Adolescent suicide attempts in three diverse island nations: patterns, contextual differences and demographic associations

Most epidemiological studies on suicidal behavior have been focused on high-income country settings. This study examine factors associated with suicidal behaviors among school-attending adolescents in three island nations. In this secondary analysis of the publicly available 2015 nationally representative GSHS data, we tested demographic, social, and behavioral attributes using multivariable logistic regression to association with suicide attempts. Within the recall period, 13.6% of participants reported having attempted suicide one or more times in the Cook Islands, 10.8% in Curaçao, and 9.8% in East Timor. In the Cook Islands, suicide ideation (AOR = 19.42, 95% CI = 9.11–41.41), anxiety (AOR = 2.51, 95% CI = 1.08–5.82), physical bullying (AOR = 3.3, 95% CI = 1.10–9.91), and cigarette smoking (AOR = 3.82, 95% CI = 1.38–10.54) were associated with suicide attempts. For Curaçaoo, suicide ideation (AOR = 7.88, 95% CI = 5.20–11.95) and suicide planning (AOR = 7.01, 95% CI = 4.24–11.60) were statistically significant. While for East Timor, suicide ideation (AOR = 4.59, 95% CI = 2.14–9.88), suicide planning (AOR = 3.36, 95% CI = 1.76–6.29), bullying victimization (AOR = 2.69, 95% CI = 1.02–7.12), and serious injuries (AOR = 2.22, 95% CI = 1.31–3.74) were statistically significant. Suicide attempt is relatively common in each of the three island nations. The socioeconomic context of adolescents might play a significant role in moderating suicidal behavior. Therefore, prevention efforts should be grounded in view of geographic, demographic, and socioeconomic contexts of the populations at risk.


Introduction
Suicide claims the lives of more than 800,000 people globally each year [1,2]. Among adolescents aged [15][16][17][18][19] years old, suicide is the third leading cause of death. While data from low-and middle-income countries (LMICs) typically underestimate the problem, available evidence suggests that the consequences of adolescent suicide attempts (SA) are prevalent in LMICs, where over 79% of global suicide deaths in 2016 occurred [2]. While numerous studies regarding suicidal behavior (SB) among adolescents have been done in mostly high-income western and Asian countries with large domestic populations [3][4][5][6][7][8], few studies exist from LMIC island nations.
The populations of the islands in our study-Cook Islands, Curaçao, and Timor-Leste are 17,564, 164,093 and 1,318,445, respectively, as of 2020 [9]. Persons aged under 24 years make up significant proportions of the overall populations-34.82% (Cook Islands), 33.06% (Curaçao), with Timor-Leste having a very young population overall with 60.28% [10][11][12]. Island nations that are not particularly subject to high rates of migration inflows In Curaçao, the number of victimizations reported by in school adolescents was associated with both mental health and health risk behaviors [13]. In Timor-Leste, suicidal behavior among school attending adolescents has been previously associated with unintentional injuries [14]. In the Cook Islands there have been calls for the prioritization of mental health and suicide prevention research both of which remain underfunded [15]. The present study aimed to examine the social and demographic correlates of suicidal behaviors among schoolattending adolescents in three island settings.

Main text
Publicly available data from nationally representative surveys conducted in the year 2015, from three island countries of Cook Islands, Curaçao, and East Timor were used for secondary analyses. Detailed information on the data collection methods, questionnaire procedures, response rates and data is available at http:// www. cdc. gov/ gshs/. Information was missing for age, sex, and both age and sex, for 3, 5, and 2 records, respectively in the Cook Islands; 10, 27, and 3 records in Curaçao; and 74, 202, and 57 records in East Timor. No cases were excluded to ensure a correct design-based analysis.
SA as a dependent variable was derived from a question in the GSHS: "during the past 12 months, how many times did you actually attempt suicide?" For this analysis, participants were classified as having attempted suicide if they reported having attempted one or more SA during the recall period. If no SA was reported, participants were classified as not having attempted suicide; for 3 records in the Cook Islands, 59 records in Curaçao, and 29 records in East Timor, this information was missing. Fifteen independent variables at the individual level were considered (age, sex, suicide ideation, suicide planning, anxiety, loneliness, bullying victimization, physical bullying victimization, involvement in physical fights, serious injury, early sexual debut, alcohol use, physical attack, marijuana use, and cigarette smoking), and six at the social level (presence of supportive parental figures, presence of helpful peers, the extent of the social network, parental smoking status, people smoked in presence, and food insecurity). Cook Islands GSHS 2015, did not ask questions about sexual experiences. Details on variable creation appear in supplement (Additional files 1, 2: Tables S1, S2).
Differences between SA involvement among the variables were screened for statistical significance using the survey version of the chi-square test, which is a designadjusted version of Pearson's chi-square test for categorical variables, and the design-adjusted version of the t-test for continuous variables (age and number of friends). We then created two binary logistic regression models. These were intended to model the ability of the selected independent variables to predict the dichotomized SA variable. The first model adjusted only for age and sex. While the second model included all variables significant at the bivariate level. We report the measures of association as adjusted (aOR) and unadjusted (OR) odds ratios along with 95% Confidence intervals (CI). Stata 16 (StataCorp, 2019) was used for analysis. All proportions-expressed in percentages-are weighted.

Results
Within the recall period, 13.6% (unweighted count: 98) of participants reported having attempted suicide one or more times in the past 12 months in the Cook Islands, 10.8% (unweighted count: 296) in Curaçao, and 9.8% (unweighted count: 377) in East Timor. In the Cook Islands and Curaçao, most suicide attempters were female (53.4% and 61.0%, respectively). While in East Timor males comprised 55.7% of suicide attempters. Table 1 shows the weighted distribution of selected factors according to suicidal behavior. The bivariate analyses show that in the Cook Islands, age, sex, food deprivation, parental tobacco use, people smoked in presence, and supportive parental figures were not statistically significantly associated with involvement in physical fights. In Curaçao, age and physical bullying were not statistically significant. In East Timor, age, sex, people smoked in presence, supportive parental figures, and helpful peers were not statistically significant. Table 2 adjusts for age and sex while Table 3 shows the final multivariable model. For the Cook Islands, suicide ideation, anxiety, physical bullying, and cigarette smoking were found to be statistically significantly associated with physical fighting status at p < 0.05; while at p < 0.01, only suicide ideation was found to be significant. For Curaçao, suicide ideation and suicide planning were statistically significant at p < 0.05; and the same was also found to be statistically significant at p < 0.01. For East Timor, suicide ideation, suicide planning, bullying victimization, and serious injury were statistically significant at p < 0.05; while at p < 0.01, suicide ideation, suicide planning, and serious injury were found to be significant.

Discussion
An inquiry into the mental health and health behaviors in nations with young populations, as in the context of the current research, has implications for economic and social advancement. The rates of SA among in-school adolescents in the Cook Islands (13.6%), Curaçao (10.8%), and East Timor (9.8%) were similar to the reports from many high-income countries [16] but higher than those from the USA and Canada [17]. The rates were, however, lower than those of sub-Saharan Africa [17]. Differences in income levels and standards of living might explain this variation [18]. Stable economic and social settings may mitigate the expression of unhealthy behavior among young people. The attributes of in-school adolescents in the three island nations are similar despite their diverse history, culture, economic, and social contexts. The islands have a similar age and sex distribution, and marijuana use and physical bullying were low. However, the individual nations had some distinctive characteristics. Tobacco use, physical attack, and food deprivation were more prevalent in East Timor and the Cook Islands than in Curaçao. Also, early sexual debut, alcohol use, and supportive parental figure were more common in Curaçao relative to the other two Islands. Curaçao is a high-income economy with a higher standard of living. It is ranked 27th in the world in terms of the Gross Nominal Domestic Product per capita (nGDP per capita). Curaçao's nGDP per capita of United States Dollars (USD) 47,020 far exceeds those of the Cook Islands (USD 17,797), and East Timor (USD 2422) [19][20][21][22]. Previous studies have suggested negative and positive adolescent attributes are disproportionately distributed in high and low-and middle-income countries (LMICs) [23][24][25][26][27]. The association of adolescent suicidal ideation and planning with suicidal attempt corroborates findings from diverse other countries and contexts [16,[28][29][30][31]. Nevertheless, suicidal ideation does not always result in suicide attempts [32,33]. With prevention in mind it is crucial to investigate mechanisms facilitating the progression from ideation to attempt. Many hypotheses on this topic exist. The three-step theory (3ST) represents contemporary understanding within the ideation-toattempt framework, pointing out that the capacity for SA, consisting of dispositional, acquired, and practical variables enabling the SA capability, is the main factor for transitioning from suicidal ideation to attempt. This is consistent with the interpersonal theory of suicide (IPTS). The IPTS additionally indicates the contribution of the exposure to painful and provocative events (PPEs), leading to habituation to pain and fear, to SA capacity. PPEs may also explain the significant association between SA and physical bullying and serious injury shown in our   14:464 results. The lack of association between suicide planning and SA in the Cook Islands is surprising and would benefit from further exploration to determine the responsible contextual factors. Also, in the Cook Islands, anxiety, physical bullying, and smoking are associated with SA, while helpful peers are protective. These findings align with reports of previous studies [16,18,31,[34][35][36].
Although there was no significant association between SA and involvement in physical fights or being physically  attacked, bullying was associated in Cook Islands. In East Timor, we found that bullying victimization and serious injuries had a statistically significant association with SA in keeping with previous findings [31,36]. Previous studies have researched different forms of bullying, including verbal harassment, physical aggression, and cyberbullying, and which were all suggested to be associated with suicidal behavior [37][38][39][40][41]. However, Kodish et al. pointed out that when only SA was considered, only verbal bullying showed association [37]. It is possible that the emotional distress caused by power imbalance and being isolated among peers, not necessarily physical injury, is linked with SA as far as bullying is concerned. Adolescents' lack of experience in coping with interpersonal conflicts and emotional fluctuations can contribute to impulsive and self-destructive behaviors as well. Furthermore, other researches indicate that mediators like depression, anxiety, low self-esteem, loneliness, and hopelessness, affect the association between bullying and suicidal behaviors during adolescence. These mediators impact directly on psychological health, but may also result from mental ill-health [34]. The victims of bullying may perceive a lack of safety and belonging with reduced social support in the school environment, potentially heightening anxiety and loneliness. In a study of young adults aged 14 to 24 years, anxiety disorder was identified as the prime risk factor for SA among various other illnesses [42,43]. Loneliness is associated with an increased risk of substance abuse [44].
In the Cook Islands, cigarette smoking had a significant association with SA. In a 2016 meta-analysis, a significant association between current smoking and suicidal behaviors was found [45]. The cumulative results indicated parental tobacco use and cigarette smoking have a low prevalence in Curaçao, while both factors showed high prevalence in the Cook Islands and Timor Leste, where almost half of the respondents have experienced people smoking in their presence in the preceding 7 days. In a way, the smoking behaviors of the youths might have been influenced by that of family members and significant others. Measures are required to strengthen awareness among the parents on how their behaviors influence young people.
Gender had no statistically significant relationship with SA. This finding contradicts existing literature demonstrating that SA and gender are associated, with most studies suggesting the males are more likely to attempt suicide and self-harm [18,46]. However, a few authors have reported the opposite [16,47,48]. Early sexual debut did not play a role in SA in the three Islands under consideration, although it showed a high correlation in Brunei and Malawi [16,31]. These differences highlight the probable effect of context in moderating the factors that underlie SA. An in-depth understanding of adolescents' characteristics and socioeconomic circumstances should precede policy-making, strategy formulation, and program implementation concerning SA.

Conclusion
Despite population and socioeconomic differences between the studied countries, SA is common with similar rates in each country. Adolescents in the three Islands Only those factors found statistically significant in bivariate analysis were used in this model. OR is odds ratios, 95% CI is 95% confidence intervals had a disproportionate distribution negative attributes potentially influenced by differences in living standards in each country. These differences highlight the need to investigate the effect of wider social, environmental and economic contexts outside of school environments.

Limitations
The cross-sectional nature of these data are not amenable to causal interpretations. Secondly, as these are selfreported responses, they are subject to social desirability bias. Lastly, the lack of responses from adolescents who were absent from school on the day of the survey.