We conducted a cross-sectional study on the demographics of students at English-speaking Canadian medical schools in 2018 through an online survey. We adapted the study methodology from previous studies on this topic[1, 2]. We coordinated the project with student leaders from the Canadian Federation of Medical Students (CFMS), which represents 14 of 17 Canadian medical schools. We excluded students from the other three Canadian medical schools, based in Quebec, for two reasons. First, previous studies have postulated that these students have distinctive demographic characteristics compared to medical students from English-speaking schools[1, 3], given their younger average age at matriculation and substantially lower tuition fees[28, 29]. Additionally, we did not have a reliable method of reaching these students as they are not represented by the CFMS.
Survey Design
Through our survey, we aimed to capture information on the following demographic characteristics: ethnicity, gender identity, sex assigned at birth, socioeconomic status, and rurality of the area respondents grew up in. Additional elements of the survey included questions regarding characteristics and behavior after entering medical school including: debt burden, preference of future specialty and practice location, and the perceived impact of demographic and financial factors on future practice. The results of those post-admission questions are not reported in this study. Instead, we focus on the demographics of students admitted to Canadian medical schools, with plans to publish further analysis of all study data. We hosted the survey on an online survey platform Simple Survey (OutSideSoft Solutions Inc., Quebec, Canada). The complete survey and explanations of questions is available as Supplementary Digital Appendix 1.
We used two previous surveys as starting points to improve content validity and allow for direct comparisons to other populations: The 2016 Canadian Census[30], and a previously validated survey addressing this research topic[2]. Most individual survey items were taken verbatim from the Canadian Census. For rurality, and parental occupation, we used classifications different from the census, which is detailed below in the section Survey Content.
We piloted the survey with 16 medical students from across Canada and subsequently altered wording for certain questions to improve clarity and applicability to the medical student population.
Survey Content
We collected data on respondents’ ages, year of medical school, and level of education prior to medical school. We also asked about ethnicity using the same terms used in the Canadian census: Aboriginal, Arab, Black, Chinese, Filipino, Korean, Japanese, Latin American, South Asian (e.g. Indian, Bangladeshi, Sri Lankan), Southeast Asian (e.g. Cambodian, Indonesian, Thai), West Asian (e.g. Iranian), and “other”. Participants could choose more than one ethnicity. We also asked students about the size of the community they grew up in, using the 2016 Statistics Canada Population Centre and Rural Area classification[31]. A rural area was defined as having a population of <1000 people, small and medium population centres as having populations of 1000–99,999, and large urban population centres as having populations of 100,000[31].
To compare participants’ socioeconomic status to the Canadian population, we asked about three commonly used and well-validated markers of socioeconomic status: parental income, occupation, and education level[32–35]. For parental income and education level, we used similar income brackets and diploma or degree classifications respectively as the 2016 census[30]. For parental education level, we used a modified version of the Pineo-Porter Occupational Scale as has been used by Dhalla et al[2, 36].
Survey delivery
We contacted medical students through class email lists. The emails included information on the purpose of the study, contact information for the research team, the nature of voluntary participation, and a link to the survey. No individual emails were collected, used, or stored at any point during the study. After the initial email, participants received three biweekly reminder emails. The CFMS also promoted the survey through their social media accounts (Facebook and Twitter), and student leaders at individual schools delivered class announcements. The survey was open for a total of 10 weeks in spring 2018 to ensure coverage of different examination and vacation schedules.
Analysis
We imported questionnaire data directly from SimpleSurvey software into SPSS Version 24 (IBM, Armonk, NY). We removed participants who declined to complete the survey at the informed consent step, and surveys which were started but not answered. When two or more consecutive surveys had identical answers and the former survey(s) had fewer questions completed, we assumed that this was the same participant who accessed the survey more than once. In these cases, we only considered the final response.
We used descriptive statistics to summarize responses to all questions and chi-squared tests to detect differences in characteristics of survey respondents and the general Canadian population via the 2016 Comprehensive Census.
Assessment of nonresponse bias
We performed post-hoc analyses to assess for nonresponse bias using two approaches[37]. First, we compared our data on age and sex to the 2017 Canadian Medical Education Statistics (CMES) report published by Association of Faculties of Medicine of Canada, a dataset which represents the entire Canadian medical student population[28]. For age, we compared our fourth-year respondents to graduating medical students in CMES, the only group for whom age was available. For sex, we compared results from our question “sex assigned at birth” to the listed sex of the CMES 2017 population from all years of medical school. We restricted the above analyses to students from English-speaking Canadian medical schools.
Second, we compared the first 100 respondents to the last 100 respondents, with the assumption that late respondents are more similar to nonrespondents[38]. We compared ethnicity, rurality, and parental income, education, and occupation between these groups.