Web Causation Spider

8/13/2019 Web Causation Spider

1/17

Pergamon 0277-9536(93)E0090-2SW. SC;. Met/. Vol. 39. No. 7. 8X7-903,p. 1994Copyri ght 8~: 1994 Elsevier Science LtdPrinted i n Great Bri tain. All rights reserved0277-9536/94 7.00 + 0.00

EPIDEMIOLOGY AND THE WEB OF CAUSATION:HAS ANYONE SEEN THE SPIDER?

NANCY KRIEGERDivision of Research, Kaiser Foundation Research Institute. 3451 Piedmont Avenue, Oakland.

CA 94611, U.S.A.Abstract-Multiple causation is the canon of contemporary epidemiology, and its metaphor and modelis the web of causation. First articulated in a 1960 U.S. epidemiology textbook, the web remains awidely accepted but poorly elaborated model, reflecting in part the contemporary stress on epidemiologicmethods over epidemiologic theories of disease causation. This essay discusses the origins, features, andproblems of the web, including its hidden reliance upon the framework of biomedical individualism toguide the choice of factors incorporated in the web. Posing the question of the whereabouts of theputative spider, the author examines several contemporary approaches to epidemiologic theory, includingthose which stress biological evolution and adaptation and those which emphasize the social productionof disease. To better integrate biologic and social understandings of current and changing populationpatterns of health and disease, the essay proposes an ecosocial framework for developing epidemiologictheory. Features of this alternative approach are discussed, a preliminary image is offered, and debate isencouragedKey Kzords+cosocial, health inequalities, philosphy of science, social class. social epidemiology

, And right spang in the middle of the web there werethe words Some Pig. The words were woven right into theweb. They were actually part of the web, Edith. I know,because I have been down there and seen them. I t says,Some Pig, just as clear as clear can be. There can be nomistake about it. A miracle has happened and a sign hasoccurred here on earth, right on our farm, and we have noordinary pig.Well, said Mrs Zuckerman, it seems to me youre a littleoff. It seems to me we have no ordinary spider

E. B. WhiteCharlofres Web [I p. SO]

Multiple causation is the canon of contemporaryepidemiology, and its metaphor and model is the webof causation. Expressed through the notion of mul-tifactorial etiology [24] and embedded in the stat-istical techniques of multivariate analysis [3-71, thebelief that population patterns of health and diseasecan be explained by a complex web of numerousinterconnected risk and protective factors has becomeone of this disciplines central concepts [3,4]. Equallyentrenched is the corollary that epidemiologys powerto improve the publics health rests upon its abilityto identify-and predict the results of breaking-selected strands of this causal web [2-4].

The widespread adoption of a multicausal frame-work would suggest the existence of a well-developedbody of epidemiologic literature that analyzes itsessential features and explores its implications forcausal reasoning about population patterns of healthand disease. Instead, theoretical work in epidemiol-ogy--especially in the United States-has recentlyfocused on methodologic issu s [3-81 and debatesabout causation have chiefly concerned the nature

and validity of causal inference [5,8-IO]. These,however, are problems for any science, not justepidemiology [I 1, 121. By contrast, relatively littlework has been devoted to developing the conceptsand framework of what might be termed epidemio-logic theory, i.e. explanations of the current andchanging health status of human societies [4, 13-201.Even less has addressed the fundamental questionposed by the webs suggestive imagery: who or whatis the spider responsible for its array of factors?[17, p. 2441.

The paucity of critical reflection is not just anacademic point, but cuts to the core of epidemiologyas a science and profession [13-211. As noted byseveral senior epidemiologists in the U.S.-most es-pecially Susser [3, 181, Terris [4, 14, 191, Stallones [15],Lilienfeld and Lilienfeld [20] and Kuller [21], modernepidemiology often seems more concerned with intri-cately modeling complex relationships among riskfactors than with understanding their origins andimplications for public health. Reflecting this trend,graduate students in epidemiology are far more likelyto be taught about study design and data analysisthan they are about how to generate epidemiologichypotheses about the societal dynamics of health anddisease. Most current U.S. epidemiologic textbooks,for example, focus almost exclusively on methodo-logic issues and devote little, if any, space to explain-ing the different theories of disease causation andetiologic concepts that help epidemiologists formu-late hypotheses in the first place (e.g. time, place, andperson, mode of transmission, herd immunity,environment, and lifestyle); even fewer discuss the

887


2/17

N NCY KRIEGER

origin of these ideas or the history of epidemiology research embodies particular ways of seeing as well asitself (Table I). knowing the world, with the express intent of analyz-

Although the importance of learning a sciences ing and improving the publics health [24, 13-211.methodology is obvious, epidemiology is more than Because the insights of any science-including epi-a mere amalgam of methods and study designs. demiology-are as much, if not more, dependentWhether explicitly articulated or not, epidemiologic upon its concepts and theories than upon its specific

Table I. Survey of U.S. epidemiologic textbooks and anthologies published since 1970: content on epidemiologic history and theory. anddiagram of web of causation

Text

Percentage of pageson epidemiologic

Total - --_ DiagramDaees Historv Theorv of Web

MacMahon B. and Pugh T. F. Epidemrology: Principles andMethods. Little, Brown, Boston, 1970.Fox .I. P., Hall C. E. and Elveback L. R. Epidemiology: Man andDi.\rax Macmillan, New York, 1970.Susser M. Causal Thinking in lhe Healrh Sciences:Concepts and Strategies o/ Epidemiology .Oxford University Press, New York. 1973.Mausner J. S. and Bahn A. K. Epidemiology: an InlroducroryText. Saunders, Philadelphia, 1974.Friedman G. Primer of Epidrnzioiogy.McGraw-Hill. New York, 1974.White K. L. and Henderson M. (Eds) Epidemiology as a FundamentalScienw: ils uses in Health Services Pluming,AdminDtmtion, und Go/m/ion. Oxford UniversityPress, New York, 1976.Lilienfeld A. and Lilienfeld D. Foundations of Epidemiology.Oxford University Press, New York, 1980.Klembaum D. G., Kupper L. L. and Morgenstern H. (Eds) EpidemiologicResearch: Principles and Quanrmticv Methods.Lifetime Learning Publications, Belmont, CA, 1982.Schlesselman J. Case-Conlrol Studies: Design, Conducr,Analysis. Oxford Umverslty Press, New York, 1982.Kahn H. A. An Inrroductmn to Epidemiologic Methods.Oxford University Press, New York, 1983.Miettinen 0. S. Theoretical Epidemiology: Principles ofOccurrmce Rexarch in Medicine. Wiley, New York, 1985.F stein A. R. Clinicul Epidemiology: the Archirecrureof C/iniu/ Reseorrh. Saunders, Philadelphia, 1985Weiss N. Clinical Epidrmiolog)~: The Sfudy of heOurcontr of l lness xford University Press,New York, 1986.Rothmam K. Modem Epidemiology. Little, Brown,Boston, 1986.Kelsey J.. Thompson W. D. and Evans A. S. Methods in ObservationalEpidemiology. Oxford University Press, New York, 1986.Hennekens C. H. and Burmg J. E. Epidemiology in Medicine.Little. Brown, Boston, 1987.Abramson J. H. Making Sense of Doto: (I Self-instrucr ionMnrmnl on the Interpreta tion of Epidemiologic Data.Oxford University Press, New York, 1988.

Winklestem W. Jr. French F. E. and Lane J. M. (Eds) BashReadings in Epidemiology. MSS Educatmnal Pub. Co.,New York. 1970.Greenland S. (Ed.) Erolurim of Epidemiologic Ideas:Annatured Readings on Concepts and Methods. EpidemiologyResources, Inc., Chestnut Hill. MA, 1987.Buck C., Llopis A., Najera E. and Terris M. (Eds) The Challengeq/ Epidemiology: hues und e/erred Readings. PanAmerican Health Organization, Washington, DC, 1988.Rothman K. (Ed.) Cuusul Inference. Epidemiology Resources,Chestnut Hill, MA, 1988.

302 0.0

339 3.5

I81 12.2377 0.0

230 0.0

235 0.9

375 6.1

529 0.0

354 0.6

166

359

XI2

I44

358 I.7 0.0

366 0.0 7.43x3

326 0.0 0.6

193

190

989

207 0.0 0.0

0.0

0.0

I.1

0.0

2.6

13.9

7.9

14.8

Il.6 +44.x

x.34.0 +

0.9 +

0.9

5.1

I.1

0.0

0.0

1.4

1.2

0.0

3.9

27.X

0.0

24.9

Epidemmlogic theory: defined as explicit discussion of theories of disease causation and/or epidemiologic concepts (e.g. time, place, person).


3/17


4/17

890 N NCY KRIEGER

with more complex models of host, agent, andenvironment. Others, under the banner of socialmedicine, argued in favor of explicitly examining thesocial determinants of health [28, 32, 331. And, aspart of meeting these challenges, epidemiologists andbiostatisticians refined and developed new study de-signs and statistical techniques, including the ap-proximation of prospectively-determined relativerisks with odds ratios derived from case-controlstudies [5, 341.

Other events in the 1950s also left their mark uponthe development of epidemiologic theory. The discov-ery of the double-helix structure of DNA by Watson,Crick and Franklin in 1953 [35] augured an explosionin biological knowledge, and raised new hopes ofimproving understanding of gene-environmentinteractions. That same year, Nordling developed thefirst mathematical multistage-mutation model of car-cinogenesis to explain the shape of cancer incidencecurves [36], thereby extending epidemiologic attemptsto link the micromechanisms of disease causationwith population patterns of disease occurrence.

Epidemiologic thought was also stimulated by thegrowing use of computers to store and analyze largedata sets, especially using multivariate techniques[3, 371. The first massive civilian demonstration ofthis new possibility involved the 1950 U.S. census andemployed the ENIAC computer. which had beenbuilt in the 1940s at the request of the U.S. WarDepartment in World War II [38]. In 1956 thecomputer language FORTRAN was developed [39],and by the end of the decade, the potential ofcomputers for cancer registries and epidemiologicstudies was clearly understood [37].

It was another aftermath of World War II, how-ever, that perhaps most strongly shaped the sub-sequent U.S. academic discourse about diseasecausation: the Cold War and its domestic corollary ofMcCarthyism. In a period when discussion of socialclass and social inequality was tantamount to heresy(even in the social sciences) [4042], and when earlycivil rights activists were branded as subversive (e.g.supporters of the 1955 Montgomery bus boycott)[40,43], it is not surprising that epidemiologists (likeother academics) generally eschewed dangerousspeculation about the social determinants of health.Instead, most pursued research based upon morebiomedical and individually-oriented theories of dis-ease causation, in which population risk was thoughtto reflect the sum of individuals risks, as mediated bytheir lifestyles and genetic predisposition to disease[3,44, 451.

It was in this context that MacMahon et al.introduced the concept of the web of causation[2, p. 181. They did so in reaction to the then prevalentnotion of chains of causation, which they arguedfailed to take into account: (1) the complex geneal-ogy of antecedents of each component in the chain[2. p. 181, and (2) how the genealogies of diversefactors or outcomes might overlap, creating a variety

of indirect as well as direct associations. Expresslychallenging the still-pervasive tendency of epidemiol-ogists to think in terms of single agents causingdiscrete diseases, the provocative metaphor andmodel of the web invited epidemiologists to embracea more sophisticated view of causality.

Conceptually, the metaphor evoked the powerfulimage of a spiders web, an elegantly linked networkof delicate strands, the multiple intersections repre-senting specific risk factors or outcomes, and thestrands symbolizing diverse causal pathways. It en-couraged epidemiologists to look for multiple causesand multiple effects, to consider interaction, and toidentify the many-as opposed to singular-routesby which disease could be prevented. With thismetaphor in the background, epidemiologists couldalso treat the web as a model to delineate theetiology of, and guide research about, specific healthproblems. More profoundly, the web tapped into anintuitive sense of interconnection, one long a part ofmany philosphical traditions [46 48] and, during the1950s increasingly incorporated into other scientificdisciplines, especially cybernetics and ecology [48,49].

To illustrate their view, MacMahon et ul. dia-grammed some components of the relationship be-tween two etiologically distinct diseases, syphilis andhepatitis, whose independent chains had no logicalreason to intersect (Fig. I). Asserting that the wholegenealogy might be thought of more appropriately asa web, which in its complexity and origins lies quitebeyond our understanding [2, p. 181. they proffereda picture that simultaneously detailed hou the hepa-titis virus might get into syringe needles used to treatsyphilis patients, producing and outbreak of jaundice(icterus), and yet left to the readers imagination(as indicated by the suggestive dots trailing off theedge of the page) the determinants of other factors,e.g. the economic status of patients, human frailty,the public provision of treatment facilities, theoccurrence of syphilis, and knowledge of therapy.

Using this model, MacMahon et al. drew severalimportant inferences about prevention and researchthat remain part of epidemiologic thinking to thisday [4, 8, 15, 181. Arguing that to effect preventivemeasures, it is not necessary to understand causalmechanisms in their entirety [2, p. 181, they statedthat [elven knowledge of one small component mayallow some degree of prevention, since whereverthe chain is broken the disease will be prevented[2, p. 181.With this image in mind of cutting strands ratherthan attempting to identify and alter the source(s) ofthe web, MacMahon et al urged epidemiologists toabandon semantic exercises aimed at hierarchicclassification of causes [2, p. 201. They insteadshould seek out the necessary (albeit rarely suffi-cient) causes most amenable to practical interven-tions and nearest (in terms of the webs configuration)to the specified outcome. For example, to preventinfectious diseases, they observed that it is a good


5/17

Epidemiology and the web of causation 891deal easier to control human water supply and toeradicate insect vectors than to breed geneticallyresistant populations [2, pp. 20-211. Noting, how-ever, that the alteration of any cause may beexpected to have multiple effects besides the oneintended, MacMahon et al. cautioned that theseside effects must be acceptable 12, p. 221although they never specified acceptable to whom.

The web of causation quickly gained widespreadrecognition as a useful concept to help orient researchregarding the multifactorial etiology of disease[3,4, 6, 15, 501. Even so, the web subsequently hasappeared in only a handful of other epidemiologictexts [51-53]-in essentially the same form, and with-out any real elaboration (Table 1). The reason for thisomission is not the webs lack of utility for epidemi-ologic thought. Instead, notions of multiple causationand multivariate analysis are so commonplace and soembedded in modern epidemiologic reasoning thatthey hardly merit discussion as a model or as anapproach to understanding disease [3, 541. The webas such is a given, and what garners attention are theanalytic problems posed by the intricate concaten-ations of its component parts [3-81.

Guided by a multicausal view, for example, epi-demiologists have greatly developed their under-standing of such critical phenomena as confoundingand effect modification [3, 5,8, 501. Among these,Rothman has called for greater attention to therelationships between necessary and componentcauses [55]. To illustrate this point, in 1976 hediagrammed several hypothetical combinations ofcausal components sufficient to produce a hypotheti-cal disease; whereas some component causes (necess-ary but not sufficient) belonged to more than onecausal pie, others (the necessary causes) belong to allthe pies (Fig. 2) [55]. Using this imagery, Rothmanhas shown how the strenth of an association betweena component cause and a given outcome depends notsimply on their specific relationship but also on theprevalence of the other component causes required tobring about the specified change [8,55]-an insightnot obvious at the time the web was first proposed.

SufficientcauseISufficientcauseII

SufficientCauseIII

Fig. 2. Rothmans conceptual scheme for the causes ofa hypothetical disease [55, p. 5891. Text accompanyingfigure: This disease (effect) has three sufficient causalcomplexes. each having five component causes. In thisscheme A is a necessary cause, since it appears as a memberof each sufficient cause. On the other hand, the componentcauses B. C, and F, which each appear in more than onesufficient cause, are not necessary causes, because they fail toappear in all three sufficient causes [55, p. 5881. Reprinted

with permission from the publisher.

As critical as these developments are, it is essentialto note what these views of multifactorial etiologyomit: discussion of the origins-as opposed to inter-actions-of the multiple causes. MacMahon et al.,for example, never explained i y they selected thecomponents that appear in their web and leftothers out, nor did they offer any specific advice asto how others might elucidate the elements of otherwebs. Having decreed the origins of the web out-side the bounds of epidemiologic enquiry, theseauthors never invoked-and essentially proscribed-the imagery of the spider. Similarly, in the case ofRothmans pies, the cook is notably absent.

What could account for the construction of aspiderless web? One answer is that the web wasnever intended to be a theory. It was not elaboratedto provide explanations of causal links, but insteadwas developed to enhance epidemiologists ability todepict and study complex interrelationships betweenspecljied risk factors and diseases.

This answer, however, is not sufficient: models donot exist independently of theories [l 1,22-251. The-ories attempt to explain i y phenomena exist and areinterrelated. By contrast, models attempt to portray/ZOMJhese connections occur and are always con-structed with elements and relationships specifiedby particular theories [l 1,22-251. In ecology, forexample, diverse predator/prey models may illustratehon the populations of two species covary, but itis evolutionary theory that directs researchers toconsider why this relationship exists and how itdeveloped, through reference to such phenomena andconcepts as consumption, reproduction, genetic vari-ation, natural selection, and historical contingency[24]. The absence of any discussion of the theoryshaping the model of the web is thus extremelyproblematic.

If the web is a model, it follows that it must bebased on or reflect some theory. Closer inspection ofthe elements of the web in turn discloses the theor-etical orientation woven into its very fabric. Onecardinal feature is that, despite claims about beinginclusive and non-hierarchic in its depiction of riskfactors, the web in fact employs a type of weightingthat levels all distinctions [54, 561. In MacMahonet al .s web, treatment in clinics and the economyoccupy the same level, and rate the same kind of boxas injection of foreign serum and epidemic hepatitisin community (Fig. 1). The web thus invevitablyfocuses attention on those risk factors closest to theoutcome under investigation, and these in turntypically translate to the direct biologic causes ofdisease in individuals and/or to lifestyles and otherrisk factors that allegedly can be addressed atthe individual level through educatton or medicalintervention.

Another important aspect of the web is that itdoes not differentiate between determinants of dis-ease in individuals and in populations. Stated anotherway, it fails to distinguish between what Rose, a


6/17

892 N NCY KRIEGER

noted English epidemiologist, has aptly termed thecauses of cases vs the causes of incidence [16, 571.As Rose has shown, these two sets of causes are notnecessarily the same and require difference researchquestions: asking why do some individuals havehypertension? is not equivalent to enquiring whydo some populations have much hypertension, whilstin others it is rare? (Fig. 4). The former emphasizesindividual susceptibility and interventions aimed athigh risk individuals, whereas the latter highlightspopulation exposures and the need to shift the distri-bution of disease in the entire population (whichwill always have its outliers) to a healthier state [57].And, by excluding any sense of history or origins, theweb sans spider discourages epidemiologists fromconsidering why population patterns of health anddisease exist and persist or change over time.

As noted by other critics [54, 561, these aspects ofthe web help identify its underlying theoreticalframework: that of biomedical individualism [44].This framework, often referred to as the biomedicalmodel, has three key features. It emphasizes biologi-cal determinants of disease amenable to interventionthrough the health care system, considers socialdeterminants of disease to be at best secondary(if not irrelevant), and views populations simplyas the sum of individuals and population patternsof disease as simply reflective of individual cases[44,45, 54, 56, 58-631. In this view, disease in popu-lations is reduced to a question of disease in individ-uals, which in turn is reduced to a question ofbiological malfunctioning. This biologic substrate,divorced from its social context, thus becomes theoptimal locale for interventions, which chiefly aremedical in nature [44,45, 541. It is essential to stressthat the web did not challenge this basic biomedicaland individualistic orientation to disease causation.What it opposed were simplistic interpretations ofthe doctrine of specific etiology, which holds thatsingle agents uniquely cause specific diseases[4,6,30,64]. The novel feature of the web was thatit emphasized the need to consider, simultaneously,how diverse aspects of the host, agent and environ-ment were implicated in the multifactorial etiology ofdisease.

THE INFLUENCE OF EPIDEMIOLOGIC THEORY UPONEPIDEMIOLOGIC THOUGHTThe hidden role of etiologic theory in the web

stands in stark contrast to the open debates aboutepidemiologic theory that frequently occurred duringearlier periods of epidemiologys development as adiscipline. Emerging in the early 1800s-alongsidethe rise of cell theory, sociology, and political econ-omy [24, 38,42, 65, 66]+epidemiology from the out-set was as much a testing ground for different theoriesabout the influence of biology and society on healthas it was a practical enterprise geared toward describ-ing population patterns of disease and influencing

health policy [14,65,66]. Much of the research of the19th century epidemiologists in both Europe and theUnited States, for example, was motivated or shapedby two central (and connected) debates: (I) whethermiasma or contagion was the principal cause ofepidemic disease [14,65-701, and (2) whether povertywas the cause or result of poor health and immorality[18, 65,66, 69, 701. Other important polemics focusedon whether womens ill health was a consequenceof too much or too little education, employment, oractivity [71], and, in the United States, on whetherthe poor health status of Black men and womenwas the result of slavery or their allegedly inferiorconstitutions [62,72]. In both cases, these disputesextended beyond narrow arguments about specificfactors associated with particular diseases to funda-mental debates about the types of causal factors-ranging from the biologic to the social-that couldlegitimately be invoked to explain population patternsof disease.

Epidemiology in the early 20th century also wascharacterized by contention between several schoolsof thought. Like their 19th century counterparts,these schools differed in their assessment of both thenumber and realm of etiologic factors and the typesof concepts required to understand these factors. Oneschool emphasized single agents (usual microbial) asthe cause of disease, as exemplified by CharlesChapins influential text. The Sources and Modes sfInfection [73, 741. Others, like Major Greenwood,used the imagery of seed, soil, and some type ofhusbandry [75] to argue that the host and environ-ment, not just the agent, needed to be taken intoaccount; this type of reasoning was clearly evident inWade Hampton Frosts trenchant analyses of epi-demic disease [76]. A third looked to basic features ofthe economy for clues about mass patterns of disease,as exemplified by Goldberger, Wheeler, and Syden-strickers classic studies on the conjoint social andbiologic etiology of pellagra [77-791. In addition toproving pellagra was a dietary deficiency disease, theyelucidated the link between pellagras prevalence inthe South and the regions cash-crop economy:whenever the cotton market crashed, the subsequentprevalence of pellagra inevitably rose.

To some degree, the prominence of one overanother of these approaches paralleled gains andsetbacks in the knowledge of related fields in thenatural and social sciences, e.g. medicine, virology,toxicology, genetics, anthropology, and politicaleconomy. As important-if not more so-were insti-tutional decisions regarding what types of researchshould be supported [28, 54, 741 and how publichealth researchers should be trained [28,74, SO]. Theinfluential Welch-Rose report of 1915 that led to thefirst U.S. school of public health being founded atJohns Hopkins University, for example, stronglyfavored its biomedical approach, as compared to themore sociological program proposed at the time byColumbia University [74]. The devastating 1918


7/17

Molecular andsubmolecular particles

Epidemiology and the web of causation 893Cells Organs Individuals Families Communities Societies

EpidemiologyClinical research

1 Pathology, PhysiologyI Cell biology IMolecular biology I

Fig. 3. Stallones figure depicting biomedical disciplines along a scale of biological organization [15, p. 701. Reprintedwith permission from the publisher.

influenza pandemic and growing concern about dis-eases like cancer, however, impelled noted epidemiol-ogists like Greenwood [75] and Sydenstricker [80] toquestion exclusive reliance upon the germ theory asa framework for understanding health. They arguedinstead that the study and prevention of crowd ormass diseases-whether infectious or chronic-would require social as well as biologic reasoning.

In light of this history, it is not surprising that thepresent-day inattention to epidemiologic theory andpreoccupation with epidemiologic methods has pro-voked several different types of concerned responseswithin the discipline. As noted earlier, several seniorU.S. epidemiologists have called for renewed atten-tion to epidemiologic theory [3,4, 18-21,8 l-841.Some, most notably Cassel [82,83] and Syme [84],have sought to develop psychosocial theories ofdisease causation to supplement or counter morebiomedical approaches. Others, like Stallones [151,have attempted to bolster the theoretical underpin-nings of the multifactorial framework by depictingthe range of factors that epidemiologists shouldconsider when constructing causal webs, rangingfrom the subcellular to the societal and the micro-scopic to the macrosystem (Fig. 3).

Criticizing these approaches, Vandenbroucke et al.[85,86] have warned against epidemiologists ten-dency of invoking a vaguely-defined but often infi-nitely complex environment to explain variation indisease rates. They contend that this approach, com-bined with epidemiologists widespread practice oftreating disease mechanisms as a black box, willsoon render the discipline incapable of making mean-ingful contributions to understanding disease etiol-ogy. Claiming that most environmental theories arebut 20th-century versions of outdated miasmaticarguments, Vandenbroucke has strongly argued thatthe future of epidemiology lies in the fast movingsearch for genetic markers [85], as brought about byspectacular advances in biotechnology and epitom-ized by the Human Genome Project [87]. Similarly,Kuller has asserted that claims about multifactorialetiology may be more a sign of ignorance thaninsight, and has urged epidemiologists to rethink aparsimonious approach to the etiology of disease andagain to consider that each disease is caused by oneagent and to evaluate confounding in terms of host,

agent, and modes of transmission [81, p. 3741. Oth-ers, however, have argued that a return to singleagent theories (and especially those favoring exclu-sively genetic explanations) will only serve to recreatethe errors made in the early period of widespreadenthusiasm for the germ theory of disease, whichresulted in artificially narrowing the scope of etiologicexplanations considered and public health inter-ventions employed and which may also have retardedthe development of research on diseases beyondits ken (e.g. cancer and cardiovascular disease)[4, 18, 54,88-911.

To date, however, few epidemiologists have at-tempted to develop systematic and theoreticallyrigorous alternatives to the web of causation. Onesuch effort has emphasized the role of evolution,adaptation, and the man-made environment indetermining current and historically-changing popu-lation patterns of health and disease [64,92-951. Thistheory is best summarized in a recent book entitledOn the Or igi ns of Human Di sease [92], written byThomas McKeown, a well-known British epidemiol-ogist.

(1)

(2)

At its cores are three claims:for most of our history as a species, humanshave lived under, and are biologically bestadapted to, Stone Age conditions, exemplifiedby the hunter/gatherer mode of existence, es-pecially its diet and level of physical activitythe transition to agriculture (ca 10,000 yearsago) brought about vast changes in humandisease, due to the concentration of largernumbers of people in more permanent settle-ments and also to the domestication of ani-mals, both of which (in conjunction with

30 r 009 0 Kenyan nomads0 London civil servants

60 80 100 120 140 160 180 200Systolic B.P.(mm Hg)

Fig. 4. Roses figure comparing distribution of systolicblood pressure in middle-aged men in two populations[57, p. 331. Reprinted with permission from the publisher.


8/17

894 N NCY KRIEGERperiodic famines) enabled numerous infectiousdiseases to flourish

(3) the transition to industrial society (ca 300 yearsago) led to a decline in infectious diseases andto a rise in non-communicable diseases, dueto improved nutrition and increased exposureto conditions for which we are geneticallyill-equipped [92, p. 1531, e.g. refined diets,cigarette smoke, sedentary lifestyles, indus-trial toxins, and other man-made hazards

Building on these premises, McKeown argues thathuman diseases can best be understood and pre-vented if divided into three fundamental etiologicgroups, defined by disease origins rather than diseasemechanisms or afflicted organs [92, p. 2221. These are:(1) prenatal diseases (determined at fertilizationor during the course of pregnancy) [92, p. 991, (2)diseases of poverty (due to deficiencies and hazardslinked to the natural environment, e.g. poor nutri-tion and poor sanitation) [92, p. 1381, and (3) dis-eases of affluence (due to maladaptation to changesin living conditions brought about by industrializ-ation) [92, p. 1401, which he nonetheless concedesnow disproportionately affect the poor in both indus-trialized and impoverished countries [92, p. 941.

McKeowns schema is thus a far cry fromMacMahon et al.s [2] web of assorted risk factors.Focusing on the health effects of agriculture, urbaniz-ation, and industrialization, McKeown simul-taneously directs us to consider broad populationshifts in disease occurrence and whether the proxi-mate causes of a particular disease operate before orafter birth and if they exert their effects via deprivingthe body of basic biological requirements or byexposing the body to types or levels of substances,hazards, or conditions not anticipated through natu-ral selection. Even so, McKeowns theory-like thatunderlying the web-fails to address an importantset of etiologic questions with important implicationsfor disease prevention policies: namely, whether allsectors of society are equally responsible for shapingthe human environment (for good or for bad), andwhy it is that specific sectors have not equallybenefited from or been harmed by these changes [96].Instead, the responsible parties are absent actors in ascenario where natural selection plays the definingrole.

In direct contrast, the other school seeking todevelop epidemiologic theory has focused on thoseissues downplayed by McKeown: specifically,the social and political determinants of health.Building on the work of 19th and earlier 20thcentury public health researchers and advocates[4, 14,26-33,65, 66, 72, 77-801, these epidemiologists[3,4, 15-19, 56,62,97-125]-along with other publichealth scholars and activists [58-61, 74, 78, 126-1301-have sought to develop and test the theory thatpast and present population patterns of health anddisease primarily stem from the social organization

of society, and especially its economic and socialactivites and inequalities.

Promoting such concepts as the political economyof health [58] and the social production of disease[54, 56, 59-62,97,98, 104, 105, 123-1301, proponentsof this alternative view reject uniform assertionsabout mans maladaptation [92, p. 951. Instead,they typically seek etiologic clues through comparingthe health status of social groups that differentiallybenefit from or are harmed by the status quo, such asemployers/employees, men/women, whites/people ofcolor, heterosexuals/homosexuals, and inhabitantsof economically developed/underdeveloped regions.The essential claim is that understanding patterns ofhealth and disease among persons in these groupsrequires viewing these patterns as the consequence ofthe social relationships between the specified groups,with these relationships expressed through peopleseveryday living and working conditions, includingdaily interactions with others [98].

This perspective fundamentally differs from astance which sees group patterns simply as the sumof individual traits and choices. It instead asks howindividuals membership in a societys historically-forged constituent groups shapes their particularhealth status, and how the health status of thesegroups in turn reflects their position within the largersocietys social structure [44]. It further implies thatchanging these population patterns requires explicitlystudying and addressing their political, economic,and ideologic determinants. According to this view,social inequalities in health are the defining problemof the discipline of epidemiology. The litmus test ofany epidemiologic theory of disease causation thusis whether it can explain past and present socialinequalities in health.

Yet, despite their emphasis on historical determi-nants of disease, few advocates of this analytic trendhave extended the range of historical processesto include, as McKeown has done, the influenceof biologic evolution, growth, and development onhuman disease processes [98]. Illustrating one suchattempt is Fig. 5, which endeavors to contrasta biomedical approach to understanding breastcancer etiology to one that incorporates develop-mental biology within the framework of thesocial production of disease [122]. The first model(Fig. 5a) displays the current predominant conceptu-alization, which stresses the role of hormonaland genetic factors. The second model (Fig. 5b)broadens the framework of exposure and suscepti-bility to consider how social factors simultaneously:(1) mediate the growth and development of breasttissue (by affecting the timing of reproductive events),and (2) determine the type and timing of exposure toexogenous agents that may contribute to breast can-cer risk.As suggested by the rudimentary nature ofFig. 5(b), however, epidemiology still lacks a theor-etical framework that truly integrates social and


9/17

(a)Social factors- Marital status- High SES

I-----Dietary fatbL

Reproductivefactors

c

- Early age atmenarche

- Late age at firstfull-termpregnancy

- Late age atmenopause

? Oral contraceptivesAbbreviation used:

SES = socioeconomic status

Technological level

Family history/genetiy factors

? Benign breast,f disease -1+ /A 4REPRODUCTIVE Breast BREAST

HORMONE LEVELS - tissue - CANCERtRadiation

Social relations

Social classes, plus race,gender and age divisions

*SOCIALLY-MEDIATED RISK FACTORS:

Type of and age atexposure to exogenous events

* Reproductivefactors

* Additionalfactors

carcinogens-- Workplace- Community- Home- Iatrogenic- Dietary

Age at firstintercourseAge at first useof birth controlAge at FETPAge at FFTPNumber, sequence,and time between ETP

Age at menarche - DietaryAge at first patternspost-partumweaningAge at menopauseAge at postmenopausalinvolution

and FTPBreast tissue

proliferation/differentiation

Exposure + Susceptibility

Altered breast tissue

Abbreviations used:BREASTCANCERI-___-----_________---- ? benign breast disease

ETP = early-terminated pregnancyFETP = first early-terminated pregnancyFFTP = first full-term pregnancyFTP = full-term pregnancy

Fig. 5. Contrast of a biomedical vs social production of disease model of breast cancer pathogenesis[122, pp. 207,209]. Reprinted with permission from the publisher.

SSM 9/7-a


10/17

896 NANCY KRIEGERbiologic understandings of health, disease, and well-being [98]. In other words, what I would call anecosocial epidemiologic theory-one that embracespopulation-level thinking and rejects the underlyingassumptions of biomedical individualism without dis-carding biology-has yet to be conceived.

MODELS METAPHORS AND THE DEVELOPMENT OFEPIDEMIOLOGIC THEORY: TOWARDS AN

ECOSOCIAL FRAMEWORK

(1) The importance of metaphors ,for cogniti on andtheory

Among the many obstacles to developing an ecoso-cial framework is one that has received relatively littleattention: the absence of a metaphor that can suc-cinctly capture the essence of this alternative view.Nothing comparable to the web exists.

The importance of metaphors for scientificthought, though rarely acknowledged, requiresgreater recognition [22, 131-1351. Often viewed aspurely evocative linguistic devices, metaphors in factplay an essential role in cognition [l36]. Like theories,metaphors attempt to produce new understandingthrough constructing novel connections betweenseemingly disparate phenomena and/or concepts,thereby enabling the unknown to be comprehendedthrough reference to the known [132, 134, 1361.To gain these insights, metaphors and theoriesboth rely upon mental imagery; indeed, the Greekroot of theory means to view [137, p. 32841 whilethat of metaphor means to overlay meaning[137, p.17811.

Although related, metaphors and theories are alsodifferent. Whereas theorizing implies delineatingstructured and testable relationships among selectedentities or ideas [l 1, 12,22-251, metaphors challengeus to experience one phenomenon or concept in termsof another [l36, p. 51. Through their deliberate disso-nance (which distinguishes them from mere analogies[134]), metaphors jar us to see relationships thatotherwise would be opaque [131ll35]. Less con-strained than formal theories, metaphors often cre-atively express intuitive ideas about how the worldworks. The metaphor of the web has certainlyplayed this role in epidemiology, as have themetaphors of production and reproduction forthose who think in terms of the social production ofdisease [54, 56, 62, 105, 1251.

Opening up new ways of seeing the world,metaphors can both help and hinder understanding.Consider, for example, the metaphor of man-as-ma-chine, as developed by Descartes. On the one hand,this metaphor led to new scientific knowledge byspurring non-vitalistic explanations of humanbiology [63, 1341. At the same time, its inherentmind/body dualism and reductionism has blockedand continues to hinder research on psychosocialdeterminants of health [45,63, 1351. The power ofmetaphors is thus fickle: they can simultaneously free

and constrain thought-and yet without them, it isunlikely that new connections will be drawn.(2) Tow ards an ecosocial metaphor f or epidemiol ogy

Perhaps one step toward developing an ecosocialmetaphor would be augmenting the metaphor of theweb with two spiders: one social, one biologic.They would certainly reintroduce the concepts ofhistory and agency, and would emphasize the import-ance of considering the origins of both social andbiologic determinants of disease. Even so, the im-agery of the spiders may be too simplistic, and mayfail to do justice to the complex origin and nature ofthe spiders themselves. It is of little help to posit thathealth and disease are socially produced within evolv-ing and socially-conditioned biologic parameterswithout offering insight into why and how this oc-curs; reducing the spiders to a new form of blackbox would only reinforce existing limitations. Norwould introducing the spiders necessarily resolve thewebs embodiment of a biomedical and individualis-tic worldview. The web never was intended to anddoes not jar epidemiologists from the long-estab-lished practice of viewing population patterns ofdisease as simply the sum of individual cases; it is farfrom obvious that adding the spiders would addressthis fundamental problem.

As an alternative, the closest image that comes tomind stems from marrying the metaphor of thecontinually-constructed scaffolding of society thatdifferent social groups daily seek to reinforce or alter[136] to that of the ever-growing bush of evolution[138], together defining the potential and constraintsof human life. This intertwining ensemble must beunderstood to exist at every level, sub-cellular tosocietal, repeating indefinitely, like a fractal object(Fig. 6) [139-1421. Different epidemiologic profiles atthe population-level would accordingly be seen asreflecting the interlinked and diverse patterns ofexposure and susceptibility that are brought into playby the dynamic intertwining of these changing forms.It is an image that does not permit the cleavage of thesocial from the biologic, and the biologic from thesocial. It is an image that does not obscure agency.And it is an image that embraces history rather thanhides it from view.

This image insists that understanding societal pat-terns of health and disease requires acknowledgingthe inextricable and ongoing intermingling-at alllevels-of the social and the biologic. Through itsfractal nature, it does not allow the individual tobecome separated from society, nor detract frompeoples irreducible individuality. A given branch ofthis intertwining structure can thus be seen as repre-senting one set of possible epidemiologic profilesproduced at a particular time by a particular combi-nation of social structures, cultural norms, ecologicmmeu, ano genetic variability and similarity (amonghumans and among other organisms in the region).At a greater level of detail, particular groups-linked


11/17

Epidemiology and the web of causation 897

Fig. 6. Computer-generated fractal structure, illustrating self-similarity on multiple scales [141, p. 18071.Reprinted with permission from the publisher.

by their membership in socially-defined categories-embody these characteristics, thus acquiring, in thememorable phrase of Cristina Laurell, their distinctclass physiognomies [ 105, p. 11841.

At an even greater level of detail, the particularitiesof health and disease among individuals becomeapparent, reflecting yet again the interplay of thesesame social and biologic influences. And, moving toyet another level, these same sets of influences can betraced through their relationship to the normal func-tioning and disturbances of organ systems, cellulargrowth and differentiation, and metabolic processes.Highlighting the inherent links between levels, thisimage thus requires considering multiple levels whenseeking to understand patterns at any given level, andlikewise highlights the need to frame questionsbroadly, regardless of the level at which any particu-lar investigation is conducted.

The proposed image accordingly begins to offer away to conceptualize the processes producing differ-ent epidemiologic profiles, both within and acrossspecific societies, at a given time and over time, at a

given location and across locales. It does not mandatea singular answer, applicable for all diseases, but doesspecify a range of questions-about social structure,cultural norms, ecologic milieu, and genetic variabil-ity-that must be systematically addressed whenanalyzing any specific situations. It requires popu-lation-thinking in its study of individuals, and recog-nition of individual variability (and similarity) in itsstudy of populations. Beyond this, it directs epidemi-ologists to think about individuals in the context oftheir everyday lives, as shaped by their interwinedhistories-as members of a particular society, and asbiological creatures who grow, develop, interact, andage [17,98,105, 112, 1231. It thus encourages askingwhether the factors or processes pertaining to ex-posure or susceptibility are exclusively or conjointlyphysical, biological, or social in nature, and alsoexogenous, endogenous, or both. Similarly, it directsattention to whether-and if so, why and how--ex-posure and susceptibility vary over the course ofpeoples lives. At issue may be social conventionsregarding who does what at which age, historical


12/17

898 N NCY KRIEGERchanges in living and working conditions and alsointeractions with other species, as well as biologicchanges in tissues, organs, and systems that bearthe imprint of evolutionary selection and contingentalterations [24, 112, 123, 1381.

This image accordingly suggests an orientationsimilar to that implied by evolutionary theories,which do not specify or predict the exact fate ofparticular species (or groups of species) and theorganisms of which they are composed, but dospecify a range of questions and a set of processesthat must be considered when attempting to under-stand their specific dynamics (extending from originto extinction) [24, 131, 1381. And, at the same time,it suggests an orientation similar to that impliedby sociological theories, which likewise do notspecify or predict the exact fate of particular socialgroups (or societies) and the individuals of whichthey are composed, but similarly specify a range ofquestions and a set of processes that must be con-sidered when attempting to understand their socialfeatures and behaviors [42,61, 821. And like bothevolutionary and sociological theories, the orien-tation implied by this image rejects a teleologicalstance (implying change in a definite direction towarda specified goal), but does not do so at the expenseof neglecting the critical role of human agency andaccountability-at the societal as well as individuallevel-in shaping population patterns of health anddisease.

Encouraging a social and ecologic point of view,this image also serves as a reminder that people arebut one of the species that populates our planet; itthus implies that the health of all organisms isinterconnected. And at the same time, by situatingsocial groups and individuals-which, by definition,include epidemiologists-in the context of particularsocieties at particular times, it demands that epidemi-ologists consider how their social position affects theknowledge they desire and that which they produce.As such, it directs attention not only to the socialproduction of disease, but also to the social pro-duction of science-that is, how a societys predomi-nant view of the world and the position of scientistsin this society influence the theories they develop,the research questions they ask, the data theycollect, the analytic methods they employ, andthe ways they interpret and report their results[45, 54,63,72,74,99, 143-1451. This stance acknowl-edges that science is at one-and indivisibly-objec-tive and partisan [99]. And, by accepting theseinherent features of science-the objectivity thatarises out of commonly employed methodologiesused by different investigators in comparable situ-ation and the subjectivity that arises out of the valuesand worldviews that promote some questions andsilence others-this image encourages epidemiolo-gists (and other scientists) to avoid the trap ofconflating scientific assumptions with reality and toembrace a humility that counters the apparent hubris

which often accompanies the biomedical approach[44,45,61, 1171.(3) What problems could an ecosocial framewo rkaddress?

One reason for pursuing the development of anecosocial metaphor and theory premised upon thisinitial image is that such a framework could offer ahelpful way of thinking about several striking fea-tures of epidemiologic data and research in theUnited States today. One is the stunning and long-standing acceptance of the absence of social classdata in U.S. vital statistics and most disease registries[99, 1161, along with paucity of good data for mostminority racial/ethnic groups [99, 1461. These gapscontrast sharply with the ever expanding knowledgeabout epidemiologic methods (Table I), genetic sus-ceptibility [87], and lifestyle risk factors [146]. It isalso both striking and telling that the U.S. NationalInstitutes of Health recently resorted to issuing direc-tives requiring epidemiologists and clinical re-searchers to include women [1471 and minorities [1481in their study populations. This change occurred onlybecause of the growing political influence of womenand people of color and their insistence upon healthresearch being relevant to their lives. That groupsoutside of epidemiology triggered these changesspeaks volumes about the provincialism and elitismsanctioned by the current framework guiding re-search today, one developed chiefly by privilegedpersons (mainly white, mainly men) trained in thebiomedical sciences.Such a framework could also spur greater precisionin epidemiologic concepts about etiology thanpresently exists. By challenging the biomedical indi-vidualism underlying the construction of the epidemi-ologic triads of race, age, and sex and of time, place,and person, it would make clear that these phenom-ena are neither simply natural nor-in the caseof personal characteristics-individually innate. Itwould promote recognition of the fact that raceis a spurious biologic concept [63,98, 115, 1491 andwould instead direct attention to how racism affectshealth-overall, and of people on both sides of thecolor line [98, 109-l 16, 1491. An ecosocial frameworkwould likewise require considering how the process ofaging cannot be separated from the social conditionsin which people are born, live, work and retire[112, 1231. It would not confuse biological sex withculturally-determined gender and would promotequestions about how the health of not only womenbut men is shaped by gender-based (and often sexist)assumptions [71,98, 109, 117-1231. Social classwould be considered a fundamental category[56, 58-61,97-108, 117, 1231, and the term personwould not be used when what really is meant is socialgroup. Such a framework would thus demand thatepidemiologists eschew terms like special popu-lations-now routinely used by U.S. federal healthagencies to describe women, the poor, and people


13/17

Epidemiology and the web of causation 899of color [146]-and would instead directly exposewhat makes these populations special: their enforcedmarginalization from positions of power, coupledwith the assumption that white, economically-securemen are allegedly the norm.

An ecosocial approach would also encourage refor-mulating, if not rejecting, other loosely-defined termsso prevalent in the epidemiologic literature-such aslifestyle [1501 and environment [15 I]-so as to endthe practice of obscuring or misclassifying agency. Itis more than a misnomer, for example, to imply thatit is simply a personss freely-chosen lifestyle to eatpoorly when supermarkets have fled the neighbor-hood, or to have a child early or late in life or not atall, without considering economic circumstances andjob demands [44,54,58-61,97,98]. An ecosocialframework would thus require situating the socialcontext of such health behaviors if they are to becomprehended, let alone changed. And, with regardto prevention, it would encourage research on notonly those factors deemed amenable to interventionthrough the medical care system, the work of publichealth departments, or the effort of solo individuals,but also on the broader determinants of health thatcan be changed only through more widespread socialaction.

An ecosocial approach would thus challenge cur-rent definitions of the environment as that which isexogenous to the organism (cf. A Dict ionary ofEpidemiology [151], as well as the uses of suchallegedly complementary phrases as the natural andsocial environment. In rejecting this analogy, itwould make clear that social conditions are notnatural but are constructed by people, with purposein mind and accountability an option. Social con-ditions are conceptually and categorically distinctfrom the natural environment, that is, the interplayof ecologies and global geologic and climactic forces,which humans can effect (and even destroy) butwhich we certainly have not created. Confoundingof concepts can muddy analyses as much as con-founding of risk factors; epidemiologists must be asrigorous about categories of thought as approachesto analyzing data.Additionally, an ecosocial framework would chal-lenge the current rigid distinction between individual-and group-level analyses. Directing attention to thehealth effects of collective phenomena that cannot bereduced to individual attributes [98, 152, 1531, it cap-tures why what has been termed the individualisticfallacy [152, 154]-i.e. the assumption that individ-ual-level data are sufficient to explain group-levelphenomena-is as much of a liability as the ecologicfallacy [152, 154, 1551, which results from confound-ing introduced by the grouping process. Tying to-gether the macro and the micro, it would encourageuse of what has been termed contextual or multi-level analysis [98, 152, 153, 156-1581, which com-bines individual- and group-level data in a clearlyspecified and theoretically justified manner. This ap-

preach has been used in other fields [156-1581, andonly recently has been explicitly introduced into theepidemiologic repertoire [152, 1531, e.g. consideringthe conjoint effects of neighborhood- and individual-level social class upon health [152]. An ecosocialframework thus has the potential to raise new con-ceptual and methodologic questions about the shap-ing of human health, much as the evolutionarysynthesis of the 1940s generated considerable intel-lectual ferment and promoted new discoveries byintegrating genetic and species-level approaches tounderstanding and studying biological evolution [24].

CONSTRUCTING EPIDEMIOLOGIC THEORY:A NECESSARY AND VITAL CHALLENGE

The field of epidemiology today suffers from theabsence of not only a clearly articulated and compre-hensive epidemiologic theory, but, it seems, even theawareness that it lacks such a theory. The scienceinstead is taught and viewed as a collection ofmethods to be applied to particular problems involv-ing human diseases and health [3,4, 15, 20,211.

To counter this state of affairs, the image proposedin this paper is intended to spur discussion aboutimportant aspects of epidemiologys purpose anddomain, as the science that seeks to explain andgenerate knowledge to improve population patternsof health, disease, and well-being. Attempting toadvance an ecosocial framework for the developmentof epidemiologic theory, this image makes clear thatalthough the biologic may set the basis for theexistence of humans and hence our social life, it is thissocial life that sets the path along which the biologicmay flourish-or wilt. As such, it emphasizes whyepidemiologists must look first and foremost to thelink between social divisions and disease to under-stand etiology and to improve the publics health, andin doing so exposes the incomplete and biased slantof epidemiologic theories reliant upon a biomedicaland individualistic world-view.

Despite its appeal, this image remains only that-an image. It is not a developed metaphor. Nor is ita substitute for a well-articulated ecosocial theory ofepidemiology. And it remains open to questionwhether this particular image could help give rise toa more concise metaphor or provide useful debateabout the current status of epidemiologic theory. Theessential point, however, is that the multivariateframework so widespread in epidemiology today, asexpressed by metaphor of the web of causation,represents an approach to epidemiologic theory thatis deeply flawed. To forge a better theory, it may stillbe worthwhile to search for the spider-whether oneor many-but this can be determined only if this taskis pursued. Otherwise, like Mr Zuckerman in E. B.Whites classic tale, Charlottes Web [l], epidemiolo-gists will continue to mistake Wilbur the pig for themiracle of the web and the work of the spider. We willthereby miss the full story.


14/17

900 N NCY KRIEGERAcknowledgements Thanks to Dr Elizabeth Fee and Dr S.Leonard Syme, for their helpful comments.

REFERENCES

4.

5.

6.

7.

8.9.

10.

Il.

12.13.

14.15.16.17.

18.19.

20.

21.

22.23.

24.

White E. B. Charlottes Web. HarperCollins, NewYork, 1952.MacMahon B., Pugh T. F. and Ipsen J. Epidemiologicmethods. Little Brown and Co., Boston, 1960.Susser M. Epidemiology in the United States afterWorld War II: the evolution of technique. Epidemiol.Rev. I, 147-177, 1985.Buck C., Llopis A., Najera E. and Terris M. Discus-sion. In The Challenge of Epidemiology: Issues andSelected Readings (Edited by Buck C., Llopis A.,Najera E. and Terris M.), pp. 149-153. PAHO/WHO,Washington, DC, 1988.Greenland S. (Ed.) Evolution of Epidemiologic Ideas:Annotated Readings on Concepts and Methods. Epi-demioloev Resources. Chestnut Hill. MA, 1987.Susser M. Causal Thinking in the Health Sciences:Concepts and Strategies of Epidemiology. Oxford Uni-versity Press, New York, 1987.Miettinin 0. S. Theoretical Epidemiology: Principles ofOccurrence Research in Medicine. Wiley, New York,1985.Rothman K. Modern Epidemiology. Little Brown,Boston, 1986.Rothman K. J. (ed.) Causal Inference. EpidemiologyResources, Chestnut Hill, MA, 1988.Susser M. What is a cause and how do we know it?A grammer for a pragmatic epidemiology. Am. J.Epidemiol. 133, 6355648, 1991.Ziman J. Reliable Knowledge: an Exploration of theGrounds for Belief in Science. Cambridge UniversityPress, Cambridge, 1978.Wisdom J. Challengeability in Modern Science. Ave-bury Press, Aldershot, 1987.Cassel J. M. Potentialities and limitations of epidemi-ology. In Health and the Community: Readings in thePhilosophy and Sciences of Public Health (Edited bvKatz A. and Felton J.), pp. 432435. The Free Press,New York, 1965.Terris M. The epidemiological tradition. Pub. HlthRep. 94, 203-209, 1979. _Stallones R. A. To advance euidemioloav. Ann. Rev.publ. Hlth 1, 69982, 1980. _ -.Rose G. The Strategy of Preventive Medicine. OxfordUniversity Press, Oxford, 1992.Krieger N. Race, Class, and Health: Studies of BreastCancer and Hypertension. Dissertation, University ofCalifornia at Berkeley, pp. 2333266, 1989.Susser M. Epidemiology today: a thought-tormentedworld. In/. J. Epidemiol. 18, 481488, 1989.Terris M. The Society for Epidemiologic Research(SER) and the future of epidemiology. Am. J. Epi-demiol. 136, 9099915, 1992.Lilienfeld A. M. and Lilienfeld D. E. Epidemiologyand the public health movement: a historical perspec-tive. J. publ. Hlth Policy 3, 140-149, 1982. _ _Kuller L. H. Euidemioloav is the study of euidemicsand their prevention. --Am. J. Epidemiol. 1341051~1056, 1991.Fleck L. Genesis and Development of a Scienttjic Fact.University of Chicago Press, Chicago (1935).Kuhn T. S. The Structure of ScienttJ?c Revolutions, 2ndedn. enlarged. University of Chicago Press, Chicago(1962).Mayr E. The Growth of Biological Thought:Diversity, Evolution, and Inheritance. The BelknapPress of Harvard University Press, Cambridge, MA,1982.

25.

26.27.

28.29.

30.

31.32.

33.

34.

35.36.37.

38.39.40.41.42.43.

44.

45.46.47.

48.49.50.

51.52.

Holton G. The Thematic Origins of Scient c Thought:From Kepler to Einstein, revised edn. Harvard Univer-sity Press, Cambridge, MA, 1988.Morris J. N. Uses of Epidemiology. E S Livingston,Edinburgh, 1957.Taylor I-and Knowelden J. Principles of Epidemiology.J A Churchill. London. 1957.Fee E. and Acheson R. M. (Eds) A History qf Edu-cation in Public Health: Health that Mocks the DoctorsRules. Oxford University Press, Oxford, 1991.Galdston I. (Ed.) The Epidemiology qf Health. HealthEducation Council, New York Academy of Medicine,New York, 1953.Galdston I. (Ed.) Beyond the Germ Theory: the Role ofDeprivation and Stress in Health and Disease. HealthEducation Council, New York Academy of Medicine,New York, 1954.Gordon J. E. Epidemiology-old and new. J. Michi-gan St. Med. Society 49, 1944199, 1950.Rosen G. What is social medicine?-a genetic analysisof the concept (1947). In From Medical Police to SocialMedicine. (Edited by Rosen G.), pp. 60-l 19. ScienceHistory Publications, New York, 1974.Galdston I. (Ed.) Social Medicine: its Derivatives andObjectives. The Commonwealth Fund. New York,1949.Cornfield J. A method of estimating comparative ratesfrom clinical data: applications to cancer of the lung,breast, and cervix. J. natn. Cancer Inst. 11, 126991275,1951.Sayre A. Rosalind Franklin and DNA. W. W. Norton

Co., New York, 1975.Nordhng C. 0. A new theory on the cancer-inducingmechanism. Br. J. Cancer 7, 68872, 1953.Skeet R. G. The impact of the computer on the cancerregistry. Gann Monograph on Cancer Res. 33, 89-95,1987.Anderson M. J. The American Census: a SocialHistory. Yale University Press, New Haven, CT, 1988.Vallis F. (Ed.) Ribbons of Time, p. 115. WeidenfeldNicolson, New York, 1988.Zinn H. A Peoples History qf the United States.Harper Colophon, New York, 1980.Blumberg P. Inequality in an Age of Decline. OxfordUniversity Press, New York, 1980.Bell D. The Social Sciences Since the Second WorldWar. Transaction Books, New Brunswick, NJ, 1982.Branch T. Parting the Waters: America in the KingYears, 1954-1963. Simon Shuster, New York,1988.Fee E. and Krieger N. Understanding AIDS: historialinterpretations and the limitations of biomedical indi-vidualism. Am. J. publ. Hlth 83, 147771486, 1993.Lock M. and Gordon D. (Eds) Biomedicine Examined.Kluwer Academic, Dordrecht, 1988.Watts A. Tao: The Watercourse Wuy. PantheonBooks, New York, 1975.Bannerman R. H., Burton J. and Wen-Chieh C. (Eds)Traditional Medicine and Health Care Coverage: aReader for Health Administrators and Practitioners.World Health Organization, Geneva, 1983.Bernal J. D. Science in History, Vol. 14. MIT Press,Cambridge, MA, (1954).Berman M. The Disenchantment of the World. CornellUniversity Press, Ithaca, NY, 1981.Weiss N. S. and Liff J. M. Accounting for the multi-causal nature of disease in the design and analysis ofepidemiologic studies. Am. J. Epidemiol. 117, 14-18,1983.Mausner J. S. and Bahn A. K. Epidemiology: anIntroductory Text. Saunders, Philadelphia, 1974.Friedman G. Primer qf Epidemiology>. McGraw-Hill,New York, 1974.


15/17

Epidemiology and the web of causation 90153.

54.

55.56.51.58.59.60.

61.

62.

63.

646566

616869

70

71

12.

73.74.

75.

16.

17.

78.

Duncan D. F., Gold R. J., Basch C. E. and MarkellisV. C. Epidemiology: Basis or Disease Prevention andHealth Promotion. Macmillan, New York, 1988.Tesh S. Hi dden A rguments: Poli t i cal Ideology andDisease Prevention Policy. Rutgers University Press,New Brunswick, NJ, 1988.Rothman K. J. Causes. Am. J. Epidemiol. 104587-592, 1916.Breilh J. Epidemiol ogia Economia M edicina y Polit ica.Distribuciones Fontamara, Mexico City, (1979)Rose G. Sick individuals and sick populations. In t. J.Epid emiol . 14, 32-38, 1985.Doyal L. The Poli t i cal Economy of Healt h. Pluto Press,London, 1979.Navarro V. Cri sis, Healt h, and Medicine: a SocialCrit ique. Tavistock, New York, 1986.Waitzkin H. The Second Sickness: Contradictions ofCapitalist Health Care. The Free Press, New York,1983.Conrad P. and Kern R. (Eds) The Sociol ogy of Healt hand Illness: Critical Perspectives. St Martins Press,New York, 1981.Krieger N. and Bassett M. The health of black folk:disease, class and ideology in science. M onthl y Rev. 38,74-85, 1986.Lewontin R. C., Rose S. and Kamin L. J. Not in ourGenes: Bi ology, I deology, and Human Nat ure. Pan-theon Books, New York, 1984.Dubos R. Man Adapting. Yale University Press, NewHaven, CT, 1965.Rosen G. A History of Public Health. MD Publi-cations, New York, 1958.Coleman W. Death is a Social Di sease: Publ ic Healt hand Poli t i cal Economy in Early Industrial France. Uni-versity of Wisconsin.Press, Madison, WI, 1982.Ackerknecht E. H. Anticontaaionism between 1821and 1867. Bul l. Hi st. M ed. 22,-5622593, 1948.Pelling M. Cholera, Fever, and Engli sh Medici ne1825-1865. Oxford University Press, Oxford, 1978.Rosenberg C. D. The Chol era Years: t he Un i t ed Stat esin 1832, 1849, and 1866. University of Chicago Press,Chicago ( 1962).Rosenkrantz B. G. Publ ic Health and the Stat e: Chang-i ng Vi ews n M assachusett s, 1842-1936. Harvard Uni-versity Press, Cambridge, MA, 1972.Apple R. D. (Ed.) Women, Health M edicine inAmerica: a Hi storical Handbook. Rutgers UniversityPress, New Brunswick, NJ, 1992.Krieger N. Shades of difference: theoretical under-pinnings of the medical controversy on black/whitedifferences in the United States, 1830-1870. Int. J. H lt hServ i ces 17, 259-278, 1987.Chapin C. V. The Sources and Modes of in fecti on.Wiley, New York, 1910.Fee E. Di sease and Di scovery: a Histo ry of the JohnsHopki ns School of Hy giene and Publi c Healt h,1916-1939. Johns Hopkins University Press, Balti-more, MD, 1987.Greenwood M. Epidemics and Crow d Di seases: AnInt roducti on to the Study of Epidemiol ogy, p, 359.Williams Northgate, London,- 1935. -_ _Frost W. H. Epidemioloav. In Papers of Wade Hamo-ton Frost, MD.a Contri bi i i on to Epidemiol ogic M ethod(Edited by Maxcy K. F.), pp. 439-542. The Common-wealth Fund, New York, 1941.Goldberger J., Wheeler G. A. and Sydenstricker E. Astudy of the relation of family income and othereconomic factors to pellagra incidence in seven cotton-mill villages of South Carolina in 1916. Publ. Hl th Rep.35, 2673-2714, 1920.Etheridge E. W. The Butt erf ly Caste: a Social H istoryof Pellegra in the South. Greenwood, Westport, CT,1972

19.80.81.

82.83.

84.

85.

86.

87.

88.

89.

9091

9293

949596.

97.

98.

99.

100.

101.

102.

103.104.

Terris M. (Ed.) Gol dberger on Pellagra. LouisianaState University Press, Baton Rouse, LA, 1964.Sydenstricker I?. Health and Envionment. McGrawHill, New York, 1933.Kuller L. H. Relationship between acute and chronicdisease epidemiology. Yal e J. Bi ol . Med. 60, 3633376,1987.Cassell J. Social science theory as a source of hypoth-eses in epidemiologic research. Am. J. publ. Hlth 54,148221488, 1964.Cassell J. The contribution of the social environmentto host resistance. Am. J. Epidemiol. 104, 107-123,1976.Syme S. L. Social determinants of disease. In Maxcy-Rosenau-L ast Publ ic Healt h Preventi ve M edici ne.13th edn. (Edited by Last J. M. and Wallace R. W.),pp. 687-700. Appleton Lange, Norwalk, CT, 1992.Vandenbroucke J. P. Is the causes of cancer a miasmatheory for the end of the twentieth century? ht. J.Epi demio l 17, 7088709, 1988.Vandenbroucke J. P., Eelkman R. H. M. and BrukersH. Who made John Snow a hero? Am. J. Epidemiol.133, 9677973, 1992.Hulka B. S., Wilcosky T. and Griffith J. D. (Eds)Biol ogical M arkers in Epidemiol ogy. Oxford UniversityPress, New York, 1990.Loomis D. and Wing S. Is molecular epidemiology agerm theory for the end of the twentieth century? Int.J. Epidemiol. 19, 1-3, 1990.Yack D. Biological markers: broadening or narrowingthe scope of epidemiology? J. clin. Epidemiol. 43,309-310, 1990.Hubbard R. and Wald E. Explodi ng the Gene M yt h.Beacon Press, Boston, 1993.Lippman A. Led (astray) by genetic maps: the cartog-raphy of the Human Genome and health care. Sot. Sci.M ed. 35, 146991476, 1992.McKeown T. The Origins of Human Disease. Black-well, Oxford, 1988.Boyden S. F. (Ed.) The Impact of Civi l i zati on on theBiology of Man. Toronto University Press, Toronto,1970.Fiennes R. N. The Envir onment of M an. St MartinsPress, New York, 1978.Cohen M. N. Health and the Rise of Civi l i zati on. YaleUniversity Press, New Haven, CT, 1989.Szreter S. The importance of social intervention inBritains mortality decline ca 1850~1914: a reinterpre-tation of the role of public health. Sot. Hi st. M ed. 1,l-38, 1988.Black D., Morris J. N., Smith C. and Townsend P.Inequali t i es in Healt h: the Black Report. PenguinBooks, Harmondsworth, 1985.Krieger N., Rowley D., Hermann A. A., Avery B. andPhillips M. T. Racism, sexism, and social class: impli-cations for studies of health, disease, and well-being.Am . J. Pr ev. Med. 9, suppl 6, 82-122, 1993.Krieger N. The making of public health data: para-digms, politics, and policy. J. pub/. Hlth Policy 13,412427, 1992.Haan M., Kaplan G. A. and Camacho T. Poverty andhealth: prospective evidence from the Alameda countystudy. Am . J. Epi demio l . 125, 989-998, 1987.Marmot M. G., Smith G. D., Stansfeld S. et al. Healthinequalities among British civil servants: the WhitehallII study. Lancer 337, 138771393, 1991.Whitehead M. The Health Divide. Penguin Books,Harmondsworth, 1987.Svensson P. G. Health inequities in Europe-intro-duction. Sot. Sci. Med. 31, 225-226, 1990.Sanders D. The Str uggle for Healt h: M edici ne and thePolitics of Underdevelopment. MacMillan, London,1985.


16/17

NANCY KRIEGER02105.106.

107.108.

109.

110.

111.

112.

113.

114.115.

116.

117.118.

119.

120.

121.

122.

123.124.

125.

Laurel1 A. C. Social analysis of collective health inLatin America. Sot. Sci. M ed. 28, 1183-l 191, 1989.Franc0 S., Nunes E., Duarte E., Breilh J., Granda E.,Yepez J., Costales P. and Laurel1 C. Debates enMedicina Social. Organization Panamericana de laSalud-ALAMES, Quito, Ecuador, 1991.Johnson J. V. Collective control: strategies for survivalin the workplace. In f. J. Hl th Serv ices 3,469-480, 1989.Karasek R. The political implication of psychosocialwork redesign: a model of the psychosocial classstructure. Int. J. Hl t h Serv i ces 3, 481-508, 1989.Krieger N. Racial and gender discrimination: riskfactors for high blood pressure? Sot. Sci. Med. 30,1273-1281, 1990.Armstead C. A., Lawler K. A., Gordon G., Cross J.and Gibbons J. Relationship of racial stressors toblood pressure and anger expression in black collegestudents. Hl fh Psychol. 8, 541-556, 1989.James S. A. LaCroix A. Z., Kleinbaum D. G. andStrogatz D. S. John Henryism and blood pressuredifferences among black men. II. The role of occu-pational stressors. J. Behav. Med. 7, 2599275, 1983.Geronimus A. T. The weathering hypothesis and thehealth of African-American women- and infants: evi-dence and soeculation. Efhnici fv Dis. 2,207-221. 1992.Dressier W..W. Stress and Adapfafi on in fhe Confexf ofCult ure: Depression in a Southern Black Communi fy.State University of New York Press, Albany, NY,1991.David R. J. and Collins J. W. Jr Bad outcomes in blackbabies: race or racism? Ethn ici fy Di s. 1, 236-244, 1991.Cooper R. and David R. The biological concept ofrace and its application to epidemiology. J. HlfhPol i t ics Pol icy Law 11, 87-l 16, 1986.Navarro V. Race or class versus race and class:mortality differentials in the United States. Lancef ii1238-1240, 1990.Roberts H. (Ed.) Womens Healfh Counts. Routledre.London, 1990. Hall E. M. Double exposure: the combined impact ofthe home and work environments on psychosomaticstrain in Swedish women and men. Inf. J. Hlfh Services22, 239-260, 1992.Bassett M. T. and Mhloyi M. Women and AIDS inZimbabwe: the making of an epidemic. Inf. J. HlfhServ i ces 21, 143-156, 1991.Krieger N. Women and social class: a methodologicstudy comparing individual, household, and censusmeasures as oredictors of black/white differences inreproductive history. J. Epidemiol. Communit y Hl fh 45,3542, 1991.Krieger N. Influence of social class, race, and genderon the etiology of hypertension among women inthe United States, Conference proceedings, Women,behavior, and cardiovascular disease, NationalHeart, Lung, and Blood Institute, Washington, DC.In press.Krieger N. Exposure, susceptibility, and breast cancerrisk: a hypothesis regarding exogenous carcinogens,breast tissue development, and social gradients, includ-ing black/white differences, in breast cancer incidence.Breasf Cancer Res. Tr eat. 13, 2055223. 1989.Arber S. and Ginn J. Gender and inequalities in healthin later life. Sot. Sci. M ed. 36. 3346, 1993.PAHO. On the concept of health and disease: descrip-tion and explanation of the health situation. Epidemiol.Bul l . 10 1-8, 1990.Organization PanAmericana de la Salud. La forma-cion en epidemiologia para el desarrollo de 10s servi-cios de salud. XIV conferencia de la AsociacionLatinoAmerican y del Caribe de education en saludpublica (ALASEP). Taxco, Mexico, 15 al 19 Noviem-bre 1987, PAHO/WHO, Washington, DC. 1988.

126

127.

128.129.130.

131.132.

133.

134.135.

136.137.138.

139.140.

141.

142.

143.144.

145.146.

147.

148.

149.

Brenner M. H. Esti mati ng the Effects of EconomicChange on Nafi onal Healfh and Social Well -being. Joi nf Economi c Comm it fee of fhe U.S. Congress.U.S. Government Printing Office, Washington, DC,1984.Eyer J. and Sterling P. Stress related mortality andsocial organization. Rev. Radical Polif. Econ. 9, 144,1977.Stark E. The epidemic as a social event. Inf. J. HlfhServ i ces 7, 681-705, 1977.Turshen M. The Poli fics of Publi c Healfh. RutgersUniversitv Press. New Brunswick. NJ. 1989.Packard k. M. Whi fe Plague, Black Labor: Tuberculo-sis and the Poli fical Economy of Health and Di sease inSoufh Africa. University of California Press, Berkeley,CA, 1989.Young R. Darwins Metaphor. Cambridge UniversityPress, Cambridge, 1985.Martin J. and Harre R. Metaphor in science. InMetaphor : Probl ems and Perspecti ves (Edited by MiallD.S.), pp. 89-105. The Harvester Press, Sussex, NJ,1982.Gould S. J. Glow, big glowworm. In Bully for Bron-fosaurus (Edited by Gould S. J.), pp. 2555268. W. W.Norton Co., New York, 1991.MacCormac E. R. A Cogniti ve Theory of M efaphor.MIT Press, Cambridge, MA, 1985.Osherson S. and Amarasingham L. The machinemetaphor in medicine. In Social Contexfs of Healfh,Illness. and Patient Care (Edited bv Mishler E. G.,Amarasingham L., HauserS. T. ef al.), pp. 2188249.Cambridge University Press, Cambridge, 198 1.Lakoff G. and Johnson M. Metaphors w e Li ae By.Chicago University Press, Chicago, 1980.The Compact Editi on of the Oxford Engli sh D icti onary.Oxford University Press, Glasgow, 197 I.Gould S. J. Wonderful Li fe: the Burgess Shales and th eNature of History. W. W. Norton Co., New York,1989.Gleik J. Chaos: M aki ng A New Science. Viking Pen-guin, New York, 1987.Glass L. and Mackey M. C. From Cl ocks t o Chaos: t heRhyt hms of Li fe. Princeton University Press, Prince-ton, NJ, 1988.Lipsitz L. A. and Goldberger A. L. Loss of complex-ity and aging: potential applications of fractals andchaos theory to senescence. J.A.M .A. 267, 18061809,1992.Schaffer W. M. Chaos in ecology and epidemiology. InChaos i n Biol ogical Systems-(Edited- by Degn H.,Holden A. V. and Olsen L. F.). oo. 2333248. Plenum_.ress, New York, 1987.Rose H. and Rose S. (Eds) Ideology of/in the Nat uralSciences. Schenkman, Cambridge, MA, 1979.Haraway D. Pri mat e Vi sions: Gender, Race, andNat ure in the Worl d of M odern Science. Routledge.ew York, 1989.Hubbard R. The Poli ti cs of Womens Biol ogy. RutgersUniversity Press, New Brunswick, NJ, 1990.U.S. Department of Health Services, Public HealthService. Healthy People 2000: National Health Pro-moti on and Di sease Preventi on Obj ecti ves. DHHS Pub.No. (PHS) 91-50123, U.S. Government Printing Office,Washington, DC, 199 1.NIH/ADAMHA policy concerning inclusion ofwomen in study populations. NI H Guide for Gr anfsand Contracts, October 24, 1986.NIH/ADAMHA policy concerning inclusion of min-orities in study populations. NI H Guide Grants andContracts, September 25, 1987.Centers for Disease Control and Prevention. Use ofrace and ethnicity in public health surveillance.M .M.W .R. 42, (No. RR-lo), l-17, 1993.


17/17

Epidemiology and the web of causation 903150.

151.

152.

153.

Coreil J., Levins J. S. and Jaco E. G. Life style: an 154.emergent concept in the sociomedical sciences. Culfure,med. psychi ar . 9, 423437, 1985.Last J. M. (Ed.) A Dicr ionary of Epidemiology-AHandbook Sponsored by the In national Epidemiol og- 155.ical Association. LX 2. Oxford University Press, NewYork, 1983. .Krieger N. Overcoming the absence of socioeconomic 156.data in medical records: validation and application ofa census-based methodology. Am . J. Publ . Hl t h 82, 157.703-710, 1992.Von Korff M., Koepsell T., Curry S. and Diehr P.Multi-level analysis in epidemiologic research onhealth behaviors and outcomes. Am. J. Epidemiol. 135, 158.1077-1082, 1992.

Alker H. R. Jr. A typology of ecological fallacies.In Social Ecology (Edited by Doggan M. andRokkan S.), pp. 69-81. MIT Press, Cambridge, MA,1969.Robinson W. S. Ecological correlations and thebehavior of individuals. Am. Social . Rev. 15 351-357,1950.Langbein L. 1. and Lichtman A. S. Ecological Infer-ence. Sage, Beverly Hills, CA, 1978.Iversen G. R. Conrextual Analysis. Sage UniversityPaper Series on Quantit ati ve Applicati ons in the SocialSciences. Seri es No. 07-081. Sage, Newbury Park, CA,1991.Van den Eeden P. and Huttner H. J. M. Multi-levelresearch. Current Social. 30, I-181, 1982.

SSM 39:7--c

Date post:	04-Jun-2018
Category:	Documents
Upload:	alexis-vidal
View:	221 times
Download:	0 times

Web Causation Spider

Documents