Most quantitative work on social inequalities in health relies heavily (and often solely) on measures of income, occupation, and education to operationalize SEP, both because those are the most common measures available in national-level datasets, and because they reflect traditional theories of how social class operates to affect health (through material and behavioural influences) [6, 8, 9, 18]. These indicators, although important, limit us in terms of our ability to theorize about additional mechanisms through which social class might operate. Recent work has accordingly begun to include additional cultural and relational aspects of SEP [19].

Following the work of Bourdieu [14], several authors elaborate on the notion that “capital” extends beyond economic assets [13, 20, 21]. Health researchers have begun to attempt to operationalize social class more fully by including responses to survey or structured interview questions on cultural and relational measures. In addition to the common measure of educational attainment, cultural capital (or dispositions and habitus) may be measured using questions on various forms of leisure participation, types of media usage, fashion preferences, and other forms of consumption tastes and preferences. Social capital as described by Bourdieu may be measured using questions about social contacts and connections as well as access to resources provided by these relationships [13, 20–22]. Others have included measures of trust, political involvement, and community participation, which operationalize social capital as a community resource, as suggested by Coleman [23].

There are several examples of articles in the International Journal of Public Health where authors have moved beyond the traditional measures of income, occupation, and education and incorporated measures of other forms of capital to understand the effects of SEP on health. Frie and Janssen [24] utilized interview data to assess the relationship between latent lifestyle dimensions and health behaviours and outcomes. Villalonga-Olives et al [25] examined the effect of both individual- and group-level social capital on the health status of the elderly in the United States, finding race and ethnic differences. A Dutch study examined the link between “highbrow” cultural participation and health, finding that measures of “distinction” were associated with healthy diet and physical activity, net of education and income [26]. In other international publications, authors have used survey data creatively to operationalize cultural capital beyond educational attainment (such as various forms of leisure participation, types of media usage, fashion preferences, other forms of consumption tastes and preferences) and social capital (social contacts and connections as well as access to resources provided by these relationships) [21, 22].

The Canadian studies that have expanded on traditional measures of SEP have, for the most part, been based on either qualitative or smaller scale survey data and are not nationally representative [27, 28]. One important source for quantitative research on SEP and health is population health surveys administered by national statistical agencies. However, whether, the extent to which, and in what ways, these sources permit researchers to employ a more contemporary theoretical understanding of social class and explore the effects of overlapping forms of capital on health is not known.

Our objective was to identify the breadth of survey questions related to economic, cultural, and social capital available through Statistics Canada surveys, and the extent to which those surveys also include health measures. The co-occurrence of these questions would suggest that it is possible to conduct theoretically-rich quantitative research on overlapping forms of capital and health using these existing datasets.

To illustrate the breadth of survey content on social inequalities and health, we first conducted a qualitative synthesis, where we grouped together similar survey questions to identify sub-categories of each indicator. Second, we quantified the co-occurrence of different forms of SEP indicators and health indicators in the same surveys. To do so, we coded each indicator as present (“1,” at least one measure) or absent (“0,” no measures) in each survey. The data are available as a supplementary file. We examined the percentage of all surveys that had measures of each indicator of SEP, and then each indicator of health.

Not surprisingly, we determined that the traditional measures of income, level of education and occupation were present in a very high proportion (70%) of surveys. Therefore, we started with the co-occurrence of income, occupation and level of education as our baseline SEP measures and then added other indicators one at a time. We systematically examined the co-occurrence of each of those iterations of SEP measures with each of the health measures.

Our initial list contained 194 distinct surveys. After applying our inclusion criteria (adults of all ages; population-based; and national scope), there were 105 distinct surveys (which included a total of 492 survey cycles); however, we were not able to access the survey documentation for three of the surveys. Ultimately, we examined the documentation from a total of 102 surveys (see Supplementary Datasheet S1). The survey dates ranged from 1969 to 2016.

Below, we summarize the breadth of survey content we found, organized using the broad categories in our extraction template (Figure 1). Under each of the broad categories, we classified our indicator questions into thematic sub-categories.

Questions about income and occupation at the individual and household levels were very common, but they varied in the level of detail (see Figure 1). Survey questions relating to wealth, which were also common, were divided into four thematic sub-categories. These included: asset-related questions (e.g., ownership of …); questions about spending and expenditures (including broad consumption habits); questions about household saving habits and debt; and questions about insurance coverage or ownership.

Survey questions about neighbourhood-level economic capital included questions about both neighbourhood geography, such as postal code or census divisions and subdivisions, and neighbourhood resources, such as the availability of various services and items in the immediate neighbourhood. Note that questions asking about sense of community in the neighbourhood were classified under Social Capital (below).

The most common question regarding cultural capital was education, which was included in most surveys. Additional cultural capital questions were diverse in nature. Many surveys contained questions around sport, leisure, or physical activity participation, as well as the frequency of involvement. In addition to general questions about activity participation, we observed questions about the specific details or nuances surrounding activity participation, such as personal preferences for how or where leisure time is spent, as well as preferred types or brands of various items.

Another dimension of cultural capital included questions about the knowledge, value, or importance of various objects, such as the importance placed by respondents on arts and cultural institutions and knowledge of guidelines around food and exercise. Finally, many surveys contained questions about the purchase, use or ownership of various items ranging from food expenditures in different types of establishments to owning a boat for recreational purposes.

Questions about social capital varied in detail and content across the different surveys, and were therefore perhaps the most difficult indicator to identify and classify. Following several iterations, we devised six sub-categories for indicators of social capital. Several surveys had questions on trust, either of the government or of other people. Many surveys asked respondents whether they felt a sense of community or a sense of belonging in their neighbourhood or local area. We also observed survey questions about informational, material or instrumental forms of support from others. We included questions about the size and composition of one’s social network in this latter sub-category.

Some surveys had questions about participating in activities with others, such as exercising and visiting. Questions about political engagement usually took the form of asking about voting behaviour, or interest in politics. And finally, in the social capital category there were questions about membership in organizations, such as churches or volunteer organizations.

All surveys analyzed (n = 102) included other demographic questions, which may be related to other aspects of social inequality. Most commonly, surveys included questions about respondent age, as well as gender or sex, and marital status. Some surveys considered family or household structure, such as the number of generations living in the respondent’s household. Another category included those asking about the respondent’s race, ethnicity or religion. Surveys also included geography-related intersections, such as province of residence of birth, census metropolitan area, and rural or urban status. Finally, there were questions about immigrant status and language usage.

Health indicator questions were common, and ranged from robust and objective measures of health to measures that captured self-reported health-seeking behaviors and perceptions. We classified the health variables into eight categories. Objective physical measures of health refer to clinical measures such as BMI or blood pressure measures. The 1986 Canadian Heart Health Survey and the 2012 Canadian Health Measures Survey (CHMS) had the most comprehensive list of objective physical measures; examples include measured height, weight, blood pressure, and grip strength. Self-reported physical measures of health refer to items such as self-reported (vs. measured) height and weight, which are then used to calculate BMI.

Then there were questions about physical disease such as allergies and chronic conditions. Sometimes these questions are framed in terms of symptoms (e.g., do you have trouble seeing, hearing, walking etc.) and other times they are framed in terms of medical diagnoses (e.g., has a medical professional ever diagnosed you with cancer). Questions about mental disease follow the same pattern, with questions about symptoms (for example, indicators of depression) and diagnoses (have you ever been diagnosed with a mood disorder?).

Several surveys contain questions regarding self-perceived health. A common version asks respondents to rate their health (either physical, mental, or both) on a scale of very poor to excellent. Another sub-category asked respondents about specific health-related behaviours or lifestyle habits, such as drug and alcohol use, smoking behaviours, food consumption, and patterns of physical activity.

A common health indicator question asked in a number of surveys was about health-related barriers or reasons for not participating in or accomplishing various tasks or activities. Most often, questions asked if an illness or disability was the reason for time lost at work or for leaving a job. Some surveys asked about financial difficulties due to long-term disability. A final subcategory included questions that referred to interactions with health professionals or the health care system. Most commonly, these questions asked about health care professional visits, such as going to a doctor or dentist, or overnight hospital stays.

One limitation of our study was the inaccessibility of master file documentation for certain surveys. However, we do not believe that this constitutes a significant threat to the validity of our findings. A larger limitation was that we only included the most recent version of each survey cycle. Therefore, we were not able to identify trends or changes over time in the types of questions asked in surveys with multiple cycles, or to determine if earlier cycles contained different or novel questions that may have been discarded in later versions. Additionally, we did not control for the length of surveys. Our findings speak to the presence of absence of certain indicators in the surveys, but not to the actual number of questions or percent of questions.

Finally, qualitative analysis of the survey documentation was a complex process. Some questions fit under multiple indicator categories, or none at all, and largely depended on how a survey question was worded. For example, some of the surveys asked about smoking cigarettes in the following ways (with our classification in parentheses): the preferred brand of cigarette (cultural capital); the amount of cigarettes smoked per day (health); the main reason you began to smoke again (friends of family smoke) (social capital); or have you bought or have you tried to buy cigarettes from a store (wealth). Categorizing the data into distinct indicator categories proved to be a complicated process, which we recognize reflects the reality of the complex interrelationships between social inequality and health. We also fully recognize that while we have identified potential indicators of cultural capital, the classification of those indicators into categories (such as “high brow” and “low brow” as used by Bourdieu) is a complicated, context-specific task.

Overall, we identified a wide breadth of question content in the national Canadian surveys in terms of indicators of SEP (including indicators that could be used to operationalize various aspects of economic, cultural and social capital), demographic variables, and indicators of health (broadly conceived). Across the Canadian surveys, there is imbalance in the co-occurrence of capital and health indicators, depending on the focus of the survey. The surveys with the highest number of health indicators do not generally contain indicators for all of the forms of capital.

National statistics agencies should be encouraged to continue to include diverse indicators of forms of capital in their surveys. In particular, it would be useful to have more questions that operationalize the economic, cultural, and social aspects of SEP in all surveys that contains health questions. The inclusion of a more robust array of SEP indicators will enable researchers to interpret their findings through a more complex understanding of how social class operates, and therefore to better advise policy makers on the mechanisms through which social inequities in health could be reduced.