In their article in this issue of the Journal (Am J Epidemiol. 2013;178(6):843–849), Galea and Link identify important heuristics for our discipline. In this commentary, I build upon their ideas by arguing that 1) social epidemiology has become an Asian, European, Latin American, and African rather than just North American endeavor, 2) realism is better suited to social epidemiology than positivism, 3) more work on social mechanisms (social class relations, racial discrimination) is needed to increase the explanatory power of social epidemiology, 4) increased attention on (social) causal models will generate more innovative social interventions, and 5) social interventions should be conducted in full partnerships with affected populations.
In their article “Six Paths for the Future of Social Epidemiology” (1), Professors Galea and Link provide a timely look at the major challenges facing social epidemiology and discuss 6 important areas on which to focus in the future. In this commentary, I build upon their ideas, giving alternative interpretations of these challenges and what might be desirable for the future of social epidemiology.
Milton Terris, the celebrated Brazilian and American epidemiologist, believed that epidemiology was a biosocial science because the process of studying the causes and patterns of population health was inherently a social one. To say he had a point would be a gross understatement. Consider this example: Deaths due to hemorrhaging are biologic facts. However, deaths caused by firearms are social facts that also implicate economic (i.e., the production and commerce of firearms), political (i.e., the laws that regulate fire arm access), and cultural (i.e., the social acceptability of problem solving via violence) relations (2). According to Galea and Link (1), social epidemiology has evolved into a valuable and established subdiscipline within epidemiology, starting in the mid-1990s around the time of President Clinton's first administration. With an increase in new sources of funding and publication forums, social epidemiologists have identified a plethora of new problems (e.g., neighborhood effects over the life course), developed novel hypotheses (e.g., impact of income inequality within and between nations), advanced methods (e.g., multilevel designs), and reached provocative findings (e.g., nongradients in health using social class concepts). Indeed, the impact of social epidemiology has reached and influenced new areas in health scholarship, including global health (3). Despite this progress, Galea and Link express concern that social epidemiology is at risk of losing its hard-won status because other subdisciplines within epidemiology are also studying social factors. In the present article, I contend that if social epidemiology is losing its identity and traction, it is not because other specialties have become more aware or interested in the social aspects of population health. On the contrary, if social epidemiology has become stagnant, it is because the field tends to ask conventional questions such as, “What social risk factors predict health outcomes?” rather than posing questions that interrogate, “Why do health inequities occur in the first place?” or “How are health inequities generated and reproduced over time?” That is, social epidemiology finds itself caught within an empiricist and pragmatist epistemology that favors descriptive summaries and observational accounts over social explanations and causal mechanisms. This state of affairs limits the depth of its inquiries, results in a dearth of causal thinking, and generates little in the way of interventions.
SOCIAL EPIDEMIOLOGY IS GLOBAL
My first reaction to the article by Galea and Link was to note the degree to which it is centered on social epidemiology in the United States. There is little doubt that the enhanced status (including name recognition) of the specialty has been achieved mainly within the United States. New research departments (e.g., Leo Syme's department at the University of California, Berkeley and his students at Harvard and the University of Michigan), theories (4), methods (5–7), and findings (8, 9) testify to the emergence of social epidemiology within the United States, with global repercussions. Social epidemiologists at the Society for Epidemiologic Research, the American Public Health Association, or the International Epidemiological Association can present their work under that label instead of the philosophically unsatisfying “Mental and Psychosocial Epidemiology.” However, there are clear instances in which US social epidemiology seems to have held general social epidemiology back. This includes social epidemiologists' reluctance to understand and measure race as a social relation rather than as the property of an organism (10) or persistence in using ranks of economic resources rather than conceptualizing and measuring the social relations that generate them (11). Failing to account for social relations has resulted in large amounts of data without much explanatory power and an overall lack of support for policy interventions that reduce social inequities in health.
Whereas US epidemiology has played it relatively safe in its social epidemiologic approaches, other research programs around the world have been more willing to advance the discipline in critical ways. For example, Europe has been a constant source of new findings in social epidemiology, at least since the Whitehall Studies (if not earlier), on the enduring effects on health of (social) class stratification (12), the population health approach (13), income inequality (14, 15), precarious employment (16), the great recession (17), and politics and welfare-state regimes on health (18). Latin America has also contributed to the development of social epidemiology, in particular with respect to theory and philosophy (19, 20). Beyond Australia and New Zealand, language barriers may prevent us from appreciating the recent contributions from Asian social epidemiologists, notably from Korea (21) but also from Japan, Taiwan, China, India, Thailand, Malaysia, Indonesia, and Sri Lanka, among other countries. Other regions also seem to be catching up and are contributing to the field (e.g., Arab nations in the Middle East) (22).
WHY EPIDEMIOLOGY IS (A BIT) SOCIAL BUT SOCIAL EPIDEMIOLOGY IS NOT (FULLY SO)
Galea and Link reviewed the recent history of social epidemiology, with special attention paid to the United States. With regard to the timing of disciplinary evolution, I would situate the emergence of multilevel analyses to test the effects of social processes on health before 2000. A multilevel ontology was provided by Krieger's theoretical papers and models (23, 24), and in a matter of a few years, innovative work using multilevel analyses to examine social mechanisms and associations appeared in the literature (5–7). The evolution of the discipline since the early 1990s has been at best incremental, with negligible advances in theory or methods. Only Nancy Krieger's pioneering work (25) between the mid-1980s and mid-1990s seemed ahead of its time, with an integration of themes that would define social epidemiology in the years to come (i.e., racism, sexism, social class, multilevel ontology, social causation, contextual data, and historical analyses).
Galea and Link claim that social epidemiology is at risk of losing its hard-won identity because of its ubiquitous status within epidemiology. Despite insights gleaned from studies on smoking and obesity that point to social mechanisms, what I think is the major weakness of the discipline is precisely what they present as one of its major strengths, namely “epidemiologists have been at the leading edge of demonstrating all these observations” (emphasis mine) (1, p. 844). Thus, social epidemiology favors an empiricist framework that yields thousands of studies, with ever larger samples and longer follow up periods yielding only associations with few causal explanations (e.g., 26). It is the avoidance of mechanisms and explanation (e.g., scientific realism; 10, 27) in favor of associations (empiricism, pragmatism) that leads to a disciplinary cul-de-sac.Most studies use socioeconomic status indicators such as educational level, occupational class, or even income without any serious consideration of the potential social mechanisms that link exposure and outcome. For example, a recent major global health study showed that educational level is inversely related to several health indicators in several countries (3). This implies that health disparities can be reduced or eliminated if population levels of education increase (e.g., everyone earned graduate degrees). Yet, this assumption is simply unrealistic: The proportion of highly educated persons occupying coveted jobs in market economies (i.e., capitalist) represents a diminutive share of the overall working population (28). Multilevel analyses provide another example. Its development and use was meant to explicitly test for social mechanisms (e.g., how Catholic schools influence student performance; 29) rather than to search for independent associations between exposures and outcomes like with traditional risk-factor regression models. Thus, there is a danger that under the banner of complex methods innovation, one may simply perpetuate the associationist status quo.
The absence of causal explanations has consequences not only for basic social epidemiology but for applied social epidemiology and public health, its associated social technology. An excess of descriptive studies and a lack of causal mechanisms leads to a very poor understanding about which interventions actually improve population health or reduce health inequities (30). Large gaps exist when it comes to knowledge of what kinds of potential interventions matter (e.g., What are the social policies that have the greatest impact on reducing health inequalities? What types of labor markets and what combination of labor market flexibility and social protection promotes worker health (31)? When do intersectoral policies and health in all policies reduce health inequalities) (32)?
IS MACRO HARDER THAN MICRO?
Galea and Link claim that the study of macrosocial determinants of health remains disproportionately small because of the intrinsic difficulty of the subject matter. There is a contemporary emergent literature on political traditions, macro policies, and welfare state policies that provides some theoretical guidance and methodological heuristics (33–35). Economic macro-analyses are even more common, stemming from a tradition of research on the effects of economic downturns and recessions (36), which continues to date (37, 38). What has become arguably the most popular topic in social epidemiology involves a macroeconomic indicator, namely income inequality (14, 39, 40). There are also new theoretic and empiric developments in neighborhood research (41) and a new generation of housing studies that do not dwell on the personal characteristics of the homeless (42). Yet, without articulating the main questions of macrosocial factors in social epidemiology (e.g., whether liberal democracies better than authoritarian regimes for population health) (43), it is difficult to know where the challenges may reside. Other social disciplines, such as economics (44), sociology (45), or political economy (46), routinely study macrosocial factors. It seems that despite its limitations, macrosocial epidemiology has started down a promising path.
Regarding methods, the trend is to rely increasingly on new and larger datasets that involve a global sample of countries. Large multilevel country datasets that contain individual-level data are powerful tools to investigate hypotheses about macrosocial determinants (e.g., whether egalitarian welfare regimes associated with better health outcomes; 47). Although these datasets are increasingly available (48–50), they do not allow researchers to understand how social mechanisms at the country level affect individual health outcomes (e.g., how specific policies are generated, implemented, and evaluated). Social sciences have developed a series of comparative historical methods that complement crossnational surveys (51) among other tools for the analysis of macro-level determinants (realist reviews, hypothesis testing case studies; 30). There is no reason why social epidemiology should not open its doors to use them when appropriate.
DO WE NEED EVER BIGGER INDIVIDUAL LEVEL DATASETS WITH LONGER FOLLOW UPS?
Galea and Link seem to assume that we will be using large (and overpowered) quantitative, individual-level longitudinal data as the sine qua non in social epidemiology. This assumption is consistent with a tradition of the parent discipline, empiricist epistemology, which prefers observations over theoretic constructs, mechanisms, and explanations, at least when it comes to social variables (e.g., socioeconomic status, race). However, (social) science needs explanations and mechanisms (scientific realism) to reveal what our senses cannot detect. (This is also why we use surveys, microscopes, and telescopes.) When individual-level data do not provide insights into the workings of society, we need to find other methods that allow us to uncover these social mechanisms. Thus, the counterfactual method is not the problem underlying the causal inferences on race in social epidemiology (10, 52, 53). The problem is that race is conceptualized as an observable property of an organism (skin color and other visible phenotypic features) rather than as a social relation that generates economic, political, and cultural inequalities as revealed with a realist epistemology (10). Similarly, the reciprocal relationships between social capital and violence can be tested using nonrecursive structural equation models (54). The problem with social capital is not one of methods but its vague theoretical construct (55) that leads to identification problems (56). What appears as a lack of methodological innovation is often a lack of good theory.
The call for complexity is well taken. Yet, advancing epidemiologic methods as a distinct branch of statistics (with no social epidemiologic theory content) is a risky proposition (57). The common practice of imposing a mathematical (e.g., additive) model without considering theory-informed social mechanisms may lead to unrealistic conclusions about the relation between social factors and health. A simple 2 × 2 table to express the theoretical ideas of social epidemiology may be preferable than the “… ad hoc models suggested by complex statistical techniques” (55, p. 239). Large individual-level datasets will not tell us why the murder rate in Chicago is dramatically higher than that in the United Kingdom (58). That would require data on arms production, distribution, laws regulating their purchase, racial segregation, and social cohesion, among other processes, and few if any of these variables are accessible from individual-level surveys. A realist epistemologist is open to a variety of methods (e.g., mixed methods) if they help uncover the social mechanisms that influence health outcomes (26). We need to augment social epidemiology by complementing it with other methods if we want to succeed in the 6 areas of social epidemiology outlined by Galea and Link.
SHOULD SOCIAL EPIDEMIOLOGY DEVOTE ITSELF MOSTLY TO THE STUDY OF BIOLOGIC MECHANISMS?
Social epidemiology being a biosocial science has led to repeated calls for uncovering how society determines proximal biologic mechanisms that lead to ill health (2). Some interesting findings have been produced in recent years (59), yet what sets social epidemiology apart is its unique attempt to uncover social mechanisms (60), whereas other specialties that also include social variables in their studies (cardiovascular, cancer, psychiatric) are in a better position to deal with the biological end of the spectrum. Moreover, health, as defined by the World Health Organization, is not just about biology. The field would veer towards biologic reductionism if studies attempted to link, say, income support policies to specific biologic processes. This is the same way in which it is reductionist to link social policies to a single disease, as is the norm in epidemiology. When we look for the association of poverty with the risk of major depression disorder, the nondiseased group is likely to include patients with substance abuse, anxiety, disorders, cardiovascular disorder, diabetes, and many other disorders affected by poverty. Estimates of attributable risk due to poverty on population health (important for applied social epidemiology) will be affected because of misclassification of the outcome. We need better models to appropriately capture the total impact of social exposures and not simply rely on the single-outcome approach (61, 62).
At the same time, some of the traditional problems of social epidemiology could be laid to rest, for example, the “selection/causation” issue, with its social Darwinist (i.e., biologic Spenglerism) connotations that naturalize inequities (63). That is, if poverty is a risk factor for mental illness among youths, society is deemed responsible and assistance is seen as legitimate. On the other hand, if mental disorders are a risk factor for poverty, the process is deemed as selection and can be attributed to genetic predispositions (the individual); therefore, society is not held responsible. Yet, there is no moral or public health justification for why patients with schizophrenia should be allowed to be poor (also an instance of social causation) if poverty follows the onset of the disorder rather than precedes it (64). Rather, the efforts of social epidemiologists may be better used to figure out the set of social protection policies (housing, active labor markets, health services) that may prevent poverty among these patients. Yet, underlying the selection/causation issue, we find conflicting views about political philosophy (i.e., individual versus collective responsibility) (65) and even scientific metaphysics (66) that are unlikely to be overcome in the near future.
SOCIAL INTERVENTIONS WITH POPULATIONS RATHER THAN ON POPULATIONS
Galea and Link acknowledge that experiments are often inadequate for social interventions. This strict approach to what constitutes evidence is compatible with scientific realism that, contrary to the dominant empiricist view in epidemiology (67), views causality as an ontologic, rather than epistemologic, category.
Another issue regarding social interventions is that social epidemiology may not realize its full potential to reduce social inequalities in health alone (29). Rather, understanding how to change the social production of health requires a multidisciplinary or transdisciplinary approach. Galea and Link assume that we should seek the solutions to social inequalities alone and remain silent on the need for such partnerships. Instead of competing with allied or disciplines, social epidemiologists would benefit by embracing and partnering with similarly minded fields to yield a fuller, more complete understanding of which social changes might bring about the greatest good and how such changes should come about. The existing literature offers few answers to these pressing questions, but partnerships with academics in other disciplines, community members in deprived locales, and decision makers in multiple levels of government can increase the relevance of the evidence generated by social epidemiologists (68, 69). More than other subfields, social epidemiology is uniquely placed to benefit from partnerships to help generate new questions and to ensure findings are used to inform population health interventions. Partnerships especially might be formed with persons implementing social policies and programs to help determine the impact of social change on health, especially given that most social programs do not measure health as an outcome (70).
WILL WE EVER REDUCE SOCIAL INEQUALITIES IN HEALTH?
Still, Galea and Link consider health inequalities a core problem for social epidemiology. Given this, the authors express “hope” and “luck” in seeing health inequalities reduced and anticipate that such a value-laden commitment will meet some opposition from epidemiologists (i.e., scientists). I see no reason for such caution, given the high stakes and consequences of health inequalities. Applied sciences such as social epidemiology are meant to produce knowledge that can be used for social change (e.g., reducing racial or class inequalities in health). Studying social inequalities in health has implicit values because inequalities are considered avoidable and unfair (71). Values are not only implicit (we study cancer because this disease has a negative value, causes immense human suffering, and we want to eliminate it); values are also explicit when social interventions and evaluations are conducted (i.e., in applied social epidemiology and its associated public health technology). This is analogous to a surgeon operating on a patient, performing this duty for the sake of procedural science, and pretending to have no vested interest in whether the patient lives or dies. Such a value-free approach to health is not only unrealistic in our social profession but also harmful to the production of meaningful and actionable science.
The ethos of value-free science that originated from Max Werber's ideas is not followed by basic scientists (whether made explicit or not, scientists have value systems and are heavily invested in their research outcomes) (72) and is inadequate for applied scientists because the goal of their research is to act upon the world. This involves making a decision on what is the proper action, which relies partially on one's values. The unfortunate effects of this ethos among social epidemiologists might be self-censoring, ignoring mechanisms, and discounting intervention research (i.e., their duty as applied scientists) for fear of appearing partial. Inevitably, this leads to more studies finding associations with race but no social mechanisms of racism or racial segregation (10, 73). Moreover, the supposedly value-free decision to circumvent possible mechanisms that link power or politics and health is anything but neutral. The decision to avoid social mechanisms such as racism or social class implies that nothing will be done to address those inequalities. Partiality is not incompatible with objectivity. For example, nothing is holding back applied social epidemiologists from on one hand designing and evaluating a housing intervention for the homeless and mentally ill while on the other hand believing in the immorality of allowing the homeless to remain on the streets (68). Applied scientific methods have a built-in corrective mechanism (74) that gives them a unique advantage over other forms of knowledge production—failed interventions will either be improved upon by others or will not be replicated at all.
A CONFIDENT DISCIPLINE IN AN UNCERTAIN FUTURE
Disciplinary developments are influenced by external social forces out of their control (e.g., funding, politics) (75). The opposition to safe injection sites in spite of the evidence of its public health benefits is a case in point (76). As long as social epidemiology continues to identify new problems and produce interesting findings, the discipline will undoubtedly thrive. Prudently, Galea and Link embrace a future in which social epidemiologists borrow insightful theories from other disciplines to augment existing frameworks. Yet, this is already happening (29, 30, 77).
Time will tell whether the social in social epidemiology will blossom, whether positivism will make room for (scientific) realism, whether the discipline will close upon itself (and risking collapse), or whether it will open up further to multidisciplinary and transdisciplinary developments. In any event, the areas identified by Galea and Link will be crucial to distinguish ourselves from the parent field.
Author affiliation: Bloomberg Faculty of Nursing, Dalla Lana School of Public Health, and Department of Psychiatry, University of Toronto and Center for Research in Inner City Health, St Michael's Hospital, Toronto, Ontario, Canada (Carles Muntaner).
The research leading to these results forms part of the SOPHIE project (Evaluating the Impact of Structural Policies on Health Inequalities and Their Social Determinants and Fostering Change), which has received funding from the European Community's Seventh Framework Program (FP7/2007–2013) under grant agreement no. 278173.
I thank Edwin Ng from the Dalla Lana School of Public Health for his comments on earlier drafts.
Conflict of interest: none declared.