Example of recruitment and response rate
A detailed presentation of the CPENS methodology is provided by Michaud et al.1. Briefly, a probability-based random sample (GPRS) of the general population from all provinces was used to recruit telephone respondents to the online survey. For this study, the sample was created using two approaches. A random dialing approach (i.e. GPRS) for the general population across the country where the sample was randomly drawn by province in proportion to their size nationally, and by First postal codes Nations and remote areas in order to oversample these specific groups. Non-respondents who did not respond to the survey received a reminder message 3 and 6 days after initial recruitment. Of the 22,892 potentially eligible participants, 11,492 were recruited for the survey, representing a recruitment rate of 50.6%. Of the 11,492 participants recruited, 6,647 completed the online survey, for an overall response rate among eligible respondents of 29.0%. To obtain a representative sample of rural, urban and suburban areas, the survey data was weighted with the most recent census data from Statistics Canada. This also fixed over and under sampled groups in certain geographies. There was no evidence of outliers in the weighted data that would indicate sampling bias. The study’s margin of error was ±1.2%, at a 95% confidence level (i.e. 19 times out of 20).
Determination of geographical sampling regions
The frame was established to target respondents from remote/rural, suburban and urban areas in the ten Canadian provinces using Forward Sortation Area (FSA) postal code information.22. Respondents indicated the geographic region that best matched the area they lived in based on population size. Because some postal codes can be both rural and urban, the geographic region in the statistical analysis was based on the self-reported geographic region.
Development of questionnaires, pre-tests and quality control
The questionnaire included content to assess noise perception, annoyance, and expectations of quiet, health-related, and socio-demographic variables. The average time to complete the online questionnaire was just under 10 minutes. The questionnaire was designed by Health Canada and pre-tested in English and French. For the pretest, 299 people were recruited by telephone (212 in English and 87 in French). This led to 72 completed online surveys (61 in English and 11 in French). Minor changes to the survey after the pre-test did not affect the pre-test data, allowing the results collected during the pre-test to be included in the final analysis. English and French versions of the survey are available from Library and Archives Canada23.
In CPENS, participants were asked to indicate how they have been personally affected by the COVID-19 pandemic with respect to physical health, mental health, environmental noise annoyance, indoor noise annoyance, stress in their lives and general well-being. The response categories for these six outcome variables were: much worse, somewhat worse, unchanged, somewhat improved, and much improved. For modeling, responses were grouped as follows: “rather/much worse” and “unchanged/somewhat/much improved”. When reporting prevalence rates, responses were grouped into the following three categories: “rather/much worse”, “unchanged” and “somewhat/much improved”. A number of other variables were collected in the CPENS that were considered potentially associated with the six outcomes assessed. These included demographic variables such as age, gender, education, income and Aboriginal status. Age in years was divided into three groups (18–34, 35–54, 55+). The following gender categories have been defined (female, male, other/prefer not to say). Education was assessed as follows: up to high school diploma or equivalent, certificate or diploma, bachelor’s degree or graduate degree. A certificate or diploma may be from a registered apprenticeship program or another trade, college, CEGEP (i.e. Quebec College), or other non-university university lower than the baccalaureate. Total household income in Canadian dollars was grouped as follows: less than $40,000, $40,000 to just under $80,000, $80,000 to just under $150,000, $150,000 and over. Aboriginal status has been grouped as follows: Identify as First Nation/Métis/Inuk (Inuit), or Do not identify. Province of residence as well as geographic region were also considered as potential predictor variables since the response to the pandemic differed by province and geographic region. Due to small sample sizes, the Prairie provinces (i.e. Manitoba and Saskatchewan) were grouped together, as were the Atlantic provinces (i.e. New Brunswick, New -Scotland, Prince Edward Island and Newfoundland and Labrador). The other provinces (British Columbia, Alberta, Ontario and Quebec) were ranked independently. Self-reported geographic area was defined as rural/remote (i.e.
A respondent’s current work or school situation was also taken into account. Respondents identified themselves as: working or attending school outside their home; work or attend school inside their home; retired; unemployed; and a portion of those who indicated “other” could be grouped as being on paid leave (ie, sick, maternity and disability). More than one option can be selected; therefore, each situation was considered separately as a “Yes/No” response.
Other variables taken into account included sleep disturbance (for any reason at home in the past 12 months), rated as highly disturbed by sleep (rating of 8 to 10) compared to slightly disturbed by sleep (rating from 0 to 7). Similarly, noise sensitivity was defined as very sensitive to noise (score of 8 to 10) versus not very sensitive to noise (score of 0 to 7). Participants were asked to rate their overall physical health relative to someone their own age, and their overall mental health (no reference to age). For both of these questions, responses included the following: poor; fair; good; very good; and excellent. These have been grouped as follows: poor/fair and good/very good/excellent. Heart disease, including high blood pressure, anxiety or depression, sleep disturbance, and hearing loss, was also rated as diagnosed by a medical professional, undiagnosed but suffering from the disease or not s not applying. The assertion of a diagnosis was supposed to indicate that the condition was current, not a condition that existed historically but was no longer current.
Weighted frequencies and cross-tabulations were used to explore the distribution of demographics and population characteristics by Aboriginal status and geographic region. Cross-tabulations of each of the health-related outcomes and noise nuisance variables affected by the pandemic with Indigenous status and geographic region were also considered. Chi-square tests of independence compared Aboriginal status to non-Aboriginal respondents, as well as geographic regions.
Initial univariate logistic regression models were used to investigate the relationship between each of the health-related outcomes, including noise annoyance variables and other variables of interest, as mentioned above. Unadjusted odds ratios (ORs) are reported for each relationship in Supplementary Material (see Table S1). Finally, a multivariate logistic regression model was developed using stepwise regression techniques with a chi-square significance level to enter an effect in the model equal to 20% and the chi-square significance level so that an effect remains in the model. by 5%. Adjusted ORs are reported for the final models for each assessed outcome affected by the pandemic. Confidence intervals (CI) of the ORs including the value 1 indicate insufficient evidence to observe an association between the outcome assessed and the variable studied.
Statistical analysis was performed using SAS Enterprise Guide 7.15 (SAS Institute Inc., Cary, NC). Unless otherwise specified, a level of statistical significance of 0.05 was applied throughout. Additionally, Bonferroni corrections were made for all pairwise comparisons to ensure that the overall type I error rate (false positive) was less than 0.05. Estimates with a coefficient of variation (CV) between 16.6 and 33.3% were designated “E” and should be interpreted with caution due to the large sampling variability associated with them; CV estimates that exceeded 33.3% were designated “F”, indicating that these data could not be published due to questionable validity. No results are reported for cell frequencies below 10.
This study was approved by the Ethics Review Board of Health Canada and the Public Health Agency of Canada (Protocol No. REB 2020-038H). Informed consent is implicit in the voluntary response to the survey questionnaire. This research was conducted in accordance with all relevant Government of Canada guidelines and regulations for conducting online surveys.