Epidemiol Rev 2004;26:124-135
© 2004 by the Oxford University Press
What’s Wrong with Social Epidemiology, and How Can We Make It Better?
George A. Kaplan1,2
1 Center for Social Epidemiology and Population Health, School of Public Health, University of Michigan, Ann Arbor, MI.
2 Department of Epidemiology, School of Public Health, University of Michigan, Ann Arbor, MI.
Correspondence to Dr. George A. Kaplan, Department of Epidemiology, University of Michigan, 1214 South University Avenue, Ann Arbor, MI 48104-2548 (e-mail: gkaplan{at}umich.edu).
Received for publication December 11, 2003;
accepted for publication February 20, 2004.
 |
INTRODUCTION
|
|---|
It is perhaps ironic that an epidemiologist who has been working
in the field of social epidemiology for over a quarter of a
century, and who directs a center focused on social epidemiology,
should coin a title suggesting that there is something "wrong"
with social epidemiology. Perhaps it is even inopportune, as
it could provide ammunition to those who believe that the practice
of social epidemiology is misguided, unscientific, ideological,
or too overreaching (
1–
3). However, this title was chosen
purposely with the hope that identifying some of the critical
intellectual, methodological, and empirical lacunae and challenges
in social epidemiology might promote continuing development
of a social epidemiology that is both scientifically enlightening
and useful, productive, and contributory to the public’s
health.
Indeed, the hope is that the "social" in epidemiology will become so integral a part of epidemiology that the term can be dropped altogether. To assert that all epidemiology is social is not an attempt at intellectual hegemony—that the problems of disease and the distribution of disease in populations over time and space can be understood from a social perspective only or that such information is in some sense more fundamental that other types of information about disease determinants. In the same way that our understanding of the etiology of chronic and infectious diseases benefits from knowledge of the pathobiologic processes involved in such diseases, increased understanding of social factors, broadly considered, may shed light on processes every bit as integral to our understanding of the etiology of those diseases.
There is no question that social epidemiology has come of age and that the term "social epidemiology" is being increasingly used to describe examination of the role of a broad array of social factors in the development and progression of many important health problems, and in the natural history of the risk factors for those diseases and conditions. While not all may agree with Geoffrey Rose’s assertion that "the primary determinants of disease are mainly economic and social, and therefore its remedies must also be economic and social" (4, p. 129), there is no question that there has been enormous growth in the study of these economic and social forces on health and disease. Figure 1, which plots growth since 1966 in use of the term "social epidemiology" in article titles, abstracts, and keywords, graphically illustrates this increased interest. Beginning in the early 1980s, growth of such publications increased rapidly, well fit by an exponential curve. In fact, this figure is likely to dramatically underrepresent the growth of social epidemiologists’ interest in the matter; a similar exponential growth pattern has been seen when studies examining socioeconomic position and health (5) and social capital and social relationships (6) were similarly totaled.
View larger version (17K):
[in this window]
[in a new window]
|
FIGURE 1. Numbers of publications included in MEDLINE, Current Contents, and PsychINFO that contain "social epidemiology" in the title, abstract, or keywords, 1966–2002.
|
|
Other contributions in this volume of
Epidemiologic Reviews take up the wide variety of topics that social epidemiologists
study. In what follows, I instead touch on a series of issues
that highlight some of the critical problems with which social
epidemiology must grapple. Many of the issues discussed are
not restricted to social epidemiology and have their analogs
in other areas of epidemiology. Thus, while the focus of this
review is on social epidemiology, it would be misleading to
suppose that similar criticisms do not apply to other lines
of research. In addition, I want to make it clear that many
of the criticisms raised apply to my own work as well as to
others’. Because this is an attempt at a form of self-criticism
of the field, I do not focus on as extensive citation as other
reviews do. Where I do cite the work of others, it should be
considered illustrative, not as singling out a particular piece
of work. The lessons to be learned hopefully apply to the many.
|
METAPHORS AND MODELS
|
|---|
Figures
2,
3,
4, and
5 present a few of the many diagrammatic
models that have been used to illustrate recent social epidemiologic
approaches to understanding the social determinants of health
and health disparities (e.g., Kaplan et al. (
7), Marmot (
8),
House (
9), Lynch (
10)). These models have many features in common,
the most prominent being an emphasis on layered, multilevel
understandings; a multiplicity of pathways; and possibilities
for reciprocal influences. Such models serve as important metaphors,
attempting to portray the component parts of complex processes,
their interrelations, and the temporal relations between components.
They act as an important caution against the potentially misleading
oversimplification that comes from focusing on one level of
influence, often the one most proximal to the outcome, and not
the flow of information and influence represented by the totality
of processes and relations. Similar to a good cartoon, they
remove extraneous factors and draw our attention to what is
believed to be most critical.
View larger version (46K):
[in this window]
[in a new window]
|
FIGURE 2. Upstream and downstream determinants of population health. Reprinted with permission from Promoting Health: Intervention Strategies from Social and Behavioral Research, by the National Academy of Sciences. Courtesy of the National Academies Press, Washington, DC.
|
|
View larger version (26K):
[in this window]
[in a new window]
|
FIGURE 3. Environmental, psychosocial, and biological pathways linking socioeconomic status to diabetes mellitus (DM), coronary heart disease (CHD), and well-being. CNS, central nervous system. Reprinted with permission from the American Sociological Association. James S. House, "Understanding Social Factors and Inequalities in Health: 20th Century Progress and 21st Century Prospects," Journal of Health and Social Behavior, volume 43, number 2 (June 2002), pages 125–142.
|
|
View larger version (32K):
[in this window]
[in a new window]
|
FIGURE 4. A conceptual framework for understanding social inequalities in health and aging. Solid arrows between boxes indicate presumed causal relations among variables; dotted arrows intersect solid arrows, indicating an interaction between the conditioning variables in the box at the beginning of the solid arrow in predicting variables in the box at the head of the solid arrow (10).
|
|
View larger version (24K):
[in this window]
[in a new window]
|
FIGURE 5. Multilevel, life-course determinants of population health. Figure 15-3 from Chapter 15: "Multilevel Approaches to Understanding Social Determinants," by Michael Marmot, from Social Epidemiology, edited by Lisa Berkman and Ichiro Kawachi, copyright " 1999 by Oxford University Press, Inc. Used by permission of Oxford University Press, Inc.
|
|
While the heuristic utility of such models can be substantial,
they do both too much and too little. Such models are common
in science, but the span of factors considered in models of
this type in social epidemiology—sometimes linking the
most macro- and microlevel phenomena—is considerably broader
than that found in many scientific pursuits (e.g., figure
6;
Kaplan and Lynch (
11)), creating considerable problems regarding
data availability and analytical methods. Information is seldom
available at the multitude of levels portrayed in such models
and, where available, is often measured cross-sectionally, making
the temporal influences that we consider so important in the
assessment of causality opaque. Against the backdrop of these
models of complex longitudinal processes, it is not unusual
to attempt to use various standard multivariate statistical
techniques to examine the relative contribution of one social
determinant versus another to the incidence or progression of
disease. Thus, the relative impact of income and race on an
outcome may be compared; the independent effect of income, education,
and occupation may be estimated; or the strength of the association
between job control and occupational social class and some outcome
may be estimated. The critical problem is that simultaneous
measurement and use of these variables in multivariate models
belies the historical, life-course, and temporally ordered social
stratification processes that they reflect. Use of the standard
toolbox of multivariate regression techniques to investigate
these complex social epidemiologic models becomes even more
difficult absent information related to measurement error (
12);
changes in exposure over time (
13); the reciprocal effects of
behavioral and social factors, for example, on each other over
time (
14); and deeper issues related to the very identifiability
of certain kinds of causal effects (
15).
View larger version (31K):
[in this window]
[in a new window]
|
FIGURE 6. Primordial prevention of cardiovascular disease (11). Dashed lines indicate processes occurring in "developing" parts of the world. Reprinted from Preventive Medicine, volume 29, by G. A. Kaplan and J. W. Lynch, "Socioeconomic Considerations in the Primordial Prevention of Cardiovascular Disease," pages S30–S35, copyright 1999, with permission from Elsevier.
|
|
Caught between this rock of inadequate data and the hard place
of analytical limitations, are such models worthwhile? The ultimate
answer will of course result from the extent to which both the
heuristic use of such models and the analytical results based
on them are illuminating. On the analytical side, new multilevel
analytical techniques (e.g., Raudenbush and Bryk (
16), Diez
Roux (
17)), extensions of recent advances in causal analysis
and simulation (e.g., Greenland et al. (
18), Wolfson (
19)),
and techniques borrowed from other fields (e.g., Zohoori and
Savitz (
20)) may prove useful.
In my opinion, the recent substantively modest, but methodologically complex contribution by Adams et al. (21) and the responses to it by economists, epidemiologists, and others in the same publication suggest a less optimistic view. While methodological rigor is always to be applauded, one is left with the impression that a number of the authors believe that the analysis of causal relations in observational data is so flawed as to potentially threaten even the conclusion that smoking causes lung cancer.
The fundamental need for better and more comprehensive data to "test" these models will not be solved by better statistical techniques, and there has been considerable lament regarding the ascendancy of technique over theory in epidemiology (22, 23) and its separation from basic foci of public health. Indeed, development of epidemiologic theory per se, separate from techniques for analyzing causal effects and partitioning sources of noise in data (24), may be required. However, even such developments will not substitute for better sources of data and for new methods to allow the stitching together of data from a variety of sources, levels, periods, and places. Quilts of such data stitched together by using such new techniques, and incorporating sensitivity analyses and other techniques, could considerably strengthen our ability to turn the complex models of social epidemiology into useful analytical models of disease processes in persons and populations.
|
THE NEED FOR SOCIAL EPIDEMIOLOGIC THEORY
|
|---|
Perhaps nowhere is the need for social epidemiologic theory
more apparent than in the study of "place" effects on health.
While some have argued that it is methodologically difficult
to identify the effects of context (
25), and discussion of the
role of context versus compositional effects continues (
26),
there is now an impressive array of studies from epidemiology,
human development, sociology, and other disciplines suggesting
an important role of place in a variety of health and developmental
outcomes (
27–
30). For example, Haan et al. (
31), using
the Alameda County Study cohort, found that residence in a federally
designated poverty area was associated with more than a 50 percent
increased risk of death over the next 9 years. These findings
were replicated at the national level by Waitzman and Smith
(
32), and, in the Alameda County Study, residence in the poverty
area was associated with greater declines in physical activity
over time and an increased incidence of depression (
33,
34).
Since this early study, there has been a rapid increase in studies
of a variety of health and mortality effects associated with
residence in areas as small as census block groups and as large
as states and nations. Although use of sophisticated multilevel
statistical techniques in such studies has increased tremendously,
numerous conceptual and methodological questions still remain
unanswered (
17,
35).
Prominent among these are questions concerning what it is about residence in a particular place that conveys increased or decreased risk. In the case of more traditional environmental exposures, it is much easier to come up with an approach to exposure measurement, modeling of exposure effects, selection of possible confounders, and plausible biologic pathways. Such areas of research also point one in the direction of a particular level or scope of spatial analysis. For example, studies of children’s exposure to lead-based paint would be more likely to focus on the spatial "grain" of the home and environment very proximate to the home; in contrast, studies of outdoor, airborne particulate matter would likely focus on much wider areas, perhaps incorporating knowledge of local geography and air circulation patterns. In the relative absence of such well-specified models of the level of influence of a particular social or economic factor on a particular health outcome, one is left with examples such as the literature relating income inequality to health outcomes, in which the "grain" varies from census block to country.
Similarly, the spatial scope over which determinants of specific outcomes such as all-cause mortality (36), infant mortality (37), or cardiovascular disease (38), or other outcomes, operate is less clear. Without some theory to suggest the ways in which the determinants of these outcomes are spatially embedded, one is left with a plethora of analyses of "area" effects that seem more often driven by the level of spatial data available, or other arbitrary factors, than by reasoned etiologic considerations.
This lack of clarity is not without its perils because it often leads to wide use of terms such as "neighborhood," which, while they may (or may not) invoke some shared implicit definition, are rather opaque with regard to the specific processes that are health damaging or enhancing. There are some exceptions; for example, work exists on the impact of informal social controls, operating at the small-scale neighborhood level, on the antisocial behavior of adolescents (39). Here, there are clearly stated social processes, which can be measured on a relatively small geographic scale, and a specified process that links these spatially distributed characteristics to the outcome in question.
|
ANALYTICAL AND METHODOLOGICAL PITFALLS IN SOCIAL EPIDEMIOLOGY
|
|---|
As in most areas of epidemiologic study, social epidemiology
makes predominant use of observational data, with all of the
attendant methodological and analytical challenges. For the
most part, advances in measurement theory and causal analysis
have not penetrated social epidemiologic analyses very deeply.
For example, the search for social factors "independently" associated
with a particular health outcome often involves use of more
traditional risk factors measured rather crudely, presumably
leading to considerable residual confounding. Along with the
crudity of measurement, most often dictated by the data available
in secondary analyses, these potential confounders are often
subject to considerable change over time, which may or may not
be related to the social factor under examination; therefore,
baseline adjustment for them results in considerable misclassification
of exposure over time.
The search for "independent" risk factors has been a dominant force in social epidemiology, and certainly risk factor epidemiology has come under fire from a number of quarters (40). As is true for much of epidemiology, claims to independent effects often conflate issues of statistical independence that are at least dependent on measurement and modeling approach with causal independence, a process not without its perils. In social epidemiology, this search for independent effects may represent more of an attempt at legitimization of a newly evolving field—identifying "new" risk factors—than something more informative. Many analyses stop with demonstrating a statistically significant association between the risk factors and outcomes in question after adjustment for known risk factors. While it could be argued that identifying new risk factors can catalyze the search for new disease mechanisms, it can also result in a plethora of new "social" risk factors that sometimes exist in an almost miasma-like fog; we observe them to be importantly related to some outcome but cannot really identify the mechanisms that explain this association.
Leaving aside the very real possibility that the measurement and analysis issues discussed earlier lead to misleading observations of independence, several examples of this problem come to mind. For example, much of the early work showing an association between socioeconomic position and mortality from cardiovascular disease or all causes statistically adjusted for potential behavioral and pathophysiologic pathways and demonstrated that adjustment for these factors did not eliminate the increased risk associated with lower socioeconomic position (41). This work has led to broad generalizations about these associations not being explained by such pathways, although some analyses do not support such a conclusion (42). Although eliminating candidate pathways can lead to identification of potentially new pathways mediating the relation, it can also represent an approach that focuses on eliminating all potential explanatory pathways. Such an approach seems less useful than one carefully demonstrating the contribution of intermediate pathways to the observed association. As a counterexample, the maturing approach to the relation between cynical hostility and cardiovascular outcomes illustrates the value (43, 44) of identifying the role of both behavioral and psychophysiologic pathways in explaining this association.
To some extent, social epidemiology, similar to much of epidemiology, would be well served by going back to careful analysis of antecedent, mediating, and confounding variables in assessing causal relations, as portrayed in Susser’s book, Causal Thinking in the Health Sciences, published over a quarter of a century ago (45). The book carefully lays out a logic of analysis based on a careful and informed biologic and social understanding of the outcomes being studied sometimes not present in more quantitatively sophisticated approaches. While promising, it remains to be seen whether recent work on counterfactual contrasts and causal analysis (46, 47), for example, will take us substantially further than this earlier contribution.
Notwithstanding the arguments about the logic of causal analysis, the focus on multilevel determinants of disease, as illustrated in figure 2, presents new methodological and data challenges to social epidemiologists. Few data sets contain the levels of information necessary to populate empirical tests of such models, and the models are usually not specified well enough to parameterize tests of them. In analyses focused on such models, it is useful to contrast explanations concentrated on "how" and "why" (42). For example, an understanding of the association between socioeconomic position and a particular outcome clearly benefits from an understanding of the specific social, behavioral, and physiological pathways that link socioeconomic position and the outcome—the "how." However, another question remains: "why" is it that these pathways are differently distributed by socioeconomic position? Answering such a question, which, paraphrasing Geoffrey Rose, might be called the "causes of the causes" (48), requires casting a net across a broad array of determinants ranging from historical influences, to public policy, and perhaps even to genes. Although difficult to address "why," it seems clear that an explanation that focuses on only the "how" is a limited one.
|
THE PURSUIT AND PERILS OF INTERDISCIPLINARITY
|
|---|
A recent Institute of Medicine report concluded that "basic
and applied research from a range of disciplines has demonstrated
the importance of reciprocal interactions over time among health
and biological, psychological, and social factors" (
49, p. 332).
In the category of biological factors, the report included "genes,
neurochemical, and hormonal processes and function of the physiological
systems"; in the psychological category, it included "behavioral,
personality, temperamental, cognitive, and emotional" factors;
and, in the social category, it included "socioeconomic status,
social inequalities, social networks and support, and work conditions"
(
48, p. 332). With similar lists, other recent reports and volumes
have come to similar conclusions, all consistent with the approaches
illustrated in figures
2,
3,
4, and
5 (
49–
52). All reflect
the widely appreciated fact that virtually all scientific endeavors
are increasingly inter-, multi-, and transdisciplinary. Although
such approaches fit well within the developing field of social
epidemiology, they are associated with certain problems.
Because many of the topics and constructs in social epidemiology come from the social sciences, they carry with them an immediacy and familiarity not found in the world of genetics, for example. Although we may not all know what a "haplotype" is, we all think we know what "neighborhood," "poverty," "community," or "depression" means. This use of familiar terms has led to a certain conceptual looseness that may not be found in other scientific areas, reminiscent of Humpty Dumpty’s statement, "When I use a word ... it means just what I choose it to mean" (53, p. 113). Thus, investigators might use the construct of "neighborhood" in describing analyses of census tracts or zip codes. For some, the term might mean an area characterized by dense, interconnected networks of people who care about each other; for others, it might mean a historically defined spatial area. In reality, what is meant is actually a spatial area constructed decades ago to represent some degree of demographic homogeneity (census tract) or an area defined by the location of postal facilities (zip codes). Similarly, the topic of "community" is often raised without any clarity as to how that community is defined or whether it has any phenomenological identity—whether the people whom the investigator locates as being part of the community actually believe they are part of it, or whether that even matters.
As in any other interdisciplinary area, studies involving many disciplines present considerable challenges. Social epidemiologists who would like to produce work that spans the biologic and social face the challenges of integrating findings from different fields, using disparate terminology and analytic methods, and understanding the typically voluminous literature in any field. To some extent, closely connected interdisciplinary groups can help to augment each other’s gaps with respect to particular areas of expertise and knowledge. However, there is a considerable danger of thin glosses on areas of expertise not one’s own. For example, one social epidemiologic explanation of socioeconomic inequalities in health places great emphasis on comparisons that persons make between themselves and others, focusing on the links between judgments of relative deprivation that they make in comparing themselves with others, and the psychosocial and neuroendocrine effects of such comparison. Thus, a critical component is this process of social comparisons. As it turns out, social comparison processes have been studied extensively for 50 years by social psychologists and sociologists, and a number of determinants and consequences of social comparisons have been studied (54–56). However, it is rare to see any of this literature in the social epidemiologic literature discussing relative deprivation and health.
As another example, consider the evidence linking dominance and social organization in nonhuman primates (mainly macaques) with coronary heart disease, or that linking dominance and rank with neuroendocrine function, often reflected by variations in cortisol levels. A recent systematic review by Petticrew and Davey Smith (57) suggests that many of the broad generalizations about the former work do not reflect the considerable heterogeneity in findings found in the studies; overall, the studies showed no systematic association between rank and disease endpoints. These authors suggest that considerably more reference is made to these studies in the literature on social determinants of health than is borne out by the inconsistency of the findings. Similarly, while there is also considerable citation of results indicating a consistent relation between rank and cortisol levels in nonhuman primates, a recent review of studies across a number of nonhuman primates indicates little consistency of findings (58). In fact, Sapolsky, a leading researcher in the area, commented that "a prime revisionist emphasis of this chapter has been how little, in fact, rank per se predicts any of those endpoints. Instead, it seems virtually meaningless to think about the physiological correlates of rank outside the context of a number of other modifiers" (59, p. 39; emphasis in original). The purpose of this discussion is not to single out particular investigators—indeed, there may or may not end up being consistent findings in these areas—but rather to highlight some of the potential perils of interdisciplinary work, particularly in areas of research that are fast moving. The lesson is not to shun integration across multiple disciplines, because it is critical, but to suggest that it be done in a way that maximizes the ability of each discipline to contribute information and nuance to the effort.
|
Personalizing populations
|
|---|
Although many of the discussions of the need for a social epidemiologic
approach cite the work of Geoffrey Rose (
60) as the rationale
for a
population-focused strategy, much of the work in social
epidemiology focuses on individuals. Surprisingly, this focus
on individuals is found even when studying factors such as social
networks and social support, where the concepts refer to exchanges
between individual persons and participation in groups, but
the data used in most studies are collected from individuals
(
61). The literature demonstrating an important association
between aspects of social relationships and health is now substantial
(
62), but, for the most part, there is relatively little study
in social epidemiology of the population distribution of these
factors or the determinants of their distribution. In some cases,
such as studies linking health disadvantage to low levels of
social capital, the lack of a population-based focus is particularly
noticeable when social capital—conceived of by most as
an attribute of a social group—is measured by individual
reports (
63). Indeed, some investigators (
64) see the same questions
about distrust and cynical hostility as reflecting individual
attributes; others (
63) consider them indicators of societal
attributes.
While Margaret Thatcher may have said that "as you know, there is no such thing as society: there are individual men and women, and there are families" (65, p. 9), social epidemiologists who discuss the importance of population-level phenomena and determinants of health have an obligation to be articulate about the difference between a population and a Thatcherian aggregate of individual persons. To some extent, this tack has been taken in studies of income inequality and health (66) and of societal influences on human development (67). In many respects, John Cassel’s (68) seminal paper, "The Contribution of the Social Environment to Host Resistance," was a step in that direction. Much more work (perhaps integrating the insights of demography, political science, sociology, and economics) needs to be done by social epidemiologists before Rose’s (69) observation—that the causes of cases within populations may be very different from the causes of differences in incidence between populations—can be fully appreciated. Social epidemiology has much to contribute to this search, but, to do so, it will need to move to a more population-based approach with respect to putative causal factors (70), notwithstanding the perils of ecologic studies (71).
|
Global social epidemiology
|
|---|
With social epidemiology’s emphasis on a broad array
of health determinants, it is surprising that so little work
takes place in the context of global or international health.
By this pursuit I mean not simply conducting studies of disease
X or risk factor
Y in various countries but instead using a
truly comparative approach. To some degree, this approach has
been attempted in research considering the association between
income inequality and life expectancy or mortality (
66) and
in that examining differences in inequalities in health across
countries (
72,
73). In both cases, the work has met with some
obstacles—in the first case because of the sensitivity
of the results to time, period, and outcomes (
74) and in the
second case because the results depended on the measures of
association used in the comparisons (
75–
77). However,
the large variation between countries, and varying trends within
countries, in many of the core determinants of health studied
by social epidemiologists—socioeconomic position, race
and ethnicity, gender, poverty and welfare policy, education,
immigration, and so forth—suggests a rich field for examining
both the impact of social determinants on health and the extent
to which they depend on context. In carrying out such analyses,
it is critical to have specific hypotheses that inform the enterprise.
Rather than simply examining the health effects of a particular
social factor in different countries, it is far more instructive
to examine variations in the levels or distribution of the factor
across countries by using between-country variations as a test
of the relation. At the same time, such studies also enable
examination of the ways in which context affects the relation.
Thus, in countries with strong economic and social supports
for the unemployed, we might expect unemployment to have less
of an effect on health than in countries where such policies
are weak. The growing number of systematic surveys and cohort
studies across the globe mean that, in carrying out such activities,
one is no longer limited to assessment of only ecologic associations.
|
Health versus health care
|
|---|
In many areas of social epidemiology, the dominant approach
to health care has been to ignore or dismiss it. Thus, numerous
articles argue that the contributions of medical care to the
health of the population are, at best, modest (
78–
80).
Reflecting the historical influences of McKeown (
81) and others,
these attempts might be seen as staking out an area of research
that is independent of medical care and as legitimizing the
enterprise. For example, the argument goes, one needs to look
beyond the contributions of medical care to explain gradients
of health by socioeconomic position (
82). Although that statement
is undoubtedly true, it ignores the fact that medical care is
provided by institutions (figure
2) and that decisions as to
who receives medical care, and the quality of that care, are
shaped by social processes. A recent Institute of Medicine report
(
83) on racial and ethnic inequalities in health care and the
consequences of lack of insurance makes this strikingly clear.
Thus, rather than health care being excluded from social epidemiologic
approaches, it needs to be reintroduced as one of the potential
social determinants of health, with the importance of its contribution
varying depending on the specific health issue being considered.
|
Moderating essentialism
|
|---|
Although Humpty Dumpty chooses to make a word mean just what
he wants it to mean (
52), use of certain constructs in social
epidemiology carries with it an almost essentialism, an observation
that has even entered the discussion of the counterfactual approach
to understanding adjustment for confounding in social epidemiology
(
84). Consider, for example, socioeconomic position. Despite
observations that the magnitude, and in some cases the direction,
of the relation between socioeconomic position and health outcomes
is not fixed over time or across places, the graded relation
between the two has become almost reified. In fact, nothing
intrinsic to income, education, or occupation, or other measures
of socioeconomic position, specifies a particular association
with any health outcomes. It all depends on the relation between
these markers of socioeconomic position and the more proximal
risk factors for the particular outcome being considered (
85).
Conceivably, in some circumstances (rare, to be sure), there
could be no relation at all, or the relation could be in the
opposite directions, as for education and breast cancer incidence.
Alternatively, the nature of the association could change over
time, as with the reversal of the socioeconomic gradient in
coronary heart disease as the association of socioeconomic position
with smoking and diet changed (
86).
Furthermore, focusing on the "gradient" per se can lead to a search for a unitary factor that explains the entire range of variation across the range of socioeconomic position. In fact, there may be a variety of factors, each contributing to a greater or lesser extent to health at various socioeconomic levels (87). Discussion of "the gradient" in a reified way has tended to imply that there is a very linear relation between socioeconomic position and a health outcome when, in fact, the relation between household income and risk of death, for instance, is decidedly nonlinear (88, 89). This striking nonlinearity has important implications for both our understanding of the causal pathways involved and the public health significance of the gradient. For example, the very small increased risk associated with increases in household income above the median level makes the observation of differences in disease risk between the highest levels much less compelling and suggests that more attention should be focused on those persons at the lower ranges of socioeconomic position.
This essentialist approach also applies, strangely enough, to many social epidemiologic uses of race and ethnicity in analyses. While most social epidemiologists would acknowledge that race and ethnicity are mainly social constructions, many studies discuss "race" or "ethnicity" effects when what they are really observing are the social, environmental, economic, and historical facts that lie underneath those terms. Indeed, notwithstanding the possible psychosocial impact of discrimination on health, socioeconomic position appears to "explain" many disparities between minority and majority group health outcomes, and there are undoubtedly other critical factors as well (90, 91). For the most part, it may not be possible to separate the multifaceted patterns of historical and contemporary exposures that have led to and underlie the social construction of race and ethnicity from the terms "race" and "ethnicity." Thus, when race effects are examined in epidemiologic studies, we risk viewing race as an essentialist category and ignoring what it stands for. A good example is the controversy surrounding the finding that college-educated African-American and White women have different birth outcomes (92). This finding has often been interpreted by some, although not by the authors, as implying that biologic differences must be responsible for the difference. This conclusion ignores the substantial differences in quality of education, differing contexts for both education and pregnancy, and the wide disparities in economic and social returns for a college education between African-American and White women. Thus, such an interpretation privileges race rather than the conditions for which it is a proxy indicator.
|
LOST OPPORTUNITIES
|
|---|
With social epidemiology’s emphasis on a broad array
of determinants of health, many of which are sensitive to economic
and social context, it would seem that social epidemiologists
would be deeply involved in examining the impact of social and
economic policy on health (
93). For the most part, however,
such pursuits are not well represented in social epidemiology,
possibly because of the difficulties involved in considering
all of the hidden sources of variation that could result in
misleading conclusions. However, one wonders whether this task
is really any more complicated than understanding the causes
of the recent decline in cardiovascular mortality (
94–
96)
or interpreting the results of community intervention on cardiovascular
disease (
97). While those studying the relation between income
inequality and health have been motivated by the large changes
in economic inequality introduced by relatively recent economic
and social changes, evaluation of the effects of social and
economic policy on health has mostly fallen to other disciplines.
For example, epidemiologists have not had a major role in assessing
major trials focused on housing relocation (
98) or poverty reduction
(
99), and little social epidemiologic work has examined the
health effects of welfare reform (
100). Some of this work can
be quite compelling, such as a recent study by economists suggesting
that welfare reform has led to a decrease in breastfeeding 6
months after birth (
101). The lack of social epidemiologic involvement
in many of these studies, particularly in their design and evaluation,
has more often than not resulted in little health data being
collected, even in the case of social and economic changes likely
to have considerable effects on health.
|
WHITHER SOCIAL EPIDEMIOLOGY?
|
|---|
It would be foolish to suggest that social epidemiology should
adopt a particular set of tools or perspectives; as in any scientific
discipline, invention and discovery are best served by using
a diversity of approaches. However, the areas highlighted in
this review are those that present considerable opportunities
for clarification and progress. The explosion of interest in
social determinants of health and multilevel approaches to health
determinants has opened up a broad arena for research. A few
years ago, a report from the Hastings Center called for an expansion
of research on social determinants of health equal to that devoted
to the human genome effort (
102). In all likelihood that will
not happen; however, this is still an era of great opportunity
for social epidemiology. Before its promise can be fulfilled,
it will be necessary to wrestle with difficult conceptual, methodological,
and analytical problems, some of which have been discussed in
this review. Social epidemiology is poised to contribute to
a new paradigm that bridges the social and the biological, the
ultimate mark of success being incorporation of this expanded
view of social determinants of health into a science of epidemiology
that sees all that it does as embedded in the social world.
|
ACKNOWLEDGMENTS
|
|---|
Partial support for preparing this article was provided by
awards from the National Institute on Aging (R37 AG11375), the
National Institute of Child Health and Human Development (P50
HD38986-05), and the Robert Wood Johnson Foundation Health and
Society Scholars Program.
|
REFERENCES
|
|---|
- Rothman KJ, Adami HO, Trichopoulos D. Should the mission of epidemiology include the eradication of poverty? Lancet 1998;352:810–13.[CrossRef][ISI][Medline]
- Zielhuis GA, Kiemeney LALM. Social epidemiology? No way. Int J Epidemiol 2001;30:43–4.[Free Full Text]
- Gori GB. Epidemiology and public health: is a new paradigm needed or a new ethic? J Clin Epidemiol 1998;51:637–41.[CrossRef][ISI][Medline]
- Rose G. The strategy of preventive medicine. Oxford, United Kingdom: Oxford University Press, 1992.
- Kaplan GA, Lynch JW. Editorial: whither studies on the socioeconomic foundations of population health? Am J Public Health 1997;87:1409–11.[Free Full Text]
- Sampson RJ, Morenoff JD, Gannon-Rowley T. Assessing "neighborhood effects": social processes and new directions in research. Annu Rev Sociol 2002;28:443–78.[CrossRef][ISI]
- Kaplan GA, Everson SA, Lynch JW. The contribution of social and behavioral research to an understanding of the distribution of disease: a multilevel approach. In: Smedley BD, Syme SL, eds. Promoting health: intervention strategies from social and behavioral research. Washington, DC: National Academy Press, 2000:31–55.
- Marmot M. Multilevel approaches to understanding social determinants. In: Berkman LF, Kawachi I, eds. Social epidemiology. New York, NY: Oxford University Press, 2000:349–67.
- House JS. Understanding social factors and inequalities in health: 20th century progress and 21st century prospects. J Health Soc Behav 2002;43:125–42.[CrossRef][ISI][Medline]
- Lynch JW. Social epidemiology: some observations on the past, present and future. Australas Epidemiologist 2000;7:7–15.
- Kaplan GA, Lynch JW. Socioeconomic considerations in the primordial prevention of cardiovascular disease. Prev Med 1999;29(6 pt 2):S30–S35.
- Prentice RL. Design issues in cohort studies. Stat Methods Med Res 1995;4:273–92.[Abstract/Free Full Text]
- McDonough P, Duncan GJ, Williams D, et al. Income dynamics and adult mortality in the U.S., 1972 through 1989. Am J Public Health 1997;87:1476–83.[Abstract/Free Full Text]
- Kaplan GA. Psychosocial aspects of chronic illness: direct and indirect associations with ischemic heart disease mortality. In: Kaplan RM, Criqui MH, eds. Behavioral epidemiology and disease prevention. New York, NY: Plenum Publishing Corporation, 1985:237–69.
- Kaufman JS, Cooper RS. Commentary: considerations for use of racial/ethnic classification in etiologic research. Am J Epidemiol 2001;154:291–8.[Abstract/Free Full Text]
- Raudenbush SW, Bryk AS. Hierarchical linear models. 2nd ed. Thousand Oaks, CA: Sage Publications, 2002:482.
- Diez Roux AV. Bringing context back into epidemiology: variables and fallacies in multilevel analysis. Am J Public Health 1998;88:216–22.[Abstract/Free Full Text]
- Greenland S, Pearl J, Robins JM. Causal diagrams for epidemiologic research. Epidemiology 1999;10:37–48.[CrossRef][ISI][Medline]
- Wolfson MC. POHEM—a framework for understanding and modelling the health of human populations. World Health Stat Q 1994;47:157–76.[Medline]
- Zohoori N, Savitz DA. Econometric approaches to epidemiologic data: relating endogeneity and unobserved heterogeneity to confounding. Ann Epidemiol 1997;7:251–7.[CrossRef][ISI][Medline]
- Adams P, Hurd MD, McFadden D, et al. Healthy, wealthy, and wise? Tests for direct causal paths between health and socioeconomic status. J Econometrics 2003;112:3–56.[CrossRef]
- Paneth N. A conversation with Mervyn Susser. Epidemiology 2003;14:748–52.[CrossRef][ISI][Medline]
- Sandler DP. A conversation with George W. Comstock. Epidemiology 2003;14:623–7.[CrossRef][ISI][Medline]
- Krieger N, Zierler S. What explains the public’s health? A call for epidemiologic theory. Epidemiology 1996;7:107–9.[ISI][Medline]
- Hauser RM. "Contextual analysis revisited." Sociolog Methods Res 1974;2:365–75.
- Diez Roux AV. Investigating area and neighborhood effects on health. Am J Public Health 2001;91:1783–9.[Abstract/Free Full Text]
- Kaplan GA, Lynch JW. Editorial: whither studies on the socioeconomic foundations of population health? Am J Public Health 1997;87:1409–11.[Free Full Text]
- Macintyre S, Ellaway A, Cummins S. Place effects on health: how can we conceptualise, operationalise and measure them? Soc Sci Med 2002;55:125–39.[CrossRef][ISI][Medline]
- Diez-Roux AV, Northridge ME, Morabia A, et al. Prevalence and social correlates of cardiovascular disease risk factors in Harlem. Am J Public Health 1999;89:302–6.[Abstract/Free Full Text]
- Brooks-Gunn J, Duncan GJ, Klebanov PK, et al. Do neighborhoods influence child and adolescent development? Am J Sociol 1993;99:353–95.[CrossRef]
- Haan M, Kaplan GA, Camacho T. Poverty and health: prospective evidence from the Alameda County Study. Am J Epidemiol 1987;125:989–98.[Abstract/Free Full Text]
- Waitzman NJ, Smith KR. Phantom of the area: poverty-area residence and mortality in the United States. Am J Public Health 1998;88:973–6.[Abstract/Free Full Text]
- Yen IH, Kaplan GA. Poverty area residence and changes in physical activity level: evidence from the Alameda County Study. Am J Public Health 1998;88:1709–12.[Abstract/Free Full Text]
- Yen IH, Kaplan GA. Poverty area residence and changes in depression and perceived health status: evidence from the Alameda County Study. Int J Epidemiol 1999;28:90–4.[Abstract/Free Full Text]
- Diez Roux AV, Stein Merkin S, Arnett D, et al. Neighborhood of residence and incidence of coronary heart disease. N Engl J Med 2001;345:99–106.[Abstract/Free Full Text]
- Lynch JW, Kaplan GA, Pamuk ER, et al. Income inequality and mortality in metropolitan areas of the United States. Am J Public Health 1998;88:1074–80.[Abstract/Free Full Text]
- Kaplan GA, Pamuk ER, Lynch JW, et al. Inequality in income and mortality in the United States: analysis of mortality and potential pathways. BMJ 1996;312:999–1003.[Abstract/Free Full Text]
- Kennedy BP, Kawachi I, Prothrow-Stith D. Income distribution and mortality: test of the Robin Hood Index in the United States. BMJ 1996;312:1004–8.[Abstract/Free Full Text]
- Sampson RJ, Morenoff J, Earls F. Beyond social capital: spatial dynamics of collective efficacy for children. Am Sociolog Rev 1999;64:633–60.[CrossRef]
- Susser M. Does risk factor epidemiology put epidemiology at risk? Peering into the future. J Epidemiol Community Health 1998;52:608–11.[Abstract]
- Marmot MG, Shipley MJ, Rose G. Inequalities in death—specific explanations of a general pattern? Lancet 1984;1:1003–6.[CrossRef][ISI][Medline]
- Lynch JW, Kaplan GA, Cohen RD, et al. Do cardiovascular risk factors explain the relation between socioeconomic status, risk of all-cause mortality, cardiovascular mortality, and acute myocardial infarction? Am J Epidemiol 1996;144:934–42.[Abstract/Free Full Text]
- Everson SA, Kauhanen J, Kaplan GA, et al. Hostility and increased risk of mortality and acute myocardial infarction: the mediating role of behavioral risk factors. Am J Epidemiol 1997;146:142–52.[Abstract/Free Full Text]
- Davis MC, Matthews KA, McGrath CE. Hostile attitudes predict elevated vascular resistance during interpersonal stress in men and women. Psychosom Med 2000;62:7–25.[Abstract/Free Full Text]
- Susser M. Causal thinking in the health sciences: concepts and strategies of epidemiology. New York, NY: Oxford Press, 1973.
- Kaufman JS, Poole C. Looking back on "causal thinking in the health sciences." Annu Rev Public Health 2000;21:101–19.[CrossRef][ISI][Medline]
- Greenland S, Brumback B. An overview of relations among causal modelling methods. Int J Epidemiol 2002;31:1030–7.[Abstract/Free Full Text]
- Schwartz S, Diez-Roux A. Commentary: causes of incidence and causes of cases—a Durkheimian perspective on Rose. Int J Epidemiol 2001;30:435–9.[Free Full Text]
- Institute of Medicine. Health and behavior: the interplay of biological, behavioral, and societal influences. Committee on Health and Behavior: Research, Practice and Policy, Board on Neuroscience and Behavioral Health. Washington, DC: National Academies Press, 2001.
- Institute of Medicine. Bridging disciplines in the brain, behavioral, and clinical sciences. Committee on Building Bridges in the Brain, Behavioral, and Clinical Sciences. Division of Neuroscience and Behavioral Health. Pellmar TC, Eisenberg L, eds. Washington, DC: National Academies Press, 2000.
- Institute of Medicine. Promoting health: intervention strategies from social and behavioral research. Committee on Capitalizing on Social Science and Behavioral Research to Improve the Public’s Health. Division of Health Promotion and Disease Prevention. Smedley BD, Syme SL, eds. Washington, DC: National Academies Press, 2000.
- National Research Council. New horizons in health: an integrative approach. Committee on Future Directions for Behavioral and Social Sciences Research at the National Institutes of Health. Commission on Behavioral and Social Sciences and Education. Singer BH, Ryff CD, eds. Washington, DC: National Academies Press, 2001.
- Carroll L. Alice’s adventures in Wonderland and through the looking glass. New York, NY: Macmillan Publishing Co, Inc, 1962.
- Festinger L. A theory of social comparison processes. Hum Relat 1954;7:117–40.
- Mussweiler T. Status and solidarity in social comparison: agentic and communal values and vertical and horizontal directions. Psychol Rev 2003;110: 472–89.[CrossRef][ISI][Medline]
- Locke KD. Status and solidarity in social comparison: agentic and communal values and vertical and horizontal directions. J Pers Soc Psychol 2003;84:619–31.[CrossRef][ISI][Medline]
- Petticrew M, Davey Smith G. Society for Social Medicine abstracts. J Epidemiol Community Health 2003;57(suppl I):A1–A21.[CrossRef]
- Abbott DH, Keverne EB, Bercovitch FB, et al. Are subordinates always stressed? A comparative analysis of rank differences in cortisol levels among primates. Horm Behav 2003;43:67–82.[CrossRef][Medline]
- Sapolsky RM. Hormonal correlates of personality and social contexts: from non-human to human primates. In: Panter-Brick C, Worthman CM, eds. Hormones, health and behavior. New York, NY: Cambridge University Press, 1999:18–46.
- Rose GA. The strategy of preventive medicine. Oxford, United Kingdom: Oxford University Press, 1992.
- Berkman LF, Glass TA. Social integration, social networks, social support, and health. In Berkman LF, Kawachi, I, eds. Social epidemiology. New York, NY: Oxford University Press, 2000.
- House JS, Landis KR, Umberson D. Social relationships and health. Science 1988;241:540–5.[Abstract/Free Full Text]
- Kawachi I, Kennedy BP, Lochner K, et al. Social capital, income inequality, and mortality. Am J Public Health 1997;87:1491–8.[Abstract/Free
TOP
INTRODUCTION
METAPHORS AND MODELS