| HOME | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|---|
| ||||||||||||||||||||||
CONCLUSIONS |
Department of Sociology and Center for the Study of Aging and Human Development, Duke University, Durham, North Carolina.
Address correspondence to Linda K. George, Department of Sociology and Center for the Study of Aging and Human Development, Duke University, Box 90088. Durham, NC 27708. E-mail: lkg{at}geri.duke.edu
IN this article, I offer some comments and suggestions that build upon the articles presented in this Special Issue. I will not comment on the individual articles. Rather, I will share some of the ideas that these articles generated as I thought about them.
Without question, the strongest "message" that the articles share and substantiate is the critical importance of studying the relationships between SES and health across the life course. An impressive volume of recent research, including that presented here, documents strong links between early socioeconomic status (SES) and health outcomes observed later in life (Blackwell, Hayward, & Crimmins, 2001
; Hayward & Gorman, 2004; Lynch, Smith, Kaplan, & Shema, 1997; Osler, Andersen, Due, Lund, Damsgaard, & Holstein, 2003). I'm a long-time advocate of the knowledge that can be gained by life course theory and research (George, 1996). Nonetheless, I am genuinely impressed by the extent to which SES during childhood creates a persisting health legacy.
Of course, a life course view of SES and health extends beyond the effects of childhood SES on later health. The entire landscape of adulthood is fertile ground. There is evidence on many fronts—ranging from transitions in and out of poverty to voluntary and involuntary midlife job change to economic shocks caused by widespread mergers and downsizing—that SES often changes multiple times during adulthood. As yet, we know very little about the effects of adults' changing fortunes on health. Obviously, we are only in the beginning stages of fully exploiting the relationships between SES and health across the life course, but the pioneering research is now in place.
CONTINUING CONUNDRUMS IN RESEARCH ON SES AND HEALTH DISPARITIES
Social Selection
An issue that probably received too little attention in the articles is social selection. The obvious form of social selection of concern when studying SES and health is the probable reciprocal relationships between the two—an issue that has received attention for many years (e.g., Fox, 1990; Hollingshead & Redlich, 1958). A substantial amount of recent research examined this issue using sophisticated statistical techniques and longitudinal data. Results consistently indicate that the dominant causal direction is from SES to health (Chandola, Bartley, Sacker, Jenkinson, & Marmot, 2003; Elstad & Krokstad, 2003; Mulatu & Schooler, 2002). And, of course, the earlier in the life course that the dynamics between SES and health are measured and monitored, the more we learn about their reciprocal effects. I have argued elsewhere that a strength of the life course perspective is that it renders the distinction between social causation and social selection essentially irrelevant (George, 1999). If we can obtain multiple observations over long periods of time, especially if those observations begin during early childhood, we can determine the dimensions of life experience that set in motion long-term patterns of vulnerability and resilience.
With regard to the articles presented at this conference, however, I am especially worried about another form of social selection: selective mortality. As clearly explicated in the article by House, as well as in his earlier work (House, Kessler, Herzog, Mero, & Breslow, 1990; House, Lepowski, Mero, Kessler, & Herzog, 1994), it is not until midlife that sufficient numbers of individuals exhibit sufficient chronic illness and/or disability to observe significant SES-related health disparities. Midlife also is the time in the life course that rates of "premature mortality" begin to increase quite rapidly (Cohen, Farley, & Mason, 2003; Horiuchi, Finch, Mesle, & Vallin, 2003; Steptoe & Marmot, 2003). SES is significantly related to survival at all ages, from infancy to late life. Because the least physically robust individuals die earliest and because they are disproportionately of low SES, research that focuses on the relationships between SES and illness or disability inevitably underestimates the total effects of SES on health. This is a complex problem that defies easy solutions.
A recent and creative approach to the problem of selective mortality was introduced by Beckett (2000). The strategy is clever and quite simple: leave deceased respondents in the analysis when predicting illness or disability outcomes. Operationally, deceased respondents are assigned a value on the health outcome that is above or below the last valid value, depending on the metric of the variable. For example, if the investigator is using a 4-point scale to measure self-rated health, with "1" representing self-reported poor health and "4" representing self-reported excellent health, deceased respondents would be assigned a value of "0." Similarly, on a disability scale that ranges from 0–7, with "0" representing no functional limitations and "7" representing disability on all the items in the scale, deceased respondents would be assigned a value of "8." This is a clever idea that is easily implemented. Unfortunately, it rests on a set of assumptions that cannot be justified. There is a qualitative difference between living and dead respondents that simply cannot be bridged by this technique. At the simplest level, for example, one can hardly assume that the "distance" between three and four functional limitations is the same as that between self-related limitations in all seven areas and being deceased. Although it was not a major focus of the article, O'Rand and Hamil-Luker report performing such an analysis.
I believe that there are two better ways of dealing with selective mortality than using this creative, but noncompelling way of coding variables. One option is to perform a competing risk model (e.g., Alberti, Metivier, Landais, Thervet, Legendre, & Chevret, 2003; Yan, Hoover, Moore, & Xiong, 2003)—a strategy more familiar to epidemiologists than to social and behavioral scientists. In situations in which the investigator wishes to use a set of independent variables, including SES and the baseline level of the dependent variable, to predict self-rated health at a later point time, multinomial logistic regression could be used to predict five outcomes: each of the four levels of self-rated health and death. The set of equations generated by multinomial logistic regression will permit the investigator to determine the degree to which the independent variables predict each of the five outcomes, relative to each other.
The second approach for handling this issue takes advantage of life course principles and generates findings that I believe to be more substantively meaningful than those generated by a competing risk model. Much of the confusion about the effects of selective mortality results from failing to distinguish between the timing of illness onset and the duration of survival after illness onset. Despite volumes of research addressing the links between SES and health, we still do not know whether SES operates primarily as a determinant of age of onset of illness (with lower SES predicting earlier onset), primarily as a determinant of length of survival after illness onset (with lower SES predicting shorter survival), or is important for both. By disaggregating timing and duration, selective mortality is a critical element of the research question rather than a methodological nuisance that must be "handled."
Cohort Effects
House's findings documented substantial cohort differences in the relationships between SES indicators and health outcomes. Other articles' findings also suggest strong cohort differences, as do a number of previously published studies (e.g., Lynch, Brown, & Harmsen, 2003; Reynolds, Crimmins, & Saito, 1998). I find this evidence interesting, but also perplexing. What are the mechanisms generating these cohort effects? It is surprising to me that these patterns are frequently observed, yet there has been little effort to either conceptualize or empirically identify the causes of those patterns. On the one hand, Link's article strongly supports the notion that SES is a "fundamental cause" of illness (see also Link & Phelan, 1995)—that is, the relationship between SES and health is robust across health outcomes and, very importantly, across time, despite dramatic changes in public health, medical technology, and improved access to health information. On the other hand, research by House and others documents differences in the relationship between SES and health across 5-year birth cohorts. How are we to reconcile these apparently divergent findings? I do not want to overstate the discrepancy across studies—after all, SES is significantly related to health across cohorts, despite cohort differences in the strength of those relationships. Nonetheless, more effort is needed to understand cohort differences in the links between SES and health. As a life course scholar, I am especially interested in the extent to which cohort differences can be explained by relatively recent or contemporaneous factors versus the degree to which they reflect the accumulated effects of long-term social structures and processes.
Other Forms of Stratification
As important as the links between SES and health are, we need to keep in mind that SES is only one form of stratification. Gender and race/ethnicity also are major forms of stratification in U.S. society. The articles by Kahn and Fazio, Williams, and Moen and Chermack provide strong evidence that, although gender and race/ethnicity are significantly related to SES, their associations with health extend beyond that in complex ways. Moreover, because gender and race/ethnicity are ascribed statuses, they are partial determinants of SES, which raises the question of whether they, even more than SES, are fundamental causes of illness.
A Caution About Specific Illnesses and Cause-Specific Mortality
I also want to offer a word of caution about research that focuses on specific illnesses and/or cause-specific mortality. Certainly there are research questions for which the appropriate focus is a specific illness or a specific cause of death. Overall, however, I believe that illness-specific and cause-specific studies can be quite misleading for the purpose of understanding the effects of SES on health. I offer two reasons for this conclusion. First, as Link and others using the "fundamental cause" framework have demonstrated, part of the power and mystery of SES is its effects across a broad range of health outcomes—especially the fact that, over time, democratic illnesses typically become the burden of the socioeconomically deprived. For example, it took less than 20 years for HIV/AIDS to move from a democratic illness to the province of the poor, the ignorant, and the addicted. My second reason rests on the insight of the late rock musician Jim Morrison, who accurately observed that "nobody gets out of here alive." We need to keep in mind that death is a zero-sum game. If deaths from cardiovascular disease decline over time, deaths from other causes must increase. An informed understanding of the links between SES and health must recognize this fact. Ultimately, the crux of the question about the effects of SES on health rests on length of life and/or length of disability-free life.
THE POTENTIAL ROLE OF BIOLOGICAL DATA
I was pleased to see part of this conference devoted to the potential uses of biological data in understanding the links between social status and health. The article by Rieker and Bird makes a strong case that, as important as social and behavioral factors are, a comprehensive understanding of sex and/or gender differences in health must also pay attention to biological factors. Horwitz focused on the limitations of the genetic lens as an explanation for health outcomes when it is used without attention to social and behavioral factors.
Although a biologist would undoubtedly find my view oversimplified, I perceive two major streams of biological data that may be important for understanding the effects of SES on health: genetic data and biomarker data. Although both types of data may offer insights into the links between SES and health, I find biomarker data to be of greater potential, at least in the short run.
Without question, biologists view human genetics as the next frontier in understanding human health and behavior. In fact, however, human genetics is in the very early stages of understanding the human genome. It appears that the major questions absorbing the energies of geneticists now focus on (a) the fact that multiple, rather than single, genes are the basis of biological and behavioral heterogeneity; (b) parceling out the magnitude of the relative contributions of genes, environment, and their interaction; and (c) understanding the processes of gene expression, including the ability to intervene in gene expression or suppression (Carey, 2003). Simply put, the field of human genetics is not ready for complex models that extend beyond the boundaries of DNA. The emerging field of behavioral genetics aspires to complex models of this sort, but at this point, the primary consideration of nongenetic phenomena is ascertaining the size of the amorphous "black box" that they label environment. It will be a long time, I think, before we can incorporate genetic profiles into our models of SES and health.
Biomarker and other preclinical data, in contrast, offer important opportunities for exploring the mechanisms that underlie the relationships between SES and health. Biomarker data are rather easily collected and analyzed via saliva, urine, and blood specimens—a feasible procedure in even large-scale epidemiologic surveys. There seems to be considerable reluctance, however, among social scientists (less so among psychologists) about integrating biological and social/behavioral data. This reluctance appears to reflect concerns about biological determinism and reductionism. I view this issue differently; to use the language of experimental research, I view social and behavioral factors as stimuli and biological markers as response. We can use biological data to demonstrate the power of social factors.
Two strategies for using biomarker data seem especially relevant to understanding the effects of SES on health. First, immune function seems to be an especially promising venue for tracing the pathways between social deprivation and health. Rieker and Bird hypothesize that immune function plays a role in explaining sex and/or gender differences in health. I hypothesize that immune function plays a role in mediating the links between SES and health as well. Evidence to date clearly documents that stressful life experiences, such as providing care to a severely ill relative, compromise immune function, which in turn predicts physical illness and emotional distress (Glaser, Sheridan, Malarkey, MacCallum, & Kiecolt-Glaser, 2000; Hadjiconstantinou, McGuire, Duchemin, Laskowski, Kiecolt-Glaser, & Glaser, 2001). More limited evidence suggests that social integration, in the form of religious participation, is associated with better immune function (Koenig, Cohen, George, Hays, Larson, & Blazer, 1997). It is important to note that research to date has linked chronic, persistent social conditions to immune function. SES, although clearly changeable across the life course, tends to remain relatively stable over significant periods of time, suggesting that it also is likely to affect immune status.
Second, unless we incorporate biomarker and preclinical data into our research, we are unlikely to observe the origins of the links between SES and health. As House's article documents, we do not observe SES differences in disease or death until midlife. To date, we have little understanding of why and how SES differences in health, which are minimal in young adulthood, become highly differentiated by midlife. Biomarker and preclinical data offer an important opportunity to trace this process. It is likely that biological parameters that precede full-blown disease are strongly associated with SES during childhood and young adulthood. For example, both immune function (a biomarker) and lung function (a preclinical factor) may be socioeconomically stratified long before their effects can be observed in the form of chronic disease or mortality. Obviously, many of the effects of SES on health result from stressors, environmental factors, and other conditions experienced during adulthood. But it is likely that deprivation early in life leaves a biological trail of vulnerability as well.
POLICY IMPLICATIONS—AND CAUTIONS
My final comments address two somewhat broader issues—issues concerned with the larger agendas that seem to underlie much of the research on health disparities. One does not have to scratch the surface of research on SES and health very deeply to see that, in addition to a scientific understanding of the origins of health disparities, investigators in this field tend to have a strong policy agenda as well. There is a strong conviction that research evidence should be used by policy makers to eliminate or substantially reduce the health effects of socioeconomic stratification. Given that health policy is typically based on factors other than research evidence, I share the conviction that informing policy makers about the origins of health disparities is an appropriate goal. But there are dangers here as well, and I cringe when research is used for moral crusades, no matter how well intentioned the motivation. Two aspects of the current focus on health disparities especially concern me.
First, studies of the mediators between SES and health have lost their luster for me. Although I understand the motivation to assess the extent to which factors such as health behaviors and stressful experiences mediate the effects of SES on health, it seems to me that these efforts ultimately distract attention from the pivotal issue. One aspect of this strategy is that it makes it very easy to "blame the victim." The factors typically viewed as mediators of the effects of SES on health invite the conclusion that, if persons with low education and income would simply "clean up their acts"—i.e., learn to avoid and/or effectively cope with stress, avoid hazardous substances, eat right, and exercise—health disparities would be substantially reduced. This conclusion conveniently ignores the facts that (a) in most research, the effects of mediators, although statistically significant, are typically quite modest (e.g., Antonucci, Ajrouch, & Janevic, 2000; Ettner & Grzywacz, 2003; Gallo & Matthews, 2003; van Doorslaer & Gerdtham, 2003), and (b) among the constraints of SES is the inability to control one's life to avoid hazards.
Research on the relationships between SES and health convincingly substantiates Link and Phelan's thesis that SES is a fundamental cause of disease (Link & Phelan, 1995). Even if all the social/behavioral mediators of the SES–health relationship identified to date were conquered, socioeconomic status would continue to predict the stratification of morbidity and mortality. Using data from well-educated civil servants, the Whitehall Studies documented that point a quarter of a century ago (Marmot, Smith, Stansfeld, Patel, North, Head, White, Brunner, & Feeney, 1991; Rose & Marmot, 1985; Smith, Shipley, & Rose, 1990). The simple fact is that the policy implications of SES–health research are not that interventions to launch behavior change among the socioeconomically deprived are needed, but rather that SES itself should be the target of public policy. As I read the research on SES and health, the public/health policy implications are simple: redistribute income and educate children to their maximum potential. The Whitehall Studies also warn us that as long as there is socioeconomic inequality, there also will be health disparities. But reductions in SES inequality, especially setting a reasonable income "floor," is the most promising route to reductions in health disparities. Admittedly, a unified assault on SES inequality is at odds with the political climate that dictates that we "fix" people rather than the structure of our social institutions. Nonetheless, a unified assault on SES inequality is the policy stance that is most compatible with what we know about SES and health.
Finally, I want to challenge an assumption held by many scholars in medical sociology, public health, and health psychology: the assumption that behaviors and choices contrary to health promotion are irrational. The way that this assumption plays out is nicely illustrated in the article by Rieker and Bird. On the one hand, these authors recognize some of the limitations of rational choice theory, noting that socioeconomic and other constraints often make it difficult or impossible for individuals to implement health-promoting behaviors. On the other hand, they also express dismay with the fact that many high-SES individuals, who presumably do not face social and economic constraints, fail to implement health-promoting behaviors. I am in complete agreement that social deprivation constrains the ability to implement health-promoting behaviors. I differ from them, however, in their dismay that individuals who apparently have the resources needed to implement health-promoting behavior fail to do so.
One of the problems with rational choice theory in general and with the perspective taken by Rieker and Bird in particular is that I believe almost all human behavior to be rational. In addition to whatever constraints of choice result from social structure and social processes, all of us operate in what I call "multiple currencies." Money is obviously a currency, but it is far from the only one that we value and use. I would argue that health is a currency, social relationships are a currency, and factors that feed our sense of self and self-esteem are currencies. Individuals juggle all of these currencies—stockpiling some, squandering others, accumulating and spending them in the ways that we believe will best serve us.
What observers view as nonrational choices or behaviors are usually those that give priority to a currency other than the one that the observers value most. A large volume of research now demonstrates that, if optimum physical and mental health is one's goal, providing long-term care to a severely ill family member is not a rational choice (e.g., Beach, Schulz, Yee, & Jackson, 2000; Burton, Zdaniuk, Schulz, Jackson, & Hirsch, 2003; Gaugler, Davey, Pearlin, & Zarit, 2000). But caregivers apparently value their relationships with loved ones above the value of optimal health. Is the man who, upon retirement, spends his nest egg on a sports car rather than long-term care insurance irrational? From the perspective of taking responsibility for his health care, probably so—but it's a good bet that the sports car is performing wonders for his self-esteem.
I am deeply disturbed by the tendency for those of us who study health to take on the role of the "health police," assuming that choices and behaviors that are contrary to maximizing health over the long-term are irrational and deserving of censure. It's been years now since I've been out to dinner and not had at least one person present comment on calories, cholesterol, or some other health-related aspect of the menu. Indeed, the justification offered for having wine with dinner is frequently its cardiovascular benefits. Obviously, I use a different currency in these situations—I go out to dinner for good food and good company, not for a lecture on health.
Beyond my personal distaste for a life lived with unremitting attention to the health effects of my choices, health researchers have no business taking on the role of "health police." I subscribe to the perspective that it is our role to produce valid knowledge and to make that knowledge available to both policy makers and the general public in an understandable form. If we choose to launch a moral crusade based on those findings, that is our right—but that is separate from our roles as researchers. Beyond that, it makes no sense to me to view health as the be all and end all of human existence. If we really want to understand the complexities of human existence, we must recognize that there are currencies beyond health and money—and I hope to respect individuals' rights to juggle their multiple currencies as constraints require and autonomy permits.
References
| ||||||||||||||||||||||
| HOME | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|---|
