This study was conducted in Alberta, a western Canadian province of ~4.8 million people, including 127,475 self-identified Métis: 20.4% of Métis people in Canada (16). The Otipemisiwak Métis Government of the Métis Nation within Alberta (MNA) represents over 70,000 citizens and is Canada’s largest Métis governance body (16).

The MNA played a central role in the study design, survey implementation, and interpretation of results. This partnership was formalized through a research agreement with a team comprising Métis, MNA representatives, and academic researchers. The study was approved by the University of Alberta Health Research Ethics Board (Pro00104745) and followed the six principles of ethical Métis research outlined by the former Métis Centre of the National Aboriginal Health Organization (17).

We conducted a cross-sectional web-based survey series in three waves from December 2020 to December 2021, aligned with distinct phases of the early COVID-19 pandemic in Alberta. The study adheres to the Checklist for RepOrting Survey Studies (CROSS) (18), and the CHecklist for Reporting Results of Internet E-Surveys (CHERRIES) (19). The first survey (December 18, 2020–January 31, 2021) was conducted during Alberta’s second COVID-19 wave, a period of province-wide lockdowns, gathering restrictions, and limited vaccine availability. The second survey (March 20–April 30, 2021) coincided with the third pandemic wave, marked by high case rates, renewed lockdowns, and expanded vaccine access. The third survey (November 1–December 20, 2021) occurred amidst mask mandates, vaccine passports, and vaccine booster doses.

Eligible participants were a convenience sample (20) of self-identified Métis individuals aged ≥16 residing in Alberta, recruited via email to registered MNA citizens and targeted posts on the MNA’s Facebook and Twitter pages. The posts outlined the survey’s purpose, eligibility, and included a secure link. Though broadly accessible, outreach focused on individuals closely connected to the MNA, limiting non-Métis participation. After reviewing an information letter, participants provided consent and completed a web-based REDCap questionnaire, piloted before launch. Each wave included a $100 gift card draw. Responses were collected via REDCap (University of Alberta); no personal identifiers were used. Data were encrypted, securely stored, and access was MNA-governed.

The Misi Yehewin surveys were administered on a single scrolling page to reduce loading time and fatigue. Each wave included items on sociodemographic characteristics, COVID-19 experiences, physical and mental health, health behaviours, community support, racism, cultural identity, and economic conditions. While core content remained consistent, item counts increased across waves: 69 (Wave 1), 78 (Wave 2), and 95 (Wave 3). Branching logic determined item visibility and reduced burden; a subset of 66 items was repeated across waves, with 28 included in the current analysis (Supplementary Table S1). Because the analysis focused on comparisons across waves, we restricted variables to items that were collected in all three waves. Some items (e.g., COVID-19 vaccination) were included only in one of the waves and were not analyzed.

Survey items were adapted from validated general and Indigenous-specific tools and reviewed by MNA team members for cultural relevance, clarity, and alignment with Métis community priorities, with wording and response options refined through discussion and agreement before each survey wave was launched. Likert-type formats and adaptive branching logic were used. Completion time was ~25 min. Responses were not editable and completeness checks were not enforced. All items included non-response options (e.g., “Prefer not to say”). To ensure privacy, no IP tracking or cookies were used. Full questionnaires are available upon request.

Analyses focused on three domains: self-reported COVID-19 events, health and wellbeing, and social, cultural, and economic wellbeing. COVID-19 events included self-reported symptoms, testing, positive results, and hospitalizations. Health and wellbeing outcomes included perceived changes in health and quality of life and symptom screening for depression (Patient Health Questionnaire-2 [PHQ-2] ≥ 3) (21), anxiety (Generalized Anxiety Disorder-7 [GAD-7] ≥ 3) (22), and stress (Perceived Stress Scale-4 [PSS-4] ≥ 6) (23). Health behaviors encompassed reductions in physical activity, outdoor leisure time, and harvesting, hunting, or gathering, along with poorer sleep quality, and increases in screen time, and fast-food consumption. Social, cultural, and economic wellbeing indicators included time with family, land-based and cultural activities, virtual interactions, community support, cultural identity, experiences of discrimination and racism, food insecurity, and financial strain.

View and participation rates were not tracked as email and social media links prevented visitor identification. Completion rate was calculated as page submissions among those who initiated the survey. Sociodemographic characteristics were summarized using frequencies and percentages. Variables included age group (16–24, 25–64, 65+), gender, relationship status, income, education, employment, and residence. Differences across waves were tested using Chi-square tests, excluding missing values. For adjusted analyses, we initially considered all measured sociodemographic characteristics as potential covariates. Those that differed significantly across waves (p < 0.05) –age group, gender identity, household income tertile, and employment status– were retained in all regression models to account for differences in sample composition across waves.

Outcomes were summarized by wave using frequencies and proportions (expressed as percentages). Response categories were collapsed (e.g., “increased a lot” and “increased a little”) and dichotomized to reflect positive or negative experiences. The Jonckheere–Terpstra test (24) was used to assess whether there were ordered sample-level differences in each outcome across the three survey waves (statistical significance set at p < 0.05). This non-parametric test is appropriate for detecting ordered differences across independent groups (24). Based on the direction of wave-specific estimates, outcomes were classified as higher in later waves, lower in later waves, or similar across waves. Outcomes with statistically significant differences across waves were further analyzed using logistic regression models adjusted for age, gender identity, household income, and employment to account for differences in covariate distributions across waves. Predicted proportions from these models were used to estimate adjusted percentage differences between waves, with Bonferroni-adjusted 95% confidence intervals (CI) reported. As samples were independent across waves, results do not reflect individual-level changes. Missing data were not imputed; analyses used complete case observations. We estimated age-related sampling bias by comparing the sample age distribution with the 2021 census data (Supplementary Tables S2, S3), and conducted age-stratified analyses for age groups 16–24, 25–64, and ≥65 years as a sensitivity analysis (Supplementary Tables S4–S7), with exact cell counts <10 suppressed to preserve confidentiality. All analyses were conducted using Stata (Release 18; StataCorp LLC, College Station, TX).

A total of 3,052 responses were received across the three survey waves: 1,508 in Wave 1, 749 in Wave 2, and 795 in Wave 3. After excluding 613 incomplete or ineligible responses, the final analytical sample included 2,439 participants: 1,170 in Wave 1, 539 in Wave 2, and 730 in Wave 3 (Figure 1). Sociodemographic characteristics are shown in Table 1. Most respondents were aged 25–64 and identified as women. Compared to earlier waves, Wave 3 included fewer older adults (p < 0.05), more women (p < 0.05) and a higher proportion of participants in the lowest income tertile (p < 0.05). Employment status also differed significantly across waves, with fewer students and more full-time homemakers in Wave 3.

Experiences with COVID-19 differed across survey waves corresponding to different phases of the pandemic (Table 2). The proportion of participants who reported experiencing symptoms was 3.8% in Wave 1 and 20.6% in Wave 3, (adjusted difference: 14%; 95% CI: 8, 19%). Attempts to get tested were reported by 54.5% of participants in Wave 1 and 64.2% in Wave 3 (adjusted difference: 8%; 95% CI 2, 14%). Confirmed positive test results were reported by 5% of participants in Wave 1 and 15.1% in Wave 3 (adjusted difference: 6%; 95% CI: 1, 11%). Hospitalizations were rare: fewer than 10 participants in Waves 1 and 2 reported being hospitalized due to COVID-19, with 2.3% in Wave 3. Age-stratified analyses of COVID-19 experiences (Supplementary Table S4) were broadly consistent with the main, particularly for participants aged 25–64 years. In strata with small cell counts (<10), some outcomes (e.g., COVID-19 symptoms and positive test results) and imprecise estimates and between-wave differences were not assessed.

Perceived declines in health were most commonly reported in Wave 1 (Table 2). The proportion of participants who reported worsened physical health was 42.4% in Wave 1and 25.8% in Wave 3 (adjusted difference: −15%; 95% CI: −21, −10%). Perceptions of worsening mental health were reported by 62.1% of participants in Wave 1 and 31.1% in Wave 3 (adjusted difference: −32%; 95% CI: −37, −26%). Reports of lower quality of life were 71.1% in Wave 1 and 54.1% in Wave 3 (adjusted difference: –17, 95% CI: −23, −11%). Despite these between-wave differences in perceived wellbeing, the proportion of participants screening positive for depressive symptoms (~41%), anxiety symptoms (~47%), and high perceived stress (~68%) was similar across waves. Supplementary Table S5 presents the age-stratified results for physical and mental health, showing consistent findings across age groups, except for perceived worsening in physical health in the ≥65 age group, in which no statistically significant differences across waves were observed.

Negative health behaviour changes were more commonly reported in earlier survey waves (Table 2). The percentage of participants reporting decreased moderate-to-vigorous physical activity was 41.7% in Wave 1 and 31.5% in Wave 3 (adjusted difference: −9%; 95% CI: −16, −3%). Similarly, spending less time on outdoor leisure activities was reported by 41.2% of participants in Wave 1 and 32.9% in Wave 3 (adjusted difference: –8, 95% CI: −15, −2%). Participation in harvesting, hunting, or gathering activities was 34.9% in Wave 1 and 25.6% in Wave 3 (adjusted difference: –8, 95% CI: −16, −1%). Increased screen time was reported by 76.2% of participants in Wave 1 and 54.0% in Wave 3 (adjusted difference: –21, 95% CI: −28, −15%). In contrast, the reporting of poor sleep quality (~45%) and fast-food consumption (~13%) was similar across waves. Stratified analyses (Supplementary Table S6) showed generally consistent patterns across age groups, except for those aged ≥65 years, where small cell counts led to imprecise estimates and limited between-wave comparisons for decreased moderate-to-vigorous physical activity and decreased outdoor leisure activities.

Social and cultural disruptions were more frequently reported in earlier survey waves (Table 3). The proportion of respondent who said they spent less time with family was 49.7% in Wave 1 and 30.4% in Wave 3 (adjusted difference: −19%; 95% CI: −26, −12%). Reduced participation in cultural activities was reported by 31.7% of participants in Wave 1 and 20.4% in Wave 3 (adjusted difference: −13%; 95% CI: −21, −5%). Increased virtual activity was reported by 52.4% of participants in Wave 1 and 33.0% in Wave 3 (adjusted difference: −18%; 95% CI: −31, −6%). Perceptions of community support and cultural identity remained relatively similar across waves. Food insecurity was reported by 39.4% of participants in Wave 1 and 52.9% in Wave 3 (adjusted difference: 12%; 95% CI: 6, 18%). The reporting of financial strain (~56%), witnessing racism (~54%), experiences of racism (~25%), and strong Métis identity (~89%) showed no significant differences across waves. Stratified analyses (Supplementary Table S7) showed these results were broadly consistent across age groups.

This study has several strengths. Its cross-sectional survey series design enabled assessment of sample-level ordered differences across distinct phases of the COVID-19 pandemic. Partnership with the MNA through a formal research agreement ensured the study was culturally grounded and aligned with Métis self-determination. This collaborative approach reflects growing evidence that Indigenous-led research is key to addressing health inequities and producing actionable insights (1, 27). The pandemic highlighted the effectiveness of Indigenous-led public health responses, reinforcing calls for greater Métis governance in health policy (30). Importantly, this study fills a critical gap by generating Métis-specific data during a period when national surveillance systems largely excluded this population.

Some limitations must be considered. The convenience sample limits the generalizability, as participants may differ from the broader Métis population. We used complete case analysis and did not impute missing values because the missing-at-random assumption was unlikely to hold for variables with substantial missingness, such as income. Taken together, these factors mean that observed differences between survey waves may reflect a combination of temporal changes in the underlying population, sampling variation, cohort effects, or selection bias inherent in the convenience sampling approach. Younger Métis people, particularly those aged 16–24, were underrepresented in our sample, and we could not calculate participation rates because the number of eligible individuals at each online survey wave was unavailable. Recruitment through MNA networks and web-based platforms required internet access and basic digital skills, so individuals with limited digital literacy or connectivity may be underrepresented. Women comprised most respondents in all waves, which may limit generalizability to men and gender-diverse Métis people. To partly address these imbalances, we adjusted all models for gender and conducted age-stratified analyses, which yielded results consistent with our main findings. Estimates for outcomes that vary by age (e.g., screen time, food insecurity, and mental health) may therefore not fully capture patterns in the broader Métis population, and the direction and magnitude of this bias cannot be determined with our data.

We did not collect comparable vaccination data across all waves, in part because vaccines were not yet available in Wave 1 and were only beginning to roll out during Wave 2, so we could not assess vaccination as a potential moderator of differences in outcomes between waves. More broadly, differences between earlier and later waves may also reflect unmeasured changes in the pandemic context, including vaccination rollout, evolving public health measures, and other time varying factors that were not captured in the study. Self-reported data may be subject to recall or social desirability bias, and COVID-19 status and related events were not verified with medical records or test documentation, so some misclassification is possible. Adjusted estimates are intended to account for major sociodemographic differences in sample composition across survey waves and should not be interpreted as causal effects of pandemic phase. Finally, the cross-sectional series design prevents causal inference or tracking of individual experiences over time; findings reflect sample-level differences across three independent survey waves.