Traditional Epidemiology, ModernEpidemiology, and Public Health

Neil Pearce, PhD

IntroductionIn this article I consider some of the

key features ofwhat has become known asmodern epidemiology, and I contrastthese with more traditional approaches. Ido not intend to present a detailedhistorical review,1-3 nor to present acomprehensive review of current ap-proaches. Rather, I will focus on some ofthe key changes in epidemiology over thepast few decades, and I will consider theconcepts of causality involved and theirideological and practical consequences. Iwill argue that the rise ofmodem epidemi-ology has been a mixed blessing and thatthe new paradigm has major shortcom-ings, both in public health and in scientificterms.4 The recent changes in epidemio-logic methodology have not been neutral,but rather (in combination with otherinfluences) they have changed-and havereflected changes in-the way in whichepidemiologists think about health anddisease.5 The key issue has been the shiftin the level of analysis from the popula-tion to the individual (S. Wing, Conceptsin modem epidemiology: population, risk,dose-response, and confounding, unpub-lished manuscript). This is typified by thecurrent lack of interest in populationfactors as causes of disease, the lack ofinterest in the history of epidemiology,and the lack of integration with otherpublic health activities. I will give particu-lar emphasis to the current neglect ofsocial, economic, cultural, historical, politi-cal, and other population factors, and Iwill refer to these using the general termof "socioeconomic factors."

Of course, traditional epidemiologywas not a monolith. A wide variety ofapproaches were used, and there is a

danger of setting up caricatures of idealtypes. It should also be emphasized thattraditional epidemiology gave rise to

modern epidemiology; therefore, theyhave many features in common. Neverthe-less, there are some important differencesbetween the traditional and the modernapproaches, particularly the loss of thepopulation perspective in recent decades.I therefore will discuss some of thereasons the population perspective hasbeen lost and the implications of thisparadigm shift. Then, I will discuss someof the key issues in developing new formsof epidemiology that restore the popula-tion perspective while making use ofrecent methodological advances. (I amtempted to use the term "postmodernepidemiology" to provide a contrast withmodern epidemiology and because somepostmodernist concepts are relevant tomy arguments; however, the use of thisterm could imply an uncritical advocacy ofpostmodernism, which has its own episte-mological and practical shortcomings.)

Traditional and ModernEpidemiology

In the first week of their training,most epidemiologists usually leam a littleabout the history of public health. Inanglophone countries they learn aboutthe work of Chadwick, Engels, Snow, andothers, who exposed the appalling socialconditions during the industrial revolu-tion, and the work of Farr and others, whorevealed major socioeconomic differencesin disease in the 19th century. At thattime, epidemiology was a branch of public

The author is with the Wellington School ofMedicine, Wellington, New Zealand.

Requests for reprints should be sent toNeil Pearce, PhD, Wellington School of Medi-cine, PO Box 7343, Wellington, New Zealand.

This article was accepted June 22, 1995.Editor's Note. See related editorial by

Winkelstein (p 621) in this issue.

May 1996, Vol. 86, No.5

Epidemiology and Public Health

health and focused on the causes andprevention of disease in populations, incontrast to the clinical sciences, whichwere branches of medicine and focusedon disease pathology and treatment ofindividuals (S. Wing, Concepts in modemepidemiology: population, risk, dose-response, and confounding, unpublishedmanuscript). Thus, the emphasis was onthe prevention of disease and the healthneeds of the population as a whole. Thedramatic decline in infectious diseasesthat has occurred since the mid-19thcentury has been attributed to improve-ments in nutrition, sanitation, and generalliving conditions,6 although it has beenargued that specific public health interven-tions regarding factors such as urbancongestion actually played the majorrole.7

There are still major socioeconomicdifferences in health, and the relativedifferences are continuing to increase.8'9Nevertheless, modern epidemiologistsrarely consider socioeconomic factors andthe population perspective, except per-haps to occasionally adjust for social classin analyses of the health effects of tobaccosmoke, diet, and other lifestyle factors inindividuals. For example, studies in mostindustrialized countries have repeatedlyfound strong associations between socialclass and cancer,10 yet social class did notfeature, except for a brief mention as aconfounder, in the most comprehensivereview of the causes of cancer in theUnited States,11 and one leading epidemi-ology text states that "social class ispresumably related causally to few if anydiseases.'2

Traditional epidemiology has be-come unfashionable and is treated some-what disparagingly in modem epidemiol-ogy texts, which have rewritten the historyof epidemiology in their own image. Inparticular, there has been a strong focuson statistical issues and paradigms'3 andan ignorance of the other modes ofthought that were integral to the work of19th-century pioneers such as Snow.'4One is left with the impression that19th-century epidemiologists used ad hocmethods that have now been placed on asounder foundation through recent devel-opments in methods of study design (e.g.,the theory of case-control studies), dataanalysis (e.g., logistic regression), andexposure measurement (e.g., new molecu-lar biology techniques). Epidemiology hasbecome a set of generic methods for themeasurement of disease occurrence, andthere has been a concomittant lack ofdistinctive theory to permit an understand-

TABLE 1-Epidemiological Paradigms

Traditional Epidemiology

MotivationLevel of study

Context of studyParadigmsEpistemological approachEpistemological strategyLevel of intervention

Public healthPopulation

Historical/culturalDemography/social scienceRealistTop down (structural)Population (upstream)

ing of the population patterns of diseaseoccurrence.15

These methodological developmentshave been paralleled by, and have re-flected, a shift in the level of analysis fromthe population to the individual. Mostmodem epidemiologists still do studies inpopulations, but they do so in order tostudy decontextualized individual risk fac-tors, rather than to study populationfactors in their social and historical con-text. McKinlayl6 argued that what is nowregarded as established epidemiology ischaracterized by biophysiologic reduction-ism, absorption by biomedicine, a lack ofreal theory about disease causation, di-chotomous thinking about disease (every-one is either healthy or sick), a maze ofrisk factors, confusion of observationalassociations with causality, dogmatismabout which study designs are acceptable,and excessive repetition of studies. Heargued that this approach diverts limitedresources, blames the victim, produces alifestyle approach to social policy, decon-textualizes risk behaviors, seldom assessesthe relative contribution of nonmodifiablegenetic factors and modifiable social andbehavioral factors, and produces interven-tions that can be harmful. These trendsare particularly noticeable in the recentrise of molecular epidemiology,'7"18 espe-cially in the renewed emphasis on issuesof individual susceptibility.

The Decline ofPopulationEpidemiology

As Vandenbroucke'9 noted some-what disapprovingly, social explanationsfor illness won't go away and are rediscov-ered with each new generation of epidemi-ologists. However, rather than ask, asVandenbroucke does, why the populationperspective keeps resurfacing, it is per-haps more useful to ask why it keeps

Modern Epidemiology

ScienceIndividual/organ/tissue/

cell/moleculeContext freeClinical trialPositivistBottom up (reductionist)Individual (downstream)

sinking. Some of the reasons for thecurrent lack of interest in the populationperspective may lie in the personal andprofessional situations of epidemiologists.In most countries the main sources offunding are government or voluntaryagencies that have little interest in, orsympathy for, studies of socioeconomicfactors and health. In the last decade,Western countries, particularly anglo-phone countries, have increasingly placedemphasis on individual responsibility, typi-fied by the famous statement by MargaretThatcher20 that "there is no such thing associety, there are only families and indi-viduals." Governments and funding agen-cies have been most supportive of studiesthat focus on individual lifestyle, andepidemiologists, either through choice orthrough necessity, have tended to go"where the money is."

A related issue is that socioeconomicfactors are "not easily modifiable" and are"too political." However, the decision notto study socioeconomic factors is itself apolitical decision to focus on what ispolitically acceptable rather than what ismost important in scientific and publichealth terms.21 Governments have repeat-edly shown that social and economicdifferences are not God-given but aredirectly affected by government policies,albeit in unexpected ways.22-24 For ex-ample, there is some preliminary evidencethat inequitable distribution of the grossnational product can have a more signifi-cant impact on overall national mortalityrates than the actual level of the grossnational product25; in some countries, alarge increase in gross national producthas been accompanied by little benefit interms of health, whereas some relativelypoor countries (e.g. China, Jamaica, CostaRica) have made major improvements inhealth care and life expectancy.26 Publichealth measures that aim to address the

American Journal of Public Health 679May 1996, Vol. 86, No. 5

Pearce

health effects of poverty may ultimatelyfind themselves in conflict with govern-ment policies, or may even have theultimate policy impact of changing thegovernment, but this does not make therole of social factors any less important orless worthy of study.

A further issue may be that epidemi-ologists tend to be most interested in riskfactors that they can relate to, or may evenbe exposed to. Epidemiologists are fre-quently at risk from tobacco smoke,alcohol, diet, viruses, and even someoccupational chemical exposures, but theyare rarely at risk of being poor. The poormay be occasionally encountered in ran-dom population surveys or after takingthe wrong exit from the autoroute; in dailylife they are mostly invisible, althoughthey are becoming harder to avoid asproblems of homelessness and exclusionincrease.

The Rise ofRisk FatorEpidemiology

However, perhaps the main reasonthat socioeconomic factors currently re-ceive little attention in epidemiology isthat they are not considered to be realcauses. Of course, many 19th-centuryepidemiologists also considered that socio-economic factors were not the real causes;rather, they studied disease at the popula-tion level because the relevant biologicagents were at that time unknown.27When these agents were discovered, atten-tion then shifted toward addressing these"real causes," and the Henle-Koch postu-lates displaced the population-based ap-proach of the 19th-century epidemiolo-gists.,

The decline in infectious disease andthe rise in relative importance of noncom-municable disease led to the developmentof a new epidemiologic paradigm in themid-20th century; this involved not only ashift in the object of study and a recogni-tion of the role of multiple causes, but alsothe development of new techniques ofstudy design and data analysis.' In acertain sense, this new paradigm repre-sented both a significant advance and astep back to the future, as it restored someof the population-based inferences thathad fallen into decline in the late 19thcentury and the first half of the 20thcentury because of the successes of thegerm theory.1 Some key figures in the newparadigm espoused a holistic view ofdisease, recognized the need for a multidis-ciplinary approach (social, biologic, statis-tical), and specified the population group

as the unit of study.1 However, the newrisk factors that were studied were oftenconceptualized in individual terms, andindividual lifestyle has received increasingemphasis during the last few decades.

Epidemiology became widely recog-nized with the discovery of tobacco smok-ing as a cause of lung cancer in the early1950s. Subsequent decades have seenmajor discoveries relating to other causesof noncommunicable disease such asasbestos, ionizing radiation, and dietaryfactors. These epidemiologic successeshave in some cases led to successfulpreventive interventions without the needfor major social or political change. Forexample, occupational carcinogens can,with some difficulty, be controlled throughregulatory measures, and exposures toknown occupational carcinogens havebeen reduced in industrialized countriesin recent decades. Another example is thesuccessful World Health Organizationcampaign against smallpox.27 More re-cently, some countries have passed legisla-tion to restrict advertising of tobacco andsmoking in public places and have adoptedhealth promotion programs aimed atchanges in lifestyle.

However, the success of risk factorepidemiology has been more temporaryand more limited than might have beenexpected (S. Wing, Concepts in modemepidemiology: population, risk, dose-response and confounding, unpublishedmanuscript).28 It is one thing to discoverthat tobacco smoke is the major cause oflung cancer, but redressing this situationis a different problem entirely. For ex-ample, smoking can be viewed as astrategy enabling women to cope withstress,29 while at the same time undermin-ing their health and that of their chil-dren.30 Any meaningful public healthintervention regarding tobacco must alsoconsider why manual workers smokemore than nonmanual workers and find itmore difficult to give smoking up3l andwhy most physicians have responded tothe epidemiologic evidence and given upsmoking, whereas nurses continue tosmoke in great numbers. Moreover, it canbe argued that the fundamental problemof tobacco lies in its production ratherthan in its consumption.32 The limitedsuccess of legislative measures in industri-alized countries has led the tobaccoindustry to shift its promotional activitiesto developing countries, so that morepeople are exposed to tobacco smokethan ever before.3334 Similar shifts haveoccurred for some occupational carcino-gens.35 Thus, on a global basis the

"achievement" of the public health move-ment has often been to move publichealth problems from rich countries topoor countries and from rich to poorpopulations within the industrialized coun-tries. Of course, this is not solely the faultof epidemiologists. However, when apublic health problem is studied in indi-vidual terms (e.g., tobacco smoking) ratherthan in population terms (e.g., tobaccoproduction, advertising, and distribution,and the social and economic influences onconsumption), then it is very likely thatthe solution will also be defined inindividual terms and the resulting publichealth action will merely move the prob-lem rather than solve it.

Levels ofAnalysisThe apparently competing explana-

tions for disease causation (e.g. tobaccosmoking in individuals vs socioeconomicfactors in populations) can be reconciledby recognizing that these explanationsoperate at different levels of analysis. Justas the occurrence of disease within apopulation can be studied at many differ-ent levels36-including populations, indi-viduals, organs, tissue, cells, and mol-ecules-the causes of disease can also bestudied at these different levels, includingsocioeconomic factors, lifestyle, the organburden of a carcinogen, and DNA ad-ducts.37 Although specific risk factors mayappear to operate at the individual level,exposure and susceptibility38'39 may occurdue to a wide range of political, economic,and social factors. For example, Millarddiscussed the factors leading to high childmortality rates in developing countriesand identified three main tiers: the proxi-mate tier includes the immediate biomedi-cal conditions that result in death (involv-ing interactions of malnutrition andinfection); the internediate tier includeschild care practices and other behaviorsthat increase the exposure of children tofactors on the proximate tier; and theultimate tier encompasses "the broadsocial, economic and cultural processesand structures that lead to the differentialdistribution of basic necessities, especiallyfood, shelter and sanitation."40

Top Down and Bottom UpThus, any meaningful analysis of the

causes of disease in populations mustintegrate the individual-biologic and popu-lation levels of analysis without collapsingone 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-sary to start somewhere.

So what is the most appropriate levelat which to commence to study the causesof disease in a population? Most research-ers will immediately answer that their owndiscipline has it right, and all of the othershave it wrong (what McKinlay16 describedas "hardening of the categories"). Molecu-lar biologists will focus on the etiologicprocess at the molecular level, muchrecent epidemiologic research has fo-cused on individual lifestyle, and someepidemiologists, demographers, and so-cial scientists have continued to conductstudies at the population level.21

These various pathways to under-standing the disease process fall into twomain approaches that mirror wider scien-tific debates in recent centuries. Thebottom-up approach is inherently reduc-tionist and positivist (i.e., it assumes thatknowledge consists only of events [facts]that come from sense perception). Thisapproach focuses on understanding theindividual components of a process at thelowest possible level and using this infor-mation as the building blocks to gainknowledge about higher levels of organiza-tion. One current example is molecularepidemiology, which attempts to under-stand disease at the molecular level andthen ultimately to use this knowledge inpublic health policy (e.g., by screeningpopulations for individual susceptibility tospecific carcinogens). This approach stemsfrom the clinical tradition and is typifiedby an emphasis on the individual, onspecific risk factors, and on the use of therandomized clinical trial as a paradigm (avariety of study designs are used, but therandomized trial is the gold standard towhich the other study designs aspire). It isimplicit in some recent definitions ofepidemiology12'42 and yields useful infor-mation about the level under study (e.g.,the molecular level), but it is debatablewhether the bottom-up approach is aneffective and efficient long-term strategyto gain knowledge or prevent disease atthe population level.18

The bottom-up approach lacks dis-tinctive theory regarding the occurrenceof disease at the population level43 (mod-em epidemiologic studies are conductedin populations, but the implicit etiologictheory is usually based at the individual-biologic level), and its products can belikened to "a vast stockpile of almostsurgically clean data untouched by humanthought."4 Although it has an air ofscientific purity, this approach is in factrarely used in other sciences or related

disciplines; for example, no one wouldattempt to predict the weather or themotion of the planets from measurementsof individual molecules. Not only is suchan approach impossible in practice be-cause of the infinitely large amount ofinformation required, but recent work inchaos theory has shown that such anapproach is also impossible in theorybecause small inaccuracies can producehuge effects in nonlinear systems.44

On the other hand, the top-downapproach (variants of which include thestructural approach,28 the dialectical ap-proach,45 and the "upstream" approach'6)stems in part from the demographictradition and starts at the population levelin order to ascertain the main factors thatinfluence health status within the popula-tion. It implicitly uses a structural modelof causation rather than a behavioralmodel or a biomedical model.46 Thisapproach is inherently realist7 (i.e., itholds that the objects of study existand-for the most part-act indepen-dently of scientists, but it differs frompositivism in that the object of scientificinquiry is not patterns of events but ratherthe underlying processes and structuresthat cause these events to occur). Causa-tion is seen as resulting from mechanismsthat are internal to the population understudy and that operate dialectically, ratherthan involving regular associations be-tween externally related independent ob-jects.48 The top-down population ap-proach is implicit in traditional definitionsof epidemiology that commonly refer to"the study of the distribution and determi-nants of health-related states or events inspecified populations, and the applicationof this study to control of health prob-lems."49 It has been supported in a recenteditorial5O that argued for the "need tomove away from the almost exclusivefocus of research on individual risk,toward the social structures and processeswithin which ill-health originates, andwhich will be more amenable to modifica-tion."51

BeyondModem EpidemiologySo how can we go "back to the

future"16 and develop new forms ofepidemiology that restore the populationperspective but that appropriately userecent methodologic advances?

The key issue is that epidemiology isfirst and foremost a branch of publichealth.52 This view is not contradictory to

the view of epidemiology as a branch ofscience; in fact, it is necessary to take a

scientific approach to discover the majorcauses of disease in populations. How-ever, if the goal is to understand andprevent the causation of disease in popula-tions, then epidemiology should start atthe population level and should addressthe major determinants of health anddisease at this level. For example, therecent Leeds declaration4 emphasized theneed to refocus upstream and to useresearch methods that are appropriate tothe level at which intervention will takeplace. Epidemiologic techniques can beused in other settings (e.g., clinical epide-miology) and for other purposes (e.g.,studies of disease progression and progno-sis), but the key contribution of epidemiol-ogy to public health is its populationfocus.

Of course, epidemiologic studies inpopulations involve individuals who havespecific exposures, but the importantdistinction is whether or not the etiologicframework is conceptualized at the popu-lation level and whether or not theseexposures are placed in their social andhistorical context.5 For example, just as avariety of health effects in various organsystems (e.g., various types of cancer) mayhave a common contributing cause (e.g.,tobacco smoking) at the level of theindividual, a variety of individual expo-sures (e.g., smoking, diet) may havecommon socioeconomic causes at thepopulation level (e.g., poverty). On theother hand, the need to analyze the causesof disease at the population level does notmean that public health action shouldonly be taken at the population level.Indeed, there is a danger of a socialengineering approach to public healththat itself reinforces the social structuresthat cause disease. Just as it is importantto understand the causation of disease atboth the population and the individuallevels, it is also essential to take action atboth the population and the individuallevels in ways that increase rather thandecrease individuals' control over theirenvironment.53

It is also important to recognize thatthe "populations" that epidemiologistsstudy are not just collections of individualsthat are conveniently grouped for thepurposes of study.2' Every population hasits own history, culture, organization, andeconomic and social divisions, whichinfluence how and why people are ex-

posed to particular factors. For example,Terris argues that

the causes of cholera in India today goback hundreds of years in India's his-tory, to the British invasion and destruc-

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

Pearce

tion of once-flourishing textile indus-tries; the maintenance of archaic systemsof land ownership and tillage; thepersistence of the caste system and theunbelievable poverty, hunger, andcrowding; the consequent inability toafford the development of safe watersupplies and sewage disposal systems;and, almost incidentally, the presenceof cholera vibrios.54

Thus, epidemiology is inevitably en-tangled with society, and it is not feasibleor desirable to study the causes of diseasein the abstract.5 To understand the causa-tion of disease in a population, it isessential to understand the historical andsocial context and to emphasize theimportance of diversity and local knowl-edge rather than only searching foruniversal relationships.55 This requires agreater involvement from the social sci-ences and a more multidisciplinary ap-proach.5657 Epidemiology is just one ofthe approaches by which the major deter-minants of health in a population can beaddressed, and it should be comple-mented by other quantitative approachesfrom the social sciences, as well asqualitative and historical studies. Theemphasis should be on using appropriatemethodology5' rather than making theproblem fit the method.

Rose58 also noted that entire popula-tions may be exposed to a particular riskfactor, and there is usually a continuum ofdisease risk (rather than a clear distinc-tion between the sick and the healthy)across the population. Small improve-ments in the health of a "sick population"may be more effective than attempts totreat or prevent illness in "sick individu-als."58 Studying these phenomena oftenrequires the use of ecologic studies, sincegroup variables may be important apartfrom individual-level variables59 (althoughthey may not be an adequate substitutefor the individual-level variables, espe-cially for confounding control6W). Forexample, analysis at the individual levelcannot explain epidemic spread at thegroup level and cannot even fully explainthe spread of infections between individu-als. The problems specific to ecologicanalysis6' arise only when one extrapo-lates downward from the population tothe individual level; many criticisms ofecologic studies are based on the question-able assumption that the individual levelof analysis is most appropriate.

It should also be emphasized that thestrength, and even the direction, of asso-ciations between risk factors and diseasewill vary between populations and overtime. For example, coronary heart disease

was at one time a disease of the affluent,but it has become a disease of the poor assmoking and eating habits have changedover time.62 Thus, appropriate preventivemeasures at the population level willdiffer widely between populations. Fur-thermore, although many specific riskfactors play an important role in anypopulation, their contribution to diseaserisk is modified by the baseline diseaserisk and the presence ofvarious cofactors,making it impossible to assume a univer-sal dose-response relationship.21 A re-lated issue is the importance of consider-ing interrelationships between causesrather than considering each cause inisolation.63

Finally, the randomized clinical trialmay be an appropriate paradigm in manyepidemiologic studies of specific riskfactors, but it often is inappropriate instudies that require a consideration of thehistorical and social context. The dangeris that attempting to eliminate the influ-ence of all other causes of diseases-in anattempt to control confounding-stripsaway the essential historical and socialcontext,2' as well as the multiple moderat-ing influences that constitute true causa-tion.63 Thus, the tendency to only studyfactors that fit the clinical trial paradigmshould be resisted, and appropriate studydesigns should be chosen (or developed)to fit the public health question that isbeing addressed.64

Epidemiology has become a set ofgeneric methods for measuring associa-tions of exposure and disease in individu-als, rather than functioning as part of amultidisciplinary approach to understand-ing the causation of disease in popula-tions. These methodologic changes havenot been neutral, but rather, in combina-tion with other influences, they havechanged-and have reflected changesin-the way in which epidemiologiststhink about health and disease. We seemto be using more and more advancedtechnology to study more and more trivialissues, while the major population causesof disease are ignored. Epidemiologymust reintegrate itself into public health,and must rediscover the population per-spective. C

AcknowledgmentsThis work was funded by a Senior ResearchFellowship of the Health Research Council ofNew Zealand.

I would like to thank Warwick Arm-strong, Robert Beaglehole, Paolo Boffetta,Sunny Collings, Fiona Cram, Peter Davis,Isabelle Godin, Phillippa Howden-Chapman,Rod Jackson, Ichiro Kawachi, Manolis Kogevi-

nas, Meri Koivusalo, John McKinlay, TonyMcMichael, Timo Partanen, Charlotte Paul,Steve Wing, and Karen Witten for theircomments on the draft manuscript.

References1. Susser M. Epidemiology in the United

States after World War II: the evolution oftechnique. Epidemiol Rev. 1985;7:147-177.

2. Lilienfeld D, Lilienfeld A. Epidemiology: aretrospective study. Am J Epidemiol. 1977;106:445-449.

3. Terris M. Epidemiology and the publichealth movement. J Public Health Policy.1987;7:315-329.

4. Directions for Health: New Approaches toPopulation Health Research and Practice.The Leeds Declaration. Leeds, England:Nuffield Institute for Health, University ofLeeds; 1993.

5. Susser M. Epidemiology today: "a thought-tormented world." IntJEpidemiol. 1989;18:481-488.

6. McKeown T. The Role ofMedicine. Prince-ton, NJ: Princeton University Press; 1979.

7. Szreter S. The importance of social inter-vention in Britain's mortality decline c.1850-1914: a reinterpretation of the role ofpublic health. Soc Hist Med. 1988;1: 1-37.

8. Smith GD, Morris J. Increasing inequali-ties in the health of the nation. BMJ.1994;309:1453-1454.

9. Pappas G, Queen S, Hadden W, Fisher G.The increasing disparity in mortality be-tween socioeconomic groups in the UnitedStates, 1960 and 1986. N Engl J Med.1993;329:103-109.

10. Logan WPD. Cancer Mortality by Occupa-tion and Social Class, 1951-1971. Lyon,France: International Agency for Researchon Cancer; 1982.

11. Doll R, Peto R. The Causes of Cancer.Oxford, England: Oxford University Press;1981.

12. Rothman KJ. Modem Epidemiology. Bos-ton, Mass: Little, Brown; 1986.

13. Stallones S. To advance epidemiology.Annu Rev Public Health. 1980;1:69-82.

14. Cameron D, Jones IG. John Snow, theBroad Street pump and modem epidemiol-ogy. IntJEpidemiol. 1983;12:393-396.

15. Krieger N. Epidemiology and the web ofcausation: has anyone seen the spider? SocSciMed. 1994;39:889-903.

16. McKinlay JB. Towards appropriate levelsof analysis, research methods and healthpublic policy. Presented at the Interna-tional Symposium on Quality of Life andHealth: Theoretical and MethodologicalConsiderations; May 25-27, 1994; Berlin,Germany.

17. McMichael AJ. "Molecular epidemiol-ogy." New pathway or new travellingcompanion? Am J Epidemiol. 1994;140:1-11.

18. Pearce N, san Jose S, Boffetta P, et al.Limitations of biomarkers of exposure incancer epidemiology. Epidemiology. 1995;6:190-194.

19. Vandenbroucke JP. New public health andold rhetoric. BMJ. 1994;308:994-995.

20. Thatcher M. Womans Own. October 1987.21. Wing S. Limits of epidemiology. Med

Global Survival. 1994;1:74-86.22. Black D. Deprivation and health. BMJ.

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.New York, NY: UNICEF, 1993.

24. Pearce N. Economic policy and health inThe Year of the Family. NZ Med J.1994;107:379-381.

25. Wilkinson R. National mortality rates: theimpact of inequality. Am J Public Health.1992;82:1082-1084.

26. Sen A. Levels ofPoverty: Policy and Chbac:e.Washington, DC: The World Bank; WorldBank StaffWorking Paper No. 401. 1980.

27. Tesh SN. HiddenArguments: Political Ideol-ogy and Disease Prevention Policy. London,England: Rutgers; 1988.

28. Loomis D, Wing S. Is molecular epidemiol-ogy a germ theory for the end of thetwentieth century? IntJEpidemiol. 1991;19:1-3.

29. Jacobson B. The Lady Killers: Why SmokingIs a Feminist Issue. London, England: PlutoPress; 1981.

30. Power C, Manor 0, Fox J. Health andClass: The Early Years. London, England:Chapman Hall; 1991.

31. Pearce N, Davis PB, Smith AH, Foster FH.Social class, ethnic group, and male mortal-ity in New Zealand, 1974-8. J EpidemiolCommunity Health. 1985;39:9-14.

32. Reid P, Pouwhare R. Te Taonga maiTawhiti: The Gift from a Distant Place.Auckland, New Zealand: Niho Taniha;1991.

33. Barry M. The influence of the U.S. tobaccoindustry on the health, economy, andenvironment of developing countries. NEnglJMed. 1991;324:917-920.

34. Tominaga S. Spread of smoking to thedeveloping countries. In: Zaridze D, PetoR, eds. Tobacco: A Major InternationalHealth Hazard. Lyon, France: InternationalAgency for Research on Cancer; 1986:125-133.

35. Pearce NE, Matos E, Vainio H, Boffetta P,Kogevinas M, eds. Occupational Cancer inDeveloping Countries. Lyon, France: Inter-

national Agency for Research on Cancer;1994.

36. Susser M. Causal Thinking in the HealthSciences. New York, NY: Oxford Univer-sity Press; 1973.

37. Potter JD. Reconciling the epidemiology,physiology, and molecular biology of coloncancer.JAMA. 1992;268:1573-1577.

38. Syme SL, Berkman LF. Social class, suscep-tibility and sickness. Am J Epidemiol.1976;104:1-8.

39. Cassel J. The contribution of the socialenvironment to host resistance. Am JEpidemiol. 1976;104:107-123.

40. Millard AV. A causal model of high ratesofchild mortality. Soc SciMed. 1994;38:253-268.

41. Rose S, Lewontin RC, Kamin LJ. Not inOur Genes: Biology, Ideology and HumanNature. London, England: Penguin Books;1990.

42. Miettinen OS. Theoretical Epidemiology:Principles of Occurrence Research. NewYork, NY: Wiley; 1985.

43. Smith A. The epidemiological basis ofcommunity medicine. In: Smith A, ed.RecentAdvances in Community Medicine 3.Edinburgh, Scotland: Churchill Living-stone; 1985:1-10.

44. Firth WJ. Chaos-predicting the unpredict-able. BMJ. 1991;303:1565-1568.

45. Levins R, Lewontin R. The DialecticalBiologist. Cambridge, Mass: Harvard Uni-versity Press; 1985.

46. Armstrong D. An Outline of Sociology asApplied to Medicine. Bristol, England:Wright; 1980.

47. Bhaskar R. Reclaiming Reality. London,England: Verso; 1989.

48. Jones K, Moon G. Health, Disease andSociety:A Critical Medical Geography. Lon-don, England: Routledge & Kegan Paul;1987.

49. Last JM, ed.A Dictionary ofEpidemiology.New York, NY: Oxford University Press;1988.

50. Anon. Population health looking up-stream. Lancet. 1994;343:429-430. Edito-rial.

51. McKinlay JB. The promotion of healththrough planned sociopolitical change: chal-lenges for research and policy. Soc Sci Med.1993;36:109-117.

52. Lilienfeld A, Lilienfeld DE. Epidemiologyand the public health movement. J PublcHealth Policy. 1982;3:140-149.

53. Adler NE, Boyce WT, Chesney MA,Folkman S, Syme SL. Socioeconomic in-equalities in health: no easy solution.JAMA. 1993;269:3140-3145.

54. Terris M. The epidemiologic tradition.Public Health Rep. 1979;94:204.

55. Kunitz S. Disease and Social Diversity. NewYork, NY: Oxford University Press; 1994.

56. Susser M, Watson W, Hopper K. Sociok*jin Medicine. 3rd ed. New York, NY: OxfordUniversity Press; 1985.

57. Connelly J. Social sciences provide aframework for epidemiology. BMJ. 1994;308:1569. Letter.

58. Rose G. The Strategy ofPreventive Medicine.Oxford, England: Oxford University Press;1992.

59. Koopman JS, Longini IM. The ecologicaleffects of individual exposures and nonlin-ear disease dynamics in populations.Am JPublic Health. 1994;84:836-842.

60. Greenland S, Robins J. Ecologic studies-biases, misconceptions, and counterex-amples.AmJEpidemiol. 1994;139:747-760.

61. Susser M. The logic in ecological: II. thelogic of design.Am JPubic Health. 1994;84:830-835.

62. Wing S. Social inequalities in the decline ofcoronary mortality. Am J Public Health.1988;78:1415-1416.

63. Dean K. Creating a new knowledge basefor the new public health. J EpidemiolCommunity Health. 1994;48:217-219.

64. McMichael AJ. Persons, populations andplanets: a full world brings epidemiologyfull circle. Epidemiology. In press.

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