Introduction
The coronavirus disease 2019 (COVID-19) pandemic and public health measures are believed to have exacerbated pre-existing concerns and inequities in youth mental health [
1‐
9]. Since March 2020, youth experienced drastic disruptions to their lives, including the widest reaching and longest school closures in history [
2]. The potential unintended consequences of school closures on adolescent mental health led to widespread concerns and public debate, with some arguing that the risks outweighed any benefits for reduced transmission [
2,
8,
9,
12]. Indeed, schools provide important structure and routine, and opportunities to socialize with peers, for adult mentorship, and to engage in extracurricular and physical activities—all factors with known links to positive youth mental health [
2,
3,
6‐
11]. Moreover, schools are often where mental health concerns are first recognized; they are key contexts for the delivery of public health interventions and the most common setting for the provision of mental health services [
13‐
15]. Acknowledging these factors, opposing parties maintain that greater illness and death due to COVID-19 transmission in schools would have been more detrimental to young people’s mental health than temporary and partial closures [
16]. Furthermore, school can also contribute to distress for some students (e.g., related to bullying and academic pressures) [
17,
18].
Literature generally indicates adverse associations between remote learning and mental health relative to in-person learning during the COVID-19 pandemic [
19,
20]. However, previous research largely relies on cross-sectional designs and retrospective accounts [
5,
20,
22]. Where prospective evidence does exist, it typically lacks baseline pre-pandemic data and robust measures of mental health and psychopathology [
5,
20,
22,
23]. Moreover, most published literature applies to the first COVID-19 wave and complete school closures; there is a need for research on the prolonged pandemic response, examining partial school closures and resultant diverse learning modes [
20,
21]. Further, research on virtual remote learning needs to consider variations by gender and students’ perceptions of their home environments. Schools can offer reprieve from difficult home lives, and financial and family distress increased during the COVID-19 pandemic [
1,
3,
24,
25]. COVID-19 and previous pandemics have also been shown to have greater impacts on girls/women, given differences in mental health (e.g., how problems are exhibited and managed), social relationships, risk of exposure to violence/abuse, and gendered delegation of sibling caregiving and household chores [
26‐
28].
Using data from an ongoing Canadian prospective cohort, the objective of this study was to test whether learning modes during the COVID-19 response were differentially associated with mental health changes from before to during the pandemic in adolescents. Results may help inform school protocols and policies to support youth mental health in the case of future events. We also tested whether learning mode associations with mental health changes were modified by gender or perceived home life. We hypothesized that experiences of learning from home would vary based on whether it was mandated or optional and that virtual learning would have more adverse associations with mental health among females and non-binary youth, and adolescents without happy home lives, than their counterparts identifying as male and with positive perceptions of their home life happiness.
Discussion
In a prospective cohort of Canadian adolescents, we examined whether mental health changes from pre-COVID-19 to the first full school year of the pandemic differed by school learning modes. In non-adjusted descriptive statistics, students learning fully in-person had less of a decline in mental health, and students learning in virtual mandated modes had greater depression increases, than their peers in other learning modes. However, our adjusted models demonstrated negligible differences in mental health outcomes by learning mode. In the main effects models, adjusting for baseline mental health, covariates (socioeconomic status indicators, household size, province, gender, and home life happiness), and school clustering, a learning mode effect was found for anxiety but not depression or psychosocial well-being; students learning in a blended mode had greater anxiety increases relative to their peers attending fully in-person. Interaction models supported our hypotheses that learning mode effects varied by students’ gender and perceived home lives. Smaller increases in depression were associated with learning fully in-person among females and learning virtually when optional among males relative to their gender counterparts in other learning modes. Conversely, among students without happy home lives, virtual learning when optional was associated with greater declines in psychosocial wellbeing relative to other learning modes.
Caution may be advisable when implementing blended learning modes, which had the only main effect and was consistent across gender and home life happiness for changes in anxiety. Blended modes divided students into alternating in-person and virtual learning cohorts, with patterns varying from alternating days to weeks. The constant back-and-forth in a blended learning format potentially requires continuing adjustment, which may have contributed to greater anxiety increases in these students relative to their peers attending school fully in-person. In a clinical sample of youth with physical illness and their parents, youth psychological distress did not differ by learning mode and parent psychological distress was lower in blended modes, but COVID-19-related worries were higher in youth and parents in virtual and blended modes relative to in-person learning [
56]. Relatively few studies have examined blended learning modes, with most research focusing on pedagogical aspects rather than mental health outcomes, or only comparing fully remote and in-person modes. Amid projections that blended learning will become the norm beyond pandemic times, these results support the need for further research prior to expanding alternative learning modes.
Learning mode associations with changes in depression differed by gender. As hypothesized, learning fully in-person appeared beneficial among females relative to learning fully or partly virtual. Conversely, in males, learning virtually when optional was associated with smaller increases in depression than other learning modes controlling for perceived home lives, baseline mental health, and other covariates. In-person contact may be more critical for girls, given their socialized greater tendency to rely on social networks for support than boys. Further, given continued gendered roles and expectations, girls learning virtually may have been more likely to be expected to contribute to caregiving and household chores than boys. It is important to recognize that females consistently report greater internalizing symptoms, whereas externalizing symptoms are typically higher in males but were not assessed in this study [
59]. It is also plausible that in-person learning was not as beneficial to males given the COVID-19 protocols, including mandatory masks, reduced breaks and lunch hours, quadmester or octomester scheduling, and reduced extracurricular and sport opportunities. Finally, the option of virtual learning, as opposed to mandated learning, may have supported their agency, an important aspect of healthy youth development. It is not clear from this research how decisions were made regarding learning mode when optional or how this time learning virtually was used. Further research is needed to explore findings regarding experiences of optional virtual learning, as opposed to mandated closures, including sustained effects on social, emotional, and physical health and development.
As hypothesized, mental health changes associated with virtual learning varied by how students perceived their home lives. Learning virtually when schools remained open was associated with relatively more adverse changes in psychosocial wellbeing in students without perceptions of a happy home life than their peers attending in other learning modes. Similarly, in the US high school students whose families had a choice of learning mode, attending remotely was associated with lower levels of social and emotional well-being than in-person learning [
19]. Learning remotely prevented respite from difficult home lives, which may have been exacerbated during the COVID-19 pandemic [
1,
3,
24,
25]. Family climate, conflict, and financial concerns have been associated with poorer mental health in adolescents over the pandemic [
57,
58]. While we controlled for pre-pandemic mental health and socioeconomic factors, families or students choosing remote learning may have experienced greater pandemic-related distress in their household. Adolescents or their families may have opted for them to continue learning online during the pandemic if they had recently experienced, or were at greater risk of, deteriorations in their mental health. In a US study, parents’ decisions around learning modes were driven by perceptions of risk, parental availability, and access to in-person education [
59]. A web survey found 9% of US adolescents reported a preference for fully online school, compared to 18% for blended learning and 65% that opted for learning fully in-person [
60]. It is plausible that students without happy home lives had limited say in the decision. Beyond the home, youth voices have been absent in pandemic-related decision-making, despite their recognized right to be involved in decisions that impact them [
61].
Results support the importance of differentiating between optional and mandated virtual learning in future research. Schools were mandated to close to in-person learning in response to worsening COVID-19 transmission rates in the community. For this reason, the virtual mandated group may have been expected to have experienced greater declines in mental health relative to students at schools that remained open. While unadjusted changes in mental health appeared more adverse in the mandated virtual group, no main effects resulted in adjusted models relative to in-person learning. Contrary to expectations, more adverse associations emerged for virtual or blended learning when schools remained open and not during mandated closures. Paradoxically, students without happy home lives and learning virtually when mandated reported less of an increase in depression symptoms relative to their peers with happy home lives learning in person; however, examining the least squares means suggests marginal differences in depression within students without happy home lives.
We were unable to rule out the possibility of unmeasured confounds and to demonstrate causality. Students were not randomly assigned to different learning modes; however, the prospective data allowed us to control for pre-existing baseline mental health and socioeconomic differences. As discussed above, mental health changes experienced by students, or other contributing factors (e.g., caregiving for a family member, poorer school experiences such as bullying victimization), may have influenced their or their family’s decision regarding attending in person, online, or in a blended mode, when provided the option. Comparisons of unadjusted and adjusted changes in mental health by learning mode, and differences between virtual optional and mandated modes, support the importance of considering self-selection, confounding, controlling for baseline mental and additional differences in students attending different learning modes. Changes in learning modes during the COVID-19 pandemic cannot be studied outside of broader lockdown measures [
20]. While we controlled for province, variations in COVID-19 case rates and implemented protocols and policies occurred across and within provinces [
50,
51].
The time between the two waves of the study used may miss changes in mental health over the dynamic pandemic response that more frequent assessments would have uncovered. Data collected between these two waves (i.e., spring 2020) were excluded due to lower response rates and higher risk of selection bias, and linkage across the additional waves would have further reduced the sample. Learning mode was assessed at the time of participation and changes in learning modes outside of data collection dates were not captured. The change from paper-and-pencil to online surveys may introduce bias. Study attrition and the relatively lower online response rates during the pandemic may also introduce selection bias; nonparticipation or non-linkage may have been more common among students with poorer mental health and experiences of new learning modes. Although attrition during the pandemic was at least partly due to random factors, such as whether schools administered the survey during class time. The attrition analysis indicated poorer mental health scores in the baseline full sample than the analytic sample. However, many non-linked students were in Grade 12/Secondary 1 and would have graduated out of the cohort; these students may partly account for baseline differences, as mental health is shown to decline with increasing grade [
62‐
64].
Our sample is also a limitation, as the majority of students identified as White. Small sample sizes in some population subgroups, including within racialized and ethnic identity groups, required categories to be collapsed and limited our ability to examine within and between group differences in learning mode effects and intersectional identities. The number of students that identified their gender differently or preferred not to say prohibited interpretation of interaction terms for this group. The gender measure itself was also a limitation; more recent waves of the COMPASS study have an improved measure of gender identity with additional response options, and a separate item assessing sex assigned at birth. Further research is needed among gender diverse youth, given evidence of disproportionately adverse impacts of the pandemic measures [
65,
66]. Remote learning may have been more detrimental, given that gender diverse youth are less likely to feel safe at home and connected to their families [
67,
68], and connections with school adults are shown to be protective for adverse health outcomes [
69]. That said, gender diverse youth experienced higher rates of bullying victimization than their cis-gender peers, with face-to-face school the primary context for bulling [
70].