- Using AI to predict suicidal behaviours in students
- AI identifies main predictors of suicidal behaviours
- Opening the door to large scale screening
- Self-esteem: a previously unrecognized predictor
- About McGill University
- Low Self-Esteem and Its Association With Anxiety, Depression, and Suicidal Ideation in Vietnamese Secondary School Students: A Cross-Sectional Study
- Study Design and Participants
- Prevalence of Low Self-Esteem and Characteristics Associated With Self-Esteem
- Relationships Between Self-Esteem and Anxiety, Depression, and Educational Stress, and Suicidal Ideation
Using AI to predict suicidal behaviours in students
How can we predict suicide risk in students, especially at a time when the COVID-19 pandemic has negatively affected many people’s mental health? According to researchers from Montreal and France, self-esteem represents an important predictive marker of suicidal risk. The team from McGill University, University of Montreal, Inserm, and Université de Bordeaux is using artificial intelligence to identify factors that accurately predict suicidal behavior in students.
“Suicide is the second leading cause of death among 15- to 24-year-olds. Early detection of suicidal behaviours and thoughts is the key to providing appropriate treatment,” says lead author Mélissa Macalli, a PhD Candidate at University of Bordeaux.
Published in Scientific Reports, their analysis is data collected from over 5,000 university students in France who were followed for at least one year between 2013 and 2019. The study shows that 70 potential predictors, four detect around 80 percent of suicidal behaviours at follow-up. These are suicidal thoughts, anxiety, depressive symptoms, and self-esteem.
AI identifies main predictors of suicidal behaviours
Using machine learning, the researchers simultaneously analyzed the many factors associated with suicidal risk and ranked them according to their importance in predicting suicidal behaviours. These factors included sociodemographic data, lifestyle, substance use, childhood trauma, and personal and family history of suicidal behaviours.
“Many known factors can contribute to the increased risk in university students, such as the transition from high school to college, psychosocial stress, academic pressures, and adapting to a new environment.
These are risks that have also been exacerbated by the health crisis triggered by the COVID-19 pandemic, although there is no clear evidence of an increase in suicide rates during the pandemic,” says co-author Massimiliano Orri, a Postdoctoral Fellow at McGill University with the McGill Group for Suicide Studies and at University of Bordeaux.
Opening the door to large scale screening
The students completed two surveys, one at the time of enrollment in the study and the other a year later, providing researchers with critical information about their health, drug and alcohol use, medical and psychiatric history, and their psychological state. This follow-up survey revealed that approximately 17 percent of the participating students, both women (17.4 percent) and men (16.8 percent), exhibited suicidal behaviours in the year elapsing between the two questionnaires.
“This research opens up the possibility of large-scale screening by identifying students at risk of suicide using short, simple questionnaires, in order to refer them to appropriate care,” explains senior author Christophe Tzourio, a Professor of epidemiology at Université de Bordeaux and Director of the Bordeaux Population Health research center. Such screening tools could provide an alternative to mental health assessments by a physician for students who are often reluctant to disclose sensitive personal information in face-to-face interactions.
Self-esteem: a previously unrecognized predictor
According to the researchers, self-esteem represents an important predictive marker of suicidal risk and should therefore be used in screening tools, even among students that do not show signs of suicidal behaviours.
In secondary analyses of 3,946 students that did not exhibit suicidal behaviours, the primary predictors that stood out for men was self-esteem. For women, they were self-esteem, depressive symptoms, and academic stress.
Low self-esteem is known to be a part of social anxiety, and to overlap with depression. Self-esteem, which is an important marker of psychological vulnerability in young adults has also been associated with suicidal thoughts. “Our study showed that self-esteem is an independent and prominent predictive marker,” say the researchers.
“The mental health specialists on our teams did not expect self-esteem to be one of the top four predictors of suicidal behaviours,” says Mélissa Macalli. “This finding would not have been discovered without the use of machine learning, which makes it possible to analyze a large amount of data simultaneously. This opens up new avenues for both research and prevention,” she concludes.
|About this study“A machine learning approach for predicting suicidal thoughts and behaviours among college students” by Mélissa Macalli, Marie Navarro, Massimiliano Orri, Marie Tournier, Rodolphe Thiébaut, Sylvana M. Côté and Christophe Tzourio was published in Scientific Reports.DOI: https://doi.org/10.1038/s41598-021-90728-z|
About McGill University
Founded in Montreal, Quebec, in 1821, McGill University is Canada’s top ranked medical doctoral university. McGill is consistently ranked as one of the top universities, both nationally and internationally.
It is a world-renowned institution of higher learning with research activities spanning two campuses, 11 faculties, 13 professional schools, 300 programs of study and over 40,000 students, including more than 10,200 graduate students. McGill attracts students from over 150 countries around the world, its 12,800 international students making up 31% of the student body.
Over half of McGill students claim a first language other than English, including approximately 19% of our students who say French is their mother tongue.
Subscribe to news releases
Visit the McGill Newsroom
Low Self-Esteem and Its Association With Anxiety, Depression, and Suicidal Ideation in Vietnamese Secondary School Students: A Cross-Sectional Study
It is well recognised that adolescence is one of the most rapid phases of human development (1). It is characterised by a rapid physical, social, and cognitive growth, as well as changes in self-esteem.
Self-esteem is reported to have a significant impact on important life outcomes including health and social outcomes during adolescence and adulthood.
For example, there is a clear connection between higher self-esteem and positive outcomes, such as occupational success, better social relationships, a sense of well-being, and positive perceptions by peers, academic achievement, and good coping skills (2, 3).
Low self-esteem is causally related to depression, substance abuse, antisocial behaviour, and suicide (4–6). The literature demonstrates that social functioning, such as acceptance by peers, is lower in children with low self-esteem (7).
Extensive research has explored risk and protective factors related to low self-esteem development during adolescence.
Reported risk factors include being a girl (6, 8, 9), the family’s low socioeconomic status (10), parents’ education level, family eligibility for public assistance, eligibility for free or reduced-cost school meals, the parents’ employment status (11), and school performance and grades (6, 12), as well as obesity (13).
Academic achievement is known to be affected by self-esteem, while self-esteem may also influence academic achievement (7). High self-esteem is reported as an important predictive factor for students’ academic achievement (14).
Another study found that while high self-esteem resulted in many positive outcomes and benefits, it did not necessarily lead to good school performance (15).
On the other hand, adolescents with poor academic results did not always have low general self-esteem (16).
There is an association between low self-esteem and negative outcomes for young people’s behavioural and mental health problems, including health-compromising behaviours such as substance abuse, early sexual activity, and eating problems (17).
A longitudinal study among a large sample of young New Zealanders found, however, that while low self-esteem significantly predicted adolescent eating and other health-compromising behaviours, it was not related to substance abuse and early sexual activity (18).
With regard to mental health, a correlation has been detected between low self-esteem and depression (19, 20), anxiety (10, 21), and adolescents’ suicidal ideation and attempts (18, 22).
Reports during the past few years indicate that there is a strong relationship between academic pressure and stress, depression, anxiety, low self-esteem, and suicidal ideation among students in secondary or high school and in young adults (23–25).
A study describing perceptions about mental illness in Hue in Vietnam noted that “studying or thinking too much” is a cause of mental health problems (26). In Vietnamese culture, the pressure from parents and schools might be expected to have an unintended effect, leading to lower self-esteem and associated outcomes seen elsewhere among high school students.
There are as yet no reports of systematic studies on this issue in Vietnam, and there are few reports of descriptive or analytical research into adolescent self-esteem.
This is particularly relevant since we found high levels of depression, anxiety, and suicidal ideation among secondary school students in Vietnam [Nguyen, D. T., Dedding, C., Pham, T. T., Wright, P., & Bunders, J. (25)].
This article reports on a study that aimed to determine (1) the prevalence of low self-esteem; (2) the characteristics associated with low self-esteem; and (3) the relationships among self-esteem and anxiety, depression, educational stress, and suicidal ideation in secondary school students.
We hypothesised that low self-esteem would be associated with a greater risk of poor mental health status.
Study Design and Participants
A cross-sectional study design was used to recruit 1,260 students at three secondary schools in urban and suburban areas in Cantho City in Vietnam. All data were collected during the first academic semester, from September to December 2011. A more detailed description of the sampling and its participants can be found elsewhere (25).
All participants were invited to provide information by self-reporting using a questionnaire. This was done after class or at home anonymously to minimise potential reporting bias and to keep information confidential. The detailed components of the questionnaire have been described elsewhere (25).
The 10-item Rosenberg Self-esteem Scale (21, 27) was used to assess global self-esteem, with higher scores indicating more positive self-regard. Each item asked for response using a 4-point rt scale ranging from 1 (strongly agree) to 4 (strongly disagree).
The scale is generally reliable, with test–retest correlations value between 0.82 and 0.88 (21). The Cronbach α of the scale in the present study was 0.77.
The scale ranges from 0 to 30: a score greater than 25 suggests high self-esteem; scores between 15 and 25 are considered to be within normal range, whereas scores less than 15 suggest low self-esteem (Morris Rosenberg).
A more detailed method description on employing the Center for Epidemiology Studies Depression (CES-D), the anxiety scale, and the Educational Stress Scale for Adolescents for this study is available (25).
To address the issue of suicidal ideation, additional questions on whether the student had ever seriously considered suicide or made a suicide plan used a 3-point scale (never, sometimes, and often). A yes/no question was also used to identify students who had attempted suicide.
To define risks and protective factors, we explore the variables listed in Table 1, including mother’s education, physical and emotional abuse by parents or other adults in the household or school, academic performance in the last semester, educational stress, attendance at a supplementary class, and use of personal tutor. The data collected using the questionnaire is available in the Supplementary Data File.
Table 1 Frequencies and percentages (%) of sociodemographic variables of participants by self-esteem status.
Demographic data were analysed descriptively to determine basic characteristics of the sample population and presented as means ± standard deviations (SD).
The χ2 test was used to assess the significance of differences in the distribution of participants by selected sociodemographic characteristics, risk factors, and outcome variables.
Associations between low self-esteem and family characteristics, educational stress, and academic achievement were explored by logistic regression analysis. Univariate independent predictors of low self-esteem with p< 0.
10 were entered in a multivariate logistic regression model, and the backward Wald method was applied to study their influence on the presence of low self-esteem.
Univariate logistical analyses were also applied to determine the relationships among low self-esteem and anxiety, depression, and suicidal ideation, and Pearson correlation coefficients were used to measure the correlations among these. All analyses were carried out using a significance level of 5%, and all tests were 2-sided. The 95% confidence intervals (CIs) of odds ratios (ORs) were also calculated. All analyses were analysed by using SPSS Inc, Chicago, the United States of America version 16.0.
Of the 1,260 students invited to participate, 111 (7.3%) were excluded from analysis because they did not adequately complete the questionnaire; for example, they did not provide answers to five or more items in the CES-D. The final sample comprised 1,149 senior high school students with a mean age of 16.1 years. Participating students were fairly evenly distributed in terms of sex: 36.
5% males and 63.5% females (Table 1). This proportion also reflects the proportion of males and females enrolled in the classes studied. Students’ participation from each grade (10–12) did not significantly differ, at around 33% each. Ninety-five percent of students were ethnically Kinh; other ethnic groups included Hoa and Khmer. This sample was in line with the school population in this area.
Prevalence of Low Self-Esteem and Characteristics Associated With Self-Esteem
The mean self-esteem score was 17.56 on a scale of 0 to 30. The scores for boys and girls were similar (17.83 vs. 17.40; t = 1.720; p = 0.086). all scores, nearly a fifth (19.4%) of the respondents reported low self-esteem according to Rosenberg’s criteria, with scores below 15 (Table 1).
univariate logistic regression analysis, among six investigated variables, two variables “having a personal tutor” and “attendance at supplementary class” were negatively associated with the risk of low self-esteem, while the other variables were positively correlated to low self-esteem (Table 2).
Students were ly to show lower self-esteem, when their mother’s education was at primary level or below, when they were often physically or emotionally abused at home or at school, when they had below average/very poor academic performance in the last semester, or when they reported experiencing high educational stress.
Table 2 Factors associated with low self-esteem: multivariate logistic regression analyses.
In the multivariate regression analysis (with the backward Wald method), not accounting for effect modification, four variables remained correlated to low self-esteem, either in a negative or positive direction.
Having a personal tutor and academic performance in the last semester were no longer a significant association, but attending supplementary classes still remained a protective factor for low self-esteem (reduced OR of 43% compared with students not attending) (Table 2).
Study results that were considerably positively associated to low self-esteem were medium and high educational stress (ORs = 2.25 and 4.02, respectively).
Relationships Between Self-Esteem and Anxiety, Depression, and Educational Stress, and Suicidal Ideation
As shown in Table 3, four variables—self-esteem, anxiety, depression, and educational stress—were related to each other. Self-esteem was negatively correlated to anxiety, depression, and educational stress, while educational stress was positively correlated to anxiety and depression.
Table 3 Pearson correlations between self-esteem, anxiety, depression, and educational stress.
One of the purposes of the study was to identify the impact of self-esteem on mental health problems.
The results of univariate logistics indicate that low self-esteem contributed significantly to anxiety, depression, and suicide among adolescents.
Compared to students who reported normal self-esteem, the students who reported low self-esteem had twice the odds of having anxiety symptoms [20.3% (187/921) vs. 34.2% (76/222)] (OR = 2.04; 95% CI OR = 1.48–2.82, p