The research ideas we pursue are often dictated by data availability or by prior literature on the research idea. Sometimes, however, we encounter a topic for which the scope of the problem has not yet been described or no suitable data have yet been collected. Although barriers to conducting research in this situation can be substantial, researchers might be able to create new knowledge from existing data to justify further analyses or data collection.

Finding appropriate data is particularly challenging when researchers examine new interdisciplinary topics, such as occupational health inequities (ie, avoidable differences in work-related health outcomes closely linked with social, economic and/or environmental disadvantages).¹ Because the close link between work and ‘social, economic, and/or environmental disadvantages’ is not well defined, relevant research fields have not been well integrated, and existing health data sets seldom have adequate data for examining the role of work in creating health inequities across population groups.² This lack of data likely contributes to the slow development of occupational health inequities as a field of investigation.

We use inequities in adequate bathrooms in US workplaces as an example of an important topic for which appropriate data are not readily available. Based on previous small-scale studies and reports, we suspected that workplace bathrooms are inadequate for some workers, but we had no suitable data sets available to directly estimate if inadequate bathrooms are patterned by type of work or worker characteristics in the US working population.^3–5 Therefore, we developed a conceptual model and used this model to guide analyses that combined multiple data sources to characterise the potential for inequities in adequate workplace bathrooms. This study had two main aims. First, to propose a conceptual model that can guide research into occupational health inequities; second, to apply the conceptual model to an inequity question for which no single data set could provide a suitable answer. To show how this conceptual model can guide research, we outline in detail our strategy and decision-making. We use inadequate workplace bathrooms as an example and investigate potential inequities across gender, race/ethnicity and education. Through this example, we also identify specific data gaps for further investigation.

We developed a conceptual model that articulates our hypotheses about the causation of occupational health inequities. Conceptual models for occupational health inequity exist, but few studies have applied these models to their investigations.^17–20 One reason might be that these models include a large number of factors and no single data set contains all variables needed to fully apply these models. Recognising the difficulty that occupational health inequity researchers face in applying conceptual models to their research, we propose a simple model, consistent with previous ones, and focusing on two interconnected mechanisms by which inequities arise: occupational segregation and occupational health inequity (figure 1).

The first mechanism, occupational segregation, represents various social processes that sort people into certain occupations based on demographic characteristics, as observed by uneven compositions in occupation by gender, race/ethnicity and educational attainment in the US workforce (in our example, we focus on these characteristics because they are the strongest axes of occupational segregation in the US workforce).^21–24 These social processes include gendered expectations and systemic discrimination, among others. For example, as a result of these social processes, men might be more likely than women to work as a roofer. The second mechanism, occupational health inequity, is the existence of avoidable differences in work-related health.¹ In our example of inadequate bathrooms, some occupations have work characteristics—such as outdoor work in a location without nearby bathrooms—that cause inadequate bathrooms for people in these jobs and subsequently affect health; other occupations do not have these characteristics.

The conceptual model demonstrates how occupational segregation and occupational health equity together cause health disparities or inequities across demographic characteristics. For example, if men were more likely to work as a roofer than women, roofers were more likely than other jobs to involve outdoor work far from bathrooms, and having far bathrooms caused bladder problems, then men would have a higher prevalence of bladder problems than women: a gendered health inequity. A key component of our conceptual model is that the second mechanism, occupational health inequity, involves avoidable differences in working conditions. In our example, we could avoid the gendered health inequity in bladder problems by providing roofers portable bathrooms on the job site and frequent breaks to use them. The model therefore suggests solutions to work-mediated health inequities: improving working conditions across occupations. This solution is fully compatible with existing occupational safety and health practices.

When researchers study work-related inequities, a conventional approach is to directly compare the prevalence of work characteristics or health outcomes across gender, race/ethnicity or educational attainment groups; for example, if working men are more likely to report inadequate bathrooms than working women (shown in figure 1 by the direct arrow from worker demographics to inadequate bathrooms). However, our conceptual model recommends against this approach because it fully ignores the true causes of inadequate bathrooms. A person’s gender, race/ethnicity or educational attainment is not the cause of inadequate workplace bathrooms. Stated another way, we are not interested in describing how inadequate bathrooms vary by demographic characteristics; rather, we are interested in the role of occupation in mediating relationships between demographic characteristics and inadequate bathrooms. Inadequate bathrooms are typically caused by work-level characteristics, not individual-level characteristics, like having a workspace without bathrooms available or work demands or workplace policies that prevent workers from accessing available bathrooms. An important exception is targeted discrimination or harassment. For example, among transgender or non-binary workers, individual characteristics such as gender identity or expression can be a source of discrimination and improper denial of access to workplace bathrooms.¹¹ Power dynamics also play a role; supervisors can control workers’ bathroom access, and fear of retaliation prevents workers from reporting inadequate bathroom access to management or OSHA.^{3 10}

One difficulty in applying conceptual models is that typically no single data set includes all variables in the model. Instead, our model identifies three steps to answer our research question using different data sets if needed (figure 2). We therefore began at inadequate bathrooms and worked through the steps that we hypothesised caused the inequities: work characteristics, occupation and, finally, worker characteristics.

We developed a conceptual model to explore how to address a research question for which appropriate data do not exist. We combined multiple data sets to generate preliminary information about inequities in adequate workplace bathrooms. Difficulties at each step prevented us from coming to firm conclusions, but our attempt highlighted ways to leverage existing information to identify opportunities for future research and gaps in available data.

By presenting our effort to investigate inequities in adequate workplace bathrooms, we hope to inspire researchers to overcome the common situation of data availability dictating what research questions we ask. Instead, researchers can leverage existing data, although imperfectly, to expand the horizon of research topics we pursue, including issues of occupational health equity. Our conceptual model echoes calls for theory-driven, data-informed research.^{37 38} By explicitly hypothesising the causal mechanisms for inequities, we identified three separate but crucial questions along the causal pathway and were able to use different data sets to address our research question. Starting a research study with a conceptual model, rather than generating a model based on available variables in a single data set, helps us to see what we could do with limited data and future directions for the field.

By breaking our larger research question into smaller ones, we could use multiple data sets to answer our question. Combining multiple data sources is often referred to as a fusion design.³⁹ Our process differed from most fusion designs because we did not integrate the data sets to create a single result; we conducted three separate analyses. In some ways, this process was similar to triangulation, in which multiple studies with different limitation are used to answer the same research question; however, our analyses answered separate smaller questions.⁴⁰ Although we hesitate to describe our strategy as either a fusion design or triangulation, we encourage researchers to think of creative ways to use multiple data sources to answer research questions.

Another strength of the conceptual model was steering us away from comparing the prevalence of inadequate bathrooms by gender, race/ethnicity and education, a typical approach if a large, diverse, individual-level data set on bathroom access existed. Such analyses focus on worker characteristics but do not inform solutions because they ignore structural causes of inequities. For work-related inequities, the workplace—not the individual—is the most effective intervention point, whether achieved by regulation or by following the hierarchy of controls.

As we reflected on each step of the analysis, we identified future directions to improve occupational health inequity research more broadly. First, incorporating information on work into population-based epidemiological studies would facilitate understanding health inequities. Although work is both a social and structural determinant of health, it is rarely included in epidemiological studies outside of occupational epidemiology.^{2 41} Studies that include industry, occupation and work characteristics would allow researchers to understand how work contributes to health inequities. Second, the opposite is true—occupational health studies would benefit from including information on non-occupational social determinants of health for the same reasons. Third, a viable way to approach understudied occupational exposures, such as inadequate bathroom access, uncleanliness and inability to take breaks, could be to launch small-scale studies. By applying the methods we used in this analysis, even limited data from these small-scale studies could be useful in addressing complex emerging issues.

We encourage researchers to start a research project not by looking at available data, but by considering which research questions are worth asking. What data are collected and what are not—what is recognised as important and what is not—is decided in specific social contexts, which can be changed. If suitable data do not exist, researchers can advance knowledge by articulating assumptions, maximising use of fragmented data and drawing conclusions with appropriate caution. Data availability does not have to limit our pursuit of important research questions.