Global incidence of suicide among Indigenous peoples: a systematic review

We systematically reviewed findings on the incidence of suicide in Indigenous populations worldwide. We searched for studies that analyzed population-based data on suicide deaths, and included papers that reported crude or standardized mortality rates. Health science librarians were consulted about the design of the search strategy with the aim to capture all peer-reviewed literature. The search combined terms related to three concept areas: population (Indigenous), outcome (suicide mortality rates), and study design (observational). Term selection was based on previous systematic reviews and combined key terms adapted for each database and also Medical Subject Headings (MeSH) as applicable. The study protocol is available in Additional file 1: Supplement 1. Additional details about the methods are reported in Additional file 1: Supplement 2, including citations for previous reviews, a list of included terms, a description of the procedures used for study selection and eligibility criteria, and a complete list of databases and hand-searched review articles.

One author (NJP) performed online text word and MeSH searches for articles indexed in PubMed, MEDLINE, Embase, Cumulative Index of Nursing and Allied Health (CINAHL), PsycINFO, Latin American and Caribbean Health Sciences Literature (LILACS), and Scientific Electronic Library Online (SCiELO). A second author (KN) replicated the search in PubMed and obtained the same number of articles as the first author. We searched for studies in any language, indexed from database inception until June 1, 2017. We conducted a secondary search with a comprehensive list of terms for specific tribal groups, nations, and populations identified in previous reviews. As no additional studies were identified, this approach validated the primary search. We also searched the WHO’s regional medical literature indexes, Indigenous-specific online research portals, and journals focused on Indigenous health. We hand-searched the reference lists of included articles and previous reviews to identify other eligible studies. Additional file 1: Supplement 2 includes a list of all databases and hand-searched sources.

One author (NJP) imported the results into a reference management program and removed duplicates. Two authors (NJP and KN) read the abstracts and screened in papers if they (1) reported a population-based crude and/or standardized suicide rate, or count and population data; (2) reported a rate for an Indigenous population; and (3) used an observational design. We excluded articles that did not include an Indigenous population, focused only on a specific age, gender, clinical subgroup, or deaths from a specific cause (for example, firearms), or were not peer-reviewed. Articles were also excluded if they were iterations, program evaluations or experimental studies, not primary studies, from the gray literature, or used identical data sources as prior studies.

Although there is no international consensus on the definition of Indigenous, we used the United Nation’s working definition to assess study population eligibility [17, 18]. The UN's conceptualization of Indigenous involves self and group identification; a special attachment to and use of traditional land, distinct knowledge, language, and culture; distinct social, economic, and political systems; common ancestry with original territorial occupants; participation in maintenance and reproduction of distinct ethnic identity; and a non-dominant socio-political status [17, 18]. A paper was eligible based on this criterion if it reported an outcome for an Indigenous population, tribe, community, nation, or group, including papers that used the geographic proxy method. For the proxy method, census data is used to detect areas where Indigenous people are a majority population [29, 30]. We considered an area to be a proxy identifier if 80% or more of the population self-identified as Indigenous.

Two authors reviewed the full text of each paper and assessed eligibility based on inclusion criteria. At this stage, we excluded papers that did not report rates for the majority of the population (aged 15–65 years), did not conduct the primary data analysis, or provided rates in figures only and did not report count and population data. If two eligible articles used the same data source with a period of overlap, we included the article with the longer study period. During screening, full text review, and data extraction, we resolved disagreements through discussion or consultation with a third author. Translators helped assess non-English language articles and assisted with data extraction for four included studies. The following data was independently extracted by two authors (AL and NJP), then compared: citation, study design, country and region/community, Indigenous population, data source, standard population, number of suicide deaths, population count, crude and standardized suicide rates (overall and by gender and age group), comparative rates for a non-Indigenous or general population, and the measure of relative effect (incidence rate ratio).

We summarized all included studies in a table and reported counts, population, crude and standardized suicide mortality rates, and rate ratios. We calculated crude suicide mortality incidence rates for articles that reported only count and population data, and we estimated rate ratios when not otherwise reported by dividing the Indigenous population rate by the comparison population rate. To identify global patterns, we presented rates and rate ratios in tables and figures grouped by WHO region, country, population, and gender; we did not pool the data due to heterogeneity. We also reported on trends in suicide mortality over time and by age group; reported time trends reflect results from included studies, not pooled and recalculated rates. We modified the Newcastle-Ottawa Scale and used it to assess the quality of included articles. Additional file 1: Supplement 2 includes a description of the quality assessment procedures and scoring, and the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) checklist is provided in Additional file 1: Supplement 4 [31].

We extracted population-based suicide mortality rates from 93 papers (Table 2) and included gender-specific incidence data from six additional studies [5, 10, 36‐39]. Overall, suicide rates among Indigenous peoples varied at all levels of aggregation in both high-income and low- and middle-income countries, and spanned from zero to 187.5 deaths per 100,000 person-years (PY; Table 2). In high-income countries, national and provincial suicide rates among Indigenous peoples ranged from 1.7 per 100,000 in Brunei Darussalam [40] to 50.4 per 100,000 among Aboriginal and Torres Strait Islanders in Northern Territory, Australia [41]. Rates in high-income countries were highest among rural Indigenous populations and in sparsely populated regions such as the Arctic. Among low- and middle-income countries, Palawan communities in the Philippines had the highest crude suicide rates (134 per 100,000) [42], while Indigenous peoples in Malaysia [43] and some Pacific small island states such as Fiji had crude rates under 7 per 100,000 population. The number of suicide deaths used for rate calculations ranged from zero to 4219 (Table 2).

Incidence rate ratios were reported or calculated for 102 Indigenous populations in 69 studies. The results showed rate disparities in the majority of studies (Fig. 2), though 22 reported rate ratios below one. The rate ratios ranged from 0.04 in China [44] to more than 20 in Brazil [45] and Canada [30] (Additional file 1: Supplement 3, Table S4). Most Indigenous populations had higher suicide rates than comparison groups; disparities were widest in studies with small populations. One study reported a suicide rate of zero for an urban Indigenous population in Brazil compared the general population rate of 4.8 per 100,000 in the same city [46].

Suicide rates appeared to increase over time, especially in the latter half of the twentieth century, though reports were limited. Among studies with reported time series (n = 24), most (83%, n = 20) had fewer than 10 data points and covered an average of 19 years. A study in Greenland was the exception; it reported longitudinal data that showed a steady suicide rate increase among Inuit that began with the near absence of suicide in the early part of the twentieth century (2.4 per 100,000) and climbed exponentially to a rate of 110.4 per 100,000 in 2010–2011; the average number of suicides per year changed from less than one to 55 during this period [12]. Aboriginal and Torres Strait Islanders in Northern Territory, Australia experienced similar rate accelerations (6.1 per 100,000 in 1981 to 50.4 per 100,000 in 2002) [41], while incidence among Alaska Natives was relatively stable, though high, from the 1980s to the early 2000s [47, 48]. Indigenous peoples in the Micronesian islands experienced a sixfold increase in suicide rates between the 1960s and the late 1980s (from 4.3 to 25.8 per 100,000) [35], and one study reported slight rate declines for both Māori and non-Māori in New Zealand from 1996 to 2002 [5]. Annual rates tended to fluctuate in studies with small populations.

Men accounted for the majority of suicide deaths in all but four studies; only two of these four studies reported a greater number of suicide deaths among women [49, 50]. Studies with gender-specific crude and age-standardized rates (n = 35) ranged from zero to 75.5 per 100,000 among Indigenous women (Additional file 1: Supplement 3, Table S3). Suicide rates were higher among Indigenous men compared to Indigenous women, though rate differences were marginal among some Pacific populations [33, 51]. Suicide rates were also higher among Indigenous men than for men in comparison populations in all countries except Israel and Fiji. Outside of the relatively low rates among Indigenous men in these countries, estimates ranged from 19.5 among Sámi [13] to 248.7 per 100,000 among Inuit [30].

This study is a comprehensive synthesis of the published evidence on the global epidemiology of suicide among Indigenous peoples. Although it is the first review of this scale, our study has several important limitations. First, included studies varied their methods of identifying Indigenous populations. Self-identification is the gold standard in administrative and registry data [67]. However, this is a recent benchmark. Its uptake has varied internationally, and some countries do not identify Indigenous populations in health data at all [53, 67]. The majority of included studies relied on linkages with census or registry data, geographic proxies, or observer-determined assessments. These procedures are useful approximations, but they use varied definitions and tend to under-count Indigenous people, especially groups without legal recognition [29, 53, 67]. This can lead to ascertainment bias and underestimation of inequities [53, 67]. A second and related limitation is the under-representation of studies from low- and middle-income countries. In our review, we may have missed studies, particularly from the Global South, due to the conceptualization of Indigenous and the search terms used, which do not necessarily apply in all contexts. We attempted to limit this bias by searching databases focused on low- and middle-income countries and including non-English language papers.

The third limitation was that it was difficult to compare suicide rates between countries. Included studies were heterogeneous in population size, number of cases, aggregation, data source and outcome assessment, method of identifying Indigenous peoples, and coverage period. Many papers provided crude estimates only and did not report numerator and denominator data by age group, gender, or ethnicity. For studies with adjusted rates, different standard populations were used, and confidence intervals were rarely reported. Differences in analytic and reporting practices made it challenging to directly and reliably compare suicide rates across studies. To address this, we examined rate ratios to assess relative differences between Indigenous and non-Indigenous/general populations. This allowed us to estimate rate disparities, which were compared globally.

The fourth limitation was that studies reporting low suicide rates may be under-represented, which is a potential publication bias. It is unclear whether the lack of low incidence populations is related to the common finding of elevated rates of suicide among Indigenous peoples compared to non-Indigenous populations or, as we suspect is more likely, to the possibility that suicide rates are rarely studied when they are low. Additional low incidence reports may exist outside of peer-reviewed studies; however, these were not identified because we did not search the gray literature. The primary reason for excluding gray literature reports was the extensive volume of sources with variable quality and also the risk of over-including data from high-income nations where public reporting of mortality data is common and vital statistics infrastructure is of high quality. Nonetheless, we identified 23 papers that reported rate parity or had a rate ratio below one, but these tended to use older data. A related problem is that case studies tended to examine suicide clusters in small populations [42, 68]. The advantage of using localized data is the ability to contextualize a complex health issue. The disadvantage is that the potential to compare health status between multiple groups, across regions, and over time is reduced.

Our results substantiate previous work [16, 52] to demonstrate that elevated suicide rates are not universal among Indigenous people and debunk notions that Indigeneity increases risk for suicide. Our results also point to several gaps in knowledge about the epidemiology of suicide in Indigenous populations globally. The lack of published suicide data on Indigenous populations in low- and middle-income countries is a glaring absence. Previous studies noted a scarcity of Indigenous-specific data in the Global South overall [18, 53]. Poor infrastructure for death registration is a key limitation [1]. In the context of suicide, this is especially problematic, because countries in Asia, Africa, and Latin and South America are the homelands for the majority of the world’s Indigenous populations [18] and, at a national level, account for more than three quarters of all suicide deaths [1]. Suicide data in high-income countries tends to be of better quality than that in low- and middle-income countries; however, many governments do not include Indigenous or other ethnic identifiers in administrative health data, and do not routinely link census or Indigenous registries with national health datasets such as vital statistics. In Canada for example, the federal government does not know how many Indigenous people die by suicide in a given year. Globally, there is a critical need to strengthen capacity for surveillance in Indigenous suicide prevention.

National governments can take several steps to improve suicide surveillance in Indigenous populations. Actions should include efforts to enhance suicide data quality and standardized classification by improving vital registration infrastructure, especially in low- and middle-income countries, and integrating mortality data with monitoring of suicide attempts [1]. Countries should adopt an equity-based approach to data collection that includes Indigenous identifiers derived from self-reported sources and linked to registries or census data to address gaps in identification, and align Indigenous identification procedures with recommendations from the International Group for Indigenous Health Measurement, adapted for each national context [1, 53, 56, 67, 69]. Building inclusive, Indigenous-centered models of data governance in suicide prevention will be a critical element of strengthened surveillance. To achieve this will require national statistical agencies to not only consult Indigenous communities, organizations, and leaders about priorities, but to respect Indigenous rights to determine the parameters of data ownership, custodianship, access, and use [29, 32, 67].

Future research and global suicide surveillance efforts will be further strengthened with longitudinal and up-to-date national and state-level datasets that allow disaggregation and comparisons of outcomes in small areas and subpopulations by ethnicity [1, 17, 53, 56]. Overall, these actions will help maintain robust public health surveillance systems in order to monitor health status, increase knowledge about the social determinants of suicide, target interventions, and evaluate strategies aimed at reducing the incidence of suicide among Indigenous peoples worldwide [1, 56]. Increasing the visibility of populations that bear the greatest burden from suicide can help drive efforts to achieve the WHO and Sustainable Development Goals of reducing national suicide rates by up to 30% [1, 69].