Prevalence of suicidal ideation and attempted suicide in the general population of Ethiopia: a systematic review and meta-analysis

Background In Ethiopia, in spite of the high burden of suicide related-adverse effect, substantial variability in the reported prevalence of individual studies about suicidal ideation and attempted suicide; there is no national level epidemiological evidence. Thus, the present study aimed to determine the pooled prevalence of suicide ideation and suicidal attempt in the general population. Methods We followed the PRISMA Guidelines to report the results of the finding. Databases including: PubMed/Medline, SCOPUS, CINAHL (EBSCOhost), African Journal Online (AJOL) and African Indexed Medicus (AIM) were searched. Heterogeneity was assessed by Cochrane chi-square (χ2) and quantified by I2 statistics test. Sensitivity test and subgroup analysis performed. Publication bias was tested by funnel plots and Egger’s test. Effect size was calculated by random effects model. Results A total of 12 studies for suicidal ideation and 10 studies for attempted suicide were included in the study. The prevalence of suicidal ideation and attempted suicide were ranged from 1 to 55% and 0.6% to 14% respectively. The 12-month pooled prevalence of suicidal ideation and suicidal attempt were 9% (5–16%), I2 = 99.64%, p < 0.001 and 4% (1–8%), I2 = 98.11%, p < 0.001] respectively. The lifetime pooled prevalence of attempted suicide was found to be 4% (3–6%). We found evidence of significant heterogeneity for suicidal ideation [I2 = 99.64%, p < 0.001] and attempted suicide [I2 = 98.11%, p < 0.001]. The subgroup analysis could not identified source of heterogeneity. The sensitivity analysis showed that none of the point estimates was outside of the overall 95%CI for suicidal ideation and attempted suicide. No evidence of publication bias from the visual inspection of the funnel plot for suicidal ideation and [Egger’s test (P = 0.174)] and attempted suicide [Egger’s test (P = 0.318)]. Conclusion High prevalence of suicidal ideation and attempted suicide were observed in the general population of Ethiopia. These suggest the need of strengthening the awareness of suicidal behaviours and evaluate the effectiveness of the national health strategy in addressing the issues of suicidal behaviours.

By its nature, suicide is a complex phenomenon that involves the interactions of factors at different level such as: system level (access to means of suicide and health care), individual level (demographic features: female gender, younger age, lower education level, unmarried status) and community level (stigma and discrimination) [4,26]. For example, systematic reviews and meta-analysis showed factors associated with suicidal behaviours include: psychological, psychiatric, biological and stressful life events [4,27,28]. Recently published systematic reviews and meta-analysis among the general population in Europe showed that female gender, age over 65 years, unemployment, low social support, adulthood adversity, childhood adversity, family history of mental disorder, affective disorder, major depression, anxiety disorders and substance use were associated with attempted suicide [27,28]. In addition to these, the lack of standardized methods for data collection, the legal status of suicide and its impact on reporting, lack of resources, stigma, cultural criticism [2,7,8,23,[29][30][31] and the association of suicide with a criminal act and insurance policies have been affected the true figure of suicidal behaviours in the general population and compared its extent to an iceberg, where only the tip is visible; while the majority hidden under the surface [4,10].
In Ethiopia despite several inconsistent individual studies with ranged lifetime prevalence of suicidal ideation (1% to 55%), attempted suicide (0.6% to 14%) [17,[32][33][34][35][36][37][38][39], death by suicide (7.2 to 8.4 per 100,000) [40] and lack of accessible health care facilities, skilled man power, stigma, taboo and cultural criticism toward people with suicidal behaviours; priorities are given for other cause of death like infectious related death and the issue of national level epidemiological evidence of death by suicide has little attention. The contextual, epidemiological understanding of suicidal ideation and attempted suicide is an important first step to design culture specific suicide interventions. Therefore, the present systematic review and meta-analysis aimed to determine the pooled prevalence of suicidal ideation and attempted suicide in the general population of Ethiopia.

Study protocol registration and publication
This systematic review and meta-analysis was conducted according to a priori registered (CRD42018112836) and published protocol [41] with slight modification to the original designed published proposal. This modification is (i) related to the populations (i. e clinical, general population, high school and university students) and (ii) suicide-related outcomes (as suicide morbidity and suicide mortality). Because of the distinct characteristics of clinical patients and students from the general populations and suicide mortality as one end of suicidal behaviours from suicide morbidity as the other end, in the present study, we presented a separate manuscript for the general population and suicide morbidity (suicidal ideation and attempted suicide).
The findings in this report presented within, the first domain of suicide prevention activity [universal activities or interventions] that targeted the general public that has not been identified on the basis of individual risk [42,43]. According to a framework of Gordon's continuum of care model, prevention, treatment and maintenance were the three model of care [44,45]. These are (i) universal activities or interventions, the focus of this study that targeted the general public has not been identified on the basis of individual risk, (ii) selected or targeted interventions that aim to reduce the risk among specific high-risk groups and (iii) indicated interventions include the treatments of targeted individuals presenting with suicidal behaviour. This model was previously adopted to guide identification of population groups and individuals with differing prevention needs, and alignment of these needs with appropriate policies and programmes for suicide prevention [45]. Results were reported in accordance with the Preferred Reporting Items for Systematic Review and Meta-Analysis (PRISMA) [46].

Search strategy
A compressive search of electronic databases: PubMed/ Medline, SCOPUS, CINAHL (EBSCOhost), African Journal Online (AJOL) and African Indexed Medicus (AIM) were searched using keywords, Medical Subject Headings (MeSH) terms. Search strings were established using "AND" and/"OR" Boolean operators for each databases. Moreover, to avoid missing of relevant studies like preprints and grey literatures, web of science (Google

Selection of studies
All studies retrieved through search strategy were imported in to EndNote X7 (Thomson Reuters, New York, USA) and removed duplicated studies. The study selection process had three stages: the first stage was screened based on the title and abstract of retrieved references against the outcome. The second stage was based on the full-text to determine its applicability to the study aim/data's clarity. Finally, studies were screened based on the eligibility criteria. Two reviewers (YDG and BBB) independently screened the studies. Disagreements were solved by discussion.

Definition of concepts
Suicidal ideation refers to thoughts, fantasies and wishes about ending of one's own life [1,29,47]. Overlapping terms in the concept of suicide attempt: (i) attempted suicide/self-harm/suicidal intent, refer to a deliberated direct destruction of body tissues with a conscious suicidal intent, (ii) non-suicidal attempt/selfinjury, refer to a deliberated direct destruction of body tissues without a conscious suicidal intent [4,11] and (iii) para-suicide, is a term that does not refer to the intention and covers behaviors that can vary from suicidal gestures to serious attempted suicide but unsuccessful to kill oneself [48]. Most standard guidelines and evidence focus on self-harm irrespective of the intent. In line with other similar studies, in this study, suicidal attempt was defined irrespective of its intent and assessment tools [4,11,48]. In the current study, for the assessment of suicidal ideation and attempted suicide, (i) majority (75% of ideation and 90% of attempted) of the included studies used Composite International Diagnostic Interview (CIDI). In Ethiopia, CIDI had established reliability and acceptability [26].
In CIDI, a hierarchy of single item questions used for the assessment of suicidal behaviours: For example, suicidal ideation was assessed by the statement: "Have you thought of taking your life in the past 12 months?"; suicide plan (asked of those who had suicidal ideation): "Did you ever make a plan for taking your own life at any time in the past 12 months?"; attempted suicide (asked to those who answered they had ideation or a plan): "Have you attempted to take your own life in the past 12 months?", (ii) The remaining 25% of suicidal ideation was assessed by Self-Report Questionnaire-20 (n = 3). SRQ-20 had established reliability and acceptability test in Ethiopia as a screening tool to identify probable common mental disorders. The scale contains an item on suicidal thoughts. Similarly, the remaining 10% of attempted suicide used Suicidal Behaviour Questionnaire-Revised (n = 1).

Eligibility criteria
According to the PICOS acronym, the following inclusion criteria were used: Participants (P): all studies carried out among the general population, Intervention (I): not applicable, Comparison (C): not applicable, Outcomes (O): prevalence of suicidal ideation and suicidal attempt, Study design (S): observational studies (crosssectional and cohort/longitudinal) were included and Study setting: all studies carried out in Ethiopia were included in the study. Exclusion criteria were: (1) case reports, conference, reviews and abstracts were excluded (2). Studies conducted among clinical population (e.g. psychiatric, physical problems including HIV/AIDs), students and specific population group such as: one sex, prisoners, refugees and street (homeless), (3) For those studies reporting the overall results of suicidal behavior and not reported separately for suicidal ideation and suicidal attempt and (4) studies not conducted in Ethiopia were excluded.

Data extraction
Data were extracted from the eligible studies using a pre-conceived and piloted data extraction Microsoft Excel format. Data were simultaneously extracted by two independent reviewers (BBB and YDG). The included data items from each article include: name of the first author, year of publication, study area, study design, data collection tool, sample size and number of cases/prevalence including for each sex.

Quality assessment
The quality of each article was evaluated using the Joanna Briggs Institute (JBI) quality appraisal criteria adapted for reporting the prevalence data for cross-sectional studies [49]. This JBI quality assessment tool has nine items with response option of "yes", "no", "unclear" and "not applicable". These items include: sample frame, sampled participants, sample size, description of study subjects and setting in details, data analysis with sufficient coverage of the identified sample, use of valid method in the identification of conditions, measurements of outcome with a standard, reliable way, appropriateness of statistical analysis and the adequate of response rate. The overall quality of each study was determined by the summed score of all items to generate a study specific global score. Studies were included in analysis if the sum of the overall individual study score is ≥ 50% of the quality assessment checklist criteria's. Finally, based on the overall mean score (8) we categorized individual study as high quality (≥ 8) and low quality (< 8). Two reviewers (YDG and BBB) independently screened the studies. Disagreements were solved by discussion.

Data synthesis and statistical analysis
The extracted data into Microsoft Excel Database were exported into Stata 14 that we installed packages for meta-analyses online. Metaprop command in Stata employed to calculate the pool point prevalence estimate with 95% CIs by a random-effects model [50] using the Dersimonian and Laird transformed inverse variance method [51]. Test for Heterogeneity was performed using Cochrane chi-square (χ 2 ) and the I 2 statistics. I 2 value greater than 50% was considered as indicative of substantial heterogeneity [52]. Two separate meta-analyses were performed for the overall outcome (suicidal ideation and suicidal attempt). To assess the possible source of heterogeneity, sensitivity analyses [53] and subgroup analyses by sex, sample size, study setting, publication years and time frame were conducted. Publication bias was tested by the visual inspection of funnel plot [54] and egger test [55]. A p-value < 0.1 was considered as indicative of statistically significant publication bias. The findings of this study summarized and presented using texts, forest plots, tables and summery of descriptive statistics.

Results
The initial database search resulted in 2759 publications. Additional, 23 studies were located from the other sources. This resulted in 2782 records. Of these records, 1329 were excluded because of duplication. Of the remaining 1453, 824 studies were excluded by their title and abstract reading which were not related to the outcome and 579 articles were not full text. Finally, 37 publications were excluded based on the eligibility criteria and objective (Fig 1).

Study characteristics
In this study, a total of 13 studies were included in the systematic review and meta-analysis. Of these, 12 studies reported suicidal ideation and 10 studies attempted suicide. These studies were carried out in five different regions of the country: South Nation and Nationality People (SNNP) (n = 5), Addis Ababa (n = 2), Oromia (n = 2) and Harar (n = 1) and Tigray (n = 1). All studies used cross-sectional study design. The sample sizes of included studies ranged from 351 to 10,203 for suicidal ideation and 351 to 10,468 for attempted suicide. Studies were published between 1999 and 2020. For the assessment of suicidal ideation, the vast majority (75%) used the Composite International Diagnostic Interview (CIDI) and the remaining three studies used Self-Report Questionnaire 20 . Similarly, 90% of attempted suicide was assessed by CIDI and the remaining one study used Suicidal Behaviour Questionnaire-Revised (SBQ-R) ( Table 1).

Quality of included studies
The quality of included articles in this systematic review and meta-analysis is shown in Table 2. Eight-five percent of the studies used an appropriate sample frame. All studies (100%) used a standard instrument, described the study subjects, setting and employed appropriate statistical analysis. The list scored quality item was observed in the sample size (31%). The overall quality score of included studies had a mean score of 7.6 (ranged 7-9). Ten articles (77%) were high-quality and three studies were low quality.

Prevalence of suicidal ideation and attempted suicide Suicidal ideation
The prevalence of suicidal ideation was ranged from 1 to 55% (Table 1). The 12-month pooled prevalence estimate of suicidal ideation was 9% (5%-16%), [I 2 = 99.64%, p < 0.001, n = 12], (Fig. 2). No lifetime reported prevalence of suicidal ideation.We found evidence of significant heterogeneity [I 2 = 99.64%, p < 0.001, (Fig. 2)]. The subgroup analysis by study setting, sample size, publication year and quality of study could not identified source of heterogeneity; however, the highest pooled prevalence was considered among studies that carried out between 1999 to 2010 [19% (95% CI 5-39%)] ( Table 3). The sensitivity analysis showed that none of the point estimates was outside of the overall 95%CI. No evidence of publication bias from the visual inspection of the funnel plot (Fig. 3) and [Egger's test (P = 0.174)].

Eligibility
Record excluded based on inclusion criteria (n=37)

Included
Studies included in systematic review and meta-analysis (n=13)

Studies included in suicidal attempt (n=10)
Studies included in suicidal ideation (n=12)    could not identified source of heterogeneity; however, the highest pooled prevalence was observed among studies used CIDI assessment tool [14% (95% CI 4-30%)] ( Table 3). The sensitivity analysis showed that none of the point estimates was outside of the overall 95%CI. No evidence of publication bias from the visual inspection of the funnel plot (Fig. 6) and [Egger's test (P = 0.745)].

Sex distribution of suicidal ideation and attempted suicide
Regarding the sex distribution, a total of seven studies (one suicidal ideation and six attempted suicide) reported data on the sex distribution of suicidal behaviours. From this, the sex distribution of suicidal ideation was 2.7% for female and 2.9 for male. On the other hand, the pooled effect of six studies for attempted suicide showed that females were around 1.19 times (AOR = 95%CI 0.41-1.96, I 2 = 0%, p = 0.987) more likely experienced attempted suicide as compared to males; yet, insignificant (Fig. 7).

Discussion
To the best of the authors' knowledge, there is no previous systematic review and meta-analysis has been reported about the pooled prevalence of suicidal ideation and attempted suicide in Ethiopia. The observed prevalence of suicidal ideation [1% to 55%]) and attempted suicide [0.6% to 14%]) in this study is within the reported prevalence of suicidal ideation and attempted suicide across the world [10].    Fig. 2 Forest plot presenting 12-month pooled prevalence of suicidal ideation using random effect models with 95% CI  [26]. The possible explanation for the variation may be due to the difference in educational level, accessibility of information, socio-economic status, and availability of resource, community's attitude toward suicidal behavior and geographical areas. In Ethiopian, for family with natural death strong social support system has been practiced; yet, fear associated with the stigma, cultural taboo, cultural criticism affect community's attitude, practice and help seeking behaviours. The observed high prevalence of attempted suicide in female; yet, insignificant is consistent with study carried out in the general populations of China [57]. This may be attributed to the high prevalence of sexual violence and depression in female as compared to males or they might have the history of abortion which is associated with higher rates of suicide. This is supported by other evidences [58][59][60]. In Ethiopia, evidence showed the association of sexual violence, early marriage (before 18 years) with suicidal attempt [60]. Another possible explanation may be due to the variation in gender specific cause of suicidal behaviours. For example, a recently published systematic review and meta-analysis showed that eating disorder, posttraumatic stress disorder, bipolar disorder, violence, depressive symptoms and previous history of abortion as female specific risks for suicidal attempts [58, where, majority of these factors are the major problems in Ethiopia. Therefore, the finding of this study implies; the existence of high burden of suicidal ideation and suicidal attempt in the general population. This suggests (i) the need of community based screening particularly for those who have difficult to access healthcare facilities (ii) the needs of strengthening community awareness toward suicidal behaviours. Suicide is a tragedy that can be prevented only when one can stand for each other.
(iii) at the moment, however, the HMIS does not have a priority indicators for intentional harm, the WHO comprehensive action plan for mental health is suggested to report suicide rate as an urgent priority and (v) suggest the needs of evaluating the effectiveness of the national mental health strategy in addressing suicidal behaviours. This is supported by studies, which showed factors such as absence of psychotic, have unmanageable debts, selfemployed, higher levels of social problem solving were associated with the difficult to prevent suicide [13][14][15]25].

Strengths and limitations of the study
This is the first systematic review and meta-analysis about suicidal ideation and attempted suicide in Ethiopia, and use of reference lists and Google Scholar to include all the available studies. However, limitations like exclusion of clinical patient, students, prisoner and the observed heterogeneity in spite of the subgroup analysis may limit the representativeness of the study.

Conclusion
High prevalence of suicidal ideation and attempted suicide were observed in the general population of Ethiopia. This suggests the need of strengthening the awareness of suicidal behaviours. It also suggests the needs of evaluating the effectiveness of the national health strategy in addressing suicidal behaviours among non-clinical population.