Commentary

Six Paths for the Future of Social Epidemiology

Sandro Galea* and Bruce G. Link

* Correspondence to Dr. Sandro Galea, Department of Epidemiology, Mailman School of Public Health, Columbia University, 722W. 168th Street,Room 1508, New York, NY 10032-3727 (e-mail: sgalea@columbia.edu).

Initially submitted July 20, 2012; accepted for publication January 11, 2013.

Social epidemiology is now an accepted part of the academic intellectual landscape. However, in many ways,social epidemiology also runs the risk of losing the identity that distinguished it as a field during its emergence. Inthe present article, we scan the strengths of social epidemiology to imagine paths forward that will make the fielddistinct and useful to the understanding of population health in future. We suggest 6 paths to such a future, eachemerging from promising research trends in the field in which social epidemiologists can, and should, lead incoming years. Each of these paths contributes to the formation of distinct capacities that social epidemiologistscan claim and use to elaborate or fill in gaps in the already strong history of social epidemiology. They present anopportunity for the field to build on its strengths and move forward while leading in new and critical areas in popula-tion health.

epidemiology; methods; population health; theory

Abbreviation: SES, socioeconomic status.

Editors note: Invited commentaries on this article appearon pages 850, 852, and 858, and the authors responseappears on page 864.

Some years ago, one of us was in conversation with a phy-sician colleague when he was asked what he did in epidemi-ology. When he answered that he considered himself a socialepidemiologist, the colleague replied, But, isnt all epide-miology social?Social epidemiology is now an accepted part of the intel-

lectual landscape in the United States and is experiencing aprominence it did not previously enjoy. As evidence of thisbroad change, consider that all of the leading epidemiologyconferences in the United States have sections dedicated tosocial epidemiology. In some of them, the number of submis-sions that fall under the label of social epidemiology sur-passes even those old epidemiologywarhorses: chronic diseaseand infectious disease epidemiology. In our own large Depart-ment of Epidemiology, social epidemiology competes evenlywith chronic disease and infectious disease epidemiology forincoming student interest.

This was not always the case. Forty years ago, there wasno institutionalized subdiscipline of social epidemiology inthe United States. Certainly, there were practitioners of whatwe would now call social epidemiology, but they linked them-selves to social medicine or medical sociology instead ofseeking to secure a recognized foothold within the discipline.However, beginning in the 1980s, a small group started todene themselves as social epidemiologists (1) and to pressfor recognition in departments of epidemiology, often to fairlyhostile reactions from colleagues in the discipline. One of uswas often told formany years that our department had reachedits quota on social epidemiologists because he was alreadythere; that is, one social epidemiologist was already enough.A request to initiate a course in social epidemiology was deniedbecause we already had a course in psychiatric epidemiologythat dealt with social issues.This started to change in the 1990s. The 1990s was a period

of high turbulence in epidemiology in general, with a seriesof inuential articles that characterized what we might callthe epidemiology wars (2, 3). These papers suggested thatthe purpose of epidemiology was to understand causes at alllevels, from individual exposures to societal inuences (4).

843 Am J Epidemiol. 2013;178(6):843849

American Journal of Epidemiology The Author 2013. Published by Oxford University Press on behalf of the Johns Hopkins Bloomberg School ofPublic Health. All rights reserved. For permissions, please e-mail: journals.permissions@oup.com.

Vol. 178, No. 6DOI: 10.1093/aje/kwt148

Advance Access publication:September 5, 2013

They offered a blueprint, a raison dtre for the eld of socialepidemiology, suggesting a multilevel framework within whichto think about causation and making it acceptable for socialepidemiologists to make their central focus the study of howsocial factors, such as social ties, poverty, and racial segrega-tion, might shape the health of populations. Key theoreticaldevelopments, such as the establishment of the concept offundamental causes (5, 6) and webs of causation (7) roundedout the picture to position the eld at the end of the decadefor what would become an empiric explosion building on theroadmap that had been charted over the preceding 30 years.The 2000s brought the rapid introduction ofmultilevelmodel-ing in the eld (8). Multilevel models were not new; theyhad been used in many other social sciences for at least 2 dec-ades, often with different names (principally random-effectsmodels), but they brought to epidemiology a ready methodo-logical approach to handle thorny issues of clustering aroundlevels of analyses. Multilevel models offered epidemiolo-gists a way to continue using the dominant tools in their eldwith suitable but relatively straightforward adaptation to theemerging search for the social factors that caused popula-tion health.This bringsus to the present,when social epidemiologyholds

appeal for our students, has contributed to a growing aware-ness of the centrality of social drivers of population health inmany health department agendas, and is a well-accepted sub-disciplinary home for epidemiologists in the leading univer-sities in the country.However, and perhaps with some irony, as social epidemi-

ology has achieved a position of primacy and found its placeamong the epidemiologic pantheon, it has also encountereda distinct dangerlosing the identity that distinguished it asaeld during its emergence. Simply put, its success is makingtrouble for itself, placing it in far more danger than at anytime in its relatively brief history.To understand this threat, we go back to the question posed

by our physician colleague interlocutor. Is epidemiology notall social?As far back as 2004, in an editorial that accompanied the

redo of their groundbreaking 1993 article Actual Causes ofDeath in the United States (9), McGinnis and Foege notedthat . . . it is also important to better capture and apply evi-dence about the centrality of social circumstances to healthstatus and outcomes. . . (10, p. 1264). Their point that socialcircumstances need to be understood even when thinking ofthe causes of death was well made; this point was recentlyquantied in an article that showed that, using comparablemethods, the number of deaths attributable to some social(e.g., low educational level) causes are of the same magnitudeas deaths attributable to physiologic (e.g., acute myocardialinfarction) and behavioral (e.g., poor diet) causes (11).Therefore, to a large degree, epidemiology is indeed all

social, and it has been social epidemiologys triumph that ithas taught this to the discipline at large. It would be hard tond an epidemiologist today in any leading academic depart-ment who would claim that social factors are not in thecausal chain of factors that lead to the production of health.Smoking of course causes lung cancer, but social networksinuence smoking (12), as do advertising, social norms, and

taxation rates (1315). Obesity is not simply a matter of energybalance. It must be understood by taking into account foodavailability, the presence or absence of walkable neighbor-hoods, the price of vegetables, and social norms about bodysize. Epidemiologists have been at the leading edge of dem-onstrating all these observations (16, 17), and research thataims to understand these phenotypes (e.g., lung cancer, obesity)must contend with the social causes as, at the very least, com-plicating confounders, if not outright causes of causes inany denitive analysis.This represents a dramatic shift in the eld and one that

well might obviate the need for social epidemiologists. Whatpurpose does it serve to label ourselves social epidemiolo-gists if everyone else is also tackling social causes from theirsubdisciplinary redoubts? Subdisciplinary labels in our eld(i.e., cardiovascular epidemiology, respiratory epidemiology)are fundamentally reinforced by the funding mechanisms atinstitutions that reward just this approach (e.g., the NationalHeart, Lung, and Blood Institute).What space is there for socialepidemiology as a distinct specialty if cardiovascular, infec-tious, and pulmonary epidemiologists are themselves study-ing social causes?We see our potential redundancy as a useful challenge to

forge a stronger and deeper social epidemiology in the timeahead.We use the dramatic change in the stature of social epi-demiology in the context of the United States as a springboardfor imagining a future for social epidemiology in the time ahead.Situated in the present, we scan the strengths of social epide-miology to imagine paths forward that will make our subdis-cipline distinct and useful to the understanding of populationhealth. We suggest 6 paths to such a future, each emergingfrom promising research trends in the eld, in which socialepidemiologists can, and should, lead in coming years.

MACRO-LEVEL DETERMINANTS OF POPULATIONHEALTH

The roots of social epidemiology lie not too far from earlyefforts to understand the development of pathology in thehuman body. Some of the earliest inuential articles in theeld were concerned with how social ties were associated withsurvival from cardiovascular disease (18). Subsequent evolu-tion in the eld moved us to neighborhoods and the exponen-tial increase during the past decade of papers that were concernedwith neighborhood characteristics (e.g., poverty, quality ofthe built environment) and how they shape population health(19). Despite some empiric work and some theoretical papersand books that have made this case, the body of work in theeld that has explored how truly large-scale (macro-level) forcesshape the health of populations remains disproportionatelylimited.We suggest that there are 2 central reasons for this short-

coming. First, thinking andworking at amacro scale are simplymuch harder than they are at amoremicro level. A concernwiththe inuence of urbanicity must, of necessity, involve samplesthat are heterogeneous across urban contexts and involve largesamples that are collected across cities. Well-designed epide-miologic studies that tackle globalization, politics, and cul-tural inuences on health are even harder to enact. Second,

844 Galea and Link

Am J Epidemiol. 2013;178(6):843849

these macro-level structures require an effort of conceptuali-zation and measurement that typically stretches the epidemi-ologists toolkit (20). Epidemiologists are taught to exercisean abundance of caution when reducing complicated macro-level factors to dichotomous, or even scalar, variables in ouranalyses. It is a caution that we laud and one of the reasonswhy we are both epidemiologists. However, this means thatour understanding of these large-scale forces remains limitedand that epidemiology is simply not at the policy table, or atbest is a junior partner, when taxation or efforts to reshape theurban environment are part of the national conversation. It alsomeans that althoughwe understand that thesemacro-level forcesstand to effectively shift the curves of disease incidence, bur-den, and duration (21), we have little empiric data that canmotivate nudging these forces in a salutary direction.We suggest therefore that social epidemiology stands to

play an enormous role in pushing the eld forward if we rede-ned social to extend to and focus on macro-level forces that,with few exceptions, other phenotype-oriented epidemiolo-gists are not readily addressing. This extends to both investi-gating the conceptualization of which macro-level factors matterand understanding how and why they might matter. It hasimplications for both how we train our future social epidemi-ologists, discussed further below, and what is normative inthe eld. We may have to accept more uncertainty, more sto-chasticity in our observations as we move further away fromthe pathophysiologic basis of disease.However, ultimatelywewould be tackling issues that are paramount to the produc-tion of population health and that are unlikely to be taken upby any other phenotype-oriented epidemiologists in the nearfuture. It would also give the eld the depth of expertise inhandling and tackling these issues and, eventually, a voice inthe policy discussions that are inevitably accompanying a rap-idly changing world.

METHODOLOGICAL INNOVATION

Social epidemiology has presented some of the thorniestchallenges with which epidemiologists have wrestled over thepast couple of decades. The counterfactual paradigm (22, 23)has played a central role in clarifying causal thinking in epide-miology. However, the concerns of social epidemiology areall too often not readily handled within this paradigm. Whatdoes it mean to enter a parallel universe wherein everythingis the same except for ones race (24)? Clearly not much clar-ies the isolated effect of race as a cause of populationhealth. Similarly, directed acyclic graphs have brought muchneeded clarity about causal thinking in epidemiology andhave rapidly gained acceptance in the eld (25, 26). However,even the best-specied directed acyclic graphs do not fullycapture the scope of social epidemiologists concerns. For one,the acyclicity of these graphs provided a challenge to many ofthe concerns of the social epidemiologist. For surely as socialcapital may inuence violence, doesnt violence also inuencesocial capital?This somewhat unhappy turn of events, nudging social epi-

demiology to the very limits of our epidemiologic armamen-tarium, may not have been an unalloyed negative. Rather, ithas pushed social epidemiologists to be leaders in innova-tion in epidemiologic methods or, at the very least, prodded

dyed-in-the-wool methodologists to innovate so that socialepidemiologists may go about their work. We have alreadynoted the rapid adoption of random-effects models coincidentwith our turn to questions about the role of neighborhoodeffects. More recently, social epidemiologists have led theway in pushing the eld to engage with methods, such as sys-tems dynamics methods, that have been used to good effectin other social and natural sciences but much less so in ourdiscipline (27, 28). These methods remain limited, their usein its infancy, and we are not necessarily convinced that theywill lead to widespread transformative use. However, theydo represent a further effort in the eld to push thinking for-ward, to expand the accepted toolkit, and to ask what can bedone better and what we can do to achieve a stronger, moreversatile set of methodological approaches.We suggest that social epidemiologists, of necessity and

hopefully by inclination, should lead the way in methodo-logical innovation in the eld.What right dowe have to claimsuch a role? Social epidemiologys focus on long chains ofcausation, operating across social, psychological, and bio-logic systems, induces a new lens that its practitioners mustadopt to engage the issues they wish to understand. The lensforces a novel perspective that needs to be used to identifynew strategies for epidemiologic understanding. For example,understanding how cultural norms inuence health behaviors,how the 2 interact reciprocally, and how jointly they shape theincidence of substance abuse takes us far from the 2 2 table(29, 30). We are not sure where this will push our methods,but with persistence, social epidemiology can lead in thedevelopment of both the novel approaches and the methodsneeded to tackle the central questions about how social cir-cumstances shape the health of populations.

UNDERSTANDING MECHANISMS

Despite the broad acceptance of the roles of causes at mul-tiple levels of inuence in epidemiology, increasing discipli-nary subspecializationhas led to theseparationofepidemiologistsinto camps framed by these same levels of inuence, perhapswith social epidemiologists and molecular epidemiologistsat the 2 poles. Although this may be due in part to the increasingspecialization required of both species of epidemiologist, it hasled social epidemiology further away from a concern with themechanisms that underlie the link between social circum-stances and the production of health.In some of our earlier writing, we suggested that risk-factor

mechanisms could not account for health disparities by socio-economic status (SES) because SES dynamically shapesexposure to new risk and protective factors over time, recre-ating the SEShealth association in different places and atdifferent times (5). Focusing only on risk factor mechanismsand assuming that they would account for SES-related dis-parities served to distract us from a core focus on the founda-tional determinants of population health, the determinantsthat if manipulated can stand to shift the disease curve mean-ingfully for whole populations. That concern remains, but thediscipline has changed. Many epidemiologists have now cometo this view and have seen the need to incorporate social factorsin any full assessment of disease risk (4, 31, 32). Therefore, weare today at a better point in the history of the eld, one that

Future of Social Epidemiology 845

Am J Epidemiol. 2013;178(6):843849

can embrace both the study of macro-level conditions and thestudy of why and how these forces may matter.We suggest that although social epidemiologists need to

keep their gaze xed on the social, they simultaneously havean enormous amount to gain from exploring biologic mech-anismsmore deeply. This has substantial implications for howwe conceive of the eld and its practitioners. Although overthe past few years there has been some exciting work publishedthat jointly considers how social inuences work together withgenetic and molecular processes, this work is sparse (33, 34).Far more dominant is social epidemiologic work that con-strains itself outside the skin, with an interest in social pro-cesses without much concern with why these processes maymatter. Of the many reasons that social epidemiology mightbenet from closer attention to mechanisms, we discuss 4.First, whatever we presume the root causes to be, in the

end, all disease is biologic; any full accounting must be cog-nizant of this fact. As a result, any brand of epidemiology (butespecially social epidemiology) needs to carry the implica-tions of the factors it studies to the pathophysiologic processesthat must be traversed for disease to occur. The more adeptwe are at showing how social factors inuence biologic pro-cesses, the stronger is our case for the health relevance of thefactors we study. Second, as social epidemiology has matured,so has the range of proposed explanations, that is, the pro-posed social drivers of disease outcomes. As one might rea-sonably expect, some of these explanations have collided toform explanations that compete for primacy in accountingfor particular disease patterns. In many instances, these havetaken the form of classic social selection versus social causa-tion explanations (e.g., neighborhood, SES) (35, 36). Althoughmany approaches are needed to resolve these debates, attend-ing to mechanisms can provide useful information because thedifferent explanations frequently imply different mechanisms.It follows that we can gather evidence about the plausibility ofcompeting explanations by determining whether the mecha-nisms each implies are in fact found to exist upon empiricinquiry. Simply put, attending to mechanisms can help socialepidemiologists sort out which of their plausible social deter-minants are the most compelling ones. Third, once a mecha-nistic pathway is discovered, the social, cultural, political,and economic context shapes the consequences of that dis-covery. Smoking is an obvious example. This signature epi-demiologic discovery has had and continues to have an impacton population health through a powerful morass of social,cultural, economic, and political factors. More recently, dis-coveries concerning infectious disease led to the human pap-illoma virus vaccine that holds the possibility to powerfullyshape cancer outcomes in the decades ahead. However, itsimpact onpopulationhealthwill be shapedbydiffusiondynamicsthat are strongly inuenced by culture, religion, and politics.Mechanisms, once discovered, become social, and social epi-demiologists should closely attend to them for this reason aswell. Fourth and nally, social epidemiology is in a uniqueposition to push the population health sciences forward atthe intersection of factors that matter inside and outside theskin. Doing so would represent perhaps the most radical ofthe prescriptions we suggest here for the eld, but it mightalso represent the most promising direction, positioningsocial epidemiology at the heart of the larger epidemiologic

concern with understanding how causes at multiple levels ofinuence jointly shape health and disease.

SOCIAL INTERVENTIONS

We shall not here retell the now well-worn arguments inepidemiology about the challenges of drawing inference fromobservational studies and about the potential role that inter-ventional research plays in clarifying causal chains (3739).That story has beenwrittenwell elsewhere. Taken to the extreme,this suggests that only experimental approaches can illuminateour understanding of causes (4042).We nd this approach unhelpful and think it is wrong.

However, there is little question that there is much that we canlearn from studying interventions and in particular social exper-iments, many of which will, of necessity, be natural experiments.Social epidemiologists, particularly social epidemiolo-

gists concernedwith large-scale social inuences and circum-stances, have little opportunity to experimentally manipulatethe conditions in which they are interested. It seems unlikelythat any social epidemiologist will be in a position to experi-mentally manipulate progressive taxation levels in differentjurisdictions to understand whether the resultant narrowingin income inequality will also be associated with improvedpopulation health. It is not, however, unlikely that such an oppor-tunity will present itself as a natural experiment, one that canwell be studied to good effect. There are indeed some veryelegant examples in the existing literature that have done justthis: capitalized onnatural experiments to understandhowmanip-ulating social circumstances does or does not shape the healthof populations. Social epidemiologists will be interested inthese natural experiments to the extent that they revealwhetherand to what extent social factors matter for health outcomes.At the same time, social epidemiologists can occasionally bemore broadly useful to the discipline because identifying andstudying the consequences of such social experiments cansometimes provide critically important information for otherareasof epidemiology.Forexample, theNaziblockagecreatedthe Dutch famine that led to discoveries relevant to the asso-ciations between neural tube defects and schizophrenia (43).Theseexamples readily convinceusthat this approach isprom-ising and fruitful. What is perhaps surprising is how few suchillustrations we can identify in the literature.Social structures change all the time. New laws and regu-

lations are introduced with regularity. Urban renewal effortsare put in place. Cultural norms shift. These changes lendthemselves to careful, systematic data gathering that can addinsights not otherwise available through observational studiesalone. Linking to some of the areas discussed here, the studyof natural experiments embeds its own conceptual and meth-odological challenges, further suggesting the inevitability ofthe social epidemiologists engagement in these areas.

THE CENTRAL ROLE OF INTERGROUP DIFFERENCESAND INEQUALITY

The roots of social epidemiology a quarter century agowere indelibly shaped by a concern over black/white differ-ences in some fundamental health outcomes in the UnitedStates (44). That many of these differences remain as they

846 Galea and Link

Am J Epidemiol. 2013;178(6):843849

were decades later well argues for the continued importanceof health inequalities as a core motivation in the eld. Weargue that it is therefore the natural mtier of social epidemi-ologists to maintain an abiding interest in intergroup differ-ences in health, or as they are also called, health inequalitiesor health disparities. We have little hope that these differenceswill vanish any time soon. With some luck and concertedeffort, we may narrow racial differences in health. However,evidence suggests that socioeconomic differences are, if any-thing, growing (4547), not to mention international differ-ences in health indicators (48, 49) that are appalling. Thisprovides sufcient motivation for several lifetimes of workbut can also be clarifying for social epidemiology as a foun-dational commitment to improving the underlying conditionsthat inuence health so as to narrow differences betweengroups.We realize that articulating a values-based commitment to

intergroup equality in health for a scientic discipline mayraise eyebrows. We suggest that what we are proposing is notparticularly different than what is de facto the case in muchof science already, but which is not as explicitly stated. Ourquestions are not unbiased. Much as we may like to think thatwe are asking questions that are informed only by prior dis-passionate knowledge, our preconceptions inevitably frameour questions and, less felicitously, our answers. It is true thatsocial epidemiology has always been more susceptible tothis than have other areas in epidemiology. Many people aredrawn to social epidemiology out of a concern for the social,a desire to understand and improve social conditions that mayinuence health. Few of us are neutral about social conditionsin our nonprofessional lifves. Our values around the impor-tance of a market-orientation to our economy determine forwhom we vote. To some degree, this sets social epidemiolo-gists apart from, for example, occupational epidemiologists.Our sentiments about ceramics are likely less a part of ourcivilian engagement than are our sentiments about incomeinequality. This indeed raises the bar for social epidemiolo-gists engagementwithour data, to force us to applya gimlet eyeto our work and that of our colleagues and tomake sure, above all,that we interpret data clearly, minimizing bias and providingreliable and valid inference that can usefully inform interven-tion. Indeed, this is the only approach that can help us achievewhat might have drawn us to social epidemiologywe willdo little to advance any values we hold if we get our sciencewrong.However, recognizing that social conditions may engage

the epidemiologist does not suggest that social epidemiolo-gists need to shed their concern about the very social condi-tions that motivated us in the beginning. Rather, we suggestthat such efforts are futile and that social epidemiologistswould do better to recognize our driving inuences in thisarea and elevate them, as appropriate, to a motivating concernthat shapes our work.

THEORYAND THE HEALTH OF POPULATIONS

Epidemiology has long ourished as a black-box disci-pline. We have applied our methods, with notable success,to isolate factors that co-vary with disease and aimed to inter-vene on these factors without a good grasp on why theymight

matter. The isolation of folic acid as a determinant of neuraltube defects is a well-recognized case in point (5052). It canbe argued, however, that these cases are not normative. Rather,particularly as we expand the scope of epidemiology to takeinto account a broad range of factors, a black-box approachmay lead us to the isolation of spurious associations or to achase for associations where there are none.As we see it, the only alternative to this conundrum is a

deeper engagement for the eld in theory, a richer groundingin an understanding of why particular factors may matter,and the condence to articulate a priori hypotheses aboutwhat social conditions might matter, leading to testing throughobservational or experimental studies. Thiswill require not onlya deeper grounding in theory borrowed from other disciplinesbut also development of a theory of the production of popula-tion health that stands on its own. With a few notable excep-tions, epidemiology remains quite poor in this regard.We see this as both a critical area for the development of

the eld and a central opportunity for social epidemiology tohelp population health science transcend some of the indi-vidualistic thinking inherited from the medical sciences thatat times holds back epidemiology as a whole. Population healthscience needs to be rooted in an understanding of the driversof the health of populations. These drivers, the factors thatshift the curve of disease incidence, severity, and duration areby denition outside the curve, exerting their inuence onwhole populations rather than on individuals. Similarly, ourmethods in the eld are population-based methods, havinglittle utility for individual diagnosis and prediction. However,conation of population-based and individual-based infer-ence is commonplace in biomedical research. We argue thata clearer articulation of the concepts that underlie social epi-demiology and how that shapes our questions and our infer-ence can further help clarify our concern with populationsand with the social drivers thereof.

IMPLICATIONS FOR THE NEXT GENERATION OFSOCIAL EPIDEMIOLOGISTS

Our prescriptions for the eld represent a tall order, onethat is too tall, in fact, for any one social epidemiologist toembrace. Perhaps it is a sign of the maturity of the subdisci-pline that it can move forward in so many directions, wellbeyond the scope of any one persons research. The approachproposed here does have clear implications for both how wepractice the eld today and, perhaps more importantly, forhow we train the next generation of social epidemiologists.Aspiring to be forward-looking, we suggest 3 particular

areas that bear careful consideration in the education of thenext generation of social epidemiologists. First, we must trainsocial epidemiologists to be comfortable across levels of inu-ence. This does not mean that all social epidemiologists couldor should move comfortably between studies of the role ofthe urban environment in shaping health and studies abouthow exogenous stressors induce epigenetic changes that canmodify gene expression and hence physiologic processes.However, we suggest that all future social epidemiologistsneed to at least understand both the social conditions that arefoundational and the causal links that result in the embodi-ment of health. So doing will lead us to ask better questions

Future of Social Epidemiology 847

Am J Epidemiol. 2013;178(6):843849

and focus on plausible answers. It also represents an opportu-nity to bring a measure of synthesis to the broader disciplineof epidemiology that currently is pursuing all these strands,often resulting in diverging and poorly cohering patterns.Second, we need a generation of social epidemiologists

that is comfortable with and can push the limits of our disci-plinary methods. This does not necessarily mean that socialepidemiologists will or should become epidemiologic meth-odologists. However, it does mean that we need a generationthat is not content with the constraints of our current methodsand that is comfortable working with peers in biostatistics,mathematics, and computer science to nd or develop bettertools. We might suggest that this should be a feature of allfuture epidemiologists. Perhaps, but, relevant to this essay, itsuggests that social epidemiologists need to be trained notsimply to adapt methods but to critically understand and chal-lenge methods. That is indeed a different orientation than ourcurrent training in the eld, and it suggests a training thatis in informed by propulsive questioning and that encouragesemerging scholars not to settle for our current methods butrather to seek their own.Third, it is incumbent upon us to embed the training of

social epidemiologists in a deeper appreciation of theory (andinevitably history of that theory) than we currently do. Becausethe eld has emerged rapidly over the past 2 decades, muchsocial epidemiology is informed by theory, but that theory isat best grafted onto the eld from other disciplines, sometimestting poorly. We aspire to a generation of social epidemiol-ogists that asks why we are asking the questions we ask, to theend of asking very different questions and identifying betteranswers than the ones we have produced thus far.

CONCLUSION

We return to the question framed by our physician colleague.If epidemiology is all social, or perhaps more accurately, ifall epidemiologists are concerned with social circumstancesas at least part of the causal chain, how then does a social epi-demiologist distinguish herself?We suggest here that we do so through clarity of focus on

understanding the macro-level factors that matter, how theyaffect biologic processes with the body, and the methods thathelp us articulate these answers. We think that well distin-guishes social epidemiologists from other epidemiologistswho are, correctly for their subdiscipline, concerned cen-trally with the interface between exposure and pathogenesis,focusing on specic disease processes. Although there are 6paths identied, they all havemuch in common. Each of thesepaths contributes to the formation of distinct capacities thatsocial epidemiologists can claim and use to elaborate or llin the gaps in the already strong history of social epidemiol-ogy. They present then an opportunity for the eld to buildon its strengths and move forward while leading in new, crit-ical areas in population health.There is a danger in our prescriptions: In attempting to do

all of this, social epidemiologists run the risk of dabbling,developing a limited understanding of social circumstances,biologic processes, theory, and methods. A vigilant eld canavoid this pitfall. We can give our emerging scholars the

foundations they need to then pursue any of the 6 paths weoutline here. We suggest that doing so will lead to a genera-tion of scholars and scholarship that can move social epide-miology in exciting and fruitful new directions.

ACKNOWLEDGMENTS

Author afliation: Department of Epidemiology, MailmanSchool of Public Health, Columbia University, New York,New York (Sandro Galea, Bruce G. Link).Dr. Galea is funded in part by National Institutes of Health

grants MH 095718, MH 082729, MH 082598, andDA034244 and Department of Defense grant W81XWH-07-1-0409. Dr. Link is funded in part by National Institutes ofHealth grant HD 058515 and Robert Wood Johnson Founda-tion grants RWJF 67952, RWJF 69247, and RWJF 70281.We thank Laura Sampson for her research assistance on

the preparation of this manuscript.Conict of interest: none declared.

REFERENCES

1. Syme SL. A conversation with Leonard Syme. Interview byW. Thomas Boyce. Epidemiology. 2011;22(6):867871.

2. Kaplan GA. The role of epidemiologists in eradicability ofpoverty. Lancet. 1998;352(9140):16271628.

3. Rothman KJ, Adami HO, Trichopoulos D. Should the missionof epidemiology include the eradication of poverty? Lancet.1998;352(9130):810813.

4. Kaplan GA. What is the role of the social environment inunderstanding inequalities in health? Ann N Y Acad Sci.1999;896:116119.

5. Link BG, Phelan J. Social conditions as fundamental causes ofdisease. J Health Soc Behav. 1995;35:8094.

6. Phelan JC, Link BG, Tehranifar P. Social conditions asfundamental causes of health inequalities: theory, evidence,and policy implications. J Health Soc Behav. 2010;51(suppl):S28S40.

7. Krieger N. Epidemiology and the web of causation: hasanyone seen the spider? Soc Sci Med. 1994;39(7):887903.

8. Diez-Roux AV. Multilevel analysis in public health research.Annu Rev Public Health. 2000;21:171192.

9. McGinnis JM, Foege WH. Actual causes of death in theUnited States. JAMA. 1993;270(18):22072212.

10. McGinnis JM, Foege WH. The immediate vs the important.JAMA. 2004;291(10):12631264.

11. Galea S, Tracy M, Hoggatt KJ, et al. Estimated deathsattributable to social factors in the United States. Am J PublicHealth. 2011;101(8):14561465.

12. Christakis NA, Fowler JH. The collective dynamics ofsmoking in a large social network. N Engl J Med. 2008;358(21):22492258.

13. Lovato C, Linn G, Stead LF, et al. Impact of tobaccoadvertising and promotion on increasing adolescentsmoking behaviours. Cochrane Database Syst Rev. 2003;(4):CD003439.

14. Buller DB, Borland R, Woodall WG, et al. Understandingfactors that inuence smoking uptake. Tob Control. 2003;12(suppl 4):IV16IV25.

848 Galea and Link

Am J Epidemiol. 2013;178(6):843849

15. Kengganpanich M, Termsirikulchai L, Benjakul S. The impactof cigarette tax increase on smoking behavior of dailysmokers. J Med Assoc Thai. 2009;92(suppl 7):S46S53.

16. Arcan C, Hannan PJ, Fulkerson JA, et al. Associations ofhome food availability, dietary intake, screen time andphysical activity with BMI in young American-Indianchildren. Public Health Nutr. 2013;16(1):146155.

17. Casagrande SS, Gittelsohn J, Zonderman AB, et al.Association of walkability with obesity in Baltimore City,Maryland. Am J Public Health. 2011;101(suppl 1):S318S324.

18. Berkman LF, Syme SL. Social networks, host resistance, andmortality: a nine-year follow-up study of Alameda Countyresidents. Am J Epidemiol. 1979;109(2):186204.

19. Kaplan GA. Whats wrong with social epidemiology, and howcan we make it better? Epidemiol Rev. 2004;26:124135.

20. Kaufman JS. Making causal inferences about macrosocialfactors as a basis for public health policies. In: Galea S, ed.Macrosocial Determinants of Population Health. New York,NY: Springer; 2007:355373.

21. Rose G. Sick individuals and sick populations. Int JEpidemiol. 1985;14(1):3238.

22. Maldonado G, Greenland S. Estimating causal effects. Int JEpidemiol. 2002;31(2):422429.

23. Greenland S, Brumback B. An overview of relations amongcausal modelling methods. Int J Epidemiol. 2002;31(5):10301037.

24. Kaufman JS, Cooper RS. Commentary: considerations foruse of racial/ethnic classication in etiologic research.Am J Epidemiol. 2001;154(4):291298.

25. Fleischer NL, Diez Roux AV. Using directed acyclic graphs toguide analyses of neighbourhood health effects: anintroduction. J Epidemiol Community Health. 2008;62(9):842846.

26. Hernan MA, Robins JM. Instruments for causal inference: anepidemiologists dream? Epidemiology. 2006;17(4):360372.

27. Galea S, Riddle M, Kaplan GA. Causal thinking and complexsystem approaches in epidemiology. Int J Epidemiol. 2010;39(1):97106.

28. Auchincloss AH, Diez Roux AV. A new tool forepidemiology: the usefulness of dynamic-agent models inunderstanding place effects on health. Am J Epidemiol.2008;168(1):18.

29. Keyes KM, Schulenberg JE, OMalley PM, et al. The socialnorms of birth cohorts and adolescent marijuana use in theUnited States, 19762007. Addiction. 2011;106(10):17901800.

30. Keyes KM, Schulenberg JE, OMalley PM, et al. Birth cohorteffects on adolescent alcohol use: the inuence of social normsfrom 1976 to 2007. Arch Gen Psychiatry. 2012;69(12):13041313.

31. Mujahid MS, Diez Roux AV, Morenoff JD, et al.Neighborhood characteristics and hypertension. Epidemiology.2008;19(4):590598.

32. Marmot MG, Bosma H, Hemingway H, et al. Contribution ofjob control and other risk factors to social variations incoronary heart disease incidence. Lancet. 1997;350(9073):235239.

33. Franceschini N, Rose KM, Storti KL, et al. Social- andbehavioral-specic genetic effects on blood pressure traits: theStrong Heart Family Study. Circ Cardiovasc Genet. 2009;2(4):396401.

34. Koenen KC, Aiello AE, Bakshis E, et al. Modication of theassociation between serotonin transporter genotype and risk

of posttraumatic stress disorder in adults by county-levelsocial environment. Am J Epidemiol. 2009;169(6):704711.

35. Dohrenwend BP, Levav I, Shrout PE, et al. Socioeconomicstatus and psychiatric disorders: the causation-selection issue.Science. 1992;255(5047):946952.

36. Diez Roux AV, Merkin SS, Arnett D, et al. Neighborhood ofresidence and incidence of coronary heart disease. N EnglJ Med. 2001;345(2):99106.

37. Kaufman JS, Kaufman S, Poole C. Causal inference fromrandomized trials in social epidemiology. Soc Sci Med.2003;57(12):23972409.

38. Stronks K, Mackenbach JP. Evaluating the effect of policiesand interventions to address inequalities in health: lessonsfrom a Dutch programme. Eur J Public Health. 2006;16(4):346353.

39. Oakes JM. The (mis)estimation of neighborhood effects:causal inference for a practicable social epidemiology. Soc SciMed. 2004;58(10):19291952.

40. Pocock SJ, Collier TJ, Dandreo KJ, et al. Issues in thereporting of epidemiological studies: a survey of recentpractice. BMJ. 2004;329(7471):883.

41. Jaffee SR, Strait LB, Odgers CL. From correlates to causes:can quasi-experimental studies and statistical innovations bringus closer to identifying the causes of antisocial behavior?Psychol Bull. 2012;138(2):272295.

42. Byar DP. Why data bases should not replace randomizedclinical trials. Biometrics. 1980;36(2):337342.

43. Susser ES, Lin SP. Schizophrenia after prenatal exposure to theDutch Hunger Winter of 19441945. Arch Gen Psychiatry.1992;49(12):983988.

44. Wallerstein NB, Yen IH, Syme SL. Integration of socialepidemiology and community-engaged interventions toimprove health equity. Am J Public Health. 2011;101(5):822830.

45. Stamatakis E, Wardle J, Cole TJ. Childhood obesity andoverweight prevalence trends in England: evidence forgrowing socioeconomic disparities. Int J Obes (Lond).2010;34(1):4147.

46. Lindert PH. Three centuries of inequality in Britain andAmerica. In: Atkinson AB, Bourguignon F, eds. Handbookof Income Distribution. Amsterdam, Netherlands: Elsevier/North-Holland; 2000:167216.

47. Wolff EN. Top Heavy: The Increasing Inequality of Wealth inAmerica and What Can Be Done About It. New York, NY:The New Press; 2002.

48. Masters RK, Hummer RA, Powers DA. Educationaldifferences in U.S. adult mortality: a cohort perspective.Am Sociol Rev. 2012;77(4):548572.

49. Montez JK, Hummer RA, Hayward MD, et al. Trends in theeducational gradient of U.S. adult mortality from 1986 to 2006by race, gender, and age group. Res Aging. 2011;33(2):145171.

50. Czeizel AE, Dudas I. Prevention of the rst occurrenceof neural-tube defects by periconceptional vitaminsupplementation. N Engl J Med. 1992;327(26):18321835.

51. Prevention of neural tube defects: results of the MedicalResearch Council Vitamin Study. MRC Vitamin StudyResearch Group. Lancet. 1991;338(8760):131137.

52. Berry RJ, Li Z, Erickson JD, et al. Prevention of neural-tubedefects with folic acid in China. China-U.S. CollaborativeProject for Neural Tube Defect Prevention. N Engl J Med.1999;341(20):14851490.

Future of Social Epidemiology 849

Am J Epidemiol. 2013;178(6):843849

/ColorImageDict > /JPEG2000ColorACSImageDict > /JPEG2000ColorImageDict > /AntiAliasGrayImages false /CropGrayImages true /GrayImageMinResolution 150 /GrayImageMinResolutionPolicy /OK /DownsampleGrayImages true /GrayImageDownsampleType /Bicubic /GrayImageResolution 175 /GrayImageDepth -1 /GrayImageMinDownsampleDepth 2 /GrayImageDownsampleThreshold 1.50286 /EncodeGrayImages true /GrayImageFilter /DCTEncode /AutoFilterGrayImages false /GrayImageAutoFilterStrategy /JPEG2000 /GrayACSImageDict > /GrayImageDict > /JPEG2000GrayACSImageDict > /JPEG2000GrayImageDict > /AntiAliasMonoImages true /CropMonoImages true /MonoImageMinResolution 1200 /MonoImageMinResolutionPolicy /OK /DownsampleMonoImages true /MonoImageDownsampleType /Bicubic /MonoImageResolution 175 /MonoImageDepth 4 /MonoImageDownsampleThreshold 1.50286 /EncodeMonoImages true /MonoImageFilter /CCITTFaxEncode /MonoImageDict > /AllowPSXObjects true /CheckCompliance [ /None ] /PDFX1aCheck false /PDFX3Check false /PDFXCompliantPDFOnly false /PDFXNoTrimBoxError true /PDFXTrimBoxToMediaBoxOffset [ 0.00000 0.00000 0.00000 0.00000 ] /PDFXSetBleedBoxToMediaBox true /PDFXBleedBoxToTrimBoxOffset [ 0.00000 0.00000 0.00000 0.00000 ] /PDFXOutputIntentProfile (None) /PDFXOutputConditionIdentifier () /PDFXOutputCondition () /PDFXRegistryName () /PDFXTrapped /False

/Description >>> setdistillerparams> setpagedevice