Background

The COVID-19 pandemic has both directly and indirectly impacted population health worldwide. As of May 2022, the cumulative number of COVID-19 confirmed cases was 520 million worldwide, with 3.83 million in Canada [1]. Patients infected with COVID-19 have various symptoms and multiple organs being impacted [2, 3]. Some patients may have post-acute COVID-19 syndrome which lasts more than 4 weeks or even chronically [4]. With massive medical resources devoted to treating COVID-19 and concerns about protecting patients from infection, people without COVID-19 may experience delayed healthcare [5], which potentially threatens people’s health. In addition, several systematic reviews have shown increasing prevalence of a variety of mental health problems during the pandemic, e.g., anxiety [6], depression [7], posttraumatic stress and psychological stress [8]. Statistics Canada reported that Canadians had worse self-perceived mental health in 2020 compared to pre-pandemic, especially among females [9]. Varied factors contributed to these mental illnesses, e.g., COVID-19 infection [10] or fear of infection [11], social and economic factors such as restrictions on social gatherings and travelling [12], and unemployment [13].

Emerging evidence suggests that general population’s health-related quality of life (HRQL) was impacted during the pandemic. Using tools like the EQ-5D-5L [14], research in China [15], Morocco [16], Portugal [17], the US [18], and Vietnam [19] has identified lower self-reported HRQL among the general population during the pandemic compared to pre-pandemic reference values (population norms). Before the pandemic, young people had the highest HRQL in many jurisdictions [20, 21], but they tended to experience greater HRQL loss during the pandemic [15, 18]. It remains unknown how the COVID-19 pandemic has impacted population HRQL in Alberta or Canada. However, in a study focusing on emergency department and primary care settings in Alberta, Canada, HRQL measured by the EQ-5D-5L among patients seeking primary care had a remarkable decline, especially for females and young people [22].

The future trajectory of COVID-19 is anticipated to be endemic and perhaps with seasonal peaks [23], so the impact of COVID-19 on HRQL is likely to exist at least for several years and potentially it will re-shape population norms. Some subpopulations, e.g., younger people, might be emerging vulnerable populations during the pandemic. Exploring HRQL of the general population and possibly identifying new population norms help governments and society better understand population health and invoke their attention on related health issues and those vulnerable populations. Besides, population norms are useful guides to interpret HRQL scores observed in individuals, groups of patients, or subpopulations [24], which further informs resource allocation and decision-making. Updated population norms as a valid reference value for comparisons are warranted. In Alberta, the use of EQ-5D-5L to measure HRQL is popular, and pre-COVID EQ-5D-5L population norms [25] were available and applied in many clinical studies. Exploring new population norms will be relevant and informative to end-users.

During the COVID-19 pandemic in Canada, there has been a lack of information on its impact on HRQL among the population. Considering the potential existence of new population norms and the associated benefits for policy-making, we aimed to examine the impact of the pandemic on HRQL, measured by the EQ-5D-5L, among the general population in the Province of Alberta.

Methods

Data

The data came from two waves of a repeated cross-sectional population-based survey conducted by the Health Quality Council of Alberta (HQCA), the COVID-19 Experiences and Impact Survey, which asked respondents questions concerning public health measures, health system access, and the support are available to stay informed, well, and protected during the pandemic [26]. This online survey invited over 15,000 Albertans to participate via several email lists and was also promoted via social media, news, online advertisement, and the HQCA website [27]. The first wave of data collection (survey 1) was between May 25 and June 29, 2020, and the second (survey 2) took place between October 1–31, 2020. HRQL was measured using the EQ-5D-5L. Questions on demographics, socioeconomic status, underlying health conditions, health behaviors, and the COVID-19 situation were included in the survey. The study procedures were approved by the Health Research Ethics Board at the University of Alberta (Pro00101660).

Pandemic and public health response during the survey period

The 2020 Alberta population was 4.4 million individuals, with 81% living in an urban area. The first confirmed COVID-19 case in Alberta was reported on March 5, 2020, and the first wave peaked on April 30, 2020, with 3022 active cases. The first wave continued until September 2020, with 18,175 cases in total. The second wave started in October 2020, and by October 19, there were 3,138 active cases, which was the highest reported in Alberta since March 2020 [28]. The second wave ended in February 2021 with a total of 115,550 positive tested cases in the province [29].

The government of Alberta initiated public health responses in March 2020, with restrictions on all gatherings, travel, and non-essential business, and the closure of schools and universities. Stage 1 of re-opening in mid-May 2020 lifted some restrictions, and with stage 2 in June 2020 more restrictions were lifted. Stage 2 continued until November 2020 when restrictions were re-instituted. The HQCA survey 1 began towards the end of stage 1 and continued into stage 2, and the survey 2 occurred near the end of stage 2.

Measures

The EQ-5D-5L is a generic preference-based HRQL measure, which includes a descriptive system and a visual analogue scale (EQ-VAS) [30]. The descriptive system has five dimensions: mobility, self-care, usual activities, pain/discomfort, and anxiety/depression, each with five levels: 1 = no problems, 2 = slight problems, 3 = moderate problems, 4 = severe problems, 5 = extreme problems, describing 3125 distinct health states. The EQ-VAS is a vertical visual analogue scale ranging from 0 “worst health you can imagine” to 100 “best health you can imagine”, and is used to assess overall respondent health [30]. The EQ-5D-5L is recommended to measure HRQL in Alberta and Canada [31]. The Canadian EQ-5D-5L value set is available to calculate EQ-5D-5L preference-based index scores, ranging from − 0. 149 to 0.949, with higher scores indicating better HRQL [32]. EQ-5D-5L population norms were published for the Alberta population in 2018, with a mean index of 0.84 and a mean VAS score of 77.4 [25]. Alberta population norms also reported dimensional proportions and HRQL distribution for age group, sex, and residence within Alberta Health Service’s five geographical zones.

The HQCA survey collected information on age group, gender, and residential location. We re-grouped the age variable based on age group categories of the Alberta norms population. For the gender variable, we only analyzed HRQL among those who identified themselves as a man or a woman. Postcodes, which could identify the residence in health zones, had many missing values. A potential substitute variable is dwelling (urban/rural) as health zone-specific HRQL in Alberta population norms showed urban health zones (Calgary and Edmonton) had slightly better HRQL compared to the rural zones (North, Central, and South zones) [25]. However, this variable was only administered in the second survey.

Other variables we used were for examining the comparability and representatives of our sample, and these included socio-demographics (levels of education, whether born in Canada, ethnicity, work sector, household income, and financial situation), underlying health conditions (diabetes, heart/vascular disease, liver/kidney disease, having or had cancer, lung disease, and autoimmune disease), health behaviors (smoking and cannabis consumption), and the COVID-19 status (had been infected or not).

Data analysis

All analyses were performed in Stata version 14. We weighted the survey sample based on age and gender within each health zone to increase the representativeness. We did not include other socio-demographic variables for weighting factors, since more variables would limit the number of observations in each stratum. For respondents with missing health zone, we treated them as a group. Our robustness analyses showed the HRQL outcomes were insensitive to our weighting strategy, when comparing with no weighting implemented and weighting on a subsample excluding respondents having missing health zones (data not shown).

We examined the sample representativeness before and after weighting. Reference data for demographics and socioeconomic status variables, except for unemployment rate, were from the most recent 2016 Canadian census [33]; unemployment rates were from the Government of Alberta [34]; underlying health conditions were from Alberta Health’s Interactive Health Data Application [35]; health behaviors were from the 2020 Canadian Community Health Survey [36] and National Cannabis Survey [37]; and COVID-19 infection data were from the Government of Alberta [29].

We explored weighted HRQL in survey 1 and survey 2 in the whole population and then in terms of age group, gender, and dwelling to see if there are any between-wave differences. EQ-5D-5L index scores and EQ-VAS were estimated with the mean and standard deviation (SD). To analyze dimensional responses, we examined the weighted proportion of respondents having no problems (level 1) versus any problems (levels 2, 3, 4 and 5), having no problems (level 1) versus having mild-moderate problems (levels 2–3) and severe-extreme problems (levels 4–5), and having problems in each level. We used a Wald test for EQ-5D-5L index scores and EQ-VAS and a design-based F-test for dimensional responses to identify between-wave differences in HRQL. If no differences were observed, we would pool the survey 1 and survey 2 data to represent HRQL during the pandemic and then compared it with the pre-pandemic Alberta population norms. For EQ-5D-5L index scores and EQ-VAS, we used the survey mean command followed by a Lincom (linear combination of estimators) command, and for dimensional responses, we used the survey proportion command followed by a Lincom command. We used the minimally important difference (MID) to examine the magnitude of the differences in the EQ-5D-5L index (MID = 0.037) [38] and in EQ-VAS (MID = 7.0). For dimension responses, we deemed the difference was a small magnitude if the absolute difference in proportions was within 5%.

Results

Sample

Observations were excluded (Survey 1: 21.49%, Survey 2: 21.43%) if data were missing for age (Survey 1: 13.15%; Survey 2: 13.37%), gender (Survey 1: 13.87%; Survey 2: 14.22%), EQ-5D-5L (Survey 1: 15.43%; Survey 2: 13.93%), or EQ-VAS (Survey 1: 15.97%; Survey 2: 17.06%). The total sample included 18,053 observations, with 8790 in survey 1 and 9263 in survey 2. Excluded respondents were a little older, had slightly more women, were a little less urban and lower educated, were more from a non-white ethnicity, had lower household income, had a poorer financial situation, and had lower HRQL compared to the included sample (Table 1).

Table 1 Data management and missing
Full size table

Survey 1 and survey 2 respondents had similar distributions across age groups, gender, place of birth, and health behaviors (smoking and cannabis consumption). Respondents in survey 1 were more likely to have higher education, to be of non-white ethnicities, to be workers in healthcare and social services sectors, to have higher household income and a comfortable financial situation, and to have higher prevalence of some underlying health conditions (diabetes, heart/vascular disease, having or had cancer, liver/kidney disease, lung disease, and autoimmune disease) compared to those in wave 2. Additionally, fewer respondents in survey 1 reported being students, working in agriculture or construction/manufacturing sectors, and experiencing COVID-19 infection (Additional file 1: Table S1).

Implementing sample weights increased the sample representativeness. Compared to the general Alberta population, respondents in both surveys were more educated, more likely to identify as “white”, more likely to be born in Canada, and had higher rates of self-reported cannabis consumption. Survey respondents reported lower unemployment levels and smoking rates. Additionally, more survey respondents were at both the upper and lower end of the household income distribution compared to the 2016 census [33]. Although there were variations in reporting chronic diseases that made direct comparison difficult, overall rates were similar between the survey and the Alberta population values. Population characteristics measured by the remaining variables were overall similar to the general population values (Additional file 1: Table S1).

HRQL in the overall population

The mean EQ-5D-5L index scores were 0.81 (0.15) and 0.82 (0.14) and mean EQ-VAS scores were 72.54 (18.57) and 71.98 (18.96) in surveys 1 and 2, respectively. Histograms of EQ-5D-5L index scores and EQ-VAS in survey 1 and survey 2 showed skewed distribution (Fig. 1). The median (interquartile range) for EQ-5D-5L index score was 0.87 (0.78, 0.91) in both surveys, and the median (interquartile range) for EQ-VAS was 76 (65, 85) in the survey 1 and 75 (65, 85) in the survey 2. The ceiling effect (respondents reporting “no problems” in all dimensions) was 6.3% and 6.7% for survey 1 and survey 2 samples, respectively. There were no significant differences in the EQ-5D-5L index score or the EQ-VAS score between the two surveys (Table 2). The dimension with the most reported problems was anxiety/depression (69.5% vs. 70.2% for survey 1 and survey 2), followed by the pain/discomfort dimension (52.5% vs. 50.7% for survey 1 and survey 2) and usual activities dimension (41.6% vs. 39.5% for survey 1 and survey 2). The dimension with the least reported problems was self-care (9.0% vs. 7.6% for survey 1 and survey 2) (Fig. 2). Dimensional between-survey differences were only significant in the usual activities dimension (p = 0.001) and anxiety/depression dimension (p = 0.019), but the magnitudes of differences were small (Additional file 1: Table S2).

Fig. 1
figure 1

Histograms of EQ-5D-5L index scores and EQ-VAS scores in survey 1, survey 2, and the pooled dataset

Full size image
Table 2 Weighted EQ-5D-5L index and EQ-VAS scores in either survey
Full size table
Fig. 2
figure 2

Proportion of respondents reporting problems in each EQ-5D-5L dimension (the Alberta norm, survey 1, survey 2, and combined). Note: Between-wave differences: p < 0.001 in usual activities, p = 0.019 in anxiety/depression dimension. Pooled data versus pre-pandemic norms: p < 0.001 in all dimensions

Full size image

HRQL in population subgroups

Some subgroups had significantly lower EQ-5D-5L index scores than their counterparts in both surveys: respondents aged 16–24 or aged 75 and older versus respondents aged 25–74, respondents who identified as women versus those who identified as men, and respondents residing in urban areas versus those residing in rural areas. In terms of EQ-VAS, respondents aged 16–24 and residing in urban areas had lower HRQL compared with their counterparts, but there were no statistically significant differences for EQ-VAS across gender subgroups (Table 2).

All subgroups except for respondents aged 65 and older had more problems reported in usual activities, pain/discomfort, and anxiety/depression dimensions and were less impacted by mobility and self-care dimensions. For respondents aged 65 and older, there were also numerous problems in the mobility dimension. Younger respondents, those who self-identified as a woman, and reside in urban areas reported higher proportions of problems in the usual activities and anxiety/depression dimensions, compared to their counterparts. Older respondents, and those who self-identified as a woman reported more problems in the pain/discomfort dimension compared to their counterparts (Additional file 1: Fig SA1–SA9).

There were no between-survey differences in EQ-5D-5L and EQ-VAS scores in each subgroup (Table 2). All age and gender subgroups had some dimensions with significant between-survey differences, but the magnitude of those differences was small (Additional file 1: Fig SA1–SA7).

HRQL in the population during and pre-pandemic

Given a few between-survey differences in HRQL, we estimated HRQL in the pooled dataset to reflect HRQL during the pandemic (Fig. 2, Table 3). Compared with the pre-pandemic Alberta norms, HRQL during the pandemic was lower for both the EQ-5D-5L index score (0.84 vs. 0.82, p < 0.001) and EQ-VAS score (77.40 vs. 72.26, p < 0.001), respectively. However, both differences were within the MID range. EQ-5D-5L and EQ-VAS scores in subgroups defined by age, gender, and dwelling were all lower than the pre-pandemic reference value. Most of these differences were significant, and such differences were greater than the MID for respondents aged 16–44, women and respondents residing in urban areas (Table 3).

Table 3 Weighted EQ-5D-5L index and EQ-VAS and comparison with Alberta norms
Full size table

People aged 16–24 had the highest HRQL in the pre-pandemic norms and HRQL decreased with increasing age, however, mean EQ-5D-5L and EQ-VAS scores for respondents aged 16–24 in the pooled dataset were even lower than those for respondents aged 65–74. Respondents residing in urban areas had higher HRQL in the pre-pandemic reference value than those residing in rural areas, and the relation is vice versa during the pandemic (Table 3).

Responses to all EQ-5D-5L dimensions during the pandemic were significantly different from pre-pandemic norms, and the differences in usual activities, pain/discomfort, and anxiety/depression dimensions were of a greater magnitude. During the pandemic, respondents reported fewer problems in the pain/discomfort dimension (64.0% vs. 51.6%) and more problems in the usual activities (26.0% vs. 40.5%) and anxiety/depression dimensions (37.2% vs. 69.9%). Especially in the anxiety/depression dimension, a lot more severe (level 4) to extreme (level 5) problems were reported (2.8% vs. 10.9%) (Fig. 2 and Additional file 1: Table S2). All subgroups reported fewer problems in the pain/discomfort dimension, especially for respondents aged 45–64 (71.4% vs. 57.1%) and self-identified as a man (62.3% vs. 47.4%). All subgroups except for respondents aged 65–74 had more problems in the usual activities dimension, especially for respondents aged 16–44 (16–24: 13.3% vs. 52.1%; 25–44: 15.6% vs. 40.5%) and residing in urban areas (23.8% (Calgary) or 23.9% (Edmonton) vs. 41.3%). All subgroups reported a significantly higher proportion of problems in the anxiety/depression dimension, especially for respondents aged 16–44 (16–24: 44.0% vs. 83.1%; 25–44: 39.1% vs. 75.8%), self-identified as a woman (39.8% vs. 75.2%), and residing in urban areas (36.7% (Calgary) or 39.9% (Edmonton) vs. 72.5%). (Additional file 1: Fig A1–A9 and Additional file 1: Tables S3, S4).

Discussion

This study showed that HRQL of the Alberta general population, as measured by the EQ-5D-5L, was lower during the COVID-19 pandemic as compared to the pre-pandemic population norms, with a significantly higher proportion of problems in the usual activities and anxiety/depression dimensions and a lower proportion of problems in the pain/discomfort dimension. This finding is robust in the overall population and across subgroups, which suggests the existence of a shifting of population norms during the pandemic.

The finding that more problems were reported in the anxiety/depression dimension during the pandemic is consistent with previous literature in other populations. In EQ-5D-5L studies from China [15], Morocco [16], Portugal [17], the US [18], and Vietnam [19], the anxiety/depression dimension had the highest proportion of problems reported. In our study, we also found the differences in anxiety/depression dimension between pandemic and pre-pandemic were greater among respondents aged 16–44, women, and those residing in urban areas. According to a September 2021 report from Statistics Canada, there was a decline in perceived mental health, especially among females, compared to pre-pandemic [9]. We further identified that respondents aged 16–24 and women suffered more in the anxiety/depression dimension. This is consistent with a systematic review on COVID-19 and mental health, which showed that females and adults aged 40 years old or younger had more risks for mental distress [39]. Potential reasons for younger people and women facing more mental health problems can be that they may be less resilient and may have less emotion regulation and coping strategies [40, 41], they are more distressed about school and work responsibilities [42], and experience uncertain conditions and financial stress [43, 44].

Usual activities is another dimension notably impacted, especially for younger groups of the population, which is similar to findings from the Chinese [15] study. Usual activities of the EQ-5D-5L refers to work, study, housework, and family or leisure activities, and the higher level of problems reported in this dimension might be explained by restrictions related to COVID-19. During the survey 1 and survey 2 periods, there were restrictions from governments and institutions related to the above activities, e.g., lockdowns, restrictions on non-essential business, and school closure. Population aged 16–44 were more impacted in this dimension when compared to the norms. People in this age group were more likely to be university students or have a child attending pre-school/daycare and K-12 schools. Existing literature has shown dramatic changes in academic and social life for students in higher education during the pandemic [45] and work-life conflicts for workers with young children at home [46]. These factors put them at greater risk of experiencing problems in terms of usual activities people compared to their counterparts [47, 48].

We observed a decreased prevalence of problems in the pain/discomfort dimension during the pandemic, which ran contrary to other studies. Many believed reducing physical activities, prolonged homestay and increased sitting time, teleworking, and e-learning would increase the risk of pain and discomfort [49, 50]. There might be several reasons that we did not observe such an effect. First, our sample might have a lower proportion of people with underlying health conditions, compared to the population participating in the study on Alberta population norms. Second, the two surveys were administered early in the pandemic, and the long-term effect on pain/discomfort may not have emerged at that point.

Respondents who aged younger, were women, or resided in an urban area tended to be more impacted during the pandemic. They were more impacted in the anxiety/depression and usual activities dimensions, as compared to population norms. This leads to lower HRQL measured by the EQ-5D-5L. Similar age and gender effects on HRQL were previously reported [15, 18, 22]. Little is known about the relationship between the dwelling and HRQL. In some jurisdictions, rural areas seemed to have more COVID-19 crises [51]. In Alberta and likely other jurisdictions with similar contexts, residents in urban areas had stricter restrictions on behaviour. Thus, there would be greater impacts on everyday life, including higher chances of infection because of more COVID-19 cases. Indeed, they showed lower HRQL compared with those living in rural areas. For these subgroups substantially impacted by the COVID-19 pandemic, it is important that policies and restrictions take into account their needs.

This study supports the existing evidence suggesting lower HRQL among the general population during the COVID-19 pandemic. Given the fact that HRQL measured by the EQ-5D-5L during the pandemic was dramatically different from the pre-pandemic norms, clinical studies conducted during the pandemic may be advised to use more applicable reference population norm values, such as those reported in this study. Although the survey only collected HRQL in the early waves of the pandemic, when COVID-19 was relatively well-controlled in Alberta, restrictions during this time were moderate compared to the rest of the pandemic. Since restrictions varied during the pandemic, HRQL collected in this survey might reflect an average level of HRQL during the pandemic. Since the beginning of 2022, many jurisdictions have been taking steps to return to normal (pre-pandemic state), but COVID-19 will still likely impact in the following months or years, with the possible emergence of other serious variants and more strict restrictions required. Population norms during the pandemic would be relevant for the next several years, and more research is needed to explore HRQL in the later waves of the pandemic.

The main limitation of this study was that the original sample was not reflective of the Alberta population as a whole. We applied sample weights to increase the comparability and representativeness. If respondents were assigned a relatively greater weight but their HRQL was not a reflection of their demographic group, this would lead to over-representing issues. Second, we collected self-reported gender identity in the survey, while the pre-pandemic norms showed HRQL by sex. Since we did not observe gender difference in HRQL, the comparison between HRQL during the pandemic and pre-pandemic is still likely valid. Third, we used the dwelling variable (urban/rural residential location) which was administered only in survey 2 as a proxy for urban/rural health zones. We were unable to observe the HRQL in terms of dwelling in survey 1, and therefore fewer observations to estimate HRQL by dwelling. Besides, the rural health zones have cities that can be categorized as urban residential areas. However, although the dwelling variable did not well represent the health zone, we still observed lower HRQL among those who reside in urban areas.

Conclusion

This study showed that HRQL was lower and the distributional characteristics were different during the COVID-19 pandemic, as compared to pre-COVID-19 population norms, with anxiety/depression and usual activities dimensions mostly affected, and younger people, women, and people residing in urban areas suffering the most. To cope with the disproportionate impacts, policy-makers should consider the health risks and HQRL problems for specific subgroups and their health and social needs.