Epidemiology, and Public Health

Traditional Epidemiology, Modern Epidemiology, and Public Health

Neil Pearce, PhD Introduction modern epidemiology; therefore, they have many features in common. Neverthe- In this article I consider some of the less, there are some important differences key features ofwhat has become known as between the traditional and the modern modern epidemiology, and I contrast approaches, particularly the loss of the these with more traditional approaches. I population perspective in recent decades. do not intend to present a detailed I therefore will discuss some of the historical review,1-3 nor to present a reasons the population perspective has comprehensive review of current ap- been lost and the implications of this proaches. Rather, I will focus on some of paradigm shift. Then, I will discuss some the key changes in epidemiology over the of the key issues in developing new forms past few decades, and I will consider the of epidemiology that restore the popula- concepts of causality involved and their tion perspective while making use of ideological and practical consequences. I recent methodological advances. (I am will argue that the rise ofmodem epidemi- tempted to use the term "postmodern ology has been a mixed blessing and that epidemiology" to provide a contrast with the new paradigm has major shortcom- modern epidemiology and because some ings, both in public health and in scientific postmodernist concepts are relevant to terms.4 The recent changes in epidemio- my arguments; however, the use of this logic methodology have not been neutral, term could imply an uncritical advocacy of but rather (in combination with other postmodernism, which has its own episte- influences) they have changed-and have mological and practical shortcomings.) reflected changes in-the way in which epidemiologists think about health and Traditional and Modern disease.5 The key issue has been the shift in the level of analysis from the popula- Epidemiology tion to the individual (S. Wing, Concepts In the first week of their training, in modem epidemiology: population, risk, most epidemiologists usually leam a little dose-response, and confounding, unpub- about the history of public health. In lished manuscript). This is typified by the anglophone countries they learn about current lack of interest in population the work of Chadwick, Engels, Snow, and factors as causes of disease, the lack of others, who exposed the appalling social interest in the history of epidemiology, conditions during the industrial revolu- and the lack of integration with other tion, and the work of Farr and others, who public health activities. I will give particu- revealed major socioeconomic differences lar emphasis to the current neglect of in disease in the 19th century. At that social, economic, cultural, historical, politi- time, epidemiology was a branch of public cal, and other population factors, and I will refer to these using the general term of "socioeconomic factors." The author is with the Wellington School of Of course, traditional epidemiology Medicine, Wellington, New Zealand. was not a monolith. A wide variety of Requests for reprints should be sent to were and there is a Neil Pearce, PhD, Wellington School of Medi- approaches used, cine, PO Box 7343, Wellington, New Zealand. danger of setting up caricatures of ideal This article was accepted June 22, 1995. types. It should also be emphasized that Editor's Note. See related editorial by traditional epidemiology gave rise to Winkelstein (p 621) in this issue.

May 1996, Vol. 86, No.5 Epidemiology and Public Health health and focused on the causes and prevention of disease in populations, in TABLE 1-Epidemiological Paradigms contrast to the clinical sciences, which were branches of medicine and focused Traditional Epidemiology Modern Epidemiology on disease pathology and treatment of individuals (S. Wing, Concepts in modem Motivation Public health Science epidemiology: population, risk, dose- Level of study Population Individual/organ/tissue/ response, and confounding, unpublished cell/molecule manuscript). Thus, the emphasis was on Context of study Historical/cultural Context free and the health the prevention of disease Paradigms Demography/social science Clinical trial needs of the population as a whole. The dramatic decline in infectious diseases Epistemological approach Realist Positivist that has occurred since the mid-19th Epistemological strategy Top down (structural) Bottom up (reductionist) century has been attributed to improve- Level of intervention Population (upstream) Individual (downstream) ments in nutrition, sanitation, and general living conditions,6 although it has been argued that specific public health interventions regarding factors such as urban ing of the population patterns of disease sinking. Some of the reasons for the congestion actually played the major occurrence.15 current lack of interest in the population role.7 These methodological developments perspective may lie in the personal and There are still major socioeconomic have been paralleled by, and have re- professional situations of epidemiologists. differences in health, and the relative flected, a shift in the level of analysis from In most countries the main sources of differences are continuing to increase.8'9 the population to the individual. Most funding are government or voluntary Nevertheless, modern epidemiologists modem epidemiologists still do studies in agencies that have little interest in, or rarely consider socioeconomic factors and populations, but they do so in order to sympathy for, studies of socioeconomic the population perspective, except per- study decontextualized individual risk fac- factors and health. In the last decade, haps to occasionally adjust for social class tors, rather than to study population Western countries, particularly anglo- in analyses of the health effects of tobacco factors in their social and historical con- phone countries, have increasingly placed smoke, diet, and other lifestyle factors in text. McKinlayl6 argued that what is now emphasis on individual responsibility, typi- individuals. For example, studies in most regarded as established epidemiology is fied by the famous statement by Margaret industrialized countries have repeatedly characterized bybiophysiologic reduction- Thatcher20 that "there is no such thing as found strong associations between social ism, absorption by biomedicine, a lack of society, there are only families and indi- class and cancer,10 yet social class did not real theory about disease causation, di- viduals." Governments and funding agen- feature, except for a brief mention as a chotomous thinking about disease (every- cies have been most supportive of studies confounder, in the most comprehensive one is either healthy or sick), a maze of that focus on individual lifestyle, and review of the causes of cancer in the risk factors, confusion of observational epidemiologists, either through choice or United States,11 and one leading epidemi- associations with causality, dogmatism through necessity, have tended to go ology text states that "social class is about which study designs are acceptable, "where the money is." presumably related causally to few if any and excessive repetition of studies. He A related issue is that socioeconomic diseases.'2 argued that this approach diverts limited factors are "not easily modifiable" and are Traditional epidemiology has be- resources, blames the victim, produces a "too political." However, the decision not come unfashionable and is treated some- lifestyle approach to social policy, decon- to study socioeconomic factors is itself a what disparagingly in modem epidemiol- textualizes risk behaviors, seldom assesses political decision to focus on what is ogy texts, which have rewritten the history the relative contribution of nonmodifiable politically acceptable rather than what is of epidemiology in their own image. In genetic factors and modifiable social and most important in scientific and public particular, there has been a strong focus behavioral factors, and produces interven- health terms.21 Governments have repeat- on statistical issues and paradigms'3 and tions that can be harmful. These trends edly shown that social and economic an ignorance of the other modes of are particularly noticeable in the recent differences are not God-given but are thought that were integral to the work of rise of molecular epidemiology,'7"18 espe- directly affected by government policies, 19th-century pioneers such as Snow.'4 cially in the renewed emphasis on issues albeit in unexpected ways.22-24 For ex- One is left with the impression that of individual susceptibility. ample, there is some preliminary evidence 19th-century epidemiologists used ad hoc that inequitable distribution of the gross methods that have now been placed on a The Decline ofPopulation national product can have a more signifi- sounder foundation through recent devel- Epidemiology cant impact on overall national mortality opments in methods of study design (e.g., rates than the actual level of the gross the theory of case-control studies), data As Vandenbroucke'9 noted some- national product25; in some countries, a analysis (e.g., logistic regression), and what disapprovingly, social explanations large increase in gross national product exposure measurement (e.g., new molecu- for illness won't go away and are rediscov- has been accompanied by little benefit in lar biology techniques). Epidemiology has ered with each new generation ofepidemi- terms of health, whereas some relatively become a set of generic methods for the ologists. However, rather than ask, as poor countries (e.g. China, Jamaica, Costa measurement of disease occurrence, and Vandenbroucke does, why the population Rica) have made major improvements in there has been a concomittant lack of perspective keeps resurfacing, it is per- health care and life expectancy.26 Public distinctive theory to permit an understand- haps more useful to ask why it keeps health measures that aim to address the

679 May 1996, Vol. 86, No. 5 American Journal of Public Health Pearce health effects of poverty may ultimately as the unit of study.1 However, the new "achievement" of the public health move- find themselves in conflict with govern- risk factors that were studied were often ment has often been to move public ment policies, or may even have the conceptualized in individual terms, and health problems from rich countries to ultimate policy impact of changing the individual lifestyle has received increasing poor countries and from rich to poor government, but this does not make the emphasis during the last few decades. populationswithin the industrialized coun- role of social factors any less important or Epidemiology became widely recog- tries. Of course, this is not solely the fault less worthy of study. nized with the discovery of tobacco smok- of epidemiologists. However, when a A further issue may be that epidemi- ing as a cause of lung cancer in the early public health problem is studied in indi- ologists tend to be most interested in risk 1950s. Subsequent decades have seen vidual terms (e.g., tobacco smoking) rather factors that they can relate to, or may even major discoveries relating to other causes than in population terms (e.g., tobacco be exposed to. Epidemiologists are fre- of noncommunicable disease such as production, advertising, and distribution, quently at risk from tobacco smoke, asbestos, ionizing radiation, and dietary and the social and economic influences on alcohol, diet, viruses, and even some factors. These epidemiologic successes consumption), then it is very likely that occupational chemical exposures, but they have in some cases led to successful the solution will also be defined in are rarely at risk of being poor. The poor preventive interventions without the need individual terms and the resulting public may be occasionally encountered in ran- for major social or political change. For health action will merely move the prob- dom population surveys or after taking example, occupational carcinogens can, lem rather than solve it. the wrong exit from the autoroute; in daily with some difficulty, be controlled through life they are mostly invisible, although regulatory measures, and exposures to they are becoming harder to avoid as known occupational carcinogens have Levels ofAnalysis problems of homelessness and exclusion been reduced in industrialized countries increase. The apparently competing explana- in recent decades. Another example is the tions for disease causation (e.g. tobacco successful World Health Organization smoking in individuals vs socioeconomic The Rise ofRisk Fator campaign against smallpox.27 More re- factors in populations) can be reconciled Epidemiology cently, some countries have passed legisla- by recognizing that these explanations tion to restrict advertising of tobacco and operate at different levels of analysis. Just However, perhaps the main reason smoking in public places and have adopted as the occurrence of disease within a that socioeconomic factors currently re- health promotion programs aimed at population can be studied at many differ- ceive little attention in epidemiology is changes in lifestyle. ent levels36-including populations, indi- that they are not considered to be real However, the success of risk factor viduals, organs, tissue, cells, and mol- causes. Of course, many 19th-century epidemiology has been more temporary ecules-the causes of disease can also be epidemiologists also considered that socio- and more limited than might have been studied at these different levels, including economic factors were not the real causes; expected (S. Wing, Concepts in modem socioeconomic factors, lifestyle, the organ rather, they studied disease at the popula- epidemiology: population, risk, dose- burden of a carcinogen, and DNA ad- tion level because the relevant biologic response and confounding, unpublished ducts.37 Although specific risk factors may agents were at that time unknown.27 manuscript).28 It is one thing to discover appear to operate at the individual level, When these agents were discovered, atten- that tobacco smoke is the major cause of exposure and susceptibility38'39 may occur tion then shifted toward addressing these lung cancer, but redressing this situation due to a wide range ofpolitical, economic, "real causes," and the Henle-Koch postu- is a different problem entirely. For ex- and social factors. For example, Millard lates displaced the population-based ap- ample, smoking can be viewed as a discussed the factors leading to high child proach of the 19th-century epidemiolo- strategy enabling women to cope with mortality rates in developing countries gists., stress,29 while at the same time undermin- and identified three main tiers: theproxi- The decline in infectious disease and ing their health and that of their chil- mate tier includes the immediate biomedi- the rise in relative importance of noncom- dren.30 Any meaningful public health cal conditions that result in death (involv- municable disease led to the development intervention regarding tobacco must also ing interactions of malnutrition and of a new epidemiologic paradigm in the consider why manual workers smoke infection); the internediate tier includes mid-20th century; this involved not only a more than nonmanual workers and find it child care practices and other behaviors shift in the object of study and a recogni- more difficult to give smoking up3l and that increase the exposure of children to tion ofthe role ofmultiple causes, but also why most physicians have responded to factors on the proximate tier; and the the development of new techniques of the epidemiologic evidence and given up ultimate tier encompasses "the broad study design and data analysis.' In a smoking, whereas nurses continue to social, economic and cultural processes certain sense, this new paradigm repre- smoke in great numbers. Moreover, it can and structures that lead to the differential sented both a significant advance and a be argued that the fundamental problem distribution ofbasic necessities, especially step back to the future, as it restored some of tobacco lies in its production rather food, shelter and sanitation."40 of the population-based inferences that than in its consumption.32 The limited had fallen into decline in the late 19th success of legislative measures in industri- century and the first half of the 20th alized countries has led the tobacco Top Down and Bottom Up century because of the successes of the industry to shift its promotional activities germ theory.1 Some key figures in the new to developing countries, so that more Thus, any meaningful analysis of the paradigm espoused a holistic view of people are exposed to tobacco smoke causes of disease in populations must disease, recognized the need for a multidis- than ever before.3334 Similar shifts have integrate the individual-biologic and popu- ciplinary approach (social, biologic, statis- occurred for some occupational carcino- lation levels of analysis without collapsing tical), and specified the population group gens.35 Thus, on a global basis the one into the other or denying the exis-

680 American Journal of Public Health May 1996, Vol. 86, No. 5 Epidemiology and Public Health tence of either.41 Nevertheless, it is neces- disciplines; for example, no one would scientific approach to discover the major sary to start somewhere. attempt to predict the weather or the causes of disease in populations. How- So what is the most appropriate level motion of the planets from measurements ever, if the goal is to understand and at which to commence to study the causes of individual molecules. Not only is such prevent the causation ofdisease in popula- ofdisease in a population? Most research- an approach impossible in practice be- tions, then epidemiology should start at ers will immediately answer that their own cause of the infinitely large amount of the population level and should address discipline has it right, and all of the others information required, but recent work in the major determinants of health and have it wrong (what McKinlay16 described chaos theory has shown that such an disease at this level. For example, the as "hardening ofthe categories"). Molecu- approach is also impossible in theory recent Leeds declaration4 emphasized the lar biologists will focus on the etiologic because small inaccuracies can produce need to refocus upstream and to use process at the molecular level, much huge effects in nonlinear systems.44 research methods that are appropriate to recent epidemiologic research has fo- On the other hand, the top-down the level at which intervention will take cused on individual lifestyle, and some approach (variants of which include the place. Epidemiologic techniques can be epidemiologists, demographers, and so- structural approach,28 the dialectical ap- used in other settings (e.g., clinical epide- cial scientists have continued to conduct proach,45 and the "upstream" approach'6) miology) and for other purposes (e.g., studies at the population level.21 stems in part from the demographic studies ofdisease progression and progno- These various pathways to under- tradition and starts at the population level sis), but the key contribution ofepidemiol- standing the disease process fall into two in order to ascertain the main factors that ogy to public health is its population main approaches that mirror wider scien- influence health status within the popula- focus. tific debates in recent centuries. The tion. It implicitly uses a structural model Of course, epidemiologic studies in bottom-up approach is inherently reduc- of causation rather than a behavioral populations involve individuals who have tionist and positivist (i.e., it assumes that model or a biomedical model.46 This specific exposures, but the important knowledge consists only of events [facts] approach is inherently realist7 (i.e., it distinction is whether or not the etiologic that come from sense perception). This holds that the objects of study exist framework is conceptualized at the popu- approach focuses on understanding the and-for the most part-act indepen- lation level and whether or not these individual components of a process at the dently of scientists, but it differs from exposures are placed in their social and lowest possible level and using this infor- positivism in that the object of scientific historical context.5 For example, just as a mation as the building blocks to gain inquiry is not patterns of events but rather variety of health effects in various organ knowledge about higher levels oforganiza- the underlying processes and structures systems (e.g., various types of cancer) may tion. One current example is molecular that cause these events to occur). Causa- have a common contributing cause (e.g., epidemiology, which attempts to under- tion is seen as resulting from mechanisms tobacco smoking) at the level of the stand disease at the molecular level and that are internal to the population under individual, a variety of individual expo- then ultimately to use this knowledge in study and that operate dialectically, rather sures (e.g., smoking, diet) may have public health policy (e.g., by screening than involving regular associations be- common socioeconomic causes at the populations for individual susceptibility to tween externally related independent ob- population level (e.g., poverty). On the specific carcinogens). This approach stems jects.48 The top-down population ap- other hand, the need to analyze the causes from the clinical tradition and is typified proach is implicit in traditional definitions of disease at the population level does not by an emphasis on the individual, on of epidemiology that commonly refer to mean that public health action should specific risk factors, and on the use of the "the study ofthe distribution and determi- only be taken at the population level. randomized clinical trial as a paradigm (a nants of health-related states or events in Indeed, there is a danger of a social variety of study designs are used, but the specified populations, and the application engineering approach to public health randomized trial is the gold standard to of this study to control of health prob- that itself reinforces the social structures which the other study designs aspire). It is lems."49 It has been supported in a recent that cause disease. Just as it is important implicit in some recent definitions of editorial5O that argued for the "need to to understand the causation of disease at epidemiology12'42 and yields useful infor- move away from the almost exclusive both the population and the individual mation about the level under study (e.g., focus of research on individual risk, levels, it is also essential to take action at the molecular level), but it is debatable toward the social structures and processes both the population and the individual whether the bottom-up approach is an within which ill-health originates, and levels in ways that increase rather than effective and efficient long-term strategy which will be more amenable to modifica- decrease individuals' control over their to gain knowledge or prevent disease at tion."51 environment.53 the population level.18 It is also important to recognize that The bottom-up approach lacks dis- BeyondModem Epidemiology the "populations" that epidemiologists tinctive theory regarding the occurrence study are notjust collections ofindividuals of disease at the population level43 (mod- So how can we go "back to the that are conveniently grouped for the em epidemiologic studies are conducted future"16 and develop new forms of purposes of study.2' Every population has in populations, but the implicit etiologic epidemiology that restore the population its own history, culture, organization, and theory is usually based at the individual- perspective but that appropriately use economic and social divisions, which biologic level), and its products can be recent methodologic advances? influence how and why people are ex- likened to "a vast stockpile of almost The key issue is that epidemiology is posed to particular factors. For example, surgically clean data untouched by human first and foremost a branch of public Terris argues that thought."4 Although it has an air of health.52 This view is not contradictory to the causes of cholera in India today go scientific purity, this approach is in fact the view of epidemiology as a branch of back hundreds of years in India's his- rarely used in other sciences or related science; in fact, it is necessary to take a tory, to the British invasion and destruc-

Public Health 681 May 1996, Vol. 86, No. 5 American Journal of Pearce

tion of once-flourishing textile indus- was at one time a disease of the affluent, nas, Meri Koivusalo, John McKinlay, Tony tries; the maintenance ofarchaic systems but it has become a disease of the poor as McMichael, Timo Partanen, Charlotte Paul, of land ownership and tillage; the smoking and eating habits have changed Steve Wing, and Karen Witten for their persistence of the caste system and the comments on the draft manuscript. unbelievable poverty, hunger, and over time.62 Thus, appropriate preventive crowding; the consequent inability to measures at the population level will afford the development of safe water differ widely between populations. Fur- References supplies and sewage disposal systems; thermore, although many specific risk 1. Susser M. Epidemiology in the United and, almost incidentally, the presence States after World War II: the evolution of ofcholera vibrios.54 factors play an important role in any technique. Epidemiol Rev. 1985;7:147-177. population, their contribution to disease 2. Lilienfeld D, Lilienfeld A. Epidemiology: a Thus, epidemiology is inevitably en- risk is modified by the baseline disease retrospective study. Am J Epidemiol. 1977; tangled with society, and it is not feasible risk and the presence ofvarious cofactors, 106:445-449. or desirable to study the causes of disease to assume a univer- 3. Terris M. Epidemiology and the public making it impossible health movement. J Public Health Policy. in the abstract.5 To understand the causa- sal dose-response relationship.21 A re- 1987;7:315-329. tion of disease in a population, it is lated issue is the importance of consider- 4. Directions for Health: New Approaches to essential to understand the historical and ing interrelationships between causes Population Health Research and Practice. social context and to emphasize the rather than considering each cause in The Leeds Declaration. Leeds, England: importance of diversity and local knowl- Nuffield Institute for Health, University of isolation.63 Leeds; 1993. edge rather than only searching for Finally, the randomized clinical trial 5. Susser M. Epidemiology today: "a thought- universal relationships.55 This requires a may be an appropriate paradigm in many tormented world." IntJEpidemiol. 1989;18: greater involvement from the social sci- epidemiologic studies of specific risk 481-488. ences and a more multidisciplinary ap- 6. McKeown T. The Role ofMedicine. Prince- factors, but it often is inappropriate in ton, NJ: Princeton University Press; 1979. proach.5657 Epidemiology is just one of studies that require a consideration of the 7. Szreter S. The importance of social inter- the approaches by which the major deter- historical and social context. The danger vention in Britain's mortality decline c. minants of health in a population can be is that attempting to eliminate the influ- 1850-1914: a reinterpretation of the role of addressed, and it should be comple- ence of all other causes of diseases-in an public health. Soc HistMed. 1988;1: 1-37. 8. Smith GD, Morris J. Increasing inequali- mented by other quantitative approaches attempt to control confounding-strips ties in the health of the nation. BMJ. from the social sciences, as well as away the essential historical and social 1994;309:1453-1454. qualitative and historical studies. The context,2' as well as the multiple moderat- 9. Pappas G, Queen S, Hadden W, Fisher G. emphasis should be on using appropriate ing influences that constitute true causa- The increasing disparity in mortality be- rather than making the tween socioeconomic groups in the United methodology5' tion.63 Thus, the tendency to only study States, 1960 and 1986. N Engl J Med. problem fit the method. factors that fit the clinical trial paradigm 1993;329:103-109. Rose58 also noted that entire popula- should be resisted, and appropriate study 10. Logan WPD. Cancer Mortality by Occupa- tions may be exposed to a particular risk designs should be chosen (or developed) tion and Social Class, 1951-1971. Lyon, factor, and there is usually a continuum of to fit the public health question that is France: International Agency for Research on Cancer; 1982. disease risk (rather than a clear distinc- being addressed.64 11. Doll R, Peto R. The Causes of Cancer. tion between the sick and the healthy) Epidemiology has become a set of Oxford, England: Oxford University Press; across the population. Small improve- generic methods for measuring associa- 1981. ments in the health of a "sick population" tions of exposure and disease in individu- 12. Rothman KJ. Modem Epidemiology. Bos- ton, Mass: Little, Brown; 1986. may be more effective than attempts to als, rather than functioning as part of a 13. Stallones S. To advance epidemiology. treat or prevent illness in "sick individu- multidisciplinary approach to understand- Annu Rev Public Health. 1980;1:69-82. als."58 Studying these phenomena often ing the causation of disease in popula- 14. Cameron D, Jones IG. John Snow, the requires the use of ecologic studies, since tions. These methodologic changes have Broad Street pump and modem epidemiol- not been neutral, but rather, in combina- ogy. IntJEpidemiol. 1983;12:393-396. group variables may be important apart 15. Krieger N. Epidemiology and the web of from individual-level variables59 (although tion with other influences, they have causation: has anyone seen the spider? Soc they may not be an adequate substitute changed-and have reflected changes SciMed. 1994;39:889-903. for the individual-level variables, espe- in-the way in which epidemiologists 16. McKinlay JB. Towards appropriate levels for For think about health and disease. We seem of analysis, research methods and health cially confounding control6W). public policy. Presented at the Interna- example, analysis at the individual level to be using more and more advanced tional Symposium on Quality of Life and cannot explain epidemic spread at the technology to study more and more trivial Health: Theoretical and Methodological group level and cannot even fully explain issues, while the major population causes Considerations; May 25-27, 1994; Berlin, the spread of infections between individu- of disease are ignored. Epidemiology Germany. must reintegrate itself into public health, 17. McMichael AJ. "Molecular epidemiol- als. The problems specific to ecologic ogy." New pathway or new travelling analysis6' arise only when one extrapo- and must rediscover the population per- companion? Am J Epidemiol. 1994;140: lates downward from the population to spective. C 1-11. the individual level; many criticisms of 18. Pearce N, san Jose S, Boffetta P, et al. Limitations of biomarkers of exposure in ecologic studies are based on the question- Acknowledgments cancer epidemiology. Epidemiology. 1995;6: able assumption that the individual level This work was funded by a Senior Research 190-194. of analysis is most appropriate. Fellowship of the Health Research Council of 19. Vandenbroucke JP. New public health and It should also be emphasized that the New Zealand. old rhetoric. BMJ. 1994;308:994-995. strength, and even the direction, of asso- I would like to thank Warwick Arm- 20. Thatcher M. Womans Own. October 1987. strong, Robert Beaglehole, Paolo Boffetta, 21. Wing S. Limits of epidemiology. Med ciations between risk factors and disease Sunny Collings, Fiona Cram, Peter Davis, Global Survival. 1994;1:74-86. will vary between populations and over Isabelle Godin, Phillippa Howden-Chapman, 22. Black D. Deprivation and health. BMJ. time. For example, coronary heart disease Rod Jackson, Ichiro Kawachi, Manolis Kogevi- 1993;307:1630-1631.

682 American Journal of Public Health May 1996, Vol. 86, No.5 Epidemiology and Public Health

23. Hewlett SA. Child Neglect in Rich Nations. national Agency for Research on Cancer; 50. Anon. Population health looking up- New York, NY: UNICEF, 1993. 1994. stream. Lancet. 1994;343:429-430. Edito- 24. Pearce N. Economic policy and health in 36. Susser M. Causal Thinking in the Health rial. The Year of the Family. NZ Med J. Sciences. New York, NY: Oxford Univer- 51. McKinlay JB. The promotion of health 1994;107:379-381. sity Press; 1973. through planned sociopolitical change: chal- 25. Wilkinson R. National mortality rates: the 37. Potter JD. Reconciling the epidemiology, lenges for research and policy. Soc Sci Med. impact of inequality. Am J Public Health. physiology, and molecular biology of colon 1993;36:109-117. 1992;82:1082-1084. cancer.JAMA. 1992;268:1573-1577. 52. Lilienfeld A, Lilienfeld DE. Epidemiology 26. Sen A. Levels ofPoverty: Policy and Chbac:e. 38. Syme SL, Berkman LF. Social class, suscep- and the public health movement. J Publc Washington, DC: The World Bank; World tibility and sickness. Am J Epidemiol. Health Policy. 1982;3:140-149. Bank StaffWorking Paper No. 401. 1980. 1976;104:1-8. 53. Adler NE, Boyce WT, Chesney MA, 27. Tesh SN. HiddenArguments: PoliticalIdeol- 39. Cassel J. The contribution of the social Folkman S, Syme SL. Socioeconomic in- ogy and Disease Prevention Policy. London, environment to host resistance. Am J equalities in health: no easy solution. England: Rutgers; 1988. Epidemiol. 1976;104:107-123. JAMA. 1993;269:3140-3145. 28. Loomis D, Wing S. Is molecular epidemiol- 40. Millard AV. A causal model of high rates 54. Terris M. The epidemiologic tradition. ogy a germ theory for the end of the ofchild mortality. Soc SciMed. 1994;38:253- Public Health Rep. 1979;94:204. twentieth century? IntJEpidemiol. 1991;19: 268. 55. Kunitz S. Disease and Social Diversity. New 1-3. 41. Rose S, Lewontin RC, Kamin LJ. Not in York, NY: Oxford University Press; 1994. 29. Jacobson B. The Lady Killers: Why Smoking Our Genes: Biology, Ideology and Human 56. Susser M, Watson W, Hopper K. Sociok*j Issue. England: Pluto Nature. London, England: Penguin Books; in Medicine. 3rd ed. New York, NY: Oxford Is a Feminist London, 1990. University Press; 1985. Press; 1981. 42. Miettinen OS. Theoretical Epidemiology: 57. Connelly J. Social sciences provide a 30. Power C, Manor 0, Fox J. Health and Principles of Occurrence Research. New framework for epidemiology. BMJ. 1994; Class: The Early Years. London, England: York, NY: Wiley; 1985. 308:1569. Letter. Chapman Hall; 1991. 43. Smith A. The epidemiological basis of 58. Rose G. The Strategy ofPreventiveMedicine. 31. Pearce N, Davis PB, Smith AH, Foster FH. community medicine. In: Smith A, ed. Oxford, England: Oxford University Press; Social class, ethnic group, and male mortal- RecentAdvances in Community Medicine 3. 1992. ity in New Zealand, 1974-8. J Epidemiol Edinburgh, Scotland: Churchill Living- 59. Koopman JS, Longini IM. The ecological Community Health. 1985;39:9-14. stone; 1985:1-10. effects of individual exposures and nonlin- 32. Reid P, Pouwhare R. Te Taonga mai 44. Firth WJ. Chaos-predicting the unpredict- ear disease dynamics in populations. Am J Tawhiti: The Gift from a Distant Place. able. BMJ. 1991;303:1565-1568. Public Health. 1994;84:836-842. Auckland, New Zealand: Niho Taniha; 45. Levins R, Lewontin R. The Dialectical 60. Greenland S, Robins J. Ecologic studies- 1991. Biologist. Cambridge, Mass: Harvard Uni- biases, misconceptions, and counterex- 33. Barry M. The influence ofthe U.S. tobacco versity Press; 1985. amples.AmJEpidemiol. 1994;139:747-760. industry on the health, economy, and 46. Armstrong D. An Outline of Sociology as 61. Susser M. The logic in ecological: II. the environment of developing countries. N Applied to Medicine. Bristol, England: logic ofdesign.AmJPubicHealth. 1994;84: EnglJMed. 1991;324:917-920. Wright; 1980. 830-835. 34. Tominaga S. Spread of smoking to the 47. Bhaskar R. Reclaiming Reality. London, 62. Wing S. Social inequalities in the decline of developing countries. In: Zaridze D, Peto England: Verso; 1989. coronary mortality. Am J Public Health. R, eds. Tobacco: A Major International 48. Jones K, Moon G. Health, Disease and 1988;78:1415-1416. Health Hazard. Lyon, France: International Society:A Critical Medical Geography. Lon- 63. Dean K. Creating a new knowledge base Agency for Research on Cancer; 1986:125- don, England: Routledge & Kegan Paul; for the new public health. J Epidemiol 133. 1987. Community Health. 1994;48:217-219. 35. Pearce NE, Matos E, Vainio H, Boffetta P, 49. Last JM, ed. A Dictionary ofEpidemiology. 64. McMichael AJ. Persons, populations and Kogevinas M, eds. Occupational Cancer in New York, NY: Oxford University Press; planets: a full world brings epidemiology Developing Countries. Lyon, France: Inter- 1988. full circle. Epidemiology. In press.

May 1996, Vol. 86, No. 5 American Journal of Public Health 683