. Some samples were limited further, e.g. to women living with a husband or child < 18, refugees or pregnant women.

Operational definitions of partners, perpetrators and forms of violence
Reflecting conservative social norms in the Arab region, almost all surveys defined a partner as a husband except for a few that included fiancés, the Comoros DHS [42], which included any cohabiting partner, and a study from Morocco that did not specify [83]. Surveys fell into four categories: those that asked about violence by the current partner; by the current/most recent partner; by any partner in life; and by any household member. Recent DHS surveys  [20,41,78,81,84,87,91] used measures that were not partner specific, i.e. they asked about violence by any family member or anyone, and relied on follow-up questions to determine whether the perpetrator was a partner. Previous work [15] suggests this approach may underestimate prevalence. The majority (n = 32) of the 46 surveys reported in journals were published with operational definitions of each form of violence measured (Additional file 1). Articles reporting on seven surveys did not provide operational definitions but did name or describe instruments used; this information provided a general idea of how violence was defined, but could not account for changes that may have been made during adaptation or translation.

Measures of physical violence
Most surveys (n = 39) from journals and all national surveys measured physical IPV. Most used measures that aligned with international guidelines [16] by asking women about a behaviourally specific list of acts (e.g. slapped, kicked, beat up, etc.). However, several studies [20,30,55,82] limited measures of physical IPV to "beaten". Surveys that used the AAS or WAST asked women about abstract concepts such as 'violence' or 'abuse'an approach known to underestimate IPV prevalence [15,94]. One study from Jordan [69] asked an open-ended question about the most serious form of violence experienced in the past year. Two surveys, one from Egypt [58] and one from Morocco [83] included forced sex in rates of physical violence. One study from Jordan [31] may have included witnessing, not just experiencing violence in prevalence rates. A study from Syria [91] required a threshold of ≥ 3 experiences in the past year and ≥ 1 in the past 30 days to classify women as having experienced IPV.

Measures of sexual violence
Just over half (n = 25) individual studies from journals measured sexual violence. Some researchers said the topic was too sensitive; researchers from Sudan removed questions about sexual violence after the first 15 respondents refused to give clear answers [90]. Most surveys defined sexual IPV as forced sexual intercourse and/or sex acts. Surveys using the WHO instrument also measured unwanted sex due to fear of a their partner might do if they refused. A few surveys measured less common acts, such as being denied sex or hurt during sex [23], being prevented from using contraception [38] or being told hurtful things during sex [57]. Studies using AAS or WAST asked about sexual 'abuse' rather than behaviourally specific acts.

Measures of emotional/psychological abuse
Two-thirds of surveys from journals measured emotional/psychological IPV, but definitions varied widely. Many studiesincluding most using DHS or WHO instrumentsdefined 'emotional' IPV to include insults, humiliation, intimidation and threats to harm her or someone close to her. Other commonly measured acts included threats to divorce, to take a second wife, to expel her from the home or to take away her children. While most studies classified threats with a weapon as physical violence, others, including from Egypt [52,58] and Saudi Arabia [89] classified such acts as emotional/ psychological or "threat" violence. Some surveys measured less common acts of emotional/psychological abuse. The WorldSAFE study from Egypt [26] asked women whether their husband had ever abandoned them without financial support for at least 6 months and whether he had ever been unfaithful. Another study from Egypt [58] asked whether their husbands had "spied" on them. One study from Jordan [62] explored "work-related harassment" by partners, such as harassing phone calls at work. Studies using the AAS asked women whether they were "afraid of someone in the household" and classified that as emotional abuse. The national survey from Tunisia [51] measured two acts that straddled emotional and sexual abuse: no longer wanting to have sex with her; and bringing other women (mistresses) to the home. Two studies from Egypt [26,91] limited emotional/psychological abuse with frequency thresholds, e.g. ≥ 3 times in a specific time period.
Eight studies from journals measured acts related to money, such as preventing women from having or earning money, taking their earnings, or denying them financial support. Three surveys [52,54,58] classified such acts as a component of 'control' or 'social' abuse; five [36,57,66,77,81] classified them as a distinct category of 'economic' abuse.

Prevalence indicator construction
Most surveys constructed prevalence indicators in accordance with international good practice guidelines [16]; however, some articles did not clarify whether timeframe was lifetime, past year, or other [34,66,67]. Some reported rates of violence by any household/family member, not just partners [41,81,84,87,88,92]. Habib et al. [23] reported the percentage of women who disclosed emotional abuse only, but no physical or sexual abuse. Similarly, Tashkandi and Rasheed [89] reported the prevalence of emotional abuse only with no physical IPV. Given that abused women often experience multiple forms of IPV, these indicators captured only a fraction of those who reported any physical or any emotional IPV. Al-Nsour and colleagues [69] reported the percentage of women who reported that the most serious violence in the past year was physical, rather than the prevalence of any physical IPV (same for emotional IPV).

Safety and ethics
While most sources described informed consent or safety procedures, only about one-fourth (12 of 46) of studies reported in journals mentioned adherence to WHO Ethical and Safety Recommendations for Research on Domestic Violence Against Women [17]. Studies from Lebanon [35] and Egypt [20] did not adhere to WHO guidelines by: a) interviewing matched members of couples about IPV; b) interviewing more than one woman in the home; and/or c) allowing children to be present during the interview. Among the 11 reports presenting findings of national surveys, four affirmed adherence to WHO guidelines; four did not but did describe field procedures that met those recommendations; two did not provide enough information to assess; and one from Morocco [49] described field procedures that did not correspond with WHO guidelines (e.g. more than one in five women may have been interviewed in front of a third person).

IPV prevalence reported in journals Physical IPV prevalence
Rates of physical IPV (ever) reported in journals ranged from 6 to 43% in population-based studies (Table 4), and from 12 to 59% in facility-based studies (Table 5). About half of the studies documented a prevalence of nearly one-third or more. Past-year rates of physical IPV ranged from about 5% to over 50% in population-based studies and from 15 to 26% in facility-based studies. Neither of the two articles reporting the lowest rates of physical IPV (5-6%) [34,41] published detailed operational definitions or clearly explained their indicator construction.

Sexual IPV prevalence
Rates of lifetime sexual IPV reported in facility-based studies ranged from 5 to 37%. Only two populationbased studies from journals measured lifetime sexual IPV, reporting rates between 7 and 8%. Past year sexual IPV ranged from 11 to 40% in population-based studies and 9 to 26% in facility-based studies.

Emotional/psychological IPV prevalence
Lifetime rates of emotional/ psychological abuse (however defined) were 19 -74% in facility-based studies and 5 -49% population-based studies; past year rates ranged from 16 to 58% in facility-based studies and 3 to 91% in population-based studies.

Prevalence of economic abuse
In five studies that measured economic abuse as a separate form of IPV, rates ranged from 12% in a study from Lebanon [81] that defined economic abuse as 'denied financial support' to 53% in a study from Jordan [66] that defined economic abuse as preventing her from knowing about/accessing family income and/or preventing her from working outside the home.

IPV prevalence from national DHS and governmental surveys
Excluding Comoros, national estimates of physical IPV ranged from 20% in Tunisia to 33% in Egypt, ever; and from 6% in Morocco to 24% in Palestine, past year ( Table 6). The 2012 Comoros DHS [42] produced the lowest national estimates of physical IPV, both ever and past year -7.3 and 4.3% respectively. National estimates of sexual IPV ever ranged from 2% in Comoros to 14% in Tunisia; while past year estimates ranged from 1% in Comoros to 12% in Palestine. Estimates of emotional/ psychological IPV ranged from 8.1% in Comoros to 33.4% in Iraq, ever, while past year estimates ranged from 6.2% in Comoros to more than half in Palestine.

IPV during pregnancy
In seven facility-based studies, currently pregnant women reported rates of physical violence during pregnancy ranging from 10.4 to 34.6% (Table 7). DHS surveys from Comoros [42], Egypt [43] and Jordan [45] measured physical violence by any perpetrator during any pregnancy in life among ever pregnant women, and found levels of 3, 7 and 7%, respectively.

Discussion
This systematic review found uneven geographic coverage and quality of evidence on IPV against women in the Arab region. Only half of the 22 Arab countries had eligible prevalence surveys, disproportionately from Egypt and Jordan. A majority of surveys were facility-based, known to have limited generalizability to the broader population of women [16]. Some governments (e.g. the State of Palestine) had invested in repeated, population-based VAW national data collection, but relatively few Arab countries had even one round of national IPV prevalence estimates, and only three countries (Comoros, Egypt and Jordan)those with DHS violence moduleshad estimates included in the SDG global database.
Evidence, however fragmented, suggests that IPV against women in Arab countries represents an important public health and human rights problem. Reported IPV rates varied widely across surveys, but generally aligned with WHO estimates of 37% of physical and/or sexual IPV against ever-     [71] d. Also, threats of harm (15%); divorce (4%); to take away children (12%); denied financial support (12%) e. Violence by any family member (ever) was 17.9% physical; 6.9% sexual; and 35.9% emotional f. Article did not report rates of any emotional violence ever; 30.9% reported emotional violence with no physical violence g. 32.8% women reported experiencing emotional but no physical IPV in the past year h. Physical and/or emotional IPV. Article did not explain how lifetime prevalence could be lower than past year i. 26% among married women partnered women in (at least a portion of) the region [7]. In DHS surveys, 3 -7% of ever pregnant women reported violence during any pregnancy in life by any perpetrator, which fell within the 2 -14% range documented by a 2010 global review [95]; but facility-based studies from the region documented substantially higher rates of IPV during pregnancy among currently pregnant women, ranging from 10 to 35%possibly due to better recall or selection bias among women receiving prenatal care compared with the general population.
Wide variations in reported prevalence likely reflects the diversity of methods and operational definitions across surveys, particularly from the journal literature, suggesting a need for caution when comparing prevalence rates across surveys, especially when measures do not conform to international guidelines [16] or were not   [71] found that prevalence rates analysed with their study-specific definitions were substantially higher than rates produced with WHO definitions: 73% vs. 50% (respectively) for lifetime psychological violence and 97% vs. 83% (respectively) for controlling behaviours. Diverse definitions of emotional/ psychological IPV are not unique to Arab countries, but have been noted by global reviews [7,96].

Gaps in the regional evidence about IPV
This review highlights gaps in the regional evidence base. One gap is a lack of data on sexual IPV. While all national surveys measured sexual IPV prevalence, only about half the surveys from journals did so, often describing the topic as too sensitive. Given that the Egypt DHS has measured sexual IPV since 2005, it should be feasible to expand data collection in this area, perhaps by encouraging researchers to share lessons learned within the region. SDG Indicator 5.2.1 metadata [10] notes that the most important sources of national IPV prevalence data are "specialized national surveys dedicated to measuring violence against women and international household surveys that include a module on experiences of violence by women," such as DHS. Most journal articles with IPV data reported findings of subnational surveys, which cannot monitor national levels of violence.
Generally there is a need for more comparable, high quality, nationally-representative IPV data to allow Arab countries to monitor levels of IPV over time and report on SDG Indicator 5.2.1. Such data would also allow researchers to examine cross-national associations between IPV prevalence and factors related to gender equality, including legal frameworks and levels of child marriage. Barriers to national data collection may include political instability, armed conflict or forced displacement, lack of national commitment to address VAW, and/or the challenges of carrying out national, household survey data collection on sensitive topics.
DHS violence modules and the WHO VAW instrument offer Arab countries a way to expand the evidence base in a comparable way. While the search did not identify any official WHO surveys in Arab countries, WHO instruments were mentioned by nearly one-fourth of surveys from journals and several national surveys. WHO is currently revising the instrument; ideally the next iteration will be translated into Arabic, adapted to the cultural context and shared in the region.

Limitations
Prevalence estimates summarized in this review may underestimate 'true' IPV prevalence due to barriers to disclosure. Widespread social norms in the Arab region support husbands' right to physically discipline wives, and abused women often face high social, economic and legal barriers to divorce and unresponsive law enforcement and health care institutions [4]. Women are often reluctant to report violence to authorities [97] and may hesitate to disclose violence to survey interviewers.
This review had other important limitations. First, it did not include a comprehensive search of the grey literature and may have missed valuable subnational studies not published in journals. Second, while basic reporting quality (clear description of research design, study population, sample size and respondent selection) was an eligibility criterion, many studies had other risks of bias. This provided a more comprehensive view of the regional evidence, but it means that these prevalence rates should be interpreted with caution. In particular, facility-based surveysespecially those with convenience samplesare not necessarily generalizable to the broader population. Additionally, field procedures, interviewer selection and training may have affected data quality and women's willingness to disclose violence [98], but these are difficult to assess from published reports alone.
Another limitation is that this review did not include a meta-analysis or a quantitative analysis of how methodological characteristics correlated with differences in rates across studies. Given the complexity of the data and lack of consensus about definitions, such analyses were not feasible given the scope of this paper, but are worthy of future attention. Finally, this review did not examine attitudes, help-seeking behaviours, consequences, risk/protective factors, or social determinants of partner violence all important topics for future research.

Conclusions
IPV against women is a global phenomenon that thrives in a culture of silence. IPV has particular importance in the Arab region, given norms and systems that reinforce male authority over women [4]. Though data are fragmented and often not comparable across sources, evidence suggests that the problem is substantial, and there is a need for greater investment in violence prevention and response. There is also need for more high quality data collection and analysis in Arab countries, to understand the magnitude of IPV, monitor change over time, and explore pathways that perpetuate violence and prevent an adequate response. Researchers also need greater familiarity with international