Rationale: Violence against migrants in insecure status

Violence shortens lives, causes harm, and has social and political implications [5]. Violence constitutes a major risk to public health [6] The effects to physical health include not only traumatic injury, but also effects on the brain, neuroendocrine system, immune response, cardiovascular disease, premature mortality and mental health conditions such as depression and anxiety [7]. While global movement of people has increased over the last three decades, so too have immigration restrictions in various contexts [8, 9]. Precarity, including and creating vulnerability to violence, has been identified in association with insecure migration status [9–11]. To date there has been no study measuring the prevalence of violence that is commonly experienced across insecure immigration statuses or how insecure migration status intersects with other social determinants of health.

People in insecure immigration status experience physical, psychological, emotional, sexual, verbal, structural and legal violence [12]. This study focuses on physical violence. It was necessary to limit the outcome so that the review was manageable, and measures of physical violence are more widely and consistently available than others [13]. Violence against migrants can be broadly categorised into four different contexts: a) Violence in transit. This is usually when a person is travelling from one place to another, usually on a journey to seek a form of protection (for example, political asylum). These journeys are usually characterised by a lack of status, as the forms of transit available are often undocumented and informal [14–21]. b) Violence in custody. This includes violence experienced by migrants during arrest—which is often the result only of crossing a border or being present in a country with no status—during detention [22–26], and during removal [27, 28]. This category also includes people in asylum reception centres who have made an asylum application or appealed an asylum decision and are awaiting the outcome of that application or appeal [24]. Due to the often-involuntary nature of asylum reception, people who are housed within asylum reception centres can be considered to be under the custody of the state [22–26]. c) Work-related violence. This often emerges as a result of work-visas abroad that are tied to a particular employer, or a particular form of employment. A lack of alternative options, along with a reduced network and often a lack of linguistic and cultural proficiency in the host state produces a vulnerability to abuse and to a form of indentured servitude or modern slavery [29–31]. Literature on human trafficking also deals with modern slavery, indentured servitude, and violence attached to both formal and informal work processes (for examples see [32–35]. d) Family violence. This predominantly includes intimate partner violence (IPV) where one or both partners is an immigrant. It includes statuses that have an included dimension of dependency where the immigrant spouse relies on the marital relationship in order to maintain status, such as in the case of spousal visas, exacerbated when they include a financial dependency element such as the UK’s no recourse to public funds stipulation, or the US’s affidavit of support requirement [36–38]. Nonetheless, this category can also be expanded to include other forms of domestic violence in mixed status or immigrant families such as child abuse [39, 40] or elder abuse [41–43].

This study aimed to estimate the prevalence of physical violence that is experienced by people in insecure immigration status.

Eligibility criteria

We included primary quantitative studies or quantitative components of mixed methods studies of any design if they reported measures of insecure immigration status and physical violence experienced by people of any age while in insecure immigration status. Only peer reviewed reports in English published since 1 January 2000 were included. See Appendix 2 [S2 Appendix] for more detailed information.

Information sources

Database selection was based on initial scoping, combined with areas of expertise across the authorship. Five databases were selected: Embase, Social Policy and Practice, Political Science Complete, SocINDEX and Web of Science Social Sciences Citation Index. All selected studies were subject to backwards and forwards citation tracking to identify additional studies for inclusion. Forwards citation tracking was carried out using the tool available in Google Scholar. We ran the searches on 22 September 2021 and updated on 31 May 2023, for records from 1 January 2000. The start date was chosen to exclude work that predated immigration reforms in the 1990s.

Search strategy

We combined three concept clusters ‘immigration’, ‘violence’ and ‘methods’ and employed a Boolean search to link the three concept clusters (AND search) while using multiple descriptive terms in each cluster (OR search) [S3 Appendix].

Selection process

The first reviewer screened all titles and abstracts against the inclusion and exclusion criteria. Studies that appeared to satisfy the inclusion criteria then underwent full-text screening. Both stages of screening took place in Rayyan. The second reviewer independently screened 20% at both stages. Discrepancies were resolved through discussion and consensus.

Data collection process

A piloted, bespoke Excel data extraction form detailed sixteen items, which were collected by the first reviewer and then checked for accuracy and completeness by the second reviewer, with discrepancies resolved through discussion and consensus. These included (a) author and year, (b) study design, (c) country of study, (d) source of participants and setting, (e) inclusion criteria, (f) timeframe and type of data collection e.g. retrospective, between date-date, (g) analysis details, (h) sample size, (i) socio-demographics, (j) exposure—how insecure status was measured, (k) timeframe of the exposure, (l) number of participants in insecure status, (m) outcome—how violence was measured, (n) time frame of violence, (o) country of violence, and (p) findings. In cases where data was not specified or disaggregated sufficiently within the study, corresponding authors were contacted to request the raw data or any available disaggregated data, with follow-up requests sent after two weeks. Of five corresponding authors, three responded. Two were able to provide the requested data.

The exposure was insecure status. The conceptualisation of insecure status was generated from a spectrum of statuses including no status, temporary statuses, and dependent statuses. There is no existing formalised definition of insecure migration status. Previous studies have demonstrated that migration is linked with precarity [9], and that power imbalances and dependencies linked to visa status produce specific vulnerabilities to violence [4]. Typically, large population level surveys will record country of birth, which means that this is frequently adopted as the indicator of immigrant status in quantitative research. Nevertheless, there is a difference between immigrant status—a binary category that relies on residing outside one’s country of birth—and immigration status, which refers to the category in which one has entered the country and the basis of their permission to remain. We adopted the conceptualisation of insecure status to link the characteristic of precarity or vulnerability to a population, allowing the effects of that precarity (rather than the specifics of each separate status) to be measured in a more robust way than is currently possible see S4 Appendix [S4 Appendix, 10].

The outcome was physical violence experienced while in insecure immigration status. The definition of violence that we adopted in this study followed that of the World Health Organization definition and typology of violence [46]. Physical violence in this systematic review included both interpersonal violence and state violence. Interpersonal violence might happen in the home, or in the community. It may be perpetrated by a stranger or an acquaintance but is affected by social relationships at the community level. Collective physical violence was also included in this review, in the context of policies that used physical coercion in their realization, such as forms of immigration enforcement. These policies are used by a collective actor (the state) against a collective that share a determined characteristic (lack of immigration status or in violation of immigration status). The WHO classifies this violence as social, political or economic. Thus, state violence in the context of immigration enforcement that is carried out by as a means of disciplining and removing people who do not meet the criteria for belonging designated by the state can be considered collective political violence. Research has identified coercive policies used at all parts of the immigration processes. This includes things like arbitrary detention; deportation [15, 28, 47, 48]; removal to unsafe locations [49–51]; torture; pushbacks to prevent border crossing even when this leaves people in particularly perilous conditions [52, 53]; maltreatment and abuse when in state custody; sexual violence and rape in state custody; and the use of restraint, assault, and brutality to achieve submission [54, 55]. If an immigration petition is rejected, state authorities might forcibly evict, detain, and remove people, and regularly leaves people without immigration status destitute. In this context the state is adopting policies that often employ physical violence or forms of coercion that result in physical violence; thus, we describe these policies as state violence. It should be noted that the academic literature finds an association between insecure migration status and structural violence as theorised by Galtung [56]. While this is important, it was beyond the scope of this research to locate, identify, and typologize all types of structural violence against people in insecure migration status. That is not to undermine the importance of recognising and addressing the relevancy of structural violence against migrants; time and space constraints required limiting the scope of the project.

Study risk of bias assessment

We conducted a detailed risk-of bias assessment of all the included studies, using an adaptation of the Joanna Briggs assessment tool for cross-sectional studies [57]. We assessed risk of bias for nine domains: definition of inclusion criteria, description of study subjects and setting, measure of insecure immigration status, identification of confounders, strategies for dealing with confounders, violence measure, reporting of raw data, reporting of association between insecure immigration status and violence. Risk of bias was assessed per study. As per the Joanna Briggs guidance, rather than attribute each study an overall score based on how many domains they met the criteria for, we reported the complete assessment [S4 Appendix]. The risk of bias was assessed independently by two reviewers and any disagreements were discussed, resolved and recorded. Because accurate data on people in insecure migration status is very difficult to obtain, we were cognizant of the difficulty in establishing accurate and representative data on violence for people in insecure migration status and assessed studies on the criterion of considering risk of bias in pre-existing data records and in interview and survey methods. For example, in studies where violence data were accessed from existing records of a health clinic or women’s shelter it is worth considering that there may be significant barriers to access, or barriers to reporting that affect people in insecure status, which has an impact on the accuracy or representativeness of data. In studies where data were collected through interview or survey, we looked for whether they reported the ways in which they considered and accounted for positionality of the researcher and potential bias imposed by the study setting. We also appraised the fitness of the measurement of insecure migration status for the study protocol, the method of measuring violence, and the strategies used to identify and account for confounding factors.

Synthesis methods

Meta-analysis was conducted using Stata 17 [58]. The raw number of participants experiencing physical violence (the numerator) and the total number of participants in the study population (the denominator) were extracted to calculate prevalence. All studies meeting the inclusion criteria and reporting disaggregated raw data (i.e., the numerator and denominator) or enough information to calculate the numerator and denominator, were included in the synthesis. Where studies reported a percentage and a denominator, the numerator was calculated by one of the reviewers.

A random effects model [59, 60] was used to determine an overall pooled prevalence estimate with 95% confidence intervals (CIs) on physical violence in those with insecure immigration status. Heterogeneity was assessed using the I² statistic. Forest plots were used to give a visual assessment of the pooled prevalence estimates, 95% CIs and weighting, produced by Stata.

Subgroup analyses [61] were not determined a-priori; several post-hoc, exploratory subgroup analyses were carried out, with the aim of exploring possible causes of heterogeneity, and of aiding interpretation of the results. These analyses included by gender as reported in the primary studies, perpetrator of the violence (community versus state versus individual), geographic region, contextual timeframe of violence, and immigration status.

Study selection

The database searches produced 14,421 records for screening before de-duplication. After de-duplication, there were 10,652 records. The abstract screening stage identified 1001 full-text reports, from which we included 31 studies [22, 31, 62–91]. All included reports and 23 relevant systematic reviews were tracked backwards and forwards for relevant citations. 27 additional texts were assessed for inclusion, of which 1 was included [Fig 1].

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. PRISMA flow diagram.

https://doi.org/10.1371/journal.pone.0300189.g001

Characteristics of studies

31 studies with 25,997 participants were included in this review (See Table 1). All were retrospective cross-sectional studies and all studies used non-probability sampling. The majority used convenience or purposive sampling, with the exception of Nakash et al. [71], which used consecutive sampling of a population of asylum arrivals. Of the 31 studies selected, 25 did not report a comparison group for insecure immigration status [22, 63, 64, 66–69, 70, 71, 73–77, 79, 80, 83–91]. Thirteen of the included reports included only female participants [63, 66, 70, 72, 73, 78, 80–82, 86, 87, 92, 93]. One study was exclusively male [76]. Four of the included studies did include female and male participants although were heavily biased towards male participants [74, 75, 84, 91]. Only three studies reported gender as a non-binary category [69, 85, 91]. While Couture-Carron et al. [92] met the inclusion criteria, because experience of violence was part of the inclusion criteria for that study, and all participants were in insecure status, it could not be included in the prevalence estimation.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Characteristics of included studies.

https://doi.org/10.1371/journal.pone.0300189.t001

Generally, there was a lot of diversity in the way violence was conceptualised and measured. Hadush et al. [86], Logie et al. [88], Ogbonnaya et al. [72], Okenwa-Emegwa et al. [73], Segneri [90], Stewart et al. [78] and Zadnik et al. [82] all adopted a validated tool to measure violence (see Table 1).

We collected data on violence occurring in 44 different countries, grouped by region: Europe [22, 62, 65, 68, 81, 91]; North America [64, 67, 69, 72, 74, 75, 78, 80, 82, 85]; Asia [31, 76, 87]; and Africa [70, 71, 77, 84, 86, 88]. Twenty-five of the included reports measured violence in a single country [64, 65, 67, 69, 70–72, 74, 75, 77, 78, 81, 82, 84–89, 91, 93, 94]. The remaining reports either specified a combination of locations in which the violence occurred or specified that the violence happened while in transit during the migration journey.

Studies measured insecure immigration status in a number of different ways. These included ‘undocumented’ [22, 64–67, 69, 72, 74, 75, 77, 80–82, 85, 89], ‘refugee /asylum seeker’ [62, 63, 68, 70, 71, 73, 83, 84, 86–88, 90, 91], ‘spousal /family visa’, ‘employment related’ [31, 76, 79]. Two studies measured more than one status but sufficiently disaggregated the data to allow for insecure status to be verified [78].

Risk of bias in studies

The majority of studies adequately defined their inclusion criteria (55%) and described their study subjects and setting (84%). Whilst some measured insecure immigration status using objective and standard criteria, the way other studies measured insecure immigration status was unclear (52%), causing us to question the validity for the purposes of our study. Most studies failed to identify and account for confounding variables (71%) and most provided insufficient or unclear information on outcome measurement for the purposes of our study (52%) [S3 Appendix].

Results of individual studies

The overall estimate of prevalence of physical violence for people in insecure migration status was 31.16% (95% CI 25.62–36.70, p < .00) [Fig 2]. However, there were high levels of heterogeneity (I² = 99.70%) which makes conclusive inferences about the prevalence of violence for people in insecure migration status difficult. Subgroup analyses were conducted to explore the heterogeneity.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. Prevalence of violence against people in insecure migration status.

https://doi.org/10.1371/journal.pone.0300189.g002

Subgroup analyses

We grouped the results to test our hypothesized expectations. We looked for prevalence of violence against people in insecure migration status by gender as a social determinant of health, and also by status type, by region, and by the time-frame in which violence occurred. Status type, region and time-frame all intersect with other social determinants of health that are not measured in this study. In each of these categories there was too much heterogeneity in the data to offer robust prevalence estimates from pooled data.

Gender.

Most of the studies reported exclusively binary gender categories, or reported only on a single gender. Only three studies reported non-binary gender categories [69, 85, 91]. The prevalence of physical violence estimate for men (35.30%, 95% CI 18.45–52.15, p < .00) was not significantly different from the estimate for women (27.78%, 95% CI 21.42–34.15, p < .00) [Fig 3], although the confidence intervals overlap.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 3. Prevalence of violence by gender.

https://doi.org/10.1371/journal.pone.0300189.g003

In total, eleven studies [63, 66, 70, 72, 73, 78, 80–82, 86, 87] were all exclusively interested in violence against women. Of these eleven studies, five were not specifically about IPV [63, 66, 70, 73, 80]. Stewart et al. [78] deals with violence associated with pregnancy, which includes but is not limited to IPV. Within the studies on IPV, it is likely that violence is underreported by people in insecure status because of the potential threat reporting poses to status.

Immigration status type.

The subgroup analysis by immigration status type was driven by the data available. We grouped research as follows: studies in which the population of interest were all in undocumented status at the time of violence; studies in which the population of interest were all in refugee or asylum seeker status at the time of violence (and therefore imply violence pre-migration as a potential confounding factor); those in a status related to their employment; and those where the status groupings are mixed or otherwise unclear while still meeting the exposure of insecure status as an inclusion criterion. The estimate of prevalence where insecure status was related to employment was 44.40% (95% CI 18.24–70.57, p < .00). The estimate of prevalence of violence experienced by people in undocumented status was 29.13% (95% CI 19.86–38.41, p < .00) and violence experienced by refugees and asylum seekers was estimated at 33.29% (95% CI 20.99–45.59, p < .00) [Fig 4]. The confidence intervals overlapped and there was no statistically significant difference between the estimates.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 4. Prevalence of violence by status type.

https://doi.org/10.1371/journal.pone.0300189.g004

Geographic region.

The subgroup of geographic region was also too heterogenous to offer a robust measure of prevalence. Indeed, the findings demonstrate diversity and complexity within geographic regions, although no sample can be considered fully representative of the regional grouping. The estimate of prevalence of violence in Asia was 56.01% (95% CI 22.47–89.55, p < .00). The three included studies involved Bangladesh [87], and several East Asian countries included in two studies [31, 76]. The sample is limited even compared to the other geographic regions. Europe 17.98% (95% CI 7.36–28.61, p < .00) and North America 19.53% (95% CI 8.30–30.77, p < .00) had lower point estimates [Fig 5], but confidence intervals were overlapping and there was no statistically significant difference.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 5. Prevalence of violence by region.

https://doi.org/10.1371/journal.pone.0300189.g005

Timing of violence.

Grouping studies by time-frame was guided by the timing of the violence recorded in the included studies. For example, migration journeys might vary in their duration, but represent a time of particular vulnerability. Similarly, the amount of time a person spends in state custody might vary, but as a period is defined by the characteristic of being incarcerated. This grouping also included ‘past 12 months’ to capture studies that shared this measurement characteristic but did not fall into one of the other categories; for example, Dias et al. [65] measured community violence, while Hadush et al., Islam et al., Logie et al., Ogbonnaya et al. and Vives-Cases et al. [72, 81, 86–88] all measured domestic violence. The ‘not specified’ category grouped all remaining studies where violence happened after arrival in the receiving country but did not include the time period or one of the aforementioned characteristics. The prevalence during the migration journey was estimated at 32.93% (95% CI 24.98–40.88, p < .00) [Fig 6].

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 6. Prevalence of violence by timeframe.

https://doi.org/10.1371/journal.pone.0300189.g006

Perpetrator.

The perpetrator groupings separated studies that specified state violence and IPV into defined categories. Prevalence of intimate partner violence attached to insecure status was 29.10% (95% CI 8.37–49.84, p < .00) and the estimate for state violence was 9.19% (95%CI 6.71–11.68, p < .00), but the data was particularly limited in the state violence category with only three included studies [64, 74, 75], and remained too heterogenous for a robust estimate [Fig 7]. As with other estimates, it should be noted that the confidence intervals overlapped.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 7. Prevalence of violence by perpetrator.

https://doi.org/10.1371/journal.pone.0300189.g007

Summary of main results

This systematic review and meta-analysis of 35 cross-sectional studies with a total of 26,116 migrants in insecure status found a prevalence estimate of physical violence 30.86% (95% CI 25.40–36.31, p < .00). When disaggregated by gender prevalence of physical violence was estimated as 35.30% (95% CI 18.45–52.15, p < .00) for men and 27.78% (95% CI 21.42–34.15, p < .00) for women. When disaggregated by insecure status type, prevalence of physical violence was estimated at 44.40% (95% CI 18.24–70.57, p) for employment-based migration, 33.29 (95% CI 20.99–45.59. p < .00) for refugee and asylum seeker statuses, and 29.13% (95% CI 19.86–38.41, p < .00) for undocumented statuses. When disaggregated by geographic region, prevalence of physical violence was estimated at 17.98% (95% CI 7.36–28.61, p < .00) in Europe, 19.53, 95% (CI 8.30–30.77, p < .00) in North America, 56.01, (95% CI 22.47–89.55, p < .00) in Asia, and 44.71% (95% CI 28.56–60.86, p < .00) in Africa. When disaggregated by the time during which the physical violence occurred, prevalence was estimated at 32.93% (95% CI 24.98–40.88, p < .00) during the migration journey, 9.19% (95% CI 6.71–11.68, p < .00) while in state custody, and 29.02% (95% CI 5.37–52.68, p < .00) during the 12 months previous to the study. When disaggregated by perpetrator, prevalence was estimated at 9.19% (95% CI 6.71–11.68, p) perpetrated by the state, 29.10 (95% CI 8.37–49.84, p < .01) perpetrated by an intimate partner, and 23.07% (95% CI 14.10–32.03, p < .00) perpetrated in the community.

Our analysis was informed by the social-ecological model of violence [102] and limited by the data on social determinants of violence reported in the included studies. The social determinants of violence that we discussed include gender and immigration status because these two risk factors were consistently reported across the studies. Other social determinants of violence (e.g., age, socio-economic conditions, social norms, laws/policy/institutions, health) might have contributed to the varying estimates, but data on these determinants was not available across studies. There was no consistency across studies in terms of variables reported, so we were unable to extract and include in our study.

While the confidence intervals for prevalence of violence r for women (27.78%, 95% CI 21.42–34.15, p < .00) and for men (35.30%, 95% CI 18.45–52.15, p < .00) in insecure status overlapped, it is important to note that this measure of physical violence does not capture structural or systemic violence. Intersectional characteristics such as socioeconomic status, ethnicity, education and community context which have a bearing on the measurement of gendered violence were not available for this study. Moreover, a scoping review [103] found that studies of gender-based discrimination and violence were experiential and focused on perceptions and opinions. This type of data was not captured within this systematic review. It is plausible to assume that prevalence of violence against women in insecure status is underestimated.

Women on spousal visas are subjects of the broader vulnerabilities that are connected with (insecure) immigrant status, such as avoiding surveillance and reporting, and with vulnerabilities connected to other co-occurring identity characteristics, such as patriarchal, racist, and gendered social structures [104, 105]. For example, Morash et al. [93] point to the important gendered disparities in status that contribute to the likelihood of abuse if immigration is sponsored by an intimate partner and particularly if the woman immigrated as a ‘picture bride’ (that is, they were selected from marriage based on a photograph rather than the development of a relationship or an in-person meeting). This indicates that a woman is being selected based on appearance or other known factors rather than her personhood, suggesting objectification. It is also worth noting that Ogbonnaya et al. [72] find no difference in the data for substantiated cases of domestic violence between Latina women with citizenship or legal residence and those who are in unauthorised status in the USA. They theorise that this indicates under-reporting on the part of those in unauthorised status because of evidence that there are alleged higher rates that remain unsubstantiated (unreported), and which cannot be explained by other cultural factors because immigrant women who are legal residents provide a control group for cultural factors [72].

This review suggests that physical violence is a widespread issue for people in insecure immigration statuses. The prevalence estimate for violence perpetrated by an intimate partner against people in insecure immigration status was 29.10 (95% CI 8.37–49.84, p < .01) and higher than the reported prevalence estimate for physical IPV in the general population (estimated at 23.1% for women in industrialised English-speaking countries [106]). It is worth noting that studies on IPV (which, in this sample of quantitative research are limited to the Bangladesh, Canada, Ethiopia, Spain, Uganda and the USA) find that there is a vulnerability to violence that can be connected to the vulnerability inherent in the dependent immigration status, and that this is an intersectional vulnerability, linked to gender and other social determinants of health such as ethnicity and community factors.

The prevalence estimate of physical violence associated with employment-based immigration statuses is high (44.40, 95% CI 18.24–70.57), yet these studies were located only in Southeast Asia among specified populations. Meyer et al. [31] estimated prevalence of physical violence among migrant workers at the Thai-Myanmar border at 24.62% (95% CI 21.31–28.25). Comparable data in Thailand data is only available for specific sectors; for example, workplace violence among nurses was estimated at 12.1% [107]. Suyanto et al.’s [79] study of Indonesian migrant workers estimated prevalence of physical violence at 37.50% (95% CI 32.90–42.34). A survey led by the International Labour Organization and the ‘Never Okay’ Project found that in Indonesia 70.93% of 1173 survey respondents has experienced violence and harassment at work. Where immigration is connected to employment, it is likely that failure to disclose violence is high due to fear of losing employment and immigration status, thus there is reason to believe even the figure of 44.40 is underestimated.

Physical violence relating to specific subgroups was subject to the same problems as the overall prevalence estimate. Overall, the research that looks at IPV in the context of insecure immigration status raised several points of vulnerability, which include increased likelihood of abuse based on immigration-related factors. These include increased stress levels, lack of community support, social isolation of victim [78, 81, 108, 109] and power disparity embedded in family-based visas [104, 105], and reduced likelihood of reporting to either victim-supporting organisations, migrant-supporting organisations, or the police [72, 93]. The reduced likelihood of reporting might be based on fear of losing status but also on other factors such as lack of knowledge of how and where to report, lack of understanding of implications of reporting, language difficulties, and social isolation. These issues are not isolated to spousal visas, but potentially affect all types of insecure statuses.

Prevalence of violence was estimated in the subgroup of legal status by the subcategories of ‘Undocumented’, refugee/asylum seeker, and ‘employment-related’. There was no sub-category for spouse or family-dependent because this data was not available as a disaggregated category in the included studies. While data within the groupings according to legal status was still too heterogenous to offer any pooled measure of prevalence, it is worth noting that this indicates that there is no meaningful pattern of violence attached to undocumented status when compared with other categories of insecure status. Similarly, it is not possible to draw a link from this data between violence pre-migration (associated with refugee and asylum statuses) and violence post migration. This suggests that it is worth investigating further what status trends and types of insecurity can be associated with high prevalence of violence. It is clear that insecure status produces vulnerability to violence, and that vulnerability is not limited only to people in undocumented status. People in other types of documented and regular immigration status that embeds a form of insecurity are vulnerable to violence.

The regional groupings again were too heterogenous to provide prevalence estimation. Furthermore, too few countries were represented in each category to say anything meaningful about the separate regions. Nevertheless, what is clear from these groupings is that there is a deficit of quantitative data on insecure migration in South America and in Asia (although this study reports only English language sources, which is a source of bias). More research is available focusing on Europe and North America, which is likely driven by data availability and research funding. Thus, we do not have a clear picture of the prevalence of violence for people in insecure migration status globally.

Conclusion

This review found that physical violence is a widespread issue for people in insecure migration statuses. It found that the prevalence of intimate partner violence against people in insecure status was higher than the rate for the population as whole. It found that people in undocumented statuses did not experience higher prevalence of physical violence than other types of insecure status.

The review suggested that better quantitative data is needed regarding insecure status and associated characteristics, and that the category of ‘foreign born’ is inadequate to measure phenomena attached to immigration status. The review suggested that qualitative review is needed to elaborate on the intersectional characteristics that may influence experience of violence when in insecure migration status, and in particular to enrich data on gender-based violence. It also suggests that there is a regional bias in available data, and that a multilingual review is necessary to better assess data deficits.