Suicidal ideation in a community-based sample of elementary school children: a multilevel and spatial analysis.
Previous studies have identified a number of individual risk factors that could contribute to suicidal ideation in adolescents. The most important clinical risk factor for suicidal ideation and suicide attempts is depression. (5) Low self-esteem, anxiety, poor family environments and impaired parent-child relationships have also been connected with suicidal thoughts. (6-8) A number of studies have shown that the suicide rates within First Nations youth are alarmingly high; however, little research has examined suicide ideation in this group, nor the extent to which different factors contribute to suicidal thoughts or attempts. (9)
Victims of bullying in children and adolescence are also recognized as being at increased risk for suicidal ideation. (10) Bullying can take various forms, (11) traditionally including physical (e.g., assault), verbal (e.g., threats), relational bullying (e.g., social exclusion) and more recently, electronic bullying (e.g., e-mail, phone). (12) Although a number of studies provided evidence that being a victim of different types of bullying increases the risk of developing mental health problems and committing suicide later in life, (13,14) few such studies have focused on elementary school children.
This study adds to the ongoing understanding of suicidal ideation among elementary school children by identifying the most relevant predictors. It has also been suggested that suicidal ideation could be more common within schools or classrooms because associating with suicidal people has been shown to "spread" suicidal thoughts to others. (15,16) According to a major psychological theory--the ecological-transactional model (15,17)--social behaviours are also affected by ecological context, such as (in hierarchical order) neighbourhood, community, school, peer and family groups. Multilevel modelling allows us to explore the importance of social context by dividing the total variation into individuals and groups to be assessed separately. Young (15) found that the contribution of school in relation to suicidal attempts and behaviours is low (< 1%), but nevertheless important given the severity of the outcome. School classroom context was also examined in a multilevel analysis, (16) which found small variation was attributed to school classes after accounting for individual-level risk factors, but suicidal ideation is significantly associated with school class composition with regard to student gender and parental education. Multilevel analysis is therefore considered in this study, including the grade within school and school-level random effects, to give an enriched picture of the complexity of suicidal ideation.
The suicide rates among neighbouring regions are often close and typically exhibit spatial dependency, since the neighbourhoods where children reside typically reflect the socio-economic status and cultural peer-groups of their parents, which in turn can influence the contact networks and mental health of children. (18) Therefore, it is also often of interest to identify high-risk regions when designing an effective prevention program with regard to suicidal ideation. However, few such studies have been conducted in Canadian literature, particularly for school children.
As such, the purpose of the current study is twofold: a) to identify the most relevant predictors for suicidal ideation in a multilevel analysis and b) to identify high-risk neighbourhoods for the prevalence of suicidal ideation among the elementary school children in Saskatoon Health Region, Saskatchewan.
Data for this study were collected during the third round of the Student Health Survey from elementary school students in grades 5-8 in Saskatoon Health Region in 2010-2011. Consent was obtained from schools, parents and children. Ethics approval was provided by the Behavioural Research Ethics Board at the University of Saskatchewan. Research assistants, from Saskatoon Health Region, administered the survey. Students completed their survey during class time, enclosed and sealed it in an envelope and returned it to the research assistant. (19) There were 12,391 students registered in grades 5-8 in participating schools and the overall response rate was 46.7% (n = 5783).
The outcome variable, suicidal ideation, was assessed using the following question: "In the past year (12 months), did you seriously consider suicide?" Potential risk factors included socio-demographic variables: gender (male vs. female), age (years) and ethnicity (Aboriginal vs. other), as well as psychological factors, including depression, anxiety levels and self-esteem. Depression was measured according to the 12-item shortened version of Centre for Epidemiological Studies Depression scale (CES-D); (20) students with scores of 0-11 were considered as having low depression, 12-20 moderate and [greater than or equal to] 21 high. Anxiety levels were measured using a 7-item scale: (19) students with scores of 0-7 were considered as having low anxiety levels, 8-14 moderate and [greater than or equal to] 15 high. Self-esteem was determined using a 5-item scale: (19) students with scores of 0-6 were considered as having low self-esteem, 7-13 moderate, and [greater than or equal to] 14 high. Quality of relationships with parents and friends was also evaluated as a potential risk factor in this study based on a set of questions, responses to which were categorized into poor, moderate and good relationship. (19) The questionnaire also includes reported experiences involving physical, verbal, social and electronic bullying (weekly, once or more vs. none).
The deprivation index reflecting the socio-economic inequalities in health at an area level was also measured based on the average income, education, employment levels, living arrangements, marital status, and proportion of single-parent families. It divides a population into five categories by quintiles. (19) Deprivation index scores were not available for the rural students; hence, all rural students were grouped into a single category in this analysis.
Multilevel logistic regression models with random intercepts were used to examine to what extent suicidal ideation varied among students clustered together within grades and schools. In this study, 34.9% of the participants were rural students and could not be assigned a neighbourhood, so we first considered adding neighbours as another hierarchical level in the multilevel analysis excluding the rural students and the resulting neighbourhood variation is close to zero. Therefore, to preserve sample size for gaining statistical power, we did not consider neighbourhood as a hierarchical level in the multilevel analysis.
In the first step, we built a model including only the random effect terms, to investigate variation in the prevalence of suicidal ideation at the grade and school levels before considering individual student attributes. The importance of each random effect was evaluated using variance partition coefficients (VPC), (21) which describe the percentage of variation in the data that is attributed to a particular organizational level.
To screen the covariates, we examined the univariable association between each risk factor and the outcome. Variables where the p-value was <0.2, based on the likelihood ratio test, were retained for consideration in building the final model. Choice of p < 0.2 as the threshold is recommended to avoid excluding important variables from the model. Manual backward selection was used to develop a main effects model, retaining only variables where p < 0.05. In the final step, all the covariates remaining in the final model were tested for significant interactions.
The association between main effects and the outcome in the final model were reported as adjusted odds ratios (AOR) with 95% confidence intervals (95% CI) and p-values. Multilevel analyses were carried out using maximum likelihood estimation in glmer from the lme4 package in the R Statistical Software (Foundation for Statistical Computing, Vienna, Austria).
Bayesian spatial analysis
To examine neighbourhood risk for suicidal ideation, we conducted a secondary data analysis based on the aggregated counts of suicidal ideation in over 62 neighbourhoods in Saskatoon. Rural students were not included in this analysis, as there was no location information available that was the equivalent of neighbourhood other than the previously considered school.
The number of cases of suicidal ideation from the sth neighbourhood is denoted as [y.sub.s], s = 1, ..., S, which is assumed to follow a Poisson distribution, i.e., [y.sub.s] ~ Poisson ([[mu].sub.s]), with mean [[mu].sub.s] = [n.sub.s][r.sub.s], where [n.sub.s] is the total number of school children from the sth neighbourhood in the study sample and rs represents the relative risk. A Poisson random effects model was used: log([[mu].sub.s]) = [[alpha].sub.0] + log([n.sub.s]) + [b.sub.s] + [h.sub.s], where [[alpha].sub.0] is the intercept, [b.sub.s] is the spatially correlated random effect term, which follows a conditional autoregressive prior distribution (22) and hs is the spatially uncorrelated random effect term that follows a normal distribution. To identify the neighbourhoods where school children are at a higher risk of suicidal ideation, posterior exceedance probabilities, P([r.sub.s] > 1), were calculated as the proportion of Markov chain Monte Carlo (MCMC) simulations for neighbourhoods that had a relative risk of suicidal ideation greater than 1.
The Bayesian computation was implemented using publicly available software (WinBUGs version 1.4). (23) Two chains with different initial values were constructed to assess convergence. To ensure the representative samples were drawn for posterior inference, the first 10,000 iterations were discarded as burn-in, and the next 10,000 iterations were used as samples. Convergence was checked by visual inspection of the iterative series using the Gelman-Rubin diagnostic plot. (23)
Of the 5,783 grade 5-8 students who completed the survey, 5,340 (92.3%) from 109 elementary schools responded to the question regarding suicide ideation; 340 (6.4%) of these students indicated they had considered suicide at least once in the previous 12 months.
In the null model including only the random intercepts for grade and school, the variation identified among grades within schools was 0.03 [+ or -] 0.18 and among schools was 0.24 [+ or -] 0.49. The VPC for grade was 0.8% and for school was 6.7%. In the univariate analysis, there was no significant difference in reported suicidal ideation between males and females or among students of different ages, or from rural and urban schools (Table 1). Almost 13% of the Aboriginal students reported suicidal ideation, compared to 5.5% of non-Aboriginal students. Children who reported higher depression, higher anxiety, lower self-esteem, poorer relationships with friends or parents, as well as those who were more frequently bullied and were from more deprived areas were also more likely to have suicidal thoughts in unconditional models.
In the final multivariable multilevel model, the variation among grades within school and among schools was almost completely accounted for by the individual-level covariates. There was almost no remaining variation among grades within schools (2.7e-10 [+ or -] 1.6e-5), and the variation among schools was also substantially reduced (0.017 [+ or -] 0.13). The VPC for grade was 0.0% and for school was 4.9%. Aboriginal status, depression, anxiety, self-esteem, parental relationship, verbal bullying and electronic bullying were associated with suicidal ideation in the final multivariable model (Table 2). There was no difference among the urban students residing in different deprivation quintiles or compared with the rural students (p = 0.73). We also tested the interactions for the variables remaining in the main effects models and whether gender modifies the effects of those covariates; none were significant and are not presented. Future studies with larger samples are warranted, so that we have enough statistical power to sort out the interaction among the variables.
After accounting for other risk factors in the final model (Table 2), Aboriginal students were more likely to report suicidal ideation than non-Aboriginal students (AOR=1.74). Students who suffered from a high level of depressed symptoms were more likely to report suicidal ideation than those with low depressed symptoms (AOR = 4.79). Likewise, the odds of suicidal ideation increased as the anxiety level increased from low to moderate (AOR=2.43) and low to high (AOR=3.36). Students who considered themselves to have an excellent relationship with their parents had a lower prevalence of suicidal ideation (AOR = 0.33). The odds of suicidal ideation were also lower for those reporting a moderate relationship with their parents, as compared to those with poor relationships with their parents (AOR = 0.46). Students who were frequently verbally or electronically bullied had higher odds of having suicidal thoughts, as compared to those who were less frequently verbally or electronically bullied. For example, students who reported being verbally or electronically bullied weekly had a significantly higher prevalence of suicidal ideation than those who had never been verbally or electronically bullied (verbal bullying: AOR=1.82; electronic bullying: AOR= 2.14).
Of the 5,340 school children who responded to the question of suicidal ideation, 3,506 (65.7%) urban school students provided their residential postal codes. Children who attended rural schools were not asked for their residential postal codes. The relative risks of suicidal ideation at the residential neighbourhood level, based on the Bayesian spatial model, were illustrated in Figures 1 and 2. The posterior exceedance probabilities show that high-risk neighbourhoods were primarily located on the west side of Saskatoon (Figure 1). The estimated posterior means of the relative risks shown in Figure 2 indicate that children from the neighbourhoods on the west side of Saskatoon are at higher risk of suicide ideation than those on the east. We also compared our model with the models including only either [b.sub.s] or [h.sub.s] term and the results indicated that the difference in the deviance information criterion (24) is negligible, so different covariance structure would not impact much on the model fit and the resulting estimates on the exceedance probability.
Results from our study support much of the existing work on suicide ideation, (6-10) but also extend the research in some important areas, which are detailed below. Our study affirmed the relationship between suicidal ideation and various psychosocial factors, such as depression, anxiety, self-esteem and parent-child relationships, after accounting for experiences with bullying. Our study also found that the association between being a victim of verbal or electronic bullying and suicidal ideation is significant. We have also identified that Aboriginal students were more likely to have suicidal thoughts, which demonstrates a need for culturally appropriate interventions to improve the mental health status of Aboriginal elementary school children.
The literature on gender effect on suicidal ideation in adolescence is complex and partly paradoxical. (25) While some studies showed either a higher rate either among boys (7) or among girls, (15,26) our study found gender difference is insignificant, which was also found in other studies. (8,27,28) Rhodes et al. (25) presented a comprehensive discussion on the reasons and timing of the gender effects in the prevalence and lethality of suicidal behaviours. The inconsistent findings might be partially explained by the shift in the interactive social network during adolescent development. Early adolescents may be influenced more by adults (parents, teachers), later adolescents more by their peers. (29) This may also help partially explain our null findings on friend relationship in the final model, but significant association between parental relationship and suicidal ideation, given our study sample focussing on early adolescents.
After adjusting for the individual-level risk factors, very little variation in suicidal ideation was attributable to grouping of students within grades or schools. To examine whether context variables at the grade within school and school level can explain the variation, we created variables summarizing the proportions of males, Aboriginal students, students reporting psychosocial problems, and students reporting different types of bullying experiences, for each grade within a school and for each school in the study sample, but none of those variables were significantly associated with suicidal ideation in the multivariable analysis, so the results are not presented.
Finally, this study investigated the geographic patterns of suicidal ideation among elementary school children. Although the neighbourhood-level analysis does not consider confounding by individual attributes, it is useful for initial targeting of programs to high-risk areas of the city that were not clearly identified by individual student attributes alone. Further research will need to consider the potential for geographic differences in the needs of rural students.
While this study had a relatively large sample size for school-based research, the study also has some limitations, including the cross-sectional nature of the survey. This limits potential inferences about causal relationships between suicidal ideation and some of the risk factors examined. Additionally, clustering within families is also worth serious consideration and such information should be collected in the student's mental health survey, as siblings from the same family typically have extensive and close contact with each other. As well, the response rate was low (46.7%), as the questionnaire was self-completed by the school children. The set of non-respondents may differ from the respondents. For example, students with mental health problems may be more likely to respond due to their interests in this topic. Alternatively, such students may be less motivated to participate because of psychological problems. To address the possibility of such biases, non-respondent adjustments (30) should be conducted in the future based on a survey using different recruitment strategy invoking greater incentives among randomly selected non-respondents. Also, in cases where children have missed reporting either intentionally or inadvertently on any of the questions asked, we have used the commonly used technique of listwise deletion, which deletes the cases containing missing data in the variables that are relevant to the analysis being carried out. This is done to avoid potential bias in our analysis. These limitations need to be taken into consideration when generalizing our results.
Our study is unique in expanding the research area on suicidal ideation in adolescence to the early adolescence period. The findings provide evidence that early adolescent peer aggression must be taken seriously. An integrated preventive intervention strategy, including parenting education programs, mental health services for youth with depression, anxiety and low self-esteem as well as bullying prevention programs, should increase the effectiveness of suicidal ideation prevention efforts. Geographical targeting could also be employed, which may be important for optimizing prevention strategies and enhancing treatment for suicidal ideation.
(1.) Varnik P. Suicide in the world. Int J Environ Res Public Health 2012; 9(3):760-71. PMID: 22690161. doi: 10.3390/ijerph9030760.
(2.) Hawton K, van Heeringen K. Suicide. Lancet 2009; 373:1372-81. doi: 10.1016/ S0140-6736(09)60372-X
(3.) Suicides and Suicide Rate, by Sex and by Age Group. Ottawa, ON: Statistics Canada, 2012.
(4.) Bridge JA, Goldstein TR, Brent DA. Adolescent suicide and suicidal behavior. J Child Psychol Psychiatry 2006; 47(3-4):372-94. PMID: 16492264. doi: 10. 1111/j.1469-7610.2006.01615.x.
(5.) Rhodes AE, Bethell J, Bondy SJ. Suicidality, depression, and mental health service use in Canada. Can J Psychiatry 2006; 51(1):35-41. PMID: 16491982.
(6.) McGee R, Williams S. Does low self-esteem predict health compromising behaviours amongadolescents? JAdolesc 2000; 23(5):569-82. PMID: 11073698. doi: 10.1006/jado.2000.0344.
(7.) Peter T, Roberts LW, Buzdugan R. Suicidal ideation among Canadian youth: A multivariate analysis. Arch Suicide Res 2008; 12(3):263-75. PMID: 18576207. doi: 10.1080/13811110802100882.
(8.) Peltzer K, Pengpid S. Suicidal ideation and associated factors among school-going adolescents in Thailand. Int J Environ Res Public Health 2012; 9(2):462-73. PMID: 22470303. doi: 10.3390/ijerph9020462.
(9.) Karmali S, Laupland K, Harrop AR, Findlay C, Kirkpatrick AW, Winston B, et al. Epidemiology of severe trauma among status Aboriginal Canadians: A population-based study. CMAJ 2005; 172(8):1007-11. PMID: 15824405. doi: 10.1503/cmaj.1040432.
(10.) Espelage DL, Holt MK. Suicidal ideation and school bullying experiences after controlling for depression and delinquency. J Adolesc Health 2013; 53(1 Suppl): S27-31. PMID: 23790197. doi: 10.1016/j.jadohealth.2012.09.017.
(11.) Wang J, Iannotti RJ, Nansel TR. School bullying among US adolescents: Physical, verbal, relational and cyber. JAdolesc Health 2009; 45(4):368-75. doi: 10.1016/j.jadohealth.2009.03.021.
(12.) Raskauskas J, Stoltz A. Involvement in traditional and electronic bullying among adolescents. DevelopPsychol 2007; 43:564-75. PMID: 17484571. doi: 10.1037/0012-1622.214.171.1244.
(13.) Kim Y, Leventhal B, Koh Y, Boyce WT. Bullying increased suicide risk: Prospective study of Korean adolescents. Arch Suicide Res 2009; 13:15-30. PMID: 19123106. doi: 10.1080/13811110802572098.
(14.) Klomek AB, Sourander A, Memela S, Kumpulainen K, Piha J, Tamminen T, et al. Childhood bullying behaviors as a risk for suicide attempts and completed suicides: A population-based birth cohort study. J Am Acad Child Adolesc Psychiatry 2009; 48:254-61. PMID: 19169159. doi: 10.1097/CHI. 0b013e318196b91f.
(15.) Young R, Sweeting H, Ellaway A. Do schools differ in suicide risk? The influence of school and neighbourhood on attempted suicide, suicidal ideation and self-harm among secondary school pupils. BMC Public Health 2011; 11:874. PMID: 22093491. doi: 10.1186/1471-2458-11-874.
(16.) Dalen JD. The association between school class composition and suicidal ideation in late adolescence: Findings from the Young-HUNT 3 study. Child Adolesc Psychiatry Ment Health 2012; 6(1):37. PMID: 23186517. doi: 10.1186/ 1753-2000-6-37.
(17.) Kidd S, Henrich CC, Brookmeyer KA, Davidson L, King RA, Shahar G. The social context of adolescent suicide attempts: Interactive effects of parent, peer, and school social relations. Suicide Life Threat Behav 2006; 36(4):386-95. PMID: 16978093. doi: 10.1521/suli.2006.36.4.386.
(18.) Pickett K, Pearl M. Multilevel analyses of neighbourhood socioeconomic context and health outcomes: A critical review. J Epidemiol Community Health 2005; 55:111-22. PMID: 11154250. doi: 10.1136/jech.55.2.111.
(19.) Modupalli K, Cushon J, Neudorf C. 2010/2011 Student Health Survey: Evidence for Action. Saskatoon: Saskatoon Health Region, 2013.
(20.) Poulin C, Hand D, Bourdeau B. Validity of a 12-item version of the CES-D used in the National Longitudinal Study of Children and Youth. Chronic Diseases in Canada 2005; 26:65-72. PMID: 16251012.
(21.) Goldstein H, Browne W, Rasbash J. Partitioning variation in multilevel models. UnderstandStatist 2002; 1:223-31. doi: 10.1207/S15328031US0104_02.
(22.) Besag J, York J, Mollie A. Bayesian image restoration, with two applications in spatial statistics. Ann Inst Statist Math 1991; 43:1-20. doi: 10.1007/ BF00116466.
(23.) Spiegelhalter D, Thomas A, Best N, Lunn D. WinBUGS Version 1.4 User Manual. Cambridge: MRC Biostatistics Unit, Institute of Public Health, Cambridge University, 2003.
(24.) Spiegelhalter DJ, Best NG, Carlin BP, Van Der Linde A. Bayesian measures of model complexity and fit. J Roy Stat Soc: Ser B 2002; 64(4):583-639. doi: 10. 1111/1467-9868.00353.
(25.) Rhodes AE, Boyle MH, Bridge JA, Sinyor M, Links PS, Tonmyr L, et al. Antecedents and sex/gender differences in youth suicidal behavior. World J Psychiatry 2014; 4(4):120-32. doi: 10.5498/wjp.v4.i4.120.
(26.) Chen PC, Lee LK, Wong KC, Kaur J. Factors relating to adolescent suicidal behavior: A cross-sectional Malaysian school survey. J Adolesc Health 2005; 37(4):337. PMID: 16182146. doi: 10.1016/j.jadohealth.2004.10.018.
(27.) Mahfoud ZR, Afifi RA, Haddad PH, Dejong J. Prevalence and determinants of suicide ideation among Lebanese adolescents: Results of the GSHS Lebanon 2005. J Adolesc 2011; 34(2):379-84. PMID: 20434762. doi: 10.1016/j. adolescence.2010.03.009.
(28.) Riesch SK, Jacobson G, Sawdey L, Anderson J, Henriques J. Suicide ideation among later elementary school-aged youth. J Psychiatr Ment Health Nurs 2008; 15(4):263-77. PMID: 18387145. doi: 10.1111/j.1365-2850.2007.01221.x.
(29.) Scherf KS, Smyth JM, Delgado MR. The amygdala: An agent of change in adolescent neural networks. Horm Behav 2013; 64(2):298-313. PMID: 23756154. doi: 10.1016/j.yhbeh.2013.05.011.
(30.) Eisenberg D, Gollust SE, Golberstein E, Hefner JL. Prevalence and correlates of depression, anxiety, and suicidality among university students. Am J Orthopsychiatry 2007; 77(4):534-42. PMID: 18194033. doi: 10.1037/000294126.96.36.1994.
Received: September 24, 2015
Accepted: December 20, 2015
Cindy Xin Feng, PhD,  Cheryl Waldner, PhD,  Jennifer Cushon, PhD,  Kimberly Davy, MPH,  Cory Neudorf, PhD 
[1.] School of Public Health, University of Saskatchewan, Saskatoon, SK
[2.] Large Animal Clinical Sciences and School of Public Health, University of Saskatchewan, Saskatoon, SK
[3.] Saskatoon Health Region, Saskatoon, SK
[4.] School of Public Health, University of Saskatchewan, Saskatoon, SK
[5.] Chief Medical Health Officer, Saskatoon Health Region, Community Health and Epidemiology, University of Saskatchewan, Saskatoon, SK
Correspondence: Cindy Xin Feng, School of Public Health, University of Saskatchewan, Saskatoon, SK S7N 5E5, Tel: 306-966-1948, E-mail: cindy.feng@ usask.ca
Acknowledgements: We thank the Saskatoon Public School Board, the Greater Saskatoon Catholic School Board, the Saskatoon Tribal Council and the Saskatoon Health Region for supporting this research. We also acknowledge the Canadian Institutes of Health Research as the funding source for the original survey.
Conflict of Interest: None to declare.
Table 1. Descriptive statistics for variables used in model building computed from the univariable analysis in grade 5 to 8 school children accounting for clustering within grade and school Characteristic Total (%) Suicidal ideation Yes (%) Socio-demographic Gender Male 2523 (47.7) 144 (43.1) (n = 5291) Female 2768 (52.3) 190 (56.9) Age, years <11 971 (18.5) 60 (18.0) (n = 5257) 11 1324 (25.2) 72 (21.6) 12 1344 (25.6) 74 (22.2) 13 1203 (22.9) 88 (26.4) >13 415 (7.9) 39 (11.7) Ethnicity Aboriginal 537 (10.5) 70 (21.7) (n = 5109) Non-Aboriginal 4572 (89.5) 253 (78.3) Type of school Rural 1355 (25.6) 85 (25.2) (n = 5299) Urban 3944 (74.4) 252 (74.8) Deprivation 1 (most 1155 (24.1) 54 (18.1) index affluent) (n = 4793) 2 810 (16.9) 37 (12.4) 3 538 (11.2) 40 (13.4) 4 499 (10.4) 42 (14.1) 5 (least 436 (9.1) 40 (13.4) affluent) Rural 1355 (28.3) 85 (28.5) Mental health Anxiety Low 1963 (41.0) 29 (9.8) (n = 4784) Moderate 1712 (35.8) 73 (24.6) High 1109 (23.2) 195 (65.7) Self-esteem Low 1103 (21.2) 193 (59.8) (n = 5192) Moderate 3059 (58.9) 103 (31.9) High 1030 (19.8) 27 (8.4) Depression Low 4039 (83.8) 124 (39.6) (n = 4821) Moderate 662 (13.7) 132 (42.2) High 120 (2.5) 57 (18.2) Social relation Parent Poor 1083 (23.3) 192 (63.6) relationship Moderate 2567 (55.3) 94 (31.1) (n = 4642) Excellent 992 (21.4) 16 (5.3) Friend Few 1051 (25.2) 151 (60.2) relationship Some 1685 (40.3) 51 (20.3) (n = 4176) Many 1440 (34.5) 49 (19.5) Bullying experiences Physical Never 4102 (78.6) 178 (53.6) (n = 5216) Once or twice 945 (18.1) 112 (33.7) Weekly 169 (3.2) 42 (12.7) Verbal Never 3111 (59.8) 99 (30.4) (n = 5202) Once or twice 1563 (30.1) 119 (36.5) Weekly 528 (10.1) 108 (33.1) Social Never 3710 (71.4) 134 (41.5) (n = 5193) Once or twice 1109 (21.4) 103 (31.9) Weekly 374 (7.2) 86 (26.6) Electronic Never 4664 (90.0) 229 (70.9) (n = 5184) Once or twice 408 (7.9) 61 (18.9) Weekly 112 (2.2) 33 (10.2) Characteristic Suicidal ideation No (%) p-value Socio-demographic Gender Male 2379 (48.0) 0.083 (n = 5291) Female 2578 (52.0) Age, years <11 911 (18.5) 0.057 (n = 5257) 11 1252 (25.4) 12 1270 (25.8) 13 1115 (22.6) >13 376 (7.6) Ethnicity Aboriginal 467 (9.8) <0.001 (n = 5109) Non-Aboriginal 4319 (90.2) Type of school Rural 1270 (25.6) 0.65 (n = 5299) Urban 3692 (74.4) Deprivation 1 (most 1101 (24.5) 0.006 index affluent) (n = 4793) 2 773 (17.2) 3 498 (11.1) 4 457 (10.2) 5 (least 396 (8.8) affluent) Rural 1270 (28.3) Mental health Anxiety Low 1934 (43.1) <0.001 (n = 4784) Moderate 1639 (36.5) High 914 (20.4) Self-esteem Low 910 (18.7) <0.001 (n = 5192) Moderate 2956 (60.7) High 1003 (20.6) Depression Low 3915 (86.8) <0.001 (n = 4821) Moderate 530 (11.8) High 63 (1.4) Social relation Parent Poor 891 (20.5) <0.001 relationship Moderate 2473 (57.0) (n = 4642) Excellent 976 (22.5) Friend Few 900 (22.9) <0.001 relationship Some 1634 (41.6) (n = 4176) Many 1391 (35.4) Bullying experiences Physical Never 3924 (80.3) <0.001 (n = 5216) Once or twice 833 (17.1) Weekly 127 (2.6) Verbal Never 3012 (61.8) <0.001 (n = 5202) Once or twice 1444 (29.6) Weekly 420 (8.6) Social Never 3576 (73.4) <0.001 (n = 5193) Once or twice 1006 (20.7) Weekly 288 (5.9) Electronic Never 4435 (91.2) <0.001 (n = 5184) Once or twice 347 (7.1) Weekly 79 (1.6) Table 2. Results of multivariable regression analysis assessing the risk factors associated with suicidal ideation in grade 5 to 8 school children (n = 3872) accounting for clustering within grade and school Independent AOR 95% CI p-value variables Ethnicity Non-Aboriginal (reference) Aboriginal 1.73 1.18-2.54 <0.001 Depression Low (reference) Moderate 2.27 1.55-3.34 <0.005 High 4.79 2.66-8.62 <0.001 Anxiety Low (reference) Moderate 2.43 1.45-4.08 <0.001 High 3.36 1.93-5.85 <0.001 Parent relationship Poor (reference) Moderate 0.46 0.33-0.65 <0.001 Excellent 0.33 0.17-0.62 <0.001 Self-esteem Low (reference) Moderate 0.48 0.34-0.68 <0.001 High 0.73 0.41-1.31 0.29 Verbal bullying None (reference) Once or twice 1.46 1.03-2.08 0.034 Weekly 1.82 1.18-2.79 0.006 Electronic bullying None (reference) Once or twice 1.38 0.92-2.06 0.12 Weekly 2.14 1.12-4.07 0.02
|Printer friendly Cite/link Email Feedback|
|Title Annotation:||QUANTITATIVE RESEARCH|
|Author:||Feng, Cindy Xin; Waldner, Cheryl; Cushon, Jennifer; Davy, Kimberly; Neudorf, Cory|
|Publication:||Canadian Journal of Public Health|
|Date:||Jan 1, 2016|
|Previous Article:||A collaboration between cousins: The Canadian Journal of Public Health and the Canadian Society for Epidemiology and Biostatistics/Une collaboration...|
|Next Article:||Contact investigation outcomes of Canadian-born adults with tuberculosis in indigenous and non-indigenous populations in Alberta.|