Measuring inequalities in COVID-19 vaccination uptake and intent: results from the Canadian Community Health Survey 2021

The data used in this study were drawn from the 2021 collection cycle of the Canadian Community Health Survey. The CCHS is an annual cross-sectional and nationally representative survey conducted by Statistics Canada. Briefly, the survey collects information on health status, health care utilization and health determinants for Canadians 12 years or older residing in the 10 provinces and 3 territories, excluding full-time members of the Canadian Forces, children in foster care, those who live in institutions, those who live on reserve, as well as those living in Nunavik and Terres-Cries-de-la-Baie-James. The survey sample was selected using a multistage stratified cluster design. Due to the complex survey design, survey weights were given to each respondent and were calibrated by province, age group, and sex to ensure that the sample data is representative of the Canadian population. More details on the survey design and sampling method are available in a previously published report [13]. In September 2020, questions about experiences related to the pandemic along with a question on COVID-19 vaccination intent were introduced, followed by a question on COVID-19 vaccination uptake in March 2021.

The CCHS has multiple non-overlapping collection periods throughout the year and this study is based on the collection period for the months of June to August 2021. For this specific period, data was collected between June 1^st and September 5^th exclusively using computer-assisted telephone interviewing (CATI). The data file used for the current study included only respondents living in the provinces for a total of n = 10,093 participants. Since adolescents 12 to 17 years old were only offered appointments for COVID-19 vaccines towards the end of May in most provinces and would not have had sufficient time to get vaccinated prior to the start of the data collection period, they were excluded from this study resulting in a final sample of n = 9,509. The response rate for the population of study, Canadians aged 18 and older living in the provinces, was 22.6%.

Respondents were asked to provide information on sociodemographic and health-related characteristics, as well as experiences during the COVID-19 pandemic.

All analyses were conducted using SAS software version 9.4 of the SAS System for Windows.¹ Descriptive statistics (frequencies and percentages) were computed to examine the distribution of vaccination status, vaccination intent, as well as sociodemographic factors in the population. To compare vaccination status among sociodemographic groups, the proportion of unvaccinated individuals was broken down by sociodemographic factors. Likewise, the proportion of those unlikely to get vaccinated by sociodemographic factor was also calculated to compare vaccination intent among different groups. The confidence interval of these proportions were then adjusted using the Wilson interval to account for the normal approximation of the binomial distribution [16]. For the final analysis, unadjusted (simple) and adjusted (multiple) logistic regression models were employed to examine associations between the response (vaccination status or intent) and predictor (sociodemographic and health-related) variables. The unadjusted models ran each of the predictor variables against each of the two outcomes, whereas the adjusted models included all predictor variables in the model.

To identify the sociodemographic determinants of non-vaccination and low vaccination intent, we examined the odds ratios (OR) generated by the logistic regression models. Specifically for vaccination status, we tested differences in the odds of being unvaccinated versus having at least one dose among sociodemographic groups. Similarly, for intent, we tested differences in the odds of being unlikely to get vaccinated versus being already vaccinated or likely to get vaccinated. A Tukey–Kramer adjustment was applied to the confidence intervals associated with the odds ratios to account for multiple comparisons.

Sampling weights were used to obtain accurate and representative estimates for the proportions and odds ratios, and their corresponding estimated variances were calculated using 1,000 bootstrap weights. All estimates were computed using a significance level of alpha = 0.05.

According to this nationally representative sample, COVID-19 vaccination coverage in the ten Canadian provinces from June to September 2021 was 89% (Table 2). Overall, the proportion of unvaccinated did not differ by region. Differences across genders, visible minority status, mother tongue, and history of COVID-19 diagnosis were also not detected although vaccination rates improved with increasing age.

After adjusting for covariates, vaccination status was associated with level of education, presence of children under 12 years old in the household, having a regular healthcare provider, and self-perceived health (Table 2 and Fig. 1). Adults with education below university were more likely to be unvaccinated than university graduates, with adjusted odds ratios (aOR) ranging from 2.4 (95% CI 1.5–3.7) to 3.5 (95% CI 2.1–6.1). Those living with at least one child under 12 years old were also more likely to be unvaccinated than those living without children (aOR 1.6, 95% CI 1.1–2.4). The risk of being unvaccinated was also higher in those without a regular healthcare provider compared to those who had one (aOR 1.6, 95% CI 1.1–2.2) and in those who perceived their health as fair or poor compared to those who perceived it as excellent, very good or good (aOR 1.8, 95% CI 1.3–2.4). As expected, due to the vaccine rollout by descending age group, vaccination decreased significantly with decreasing age.

Only 5% of the population did not intend to get vaccinated (Table 3). The proportion of those unlikely to get vaccinated were slightly higher in the Prairies compared to the other regions. Vaccination intent did not differ across Indigenous identity, gender, employment status, marital status, country of birth, mother tongue, dwelling ownership status, having a regular healthcare provider, or status of COVID-19 diagnosis. Based on the adjusted model, vaccination intent was associated with region of residence, age, level of education, visible minority status, presence of children under 12 years old in the household, household composition, and self-perceived health (Table 3 and Fig. 2). The risk of being unlikely to get vaccinated was greater in the Prairies compared to the risk in the province of Quebec (aOR 2.2, 95% CI 1.2–4.1). Additional comparisons among the regions also showed that risks were greater in the Prairies than in the Atlantic region. Greater risks were also observed in those less than 40 years old and between 50 to 59 compared to those 70 and over (aOR between 2.8, 95% CI 1.2–6.8 and 4.0, 95% CI 1.3–12.3). The intent not to get vaccinated was more frequent in individuals with no university degree than in university graduates (aOR between 2.5, 95% CI 1.5–4.3 and 3.8, 95% CI 1.9–7.6). The intent not to get vaccinated was also more frequent in those living with at least one child under 12 years old than in those with no children (aOR 1.8, 95% CI 1.1–2.9), more frequent in unattached individuals than in those living in a couple (aOR 2.6, 95% CI 1.1–6.1), and more frequent in those who perceived their health as fair or poor than in those perceiving it as excellent, very good or good (aOR 2.0, 95% CI 1.3–2.9). However, decreased risks (aOR 0.3, 95% CI 0.2–0.7) of being unlikely to get vaccinated were observed in those who are part of a visible minority group compared to those who are not a visible minority.

In this study, vaccination coverage among Canadian adults was high, with 89% having received at least one dose of a COVID-19 vaccine. Only 11% of the population were unvaccinated. This coverage is somewhat higher than what was reported by the Canadian COVID-19 vaccination coverage surveillance system where by beginning of September 2021, 84% of those 18 and older received at least one dose, which may be explained by differences between the CCHS respondents and the general population [17]. Among all the sociodemographic factors included in the vaccination status model, age, level of education, presence of children under 12 years old in the household, having a healthcare provider and self-perceived health were identified as significant determinants of COVID-19 vaccine uptake.

The proportion of unvaccinated individuals decreased with increasing age ranging from 17% in 18–29 years old to 4% in those aged 70 years or over. This can be partially explained by the vaccination rollout, as the elderly population was eligible to get vaccinated earlier in the campaign [18]. Analogously, some studies highlighted that risk perceptions toward COVID-19 differ by age [19, 20]. Given that the severity and mortality of COVID-19 increases with age [21], older adults might feel more at risk, thus more motivated to be vaccinated. In fact, increased risk perceptions of COVID-19 have also been shown to be a strong predictor of COVID-19 vaccine acceptance [22].

In this study, only one of the three proxy variables for SES, namely education, was significantly associated with vaccination status. Individuals with less than university education had a higher risk of being unvaccinated than university graduates. Similarly, a recent study in the US looked at patterns in COVID-19 vaccination coverage among adults and reported that vaccination uptake was lower among adults with low educational attainment, which is consistent with our results [23]. Education is proven to be associated with greater engagement in pro-health behaviors which could be seen as a factor in favor of vaccination [24].

According to the CCHS results, individuals who lived with at least one child under 12 years old had a higher risk of being unvaccinated. One might postulate that this association is driven by age since individuals living with young children are, for the vast majority, younger adults. However, it should be noted that this association is still significant when adjusted for age groups. A study in the US found similar results where the presence of children was negatively associated with COVID-19 vaccination [23]. As postulated by Bell et al., this may be related to access barriers to vaccination; parents of younger children may face obstacles to schedule and attend vaccination appointments due to competing priorities [25].

The risk of being unvaccinated was also significantly higher in those who had fair or poor self-perceived health compared to those with excellent, very good or good self-perceived health (aOR 1.8). This could be explained by the fact that individuals with poor health had less intent to get vaccinated, and therefore the coverage was lower. A study in the U.S found that those with underlying medical conditions and BMI > 40 were not more willing to get vaccinated than those without these risk factors [26]. This could be associated with different perceptions on vaccine safety, side effects and effectiveness among those with poor health. One study demonstrated that people cared more about the vaccine’s health risk than its effectiveness [27]. Therefore, those with poor health may be more concerned with health risks associated with getting vaccinated than being immunized against COVID-19. On top of that, reduced COVID-19 mortality risk following immunization may in part explain the current finding as it suggests substantial “healthy vaccinee effects” which refers to a situation when vaccinated individuals tend to be healthier than unvaccinated individuals [28]. This pattern is the opposite of what was observed for influenza vaccination in Canada in previous cycles of the same survey (CCHS) where excellent self-perceived health was associated with non-vaccination among adults aged 18 to 64 years with a chronic medical condition and in adults aged 65 years and older [29].

Finally, having a regular healthcare provider was positively associated with COVID-19 vaccine uptake. It is conceivable that individuals with a regular healthcare provider may have easier access to health-related resources and may be more willing to get vaccinated in order to protect themselves from the disease. Additionally, multiple studies have demonstrated that those who are hesitant to get a COVID-19 vaccine are concerned about the safety of the vaccines and the risks and side effects attached to it [22, 30‐33]. Considering healthcare providers as being one of the trusted sources of information on vaccination, they may help to soothe the fear and concerns over COVID-19 vaccines, reassure their patients on the safety and effectiveness of the vaccines, and promote vaccination during consultations, which can therefore improve vaccine uptake [8, 34].

The proportion of the population who did not intend to get vaccinated against COVID-19 was as low as 5%. When adjusting for all other predictor variables, lower vaccination intent was significantly associated with region, younger age, lower educational attainment, not being part of a visible minority group, presence of children under 12 years in the household, unattached individuals and poor self-perceived health.

Vaccination intent differed between the Canadians provinces. Individuals living in the Prairies had higher risks of not intending to get vaccinated compared to other provinces. In the same vein, another Canadian study also found that these three provinces had higher proportions of individuals who did not intend to get vaccinated [7]. However, once adjusted for other sociodemographic factors, these differences were not significant despite individuals from Alberta having higher predicted probability of not intending to get vaccinated [7].

Adults younger than 40 years old and between 50 to 59 had lower COVID-19 vaccination intent than those aged 70 or over. Other Canadians studies had similar results where individuals below 60 years of age demonstrated lower intent to get vaccinated [5, 7]. A systematic review of 45 studies conducted in various countries hypothesized that older individuals have a greater sense of responsibility and accountability for themselves and their surroundings compared to the younger population which may also explain why older individuals were more likely to be vaccinated [35].

Following the pattern of vaccination status, the intent not to get vaccinated was more frequent in individuals with no university degree than in those who held such a degree. This is in agreement with another Canadian study that reported having a university education level as one of the strongest predictors of COVID-19 vaccine intention [22]. Education level plays an important role in vaccination acceptance as it highly correlates with belief in COVID-19 vaccine safety [36]. According to Kricorian et al., individuals who believed COVID-19 vaccines to be unsafe were likely to have difficulty understanding scientific information, higher mistrust in scientific research, and not to follow scientific recommendations. This could contribute to the lower intent of receiving the vaccine.

Moreover, visible minorities overall were found to be more eager to get vaccinated than the rest of the population. A major caveat to this finding was that it applies to visible minorities as a whole, as the number of participants from these groups in this study was not sufficient to analyse them separately. Supporting evidence from a US study indicates that some visible minority groups such as Asians and Hispanics are less likely to have vaccine hesitancy than Whites across all hesitancy measures [37]. Nevertheless, according to other studies in other countries, higher vaccine hesitancy was observed in most minority ethnic groups compared to the White British or Irish group; and identifying as Black or African American was associated with lower vaccination likelihood as opposed to identifying as White [38, 39]. The association between visible minority status and vaccination intent observed in the current study may in part be explained by the multi-ethnic characteristic of the Canadian healthcare workers. In Canada, visible minorities account for approximately one third of nurse aides, orderlies and patient service associates, with higher proportions of Black, Filipino and South Asian workers in these occupations [40]. In addition to having been prioritized for vaccination, being at increased risk of COVID-19 infection and transmission may contribute to healthcare workers increased willingness to get vaccinated. Further exploration is essential to better understand the association between the various visible minority groups and COVID-19 vaccination intent.

Presence of children under 12 years old in the household was negatively associated with COVID-19 vaccination intent (aOR 1.8). A Canadian study on predictors of vaccine hesitancy for COVID-19 public health messaging implications revealed that having more than three children in the family is a strong determinant of immunization noncompliance [41]. This finding is also consistent with other research where presence of children in the household increased the odds of vaccine hesitancy [37, 42]. It could seem counterintuitive given that having multiple children ought to encourage parents to vaccinate in order to prevent cross-infection within the household. Nonetheless, relations between family size and vaccination intent may be explained through other socioeconomic factors.

Additionally, unattached individuals had lower COVID-19 vaccination intent than coupled individuals. A US nationwide study on predictors of intention to vaccinate against COVID-19 also demonstrated that having a spouse or partner was associated with higher anticipated likelihood of vaccination [43]. Unattached individuals might not have as much collective family responsibilities as married individuals and those with children, which could explain their lower vaccination intent [35].

Lastly, the current study showed that individuals with fair or poor self-perceived health had lower intention to vaccinate against COVID-19 than those with excellent, very good, good self-perceived health. This is in agreement with a study on COVID-19 vaccine hesitancy associated factors in Saskatchewan where individuals with very good or excellent health status were more likely to vaccinate than those with poor of fair health status [6]. However, more research is warranted to examine the association between self-perceived health and COVID-19 vaccination intent, especially since many perceived that aspects of their overall health had deteriorated during the pandemic [44].

In the present study, some sociodemographic factors such as gender, immigration status, Indigenous identity, and employment status were not significantly associated with COVID-19 vaccination uptake or intent. However, previous studies conducted in Canada showed that being a male was positively associated with vaccination intent [22, 45]. Although no difference in uptake for Indigenous status was observed in the adjusted model, Indigenous groups might still differ from non-Indigenous. It should be noted that statistical non-significance is not proof of absence of an association. Sometimes, the non-significant result is due to lack of power rather than lack of effect; the sample size of Indigenous respondents might be too small to provide sufficient power to detect an association. This can also be true for small groups of other variables such as immigration status or COVID-19 status. Other Canadian studies with various sample sizes of Indigenous respondents and somewhat different target populations found that Indigenous groups and individuals born outside of Canada had lower odds of getting vaccinated [5‐7].

Given the paucity of studies assessing inequalities in COVID-19 vaccination uptake and intent, further work is needed for a deeper understanding of the contributing factors in the Canadian context.

Factors other than sociodemographic can also play a role in vaccination uptake and intent. Health inequities, vaccine hesitancy as well as knowledge, attitudes and beliefs (KABs) are among the multitude of factors that can have an impact. Although some KABs have been associated with sociodemographic characteristics, including only sociodemographic variables in the models might not provide a comprehensive picture. Unfortunately, information on KABs was not collected in the CCHS. Nonetheless, the assessment of sociodemographic factors can inform interventions by identifying target groups. Some people do not intend to get vaccinated due to concerns about the safety and effectiveness of the vaccine [22]. The novelty of COVID-19 vaccines could also play a role in Canadians’ intent to get vaccinated, as well as their lack of knowledge about vaccination [22, 45].

As with any large scale survey, the CCHS has several strengths and limitations that must be carefully considered when interpreting the results. A major strength of the survey was the sufficiently large sample size to allow for analysis by several sociodemographic and health-related factors. Additionally, given the complex survey design and the use of survey weights, the findings are nationally representative and allows us to make inferences to the Canadian population. This study can also be a catalyst to potential additional works to examine hypotheses on changes of vaccination status and intent over time, on intent at the provincial level, and on the impact of additional sociodemographic indicators such as household income and rural/urban living area. Most importantly, this study is one of few that examine vaccination status and intent at the national level in Canada, contributing to the growing body of research on COVID-19 vaccine acceptance or hesitancy.

Some study limitations need to be acknowledged. The CCHS shares the usual limitations of surveys based on self-reporting which may be subject to recall bias given that the data was collected more than 7 months after the beginning of COVID-19 vaccination. However, recall bias is less likely to occur in the present study due to high media coverage surrounding the COVID-19 vaccination campaign and the proof of vaccination credentials issued by many jurisdictions across Canada. There are also some limitations to collecting data only through telephone interviews [46]. As a result of the COVID-19 pandemic, no computer-assisted personal interviewing (CAPI) was conducted in 2021; only CATI was used to collect data. Consequently, CATI is limited by the fact that participants have the possibility to not answer the phone whereas they are a lot less comfortable refusing an interview when they are facing the interviewer in person often resulting in lower response rates for CATI compared to CAPI. Indeed, the CCHS response rates significantly decreased in 2021. As was done for previous CCHS cycles, survey weights were adjusted to minimize any potential bias that could arise from survey non-response; non-response adjustments and calibration using available auxiliary information were applied. Despite these rigorous adjustments and validations, the higher non-response rate increases the risk of a remaining bias and increases the magnitude with which such a bias could impact estimates produced using the survey data. Moreover, selection bias cannot be ruled out since individuals with greater interest in the topic could be more likely to respond to the survey. In addition, as with all surveys, the social desirability bias needs to be considered.

In addition, the small number of observations among visible minority groups prohibited a more detailed breakdown of the visible minority status variable by individual visible minority group. This may explain why our finding on visible minority is inconsistent with other studies conducted in Canada or elsewhere. As with many other Statistics Canada surveys, the CCHS excluded First Nations on-reserve communities. Moreover, for the Indigenous status variable, the small number of observations did not allow a further analysis broken down by First Nation, Métis and Inuit. Future research should strive to include a sufficient number of visible minority and Indigenous participants to allow more detailed analyses of intent to get vaccinated and vaccination coverage in these populations. Continued collection would allow for data pooling to increase the sample size and further explore sub-populations.