Survey design and study population
The present study is based on data collected as part of the COMPASS longitudinal study (https://uwaterloo.ca/compass-system/). COMPASS is a multicentre study of adolescent health in Canada. Each year, youth in participating high schools are asked to complete a questionnaire about their lifestyle and behaviors. In the province of Quebec, school surveys have been conducted since the spring of 2017, in partnership with the school communities and the regional public health departments. Since 2020, the data collection is conducted online using Qualtrics XM (Seattle, WA, USA).
The current analysis is based on data collected between March and May 2022. The study population includes all students in the 5 years of the high school level (equivalent to Grades 7 to 11 in the USA and the rest of Canada) of 113 secondary schools in the province of Quebec, all localized in its eastern part. Of the 58 792 adolescents who were solicited, 48 289 (82.1%) answered the online questionnaire. Active parents’ refusal rate was less than 0.56% (327 participants).
Measures
Measures of climate change anxiety were self-reported and taken from Clayton and Karazsia (Clayton & Karazsia, 2020; Mouguiama-Daouda et al., 2022). Two questions addressed youth concerns. The first question assessed cognitive-emotional impairment related to climate change. Using a 5-point scale, participants reported how often thinking about climate change makes it difficult for them to sleep. The second question was a measure of functional impairment. Using the same scale, participants reported the extent to which their concerns about climate change undermine their ability to work their potential. An indicator variable was constructed to combine the youth who responded sometimes, often, or almost always to either one of the two questions. These adolescents were considered to have more climate anxiety than those who never or rarely had these impairments. Two other questions dealt with behavioral engagement. One asked people how often they try to reduce their behaviors that contribute to climate change; the other asked how often they believe they can do something to help address the problem of climate change. The latter served as a measure of psychosocial self-efficacy which refers to one’s perceived ability to engage in various situation-specific self-management tasks (13).
We identified adolescent characteristics and environmental factors that could influence climate anxiety based on the literature. Adolescent characteristics included age and gender. Because of the difficulties that youth have in reporting family income, we created a score of family affluence, inspired by the Family’s Affluence Scale (14). The score was based on the adolescents’ responses regarding1) the average amount of money the adolescents receive each week for personal spendings or savings; 2) skipping breakfast because there is nothing to eat at home; 3) going to bed hungry at night because there is not enough money to buy food; 4) having the feeling that their family and themselves are less financially comfortable than the average student in their class; 5) having their own bedroom; 6) the number of people in their household; and 7) being worried about their family being able to pay bills and expenses. Composite scores were dichotomized into more or less affluent. Adolescents from the last two quintiles were considered to be less affluent.
We included 4 measures of well-being; (i) the presence of moderate or severe anxiety; (ii) adaptation to the pandemic; (iii) school connectedness; and (iv) being able to talk about their problems with family.
Anxiety symptoms were measured using the 7-item Generalized Anxiety Disorder scale (15). Students were asked how often they experienced each symptom in the last 2 weeks. The level of anxiety is considered to be moderate to severe when the score is greater than or equal to 10 out of 21. People scoring 9 or less are considered to have low levels of anxiety.
The participants’ adaptation to the pandemic was measured using a set of 5 questions to report their worries about the current circumstances; their personal health; their family members’ health; and their stress level (16). Composite scores were dichotomized into less well adapted vs better adapted.
School connectedness reflects the youth's sense of closeness to people at school and attachment school environment. It was collected by utilizing a modified version of the National Longitudinal Study of Adolescent Health School Connectedness scale (17, 18). The items within this scale attempt to tap into global school connectedness by capturing the belonging, liking/enjoyment, closeness, fair treatment, and safety felt by the child (19). Composite scores were dichotomized into strong school connectedness vs weak school connectedness.
Participants used a 5-point Likert scale to rate the extent to which they can talk about their problems with their family. Respondents were divided into two groups: those who somewhat agreed versus those who somewhat disagreed or were neutral.
School characteristics included the type of school attended (public vs private), and the location (urban vs rural).
Statistical Analyses
This study is based on a cross-sectional design. Descriptive statistics were calculated using survey sampling weights. The three binary outcomes were 1) concerns about climate change that interfere with sleep or ability to work, 2) the belief that they can do something to help address the problem of climate change, and 3) trying to reduce behaviors that contribute to climate change. We performed poisson regressions to examine associations between the binary outcomes and the participants’ characteristics as well as indicators of wellbeing. We used robust estimators to account for overdispersion. We also examined associations between climate anxiety (interference with sleep or ability to work), self-efficacy (the belief that they can do something to address the problem) and behavioural engagement (trying to reduce behaviours). The poisson regressions were adjusted for potential confounders (age, gender, school type, family affluence, and location) to estimate risk ratios. Robust estimators accounted for school clustering. The margins command in Stata was used to obtain estimated proportions (20). All analyses were performed with STATA version 17.