The web- and mobile-based Canadian COVIDiet cohort study was initiated on May 29, 2020, toward the end of the COVID-19 pandemic’s first wave of infections. The primary objective was to provide longitudinal data on the impact of the pandemic on eating behaviors, diet quality, and related lifestyle behaviors of Canadian adults. Based on the Canadian population estimated at 31 million adults and considering a 5% margin of error, 95% CI, and 50% response distribution [18], 400 participants were required per age strata (18-45, 45-65, ≥65 years). Following power allocation relative to population size [19], the target sample size was 1920 participants. Until December 5, 2020, 2959 adults from across Canada were recruited using social media advertisement, professional societies, and institutional email lists and from community centers, including food banks. Efforts were made to recruit across age groups, regions, and gender (men and women). Regions were delimited as Ontario, Québec, British Columbia, Prairies (Alberta, Manitoba, Saskatchewan), Atlantic provinces (Nova Scotia, New Brunswick, Newfoundland and Labrador, Prince Edward Island), and territories (Northwest Territories, Yukon, and Nunavut). Interested individuals were invited to contact the study coordinator by email or visit the study website for more information and for eligibility screening.

Eligible participants were 18 years of age or older, were living in Canada at the time of study entry, were able to read English or French, owned a smartphone or tablet, and had access to the internet. Individuals living in care homes or other institutional environments; hospitalized, pregnant, or breastfeeding; or with an active and uncontrolled acute disease that interfered with usual food intake were excluded. The COVIDiet study participants were followed every 3 months for 1 year to capture the impact of the pandemic across waves of infections and for 1 year thereafter to evaluate the longer-term effect. The current analysis was conducted on baseline (May-December 2020) and retrospective data from the baseline questionnaire. Individuals who did not fill out the baseline questionnaire (n=904), provided implausible responses (n=8), and had missing data on changes in weight and lifestyle behavior (eating habits, alcohol consumption, physical activity level, sleep quality and smoking; n=442) were excluded. The final sample included 1609 participants (Multimedia Appendix 1, Figure S1).

Data were collected with online encrypted surveys using AirTable [20] and deidentified using generated ID codes. Participants were eligible to win a CA $250 (US $182) gift card when they enrolled in the study; 1 draw occurred at every 500 participants enrolled. All participants provided online informed consent. This study was approved by the McGill University Research Ethics Board (20-04-064) and was registered on ClinicalTrials.gov (NCT04407533).

Baseline (during) versus prepandemic changes in weight and lifestyle factor status were categorized into 3 levels. For weight, a difference greater than +/–3% was considered an increase or a decrease in weight, respectively; participants with 0%-2.9% or insignificant weight change were categorized as having unchanged weight. For alcohol consumption, an upward change in 1 or more categories (<1 drink/week, 1-3 drinks/week, 4-8 drinks/week, 9-14 drinks/week, ≥15 drinks/week) was categorized as increased consumption, a downward change as decreased consumption, or no change as unchanged consumption. For physical activity level, a change in adherence to the WHO recommendations (yes or no) was categorized as increased (no to yes), decreased (yes to no), or unchanged (yes to yes or no to no). For eating habits, a perceived change was reported as improved, unchanged, or worsened. For sleep quality, an increase or decrease in 1 or more categories (very bad, fairly bad, fairly good, very good) was categorized as increased or decreased sleep quality, respectively, or otherwise categorized as unchanged. For smoking status (daily or occasionally/never), a change was categorized as increased (no to yes), decreased (yes to no), or unchanged (yes to yes, or no to no).

In total, 942 (46%) of 1609 participants were recruited between the end of the first wave of infections and the beginning of the second wave, that is, when lenient restrictive measures were in place. The other participants (1105/1609, 54%,) were recruited during the second wave of the pandemic, during strict restrictive measures. Table 1 shows prepandemic sociodemographic and lifestyle characteristics of participants. Participants mostly self-identified as women, were aged 18-89 years (mean 40.2, SD 15.5 years), and had a mean BMI of 26.1 (SD 6.3) kg/m². The majority were White, had at least a bachelor’s degree, and reported having a household income > CA $50,000 (US $36,400) per year.

Since the pandemic, 35.1% (565/1609) participants had a decreased income, 49.0% (788/1609) changed their work arrangement, and 94.3% (1518/1609) kept the same living situation (1295/1609, 80.5%, were living with others, and 223/1609, 13.9%, were living alone). Regarding mental well-being, 52.2% (840/1609) participants reported a worsened mood and 63.8% (1026/1609) reported increased stress levels compared to before the pandemic. Among lifestyle factors, most of the population reported either no change or insignificant weight gain or loss and no change in alcohol intake, physical activity, smoking, and sleep quality, yet 44.1% (709/1609) reported a decreased perceived quality of eating habits (Table 2).

Among the tested LCA models, the model with 2 emerging lifestyle behavior change classes had the best fit despite a slightly lower AIC compared to the model with 3 classes (Multimedia Appendix 2, Table S1) and was therefore selected for use. Participants with the class 1 membership (974/1609, 60.5%, population share probability) were more likely to report unchanged weight, sleep quality, smoking and alcohol intake, unchanged/improved eating habits, and increased physical activity. Participants with the class 2 membership (635/1609, 39.5%, population share probability) were more likely to report significant weight gain, deteriorated eating habits and sleep quality, unchanged/increased alcohol intake and smoking, and decreased physical activity. Class 1 was referred to as the healthy and class 2 as the less healthy pattern. Results from the LCA are presented in Figure 1.

Participants with the less healthy pattern were more likely to self-identify as women or a gender minority, to have a lower income prepandemic, and to live with 2 or more chronic diseases. They were also more likely to report a decreased income since the pandemic. Age, ethnicity, residential area, and change in the living situation did not differ between classes.

Among risk factors, body image dissatisfaction (odds ratio [OR] 8.8, 95% CI 5.3-14.9), increased stress level (OR 3.4, 95% CI 2.0-5.8), depression (OR 1.7, 95% CI 1.2-2.4), and self-identifying as a gender minority (OR 5.5, 95% CI 1.3-22.3) were positively associated with adopting less healthy lifestyle behaviors in adjusted models (Figure 2 and Multimedia Appendix 2, Table S2). In addition, the province of Québec was inversely associated with less healthy lifestyle behavior changes (OR 0.51, 95% CI 0.33-0.80).

Gender-weighted sensitivity analysis showed a stronger positive association between women and the less healthy lifestyle behavior changes, and the association for gender minorities disappeared (Multimedia Appendix 2, Table S3). The results were otherwise consistent with the nonweighted analysis.

Stratified analysis by age group showed positive associations of body image dissatisfaction, symptoms of depression, increased stress, and women and minority genders with less healthy lifestyle behavior changes in younger adults (<55 years old) and positive associations of symptoms of anxiety and body image dissatisfaction with less healthy lifestyle behavior changes in older adults (Multimedia Appendix 2, Table S4). However, depressive symptoms, gender, and stress level were not determinants of lifestyle behavior change patterns in this subgroup. Living with a chronic disease was predictive of less healthy lifestyle behavior changes in older adults.

Stratified analysis by BMI group showed positive associations between body image dissatisfaction, symptoms of depression, increased stress levels, lenient restrictive measures at the time of enrollment, living with ≥2 chronic diseases, living situation change (lived with people and moved alone), and less healthy lifestyle behavior changes in those with a usual BMI of <25 kg/m² (Multimedia Appendix 2, Table S5). In participants with a usual BMI of ≥25 kg/m², increased stress levels, maintenance of the same work arrangement, and body image dissatisfaction were positively associated with less healthy lifestyle behavior changes. Living with ≥2 chronic diseases was predictive of less healthy lifestyle behavior changes in this population subgroup. Gender was not a determinant of the outcome within either BMI group.

Stratified analysis by the burden of restrictive measures at the time of enrollment showed body image dissatisfaction, symptoms of depression, and increased levels of stress to be positively associated with less heathy lifestyle behavior changes in both subgroups (Multimedia Appendix 2, Table S6). Living with ≥2 chronic diseases during lenient restrictive measures was predictive of less healthy lifestyle behavior changes. None of the participants recruited during lenient restrictive measures self-identified as a gender minority; gender was only significant within the subgroup of participants recruited during strict restrictive measures.

During the COVID-19 pandemic, more than half of the Canadian COVIDiet study adults reported no change in body weight, smoking status, exercise level, alcohol intake, and sleep quality; however, a dominant proportion (44%) reported worsened eating habits. Considering these lifestyle factors altogether, we identified 2 distinct lifestyle behavior change patterns: healthy and less healthy. The significant and independent determinants of a less healthy lifestyle behavior change pattern were body dissatisfaction, depression, increased stress levels since the pandemic, self-identifying as a gender minority, and the province of residence.

Several surveys have reported on changes in individual lifestyle factors during the pandemic in adults worldwide [10,29-40], but few observational studies have examined the overall lifestyle behavior change patterns and associated risk factors [41,42]. Similar to our findings and using comparable data-driven statistical approaches, one study found “favorable,” “unfavorable,” and “no change” lifestyle change groups in a French cohort [42], while another study identified “prohealthy,” “unhealthy,” and “constant” groups in a Polish population [41].

In this study, the healthy lifestyle behavior change pattern (974/1609, 60.5%) was mainly characterized by an increased physical activity level, unchanged or improved perceived eating habits, and stable weight, sleep quality, and smoking status. An increased physical activity level has also been reported in other countries [29,42]. It may have resulted from increased free time due to telework or from a desire to improve personal health in facing the disease outbreak. In addition, participants allocated to the healthy class were more active at baseline, which may have led them to coping with the pandemic-related stress with exercise [43], thus increasing their physical activity level, while decreasing stress levels. With regard to diet, stable and improved eating habits were observed in our study and elsewhere [4,29,42]. Staying at home may have favored the avoidance of less healthy food, such as fast food, and more frequent cooking for some individuals [4,29,42,44]. One study in the Canadian province of Québec (NutriQuebec study; N=853, with 87.2% women, 52.5% aged 50-69 years) showed a slight overall increase in diet quality (+1.1/100 points in the Healthy Eating Index-2015; 95% CI 0.6-1.5) during early lockdown (April-May 2020), especially in younger adults, in those with lower education, or in those with obesity [4]. Although not directly comparable, only 22.9% (369/1609) of the overall population reported improved eating habits and a larger proportion reported worsened eating habits (709/1609, 44.1%) in our study. In addition, only 35.5% (346/974) reported improved eating habits and the majority reported unchanged habits in the healthy lifestyle behavior change group. This suggests that the diet quality of Canadians may not have improved at a population level; in-depth dietary assessment will permit more robust conclusions. Lastly, the COVID-19 pandemic appears to have had disproportionate impacts on populations living with social inequalities [45]. Participants in the healthy pattern group were more likely to self-identify as men, to report higher and unchanged household incomes, to be satisfied with their body image, and to be living with less than 2 chronic diseases. The lifestyle behavior changes observed in this group may be reflective of social privilege.

Contrastingly, the less healthy lifestyle behavior change pattern (635/1609, 39.5%) was characterized by perceived worsened eating habits, decreased physical activity, decreased sleep quality, and weight gain. Similar patterns of lifestyle habit deterioration were observed in Canada, predominantly in British Columbia [46], and in other countries [47]. Another Canadian study (N=2338, with 90.2% women, February-April 2020) showed that although moderate-to-vigorous activity tracked with wearable devices returned to the prepandemic level 6 weeks after an initial decline in March 2020, the step count and light physical activity remained lower [48]. The lockdown imposed important modifications on daily physical activities, including work-related commuting interruption due to telework or unemployment, the prohibited practice of team sports, and the closure of fitness centers. Those at higher risk of complications and particularly worried of contracting COVID-19 possibly took extra isolation precautions and reduced their physical activity level. In addition, our data show that 1 independent determinant of the less healthy classification was increased levels of stress. Both objective and subjective measures of stress have been shown to hinder physical activity levels, especially in less active individuals [43]. Worsened eating habits may be explained by modified food supply and food choices due to more limited access to grocery stores, fresh food, and working from home. Increased snacking and sweet consumption have been reported in numerous global studies (reviewed in [49]) and may have contributed to participants’ perception of worsened eating habits. In addition, the less healthy lifestyle behavior change class more likely reported increased alcohol intake. Altogether, these factors probably explain the observed body weight increase in this group. Comparably, in the NutriNet-Santé cohort study (France; N=37,252; weighted proportions: 52.3% women, 42.5% aged 25-50 years), among the 3 clusters identified, 1 reflected unfavorable changes that included decreased physical activity, increased snacking and sweet consumption, decreased vegetable and fruit consumption, and weight gain between March and May 2020 [50]. These changes were associated with being a woman, working from home, the presence of children at home, a lower income, and more depressive symptoms. In our study, individuals in the less healthy lifestyle behavior change group also more likely reported increased stress levels and anxiety and depressive symptoms and a greater proportion had more than 2 chronic diseases and reported a decreased income. A healthy lifestyle, including regular physical activity and high diet quality, were associated with reduced risk of chronic diseases [51] and mortality [52]. As a strategy to promote health during self-quarantine, the WHO regional office for Europe released physical activity and dietary guidelines [52]. However, none were made available in Canada in the early months of the pandemic. The impact of such guidelines on public health during the pandemic remains to be assessed.

Among potential risk factors, body image satisfaction was the strongest determinant of lifestyle behavior change patterns during the pandemic, independent of several other factors, including gender, age, and income. Close to half of the studied population (707/1609, 43.9%) reported being not satisfied with their body image during the pandemic, a proportion higher than that observed in a population survey in Canada before the pandemic (15%) [53]. Body image dissatisfaction was also more frequent in those self-identifying as women and gender minorities, compared to men, which is typically observed in Western societies [54,55]. Evidence supports that individuals dissatisfied with their body image may have poorer mental health [56] and may engage in less healthy lifestyle behaviors, such as decreased physical activity [57]. A study showed an association of worsened body image dissatisfaction with increased psychological distress during the pandemic [58]. Yet, in our study, body image dissatisfaction was associated with less healthy lifestyle behavior changes independent of stress, anxiety, and depression. The reverse association is also possible, that is, those who adopted less healthy lifestyle behavior changes, including weight gain, may have experienced a greater level of dissatisfaction toward their body image. In our study, the majority of participants also reported increased stress levels since the pandemic (1026/1609, 63.8%), which was associated with less healthy lifestyle behavior changes. Compelling prepandemic evidence has supported an association of stress, anxiety, and depression with poorer lifestyle behaviors, including weight gain [59-63]. However, the pandemic appears to have exacerbated negative mental health in some individuals, putting them at risk of adopting poorer lifestyle behaviors [17] or vice versa [64].

Limited data of the impact of the COVID-19 pandemic on gender minorities have been collected and are available. In our study, self-identifying as a gender minority was associated with adopting less healthy lifestyle habits; although the sample size was small for this group (24/1609, 1.2%), the proportion was comparable to that reported in Canada (0.33% from Census data) [65]. Current evidence supports high rates of worsening physical and mental health outcomes during the pandemic among individuals belonging to gender minorities [66,67]. From stratified analyses, both self-identifying as a woman and self-identifying as a gender minority were associated with less healthy lifestyle behavior changes among younger participants (<55 years) only. Before the pandemic, women (compared to men) and gender-diverse individuals (compared to men and women) already reported poorer mental health in Canada. Since the beginning of social distancing, negative mental health has been reported in women and gender minorities in Canada [68] and elsewhere [58,69-71]. In people belonging to gender minorities, the lack or loss of access to community groups, stigmatization, and negative psychological outcomes of the pandemic may have contributed to the adoption of less healthy lifestyle behavior changes [66]. In younger women and gender minorities, the telework and at-home childcare arrangement may have further adversely impacted lifestyle behaviors [42,71].

Living in the province of Québec was significantly associated with adopting healthier lifestyle behaviors during the pandemic compared to Ontario and other large provinces. Yet, restrictive measures have been comparable across these provinces throughout the pandemic [1]. Some evidence suggests that the Québec population may have experienced healthier changes [4]. However, the association with healthier lifestyle behavior changes during the pandemic is most likely attributable to greater recruitment in this province during less stringent restrictive measures in our study (Québec: 301/576 during lenient and 71/1033 during strict restrictions; Ontario: 139/576 during lenient and 464/1033 during strict restrictions).

The strengths of this study include the overall large sample size and adequate representation across provinces. Another strength is the use of LCA, as opposed to using measures of central tendency, that allowed us to identify 2 classes and informed on distinct patterns within the population.

The study also has limitations. First, variations in restriction measures were observed across provinces between the start and the end of recruitment, which may have contributed to confounding residuals, although our models were adjusted for the pandemic phases during which participants responded to the questionnaire. Second, the use of self-reported lifestyle questionnaires may have introduced subjectivity to measurements. Third, the use of advertisement strategies combined with the nature of our study may have led to selection bias of more educated and healthier individuals. In addition, the over-/undersampling of specific groups of individuals, namely gender and age, may have limited the generalizability of our results to the Canadian population; however, our sensitivity analysis using gender-based calibration weights showed consistent results with our main analysis (Multimedia Appendix 2, Table S3). Fourth, although subgroup analysis by age, baseline BMI, and pandemic wave group was successfully conducted, it was not possible to examine the associations between risk factors and lifestyle behavior change patterns across ethnic and gender groups; further studies with a larger sample size within each, especially ethnic and gender minorities, are warranted.

Findings from this observational study showed that 39.5% (635/1609) of participants adopted less healthy lifestyle behaviors during the COVID-19 pandemic but that the majority (974/1609, 60.5%) adopted healthier behaviors. Factors such as body image perception, depression status, change in the stress level, and gender identity may have motivated these lifestyle behavior changes. It is yet to be determined whether the latter will sustain in the long term. In the postpandemic period, multidisciplinary strategies involving dietitians, psychologists, and kinesiologists may be considered to assist adults in (re)gaining mental and physical health. Lockdown-adapted public health recommendations and resources related to nutrition, physical activity, sleep, and smoking should be promoted during future outbreaks.