Despite a mounting consensus among researchers, advocates, and health professionals that social determinants are the primary drivers of the health of Canadians, the empirical basis for this claim remains thin. Studies from several European countries have shown that income and social security are primary drivers, with social determinants of health (SDOHs) explaining up to 89% of the variation in self-reported health in these countries [1]. A similar trend is observed in the United States, where socioeconomic factors tend to be the strongest predictors of health outcomes [2]. One of the very few flagship reports describing health disparities across Canadian cities, Reducing the Gaps in Health: A Focus on Socio-Economic Status in Canada, championed by the Urban Public Health Network (UPHN), found that income-related health inequalities have remained largely unchanged since the early 2000s and vary considerably between cities and indicators [3-5]. Many of these inequalities persist at municipal, provincial, and national levels and are direct consequences of individuals’ socioeconomic disadvantages [3,6]. As SDOHs are many and their interactions are complex [7,8], a more in-depth account of how social determinants impact health is needed to identify the highest-impact targets for interventions that can reduce these inequalities and improve health outcomes.

Considerable data are required to measure different aspects of the SDOHs, and these data have to be combined, motivating calls for more significant investments in linkages between health and social data [9-12]. During the past 2 decades, Canadian research data access centers have developed a new way of conducting health and social research using nationally representative linked surveys and administrative data [13]. However, despite offering a significant amount of data on the various SDOHs, the advanced knowledge of these data sources that is needed to work with them is limited among researchers and academics [14]. Considerable disparities also exist in the capacity of local health units to survey regional differences in population health and health equity, with some units having no capacity at all [15,16], and becoming familiar with large and complex data sources such as those in the Research Data Centres (RDCs) is costly. Even among units that do have the capacity to make these investments, work tends to be carried out in isolation, and no national standards exist to ensure that research is carried out in ways that support comparative learning [17-19].

This paper presents an innovative protocol for the world’s first fully integrated virtual, multilevel, and longitudinal social determinant laboratory environment, entitled Canadian Social Determinants Urban Laboratory (CSDUL). CSDUL aims to facilitate multilevel statistical analyses of how social and structural determinants of health impact the overall health of Canadians―achieved by creating a set of programs, algorithms (statistical codes and syntaxes), and data components accumulating information from at least 15 primary data sources within the secure research environment of the RDCs. The project will further use CSDUL to replicate the decomposition analysis of income-related inequalities in self-reported health carried out by the World Health Organization (WHO) Regional Office for Europe (WHO/Europe) [20] for urban Canada. Decomposition analysis is a method used to investigate changes in specific indicators by breaking them down into various contributing factors, enabling us to identify the underlying drivers of these changes [19].

In short, this project will allow us to determine how much of the variation in income-related health inequalities in Canada can be attributed to the various SDOHs (ie, report on the relative impact of social determinants on health inequities). In addition to replicating previous studies that have illustrated the contribution of SDOHs in determining health status, the CSDUL environment will allow us to examine how different categories of SDOHs operate on health at different levels and bridge the gap between the WHO’s conceptual and action frameworks. Our research will adopt a slightly modified version of the WHO Conceptual Framework for Action on the Social Determinants of Health [21]. The WHO SDOH framework illustrates how social, economic, and political mechanisms give rise to environmental and socioeconomic conditions and how these, in turn, shape access to important determinants of health status [21]. One of the innovations of the SDOH framework is that it also conceptualizes the health system as a social determinant. However, the WHO’s separate but complementary framework for action on tackling SDOH inequalities identifies 4 levels of mechanisms—global, macro, meso, and micro—that motivate different kinds of policy actions to improve population health and reduce inequalities [21].

The specific research aims of this project will be to work with Statistics Canada’s integrated Canadian Population Health Survey (CPHS) data [22] and other leading social determinant data to (1) develop and distribute the world’s first fully integrated virtual, multilevel, and longitudinal social determinant laboratory environment, which we will call CSDUL; and (2) replicate the decomposition analysis of income-related inequalities in self-reported health carried out by the WHO/Europe for urban Canada and report on the relative impact of social determinants.

Our study will follow the WHO/Europe [20] and use a 2-fold Oaxaca-Blinder decomposition [70-72] to identify the primary drivers of self-reported health. We are mindful that the CCHS has been redesigned multiple times (most recently in 2015); we will follow best practices [73] and only work with variables that have been measured in the same way over time as we have done in previous work [74]. As was done by the WHO/Europe, we will also report on the distribution of social determinants and health outcomes in the cities and towns we include and replicate our analysis for additional self-rated outcomes―mental health and life satisfaction.

The decomposition analysis will be conducted as follows. If y_q is a vector of observations of dichotomized self-reported health in quintile q and X_q is a matrix of observed social determinants and a constant, then the difference between average self-reported health in the lowest quintile, ȳ₁, and the highest quintile, ȳ₅, can be fit using linear regression and broken down into 2 components: first, a component that is attributable to differences in X_q between the quintiles (ie, differences in their composition) and, second, a component that is attributable to differences in associations between y_q and X_q, captured by the coefficients in the linear model, β_q. If we use the associations of the fifth and richest quintile, β₅, as our benchmark for best health, then the estimated difference in average self-reported health between that quintile and the worst-off quintile, β₁, is decomposable as follows:

If social determinants in X_q are indexed by m, then the estimated proportional contribution of determinants m ∈ G is as follows:

We will estimate equation 2 for each group of categories of social determinants and levels. Our baseline estimates of β_q will use pooled multilevel logistic regression with mixed effects and include all predictors in X_q identified by our nodes (use of logistic regression will slightly modify the aforementioned notation, but the interpretation will remain the same [75,76]).

Data will be pooled to ensure sufficiently large local sample sizes (in previous work [74], we have found that pooling ≥5 years is needed—at this time, we only have 7 years, which is not enough for two 5-year pooled intervals). We will include indicators for year to control for period effects. Provincial effects will be set to fixed, and meso-level variable effects (ie, health region and metropolitan area or municipality) will be set to random. As this is the first study of its kind, we will explore different model specifications and document the implications of these choices for our decomposition results [77,78].

All statistical estimates will be calculated using probability weight, and SEs will be calculated using bootstrapping to account for the complex survey design of the CCHS [79].

We will also examine the impacts of accommodating interaction and effect modification [80] between certain predictors—specifically, self-reported gender (ie, male; female; and, when possible, nonbinary) and different categories of non–White-identifying (eg, Black-identifying) individuals—and fitting random effects at lower levels (eg, aggregate dissemination area or census tract). This would allow us to see, for instance, whether the impact of education is more significant for women than for men [81,82].

The University of Saskatchewan Research Ethics Board deemed this study exempt (E486). All datasets used in this study are covered under existing data sharing agreements within the RDC network and provided to researchers in a controlled research environment. The required data linkages have already been completed by Statistics Canada. Furthermore, the datasets are internally deidentified, harmonized, and thoroughly documented by Statistics Canada and its partners, ensuring secure and consistent use for research purposes. No monetary compensation was paid.

The WHO Commission on Social Determinants of Health has called for a third wave in population health research, which “makes explicit that health systems and the people who use them exist within a social context that can powerfully determine people’s chances to be healthy—not only through access to health services, but also through access to a range of other resources, opportunities, and rights” [83]. There are numerous social determinants that can impact health outcomes, and the way in which these factors interact with each other is complex [8,84]. For example, political and economic context refers to a broad set of structural and functional aspects of society that cannot be directly measured at the individual level but have an immense influence on the patterns of social stratification and, therefore, on the health of populations [21]. The environmental conditions in which populations live, work, and play have a profound influence on their health and well-being [21]. Exposure to harmful elements of the built environment—pollution, low-quality housing, and low amounts of green space, for example—can differ within cities and across subpopulations, leading to environmentally mediated health disparities in specific conditions and overall health status [85]. Biological factors, which include age and sex, also need to be considered when examining the impact of SDOHs in health inequality research [21]. Material circumstances are the direct factors that determine an individual’s ability to live a healthy life [21]. These factors include income, education, employment status, and food security, among others. These factors are proximate measures of socioeconomic position that provide insights into the level of resources available to individuals to enable health-promoting conditions and behaviors [21]. In contrast, contextual determinants such as the socioeconomic environment provide individuals with the opportunity to live a healthy life [21]. The health system is considered an intermediary SDOH, and improving equity in health care requires addressing not only access but also the appropriateness and acceptability of care. Promoting intersectoral action to enhance these aspects can improve overall health outcomes and reduce differential quality of care [21,86]. The WHO SDOH framework has evolved, drawing on various models developed for different purposes, including public education and advocacy, to better understand the complex interrelationships between social factors. It is important to note that these models, including the aforementioned Commission on Social Determinants of Health, have limitations. For instance, they often fail to account for critical aspects emphasized in Indigenous communities, such as the roles of culture and land. Our work aims to contribute to the ongoing evolution of SDOH frameworks, offering a more comprehensive and inclusive perspective that reflects current research.

CSDUL will be the first of its kind that will bridge the gap between the WHO’s conceptual and action SDOH frameworks in a data-driven way and clarify our understanding of what are the factors most likely to drive health outcomes in Canada (ie, what makes Canadians sick or healthy). We will operationalize the WHO SDOH framework, which considers the health system as a social determinant. In addition, by linking multiple leading databases that provide information at the micro, meso, and macro levels, CSDUL will also help researchers see how the various categories of determinants operate on health at different levels, as conceptualized by the WHO’s separate but complementary framework for action on tackling SDOHs. It will also allow us to explore interaction effects with identifiable groups (including by gender) to understand how social determinants impact their health differently. Moreover, the data and decomposition methods used in our study will allow us to combine or differentiate among associations between social and structural determinants and more effectively link their conceptualization with the levels at which decisions are made and actions are taken.

CSDUL will also provide a framework and data components that can augment existing microsimulation tools such as Statistics Canada’s Population Health Model [87], which are used to forecast the impacts of health policy decisions. Crucially, the development and future maintenance and dissemination of CSDUL will ensure that Canadian researchers are able to collaboratively use, build on, and learn from the work of others in the RDC. Researchers who are new to the RDC, including health system researchers, will no longer have to start from scratch, thus dramatically reducing the costs of conducting population health and health equity analysis using Canada’s linked health and social data.

Health inequities and inequalities are common in Canadian cities, but they are not equally distributed across various population groups. Due to data limitations, it has been difficult to conduct large-scale comparative research on health disparities in Canadian cities in the past. However, this situation is rapidly changing, and with the introduction of CSDUL, it will be possible to investigate SDOHs at both the individual and area levels simultaneously and seamlessly. Decomposition analysis is a method for understanding what contributes to health inequities by estimating associations among health inequalities, populations, and correlated determinants, such as those that our hub and nodes will identify, and analyzing the relative impact of each. Our work will go beyond the WHO/Europe analysis by providing data-driven relative contributions of individual-level variables, political and economic context, environmental context, socioeconomic context, material circumstances, and health systems.