Suicide rates and suicidal behaviour in displaced people: A systematic review

Background Refugees, and other forcibly displaced people, face mental distress and may be disproportionately affected by risk factors for suicide. Little is known about suicidal behaviour in these highly mobile populations because collecting timely, relevant, and reliable data is challenging. Methods and findings A systematic review was performed to identify studies of any design reporting on suicide, suicide attempts, or suicidal ideation among populations of displaced people. A sensitive electronic database search was performed in August 2020, and all retrieved studies were screened for relevance by two authors. Studies were categorised by the population being evaluated: refugees granted asylum, refugees living in temporary camps, asylum seekers, or internally displaced people. We distinguished between whether the sampling procedure in the studies was likely to be representative, or the sample examined a specific non-representative subgroup of displaced people (such as those already diagnosed with mental illness). Data on the rates of suicide or the prevalence of suicide attempts or suicidal ideation were extracted by one reviewer and verified by a second reviewer from each study and converted to common metrics. After screening 4347 articles, 87 reports of 77 unique studies were included. Of these, 53 were studies in representative samples, and 24 were based on samples of specific target populations. Most studies were conducted in high-income countries, and the most studied population subgroup was refugees granted asylum. There was substantial heterogeneity across data sources and measurement instruments utilised. Sample sizes of displaced people ranged from 33 to 196,941 in studies using general samples. Suicide rates varied considerably, from 4 to 290 per 100,000 person-years across studies. Only 8 studies were identified that compared suicide rates with the host population. The prevalence of suicide attempts ranged from 0.14% to 15.1% across all studies and varied according to the prevalence period evaluated. Suicidal ideation prevalence varied from 0.17% to 70.6% across studies. Among refugees granted asylum, there was evidence of a lower risk of suicide compared with the host population in 4 of 5 studies. In contrast, in asylum seekers there was evidence of a higher suicide risk in 2 of 3 studies, and of a higher risk of suicidal ideation among refugees living in camps in 2 of 3 studies compared to host populations. Conclusion While multiple studies overall have been published in the literature on this topic, the evidence base is still sparse for refugees in camps, asylum seekers, and internally displaced people. Less than half of the included studies reported on suicide or suicide attempt outcomes, with most reporting on suicidal ideation. International research networks could usefully define criteria, definitions, and study designs to help standardise and facilitate more research in this important area. Registration PROSPERO CRD42019137242.

understanding the scope of the problem so that interventions can be tailored to meet the needs of specific populations.
This systematic review aims to synthesise what is known about the rates and prevalence of suicide and suicidal behaviour among displaced people from published literature.

Methods
The protocol for this systematic review was registered with PROSPERO (CRD42019137242). In the review protocol it was planned to include studies of people who have been displaced because of natural disasters, but due to the large number of studies, estimates from these populations will now be reported in a separate systematic review. Initially, studies on populations affected by conflict but who have not been displaced were also considered for inclusion. However, the difficulty in defining these populations and the variable level of exposure to conflict, led the author team to decide against including these studies. Studies were also excluded if they reported prevalence estimates for returned asylum seekers or displaced populations (i.e. people not currently displaced), or for economic migrants.

Inclusion criteria
Study designs. Studies of any design that could provide an estimate of the rate or prevalence of suicide or suicide attempts, or prevalence of suicidal ideation, in populations of displaced people were considered for inclusion. This included observational studies (longitudinal population-based cohort studies, case-control, and cross-sectional surveys) and baseline rates or prevalence in randomised controlled trials of interventions targeted at the populations of interest. Case reports, reviews, and case series were excluded. Publication status was not used to determine eligibility.
Populations. Studies were included if they reported on populations (of any age) of forcibly displaced people, which we defined as people who have had to leave their homes in the context of an emergency because of a deliberate event such as conflict or war. This includes those who identified as refugees or asylum seekers (i.e. displaced people who cross international borders) as well as internally displaced people (i.e. people who remain in their own country). A list of common definitions for terms used in this review are provided in Box 1.
Outcomes. All studies reporting data on rate or prevalence of suicide, suicide attempts, and suicidal ideation among populations of displaced people were considered for inclusion. For this review we consider suicidal ideation to mean having thoughts or ideas about the possibility of ending one's life, a suicide attempt to be an attempt to die by suicide that results in survival, and suicide to be intentionally causing one's death. These outcomes were defined in each study by the authors of the primary studies. We considered suicide estimates measured by death registers, surveillance systems, autopsies (verbal and recorded), and medical records, and estimates of suicidal attempts and ideation measured using validated tools or survey questions.
As the design of the included studies varied, the source and type of data reported for these outcomes also varied. All relevant data for these outcomes was extracted and presented according to data source, host country, and prevalence period (i.e. lifetime prevalence, point prevalence). Where available, data were also extracted on comparative estimates between displaced people and host or general populations.
Search methods. A broad search strategy informed by experts in this area was developed to maximise sensitivity, combining medical subject headings and free text terms (see S1 Table) to identify relevant studies in the following databases: Searches were conducted in April 2019 and updated in August 2020. Searches were not limited by publication date, geographical location, or language. The reference lists of relevant articles and reviews were screened to identify additional studies.
Selection of studies. Two reviewers independently screened the abstracts of all studies identified by the search strategy. Following this stage, two reviewers also independently screened full texts of studies appearing to meet the inclusion criteria. Discrepancies were discussed among the reviewers, and where necessary a third reviewer assessed the study under discussion until a consensus was reached.
Data extraction and management. One reviewer extracted study characteristics and relevant outcome data from all included studies. A second reviewer cross-checked all extracted data and any discrepancies between the two reviewers were discussed. Relevant information extracted from each study included: population characteristics (such as ethnicity, age, religion, trauma exposure, and duration of resettlement), setting, study design, study dates, data Box 1. Definitions of subgroups of displaced people used in this review.
Refugees People who have fled their country because they are at serious risk of human rights violations and/or persecution. They include individuals recognised under the 1951 Convention relating to the Status of Refugees; its 1967 Protocol; the 1969 OAU Convention Governing the Specific Aspects of Refugee Problems in Africa; those recognised in accordance with the UNHCR Statute; individuals granted complementary forms of protection; or those enjoying temporary protection. Since 2007, the refugee population also includes people in a refugee-like situation.

Internally displaced people
People or groups of individuals who have been forced to leave their homes or places of habitual residence, in particular as a result of, or in order to avoid the effects of armed conflict, situations of generalised violence, violations of human rights, or natural disasters, and who have not crossed an international border.

Asylum seekers
Individuals who have sought international protection and whose claims for refugee status have not yet been determined, irrespective of when they may have been lodged.
collection method and source of data. For each of the three outcomes of interest, the number of participants analysed as well as summary statistics were extracted where available. Host countries/territories (where studies were conducted) were classified into the four World Bank income levels (i.e., low, lower middle, upper middle, high). Methodological quality. The search strategy was designed to capture the breadth of the available literature in this population. As a result, a variety of study designs were considered for inclusion in this review. This prevented the use of a single checklist approach to assessing methodological quality and required a more pragmatic approach. The assessment in this review was based on the study design and the source from which data were obtained. Using this pragmatic approach, we judged studies drawing on vital events registration, specific demographic surveillance, and longitudinal population-based cohorts to be of higher quality while studies drawing on data from retrospective cohorts/analysis and cross-sectional surveys were considered low quality.
Data synthesis. The following common metrics were calculated using the extracted data for each study, when possible, to facilitate comparison across studies, population groups, and settings: suicide rate per 100,000 person-years, prevalence of suicide attempt in percentage, and prevalence of suicidal ideation in percentage with 95% confidence intervals (CIs). Where data were also available comparing the rates or prevalence of the three outcomes in displaced people with the general population of the host country, relative estimates (relative risks or hazard ratios) with 95% CIs were calculated and presented.
We chose not to pool rates of suicide across studies due to the heterogeneity of study designs and used narrative syntheses with tabulated and textual descriptions systematically reporting results for the three aspects of suicidal behaviour. Comparative results with host populations were synthesised using the vote counting based on the direction of effect method [14,15].
The reporting of studies and results in this review are grouped into two main categories, based on their likely representativeness to the target population or to a target subgroup (for example, adolescents, pregnant women): 1. studies using representative or non-selected samples of the target group (that is, evaluating where possible the whole population or whole subgroup, as applicable), referred to below as "general samples", and 2. studies using selected samples of the target group based on their inclusion criteria that may confound the outcomes (for example, known high risk groups within the populations of displaced people, such as those already diagnosed with mental illness), referred to below as "specific samples".
Furthermore, within each category results are presented separately for the following five population subgroups: 1. refugees granted permanent asylum status, 2. refugees living in refugee camps or with only temporary protection, 3. asylum seekers, 4. mixed samples of refugees & asylum seekers, and 5. internally displaced people.
Due to possible overlap, if a study reports the population as asylum seekers living in camps, this was categorised under refugees living in camps.

Included studies
A total of 6899 records were identified by the electronic searches. Following removal of duplicates, 4347 abstracts were screened by two reviewers. Of these, 1107 full-text articles were considered potentially relevant. Finally, 87 articles reporting on 77 studies  plus 10 companion articles [93][94][95][96][97][98][99][100][101][102] were included in this review. Reasons for study exclusions are outlined in the PRISMA flow diagram (Fig 1). Five of the included articles were published in languages other than English. These were Danish, Dutch, two in German, and Spanish [50,79,83,90,99].
Of the 77 included studies, 53 were in general samples and 24 used specific samples. The types of data sources used for each outcome are presented below. Characteristics of the 53 included studies and their populations that used general samples are presented in Table 1, grouped by population subgroup. Additional characteristics are also reported in the corresponding outcome results tables. Study characteristics and results for the specific samples (n = 24 studies) are reported in S2 Table. The various samples evaluated by these studies included people diagnosed/referred for mental health conditions, intoxication-related emergency department admissions, forensic autopsies ordered by attorney's office, and emergency department admissions that required a forensic exam.

Suicide
Among the studies using general samples, 11 reported on suicide rates [20,33,37,42,49,57,62,72,73,79,89]. Suicide results, including a visual bar chart display and some characteristics are presented in Table 2. Six of these studies were in refugees granted asylum, one was in refugees living in a refugee camp, three were in asylum seekers, and one was in internally displaced people. Nine studies were conducted in high income level host countries (HIC; as per 2021  World Bank classification), while the study in a refugee camp was in Thailand (an upper middle-income country, UMIC) and the one study on internally displaced people was in Sudan (a low income country, LIC). The sample size of displaced people in these studies ranged from 4164 to 196,941. All six studies in refugees granted asylum and one of the studies on asylum seekers used vital events registration as the data source. The other studies on asylum seekers used specific monitoring data as the data source or a cross-sectional study design. The study in a refugee camp was a retrospective cohort, and the one on internally displaced people used a cross-sectional study design.
Suicide rates ranged considerably from 4.0 to 290 per 100,000 person-years across the 11 studies in general samples ( Table 2). In refugees granted asylum, suicide rates ranged from 4.1 to 25.5 per 100,000 person-years, while in asylum seekers suicide rates ranged from 4.0 to 51.2 per 100,000 person-years. The study in a refugee camp found a suicide rate of 16 per 100,000 person-years, while the study on internally displaced people (in Sudan, a LIC) had the highest suicide rate of 290 per 100,000 person-years.
Only eight studies with general samples also compared suicide rates in displaced people to the host population, including five studies in refugees granted asylum and three conducted in asylum seekers. Seven studies [20,37,42,57,62,72,89] reported data in formats that could be presented visually in forest plots, which were grouped by outcome measure and then subgrouped by population type (Fig 2).
Among the studies in refugees granted asylum, there was evidence of a lower risk of suicide compared with the host population in four studies [20,42,62,72]. The final study on refugees granted asylum [89] reported lower, higher, and inconclusive risk compared with the host population depending on gender of the refugees and their birth region. In contrast, among the studies in asylum seekers there was evidence of a higher suicide risk compared with the host population in two studies [50,57], including one study that reported age-standardised mortality ratios (SMR) of 2.8 [95% CI 1.5, 4.1] in males, and 1.5 [95% CI 0.0, 3.6] in females [50]. The third asylum seekers study was inconclusive about whether the risk was higher or lower, as it varied based on gender [37].
An additional study [31] on asylum seekers in detention in the UK estimated 211 suicides per 100,000, however, a rate estimate could not be calculated as total person-year was not reported.
Only two studies using specific samples were identified that reported on suicide rates; one study described forensic autopsies of deceased refugees [36], and the other study only reported on the proportion of suicides amongst causes of death in refugees [41] (S3 Table).

Suicide attempt
Among the studies using general samples, 15 [17, 18, 20, 30, 39, 47-49, 54, 55, 77, 79, 83, 85, 88] reported on the prevalence of suicide attempts ( Table 3). The results are ordered by timeframe for the prevalence estimates and then sub-grouped by population type. Six of these studies were in refugees granted asylum, three were in refugees living in refugee camps, one was in mixed refugees & asylum seekers, two were in asylum seekers, and three were in internally displaced people. Ten studies were from high income countries, and the other five were conducted in India (a lower middle-income country, LMIC), Jordan (UMIC), Sudan (LIC), and two in Colombia (UMIC). The sample sizes of displaced people ranged from 53 to 196,941.
Two of the 15 studies had a prospective cohort design, one used specific monitoring data as the data source, one used national statistics, while 11 studies had a cross-sectional design. The prevalence timeframe of the studies (i.e. either how far back participants were asked about the occurrence of a suicide attempt or the follow-up time) ranged from 1 week to 9 years.
The prevalence of suicide attempts ranged from 0.14% to 15.1% across the studies in general samples (Table 3). In refugees granted asylum, the prevalence of suicide attempt ranged from 0.14% to 15.1%, and in the refugees living in camps it ranged from 0.18% to 7.3%. In the studies on asylum seekers the prevalence was 1.0% and 2.9%, and in internally displaced people it ranged from 2.2% to 9.1%. The study in mixed refugees & asylum seekers found a suicide   attempt prevalence of 0.78%. The one study in a LIC (Sudan) did not find a higher prevalence than average.
Six studies in general samples compared the prevalence of suicide attempts in displaced people to the host population. Only five of these studies [20,30,54,77,81] reported data in formats that could be presented visually in forest plots (Fig 3). Two of these studies were in refugees granted asylum, one was in mixed refugees & asylum seekers, one was in internally displaced people, and one was a pooled report of mixed refugee populations across 21 countries [81].
In the two studies on refugees granted asylum, one reported a lower risk of suicide attempt than in the host population [20] while the other reported a higher risk [30]. The study of mixed refugee types [81] reported an odds ratio of 1.00 [95% CI 0.60, 1.67] indicating little to no difference between the refugees and the host populations. In one study [77] on mixed refugees & asylum seekers, they found a higher risk of suicide attempts. One study [54] on internally displaced people found a 4-fold greater risk of suicide attempt compared to the host population, RR 4.43 [95% CI 2. 18, 9.00]. In addition, one study [79] found asylum seekers had a 3.4 times higher rate of suicide attempts compared to the host population (unadjusted).
One study [87] reported a composite outcome of self-harm or suicide. Refugees granted asylum were evaluated over a one-year timeframe and prevalence was reported as 0.10%, the rate was 101.2 per 100,000 person-years, and the comparative result was RR 0.80 [95% CI 0.58, 1.08] (unadjusted) for self-harm or suicide compared to the host population.
An additional three studies reported only a composite outcome of self-harm or suicide attempt. One [72] was conducted in refugees granted asylum, and the study timeframe ranged from 5-14 years. Prevalence was 0.68%, and the rate was 8 [80] in asylum seekers assessed 10 detained families and found the prevalence of self-harm or suicide attempt was 29.4%.
Nine studies using specific samples of the target populations reported on suicide attempts (S4 Table).

Suicidal ideation
Among the studies with general samples, 37 [16-18, 21, 25, 30, 32, 35, 44, 45, 47-49, 52, 54, 55, 60, 61, 64, 66, 70, 71, 75, 77-80, 82-84, 86, 88, 90] reported on the prevalence of suicidal ideation ( Table 4). The findings were ordered by prevalence timeframe and sub-grouped by population type. 11 studies were in refugees granted asylum, 11 were in refugees living in refugee camps, eight were in asylum seekers, three were in mixed refugees & asylum seekers, and four were in internally displaced people. 22 studies were from high income countries, and 15 were Two of the 37 studies used specific monitoring data as the data source, two used national statistics, and 33 had a cross-sectional design. The timeframe of the evaluation ranged from the 'current state' to 1 year. Sample sizes of displaced people ranged from 33 to 8526.
Prevalence of suicidal ideation ranged from 0.17% to 70.6% across the studies in general samples (Table 4). In refugees granted asylum, the prevalence of suicidal ideation ranged from 0.24% to 35.0%, and in refugees living in camps it ranged from 1.2% to 32.3%. In asylum seekers the prevalence ranged from 0.17% to 70.6%, and in internally displaced people from 0.48% to 19.8%. The mixed refugees & asylum seekers studies reported prevalences of 3.1%, 23.8% and 33.9%. There was not a pattern of generally higher prevalence in low income countries across the 4 LIC studies.
Only nine studies [16,30,45,54,71,75,77,81] in general samples compared suicidal ideation in displaced people to the host population as displayed in Fig 4. Two of these studies were in refugees granted asylum, four were in refugees living in refugee camps with only temporary protection, one was in mixed refugees & asylum seekers, one was in internally displaced people, and one was a pooled report of mixed refugees across 21 countries [81].
Among refugees living in refugee camps, there was evidence of a higher risk of suicidal ideation compared with the host population from three studies [16,45], while one study reported a lower risk [75]. There was evidence of a higher risk of suicidal ideation among refugees granted asylum in two studies [30,71]. Individual studies reported higher risk of suicidal ideation in mixed refugee and asylum seeker populations [77], and internally displaced people [54]. The report [81] of mixed refugees across 21 countries reported little to no difference in risk of suicidal ideation compared with the host population (OR 1.00 [95% CI 0.80, 1.25]). An additional two studies in general samples reported a composite outcome of suicidal ideation or suicide attempt. One study [26] evaluated mixed refugees & asylum seekers, and the study timeframe was 1 month. Prevalence was 9.1% [95% CI 4.9, 15] for suicidal ideation or suicide attempt. The other [29] was conducted in asylum seekers, and the prevalence was 55.6% for this composite outcome. This study's timeframe was not reported.
Three studies only reported continuous outcome data on suicidal ideation. One [76] was conducted in refugees granted asylum but the study's timeframe was not reported. Mean score using the Suicidal Ideation Scale was 20.34 (SD 7.39). A study [91] in refugees with only temporary protection reported mean scores using the Personality Assessment Index's Suicide Ideation subscale were 49.5 (SD 10.9) in males and 53.8 (SD 17.0) in females (timeframe was not reported). The other [38] assessed asylum seekers, and the timeframe was 2 weeks. Mean score using the PHQ-9 symptom severity Item 9 for suicidal ideation was 0.98 (SD 1.11).
Sixteen studies using specific samples reported on suicidal ideation (S5 Table).

Discussion
This systematic review identified 77 studies that reported on suicide, suicide attempts, or suicidal ideation in displaced people. Estimates of suicide rates from the studies which are more representative of refugee and asylum seeker populations (i.e. general samples) ranged widely and appeared to differ based on the source of the data and the population subgroup being assessed. Estimates of the prevalence of suicide attempts or suicidal ideation also ranged widely and were inconsistently reported across studies, making comparisons across groups challenging. While numerous studies have been published on this topic, the evidence base is still sparse for the differing contexts of refugees in refugee camps, asylum seekers, and internally displaced people. Additionally, fewer than half of the studies reported on suicides or suicide attempts and almost one third of the studies were not representative samples of the target population. This systematic review has provided an overview of the populations, contexts, and study designs that have been used for research on suicide and suicidal behaviours in these populations, but further research is warranted to fill the gaps in knowledge around these specific populations.
The heterogeneity in study populations, designs, analytical approaches, and results mean that it is not possible to draw firm conclusions on the comparative rate of suicide in refugees and host populations. The results from different studies are inconclusive, with several studies suggesting that rates of suicide may be lower in refugees granted asylum than in the host populations [20,42,72,89]. These results should be interpreted with caution because the variability of the available data may be due to effect estimates being mediated by a range of mechanisms. These factors might include the local context (such as rates or prevalence of suicidal behaviour in the host population, as well as possible confounders such as poverty and deprivation), as well as possible mechanisms of action, such as the extent of physical injury, loss of housing, financial distress, and the level of social support that is provided. The length of stay in a host country has also been suggested as a potential mediator, with refugees having a lower risk once they have lived in a host country for longer than 10 years [42].
The strengths of this systematic review included developing a sensitive search strategy and screening all results in duplicate, as well as careful evaluation and reporting of the data according to study design and population. However, there are some limitations to this review, particularly around the sparse international evidence base which prevents us from drawing firm conclusions. The wide variety of study designs and approaches to reporting risks prevented meta-analysis of the results. This variety in study designs also means that we were only able to do a relatively high-level assessment of methodological quality based on the source of data for the studies included in this review. Further data on mediators or risk factors for suicidal behaviour in these populations could add to the evidence base and strengthen the development of prevention or management interventions. In addition, there is potential for overlap in the five population subgroups used in this review. It is often difficult to distinguish which of these subgroups a study population belongs to without access to further population characteristics from the primary studies.
Despite the sparsity of studies, there is evidence of a high risk of suicide and suicidal behaviour among refugees in camps and asylum seekers, which suggests that these groups require additional support and monitoring. Strengthening the evidence base around the risk of suicidal behaviour in these vulnerable groups is needed to further develop understanding of the factors that influence them. Many factors have been suggested to influence the risk of suicide and suicidal ideation in these populations, including a combination of socioeconomic disadvantage [93], exposure to potentially traumatic events [103], increased depression and anxiety [104], or a lack of appropriate and accessible care [39,105]. Improvements in understanding the suicide risk in these populations can be made by better routine data collection to enable consistent surveillance across countries. Countries hosting people affected by conflict, including refugees and asylum seekers, should ensure that these groups are identifiable in routine data [12,104]. Agreed methods for analysing and presenting data would support comparisons between exposed populations in different countries and settings and would aid in quantitative synthesis of the risk of suicidal behaviours.
Finally, a better understanding of mediating factors may also help to inform policy responses and potential interventions for refugees or displaced people. A recent review found only a limited number of suicide prevention or response programs were implemented for refugee or displaced populations [4]. The review recommended programs that were multi-tiered, focussing on multiple levels of suicide prevention. A comprehensive approach to suicide prevention should include adequate surveillance, identify risk and protective factors, and ensure evaluation of programs following implementation. Studies on prevalence are well placed to identify risk and protective factors that might explain some of the variability across populations and contexts. For example, studies should consider including measures of local context as well as potential mediators such as experience of physical injury, loss of housing, financial hardship, levels of social support, legal context, and access to mental health services [106,107].
Given the high risk of suicide in refugees in camps and asylum seekers, host countries should increase efforts to prevent suicide in asylum seekers, with actions that may span from speeding up asylum processes and improving social conditions of asylum seekers to offering better mental health prevention and management interventions. Given that prevalence of mental disorders is high in displaced people and that that these conditions appear associated with suicide attempts, there is a need to invest in prevention and care of mental disorders in this vulnerable group.
Supporting information S1