Prevalence of suicidal behavior in young university students: A systematic review with meta-analysis

Objective: to identify the prevalence of suicidal behavior in young university students. Method: a systematic review with meta-analysis of cross-sectional studies based on the Joanna Briggs Institute proposal, and carried out in the PubMed, Web of Science, Scopus, PsycINFO and LILACS databases and in the Brazilian Digital Library of Theses and Dissertations, with no language or year restrictions. A total of 2,942 publications were identified. Selection, data extraction and methodological evaluation of the studies were performed by two independent researchers. The meta-analysis was performed considering the random effects model. Results: eleven articles were included in this review. The prevalence variation for suicidal ideation was from 9.7% to 58.3% and, for attempted suicide, it was from 0.7% to 14.7%. The meta-analysis showed a 27.1% prevalence for suicidal ideation in life, 14.1% for ideation in the last year, and 3.1% for attempted suicide in life. Conclusion: the high prevalence of suicidal behavior, even with the considerable heterogeneity of the studies, raises the need to implement interventions aimed at preventing suicide and promoting mental health, especially in the academic environment.


Introduction
The University is a space for political, social and professional training that enables the construction of fundamental knowledge about a given area and favors the development of skills inherent to the professional's performance. The academic study phase requires making important decisions, in addition to being a time for new experiences, discoveries and friendships cycles that will require the student to be able to deal with new situations.
However, unsatisfactory adaptation to this context can lead to mental distress in the life of this young person (1)(2) .
In addition to these issues, it is highlighted that the academic setting can generate a competitive environment among the students. The requirement to excel, the excessive hour load of academic subjects and work, and unharmonious relationships with the professors can generate tension and overload, as well as trigger mental illness in the student, in addition to contributing to an increased suicide risk (2) .
Suicidal behavior is a complex phenomenon that includes ideation, attempted and consummated suicide, and is related to biological, psychological, social and environmental factors (3) . Attempted suicide is the manifestation of a process that develops gradually (4) . It is necessary to know the circumstances in which suicidal behavior arises to prevent the factors that cause it (5) .

Data from the World Health Organization (WHO)
evidenced approximately 800,000 deaths due to suicide worldwide in 2016, which represents an annual rate of 10.6 suicides per 100,000 inhabitants. In addition, among young people, suicide is the second leading cause of death in the 15-29 year-old age group worldwide, and one of the 10 leading causes of death in North America (6)(7)(8) . Thus, suicide is a social phenomenon (9) , a serious public health problem that needs to be addressed by civil society and by the public administration. It is an avoidable event through the early identification of suicidal behavior and efficient intervention strategies (8) . Therefore, it becomes necessary to know the prevalence of this behavior and the factors that are associated with its occurrence.

Identification of the suicide and suicidal behavior
rates favors the implementation of strategies to reduce this event (10) , which, combined with the recognition of the determining or risk factors for this phenomenon, enables action from a prevention perspective (11) , in addition to providing subsidies to devise evidence-based strategies (12) .
Several risk factors predispose the individual to a greater chance of developing suicidal behavior (8)(9) , the following among them: individual (childhood adversities, severe mental disorders, depression, personality disorders, drug abuse, physical health problems) and environmental (violence, socioeconomic inequality, lifestyle, lack of social support, media effects and access to lethal means), which combine and generate greater vulnerability to suicidal behavior (8)(9) . Therefore, this phenomenon must be understood from a multi-factorial perspective, since considering it only in its biological context makes it impossible for intervention measures to be effective (12) . Given the above, the objective of this study is to identify the prevalence of suicidal behavior in young university students. Web of Science 11/06/2019; 11/18/2020 (All Fields: ("university students" OR "undergraduate" OR "college students" OR "academics") AND All Fields: ("suicidal ideation" OR "attempted suicide" OR "suicide" OR "suicidal behavior") AND All Fields: ("crosssectional studies" OR "prevalence" OR "frequency") Scopus 11/06/2019; 11/18/2020 (TITLE-ABS-KEY ("university students" OR "undergraduate" OR "college students" OR "academics") AND TITLE-ABS-KEY ("suicidal ideation" OR "attempted suicide" OR "suicide" OR "suicidal behavior") AND TITLE-ABS-KEY ("cross-sectional studies" OR "prevalence" OR "frequency")) Brazilian Digital Library of Theses and Dissertations 11/19/2020 (All Fields: suicidal ideation AND All Fields: university students) (All Fields: suicidal ideation AND All Fields: prevalence) (All Fields: attempted suicide AND All Fields: university students) (All Fields: attempted suicide AND All Fields: prevalence)

Publication selection criteria
The publication selection criteria were designed to answer the aforementioned review question. Thus, the following inclusion criteria were adopted: original articles characterized as cross-sectional studies, carried out with undergraduate students aged 18 years old or over, with no restrictions on gender and geographic location; and studies that evaluated the occurrence of suicidal ideation and/or attempted suicide and/or suicide, using a probabilistic sampling technique.
As for the exclusion criteria, the following were adopted: studies that did not detail the age group or that included an age group under and over 18 in the same sample, but did not detail the results by age group that would support the extraction of suicidal ideation and/or attempted suicide for those over 18 years of age; research studies with graduate students; and surveys that did not use a validated instrument to assess suicidal behavior or that presented an incomprehensible methodology.
To expand the number of studies, no date or language restrictions were applied; in addition to that, the reference lists of the eligible studies and in the Brazilian Digital Library of Theses and Dissertations were consulted, whose search strategy is described in Figure 1.
After searching the databases, the publications were exported with the aid of the Zotero reference manager and duplicates were removed.

Instruments used and research variables
The JBI standardized instrument was used to extract data from prevalence and incidence studies, which contains information about the following: condition measured; measurement method; and characteristics of the participants and of the study (13)(14) .
In addition, to assess the methodological quality, the JBI checklist for prevalence studies (13-14) was used, composed of nine items whose answers can be: yes, unclear, no and not applicable. It was decided not Rev Latino-Am. Enfermagem 2021;29:e3495.
to exclude articles during the methodological quality assessment stage. The review team discussed each assessment item for each study design included in relation to what was considered acceptable for the review. In order for the process to be transparent, the decision not to establish a cutoff point for inclusion was made before starting the critical assessment and was agreed upon among all the reviewers. This decision is consistent with the JBI guidance in noting that cutoff scores are generally advised against. The JBI recommends presenting the critical evaluation results for all the questions using a table rather than summarizing them with a score, which was presented in the manuscript in the Results section to ensure quality and transparency of the writing.
In this study, in order to assess item 3 of the aforementioned checklist that deals with the adequate minimum sample size for a prevalence study, the JBI recommends the use of sample calculation by means of the following formula: , where Z 2 α : 1.96; P (prevalence of the phenomenon of interest): 20.0% or 0.20; and 5.0% (0.05) sample error (d). Thus, studies with a minimum sample of 246 participants were considered adequate.
As for item 5 (Was the data analysis conducted with sufficient coverage of the identified sample?), a minimum sample size of 246 was considered; as well as probabilistic or census selection of the participants in order to minimize selection bias; and a 70.0% (18-19) minimum response rate which, in turn, was also a parameter for item 9.

Data treatment and analysis
The studies were critically evaluated by two independent researchers. In case of disagreement, a third reviewer was consulted. Initially, titles and abstracts were read and then the text was read in full. Subsequently, the eligible studies were evaluated for their methodological quality, using the JBI Critical

Appraisal Checklist for Studies Reporting Prevalence
Data instrument (13)(14) . The instrument evaluates the following criteria (13)(14) : appropriate recruitment of the participants; adequate sample size; detailed description of the subjects and of the study setting; data analysis with sufficient sample coverage; use of valid methods to identify the condition; measurement of the phenomenon in question in a standard way for all the participants; satisfactory response rate; and adequacy of the statistical analysis.
The evaluation, extraction, synthesis and metaanalysis stages were carried out with the aid of JBI's System for the Unified Management, Assessment and Review of Information (SUMARI) software. After evaluating the methodological quality, the data were extracted with the aid of the JBI instrument (13)(14) . In addition, information was collected about the following: data collection locus, gender and age group of the participants, instrument used to assess suicidal ideation, attempted and consummated suicide and prevalence of suicidal behavior, in addition to the statistical analysis technique. It is noted that the first two stages were performed by two independent researchers, previously trained to minimize evaluation and extraction errors.
To perform the meta-analysis, data on the prevalence of suicidal ideation and attempted suicide were organized as follows: ideation in life, during the year and over the last week; attempted suicide in life, during the year and over the last week. Subsequently, the meta-analysis was carried out considering the random effects model.

Ethical aspects
As this is not a research study involving human beings, it was not necessary to submit it to any Committee of Ethics in Research with human beings.

Results
A total of 2,942 studies were identified in the databases, records and other methods, of which 860 were excluded for being duplicates and 1,837 for not meeting the eligibility criteria, after reading the titles and/or abstracts. Thus, 245 articles were selected for full-text reading, of which two were not available for access, resulting in 243 articles for full reading. After analyzing the eligibility criteria, 232 studies were excluded, resulting in 11 articles that comprised this review (20)(21)(22)(23)(24)(25)(26)(27)(28)(29)(30) .   showed a response rate below 70.0%. It is noteworthy, however, that all the participants were selected in a random/probabilistic or censused manner. Further details are described in Table 1.
The studies included met most of the JBI checklist's criteria for prevalence studies. However, two research studies did not meet the minimum sample size criterion (n<246); one study did not clearly present the research participants and locus; three presented data Abdu, 2020 (20) Menezes, 2012 (26)  According to data in Table 2, most of the research studies (n=8; 72.7%) presented a high response rate, with the exception of three surveys whose rates were from 55.0% to 61.0% (21) , 8.0% (24) and 11.0% (25) , respectively.
Furthermore, two surveys did not report the percentage of participants who answered the research instruments (27)(28) .
Therefore, calculation for these two studies was performed, based on the expected and achieved sample size disclosed by the authors.
As for the undergraduate course, four studies (21,26,29,30) evaluated suicidal behavior in health sciences students, of which two estimated the prevalence in undergraduate Nursing (21,29) , two in medical students (21,26) and one in Pharmacy students (21) .
The suicidal ideation rate in the Nursing course varied from 12.3% (in the last week) (21) to 21.3% (in the last year) (29) ; for the Medicine course, the variation was from 13.4% (in the last week) (21) to 18.4% (in life) (26) ; and the ideation rate found in the Pharmacy course (in the last week) was 12.3% (21) .
Considering the variability of instruments used, the different sample sizes and the different countries where the studies were developed, it is expected that there will be wide heterogeneity across the studies. However, the main objective of this review was to provide a general overview of the prevalence of suicidal behavior (ideation and attempted suicide); thus, even in the presence of high heterogeneity, it was decided to perform and present the meta-analysis.
The prevalence of suicidal behavior varied widely. The study carried out in Ethiopia (20) presented the highest rate of suicidal ideation, with a quite expressive percentage, and the one conducted in Mexico (22) showed a lower percentage. However, when the recall period included the presence of suicidal behavior in life, the rate in Mexico increased significantly.
The ideation rate presented in the study conducted in Ethiopia was higher than the one found in other surveys that evaluated suicidal ideation in young individuals (34)(35) . The prevalence found was close to studies that evaluated suicidal behavior in patients with mental disorders (36)(37) , whose ideation rate tends to be higher due to mental illness. There are few research studies on suicidal behavior in Sub-Saharan Africa, which can be related to limited resources, as well as to the stigma associated with suicide in some countries of this region (38)(39) . In these countries, among them Ethiopia, there is difficulty accessing mental health services, with shortage of professionals and mental health care institutions centralized in capital cities (38,40) .
The scarcity of data on suicidal behavior leads to underreporting, increased stigma and difficulty accessing the health system (38) . Middle-and low-income countries account for approximately 75.0% of the number of suicides in the world (7) ; therefore, it is important to recognize the role of the sociocultural and economic factors in suicide, which can be expressed in different ways according to gender, reflecting social roles (41) .
validity of these instruments in the Brazilian context can be relevant.
Most of the studies in this review included students from the health sciences, most notably Medicine and Nursing. However, there is evidence of high prevalence of suicidal ideation in students from different knowledge areas, with a percentage of 9.9% in the last 30 days.
The associated factors were sexual orientation, suicide attempts in the family and presence of depressive symptoms (53) . A study conducted with students from two universities in South Africa found a 3.9% prevalence of attempted suicide, with an increased risk associated with students who identified themselves as "black-skinned" and "female" (54) . The publication bias stands out as a limitation of this review, since not all the scientific databases were investigated, as well as the fact that only studies available for free were included. However, some strategies were used to try to reduce the bias regarding the design of the studies that comprised the review, such as the inclusion of research studies that used validated instruments to minimize overestimations or misclassification errors of single-item self-report questions; as well as the inclusion of those that used probabilistic samples to obtain more valid estimates.

Conclusion
The results of the meta-analysis showed high heterogeneity in the studies included in this review.
However, the descriptive findings pointed to a high prevalence of suicidal ideation and attempted suicide in young university students, more pronounced in women.
The results show the need to implement health promotion public policies that consider the mental health approach in university spaces.
It is necessary to emphasize that, although suicidal ideation is a risk factor associated with suicide, research studies are needed with larger samples and with application of more than one instrument to measure suicidal behavior, both for comparison purposes and to better understand the phenomenon in this population.