Does Epidemiologic Theory Exist? On Science, Data, and Explaining Disease Distribution - Oxford Scholarship Jump to ContentJump to Main Navigation
Epidemiology and the People’s HealthTheory and Context$

Nancy Krieger

Print publication date: 2011

Print ISBN-13: 9780195383874

Published to Oxford Scholarship Online: May 2011

DOI: 10.1093/acprof:oso/9780195383874.001.0001

Show Summary Details
Page of

PRINTED FROM OXFORD SCHOLARSHIP ONLINE (www.oxfordscholarship.com). (c) Copyright Oxford University Press, 2015. All Rights Reserved. Under the terms of the licence agreement, an individual user may print out a PDF of a single chapter of a monograph in OSO for personal use (for details see http://www.oxfordscholarship.com/page/privacy-policy). Subscriber: null; date: 28 August 2016

Does Epidemiologic Theory Exist?

Does Epidemiologic Theory Exist?

On Science, Data, and Explaining Disease Distribution

Chapter:
(p.3) 1 Does Epidemiologic Theory Exist?
Source:
Epidemiology and the People’s Health
Author(s):

Nancy Krieger (Contributor Webpage)

Publisher:
Oxford University Press
DOI:10.1093/acprof:oso/9780195383874.003.0001

Abstract and Keywords

Chapter 1 opens with a series of figures displaying epidemiologic data about trends and disparities in diverse measures of health status in the US and worldwide—and asks what assumptions, and hence underlying theories, guide generating and interpreting these data. Tackling the question of what constitutes a scientific theory, it then describes domain-specific features of epidemiologic theory—which by definition must exist if epidemiology is a science, and not simply a set of methods. Documenting, however, that epidemiologic textbooks from 1980–2000 are silent about epidemiologic theory—and earlier and later textbooks offer only little discussion—the chapter considers alternative arguments about why this gap exists. It argues the gap reflects the dominance of implicit, rather than explicit, use of epidemiologic theory to inform epidemiologic research. The chapter concludes with a call for the explicit analysis of these theories, so as to improve the intellectual rigor of the field.

Keywords:   disparities, epidemiologic research, epidemiologic textbooks, epidemiologic theory, epidemiology, methods, science, scientific theory, theory, trends

Theory. Traced to its Greek roots, “theory” means to see inwards; to theorize is to use our mind’s eye systematically, following articulated principles, to discern meaningful patterns among observations and ideas (Oxford English Dictionary [OED] 2008). The implication is that without theory, observation is blind and explanation is impossible.

In this chapter, I will make the argument that epidemiologic theory is a practical necessity for thinking about and explaining disease distribution. What could be more obvious?

Yet, apparently refuting what ought to be this simple self-evident claim is the curious fact that epidemiologic textbooks have, for the past several decades—as I discuss below—offered little or no guidance on what an epidemiologic theory is, let alone why such theory is important or how it can be used (Krieger, 1994). Sorting out this conundrum requires considering what scientific theory is—and what place it might have in epidemiologic thinking and research.

Figuring Out the People’s Health: Theory and the Stories (About) Data (that People) Tell

First: Why even posit that epidemiologic theory is a practical necessity? Consider the epidemiologic data shown in Figures 1–1 through 1–7. Together, they illustrate population distributions of disease—over time, space, and social group—in the United States and globally.

Figure 1–1 presents data from a study titled “The fall and rise of US inequities in premature mortality: 1960–2002” (Krieger et al., 2008). These data show that between 1960 and 2002, as rates of U.S. premature mortality (Figure 1–1a, deaths before age 65 years) and infant death (Figure 1–1b, deaths before age 1 year) declined in all county income quintiles, socioeconomic and racial/ethnic inequities in premature mortality and infant death (both relative and absolute) shrank between 1966 and 1980, especially for U.S. populations of color, but from 1980 onward, the relative health inequities widened and the absolute differences barely changed. Why? (p.4)

                   Does Epidemiologic Theory Exist?                   On Science, Data, and Explaining Disease Distribution

Figure 1–1. The fall and rise of U.S. inequities in premature mortality, 1960–2002, by county median income quintile. (Krieger et al., 2008)

Figure 1–1a. The fall and rise of U.S. inequities in premature mortality: deaths before age 65 years, 1960–2002, by county median income quintile, for: (A) total population by county median income quintile, and (B) the US White population (dashed lines) and populations of color (solid lines).

(p.5)
                   Does Epidemiologic Theory Exist?                   On Science, Data, and Explaining Disease Distribution

Figure 1–1b. The fall and rise of U.S. inequities in premature mortality: infant deaths, 1960–2002, by county median income quintile, for: (A) total population, and (B) the US White population (dashed lines) and populations of color (solid lines).

(p.6) (p.7)

Age group (years)

Population 1

Population 2

US 2000 standard million population

Population 1

Population 2

Deaths(N)

Population(N)

Age-specific death rate per 100,000

Deaths (N)

Population(N)

Age-specific death rate, per 100,000

Number of deaths if apply their death rates to the same standard population

(A)

(B)

(C) = ((A)/(B)) * 100,000

(D)

(E)

(F) = ((D)/(E)) * 100,000

(C) * standard population

(F) * standard population

〉1

99

17,150

577.3

202

15,343

1,316.6

13,818

79.8

181.9

1–4

22

67,265

32.7

27

64,718

41.7

55,317

18.1

23.1

5–14

32

200,511

16.0

51

170,355

29.9

145,565

23.3

43.5

15–24

134

174,405

76.8

200

181,677

110.1

138,646

106.5

152.6

25–34

118

122,567

96.3

296

162,066

182.6

135,573

130.6

247.6

35–44

210

113,616

184.8

421

139,237

302.4

162,613

300.5

491.7

45–54

426

114,265

372.8

895

117,811

759.7

134,834

502.7

1,024.3

55–64

784

91,481

857.0

1,196

80,294

1,489.5

87,247

747.7

1,299.5

65–74

1,374

61,192

2,245.4

1,471

48,426

3,037.6

66,037

1,482.8

2,005.9

75–84

1,766

30,112

5,864.8

1,117

17,303

6,455.5

44,842

2,629.9

2,894.8

85+

1,042

7,436

14,012.9

360

2,770

12,996.4

15,508

2,173.1

2,015.5

Total

6,007

1,000,000

6,236

1,000,000

1,000,000

8,195.0

10308.04

Population

Death rates (per 100,000)

Ratio of death rates: Population 2/Population 1

Crude

Age-standardized

Crude

Age-standardized

Population 1

600.7

819.5

1.04

1.27

Population 2

623.6

1,038.0

(p.8) Figure 1–2 depicts age-specific trends in U.S. breast cancer incidence rates among U.S. White women from 1937 to 2003 (Krieger, 2008). It reveals a marked jump in incidence among women age 55 years and older starting in 1980, with rates then falling after 2002. Why?

                   Does Epidemiologic Theory Exist?                   On Science, Data, and Explaining Disease Distribution

Figure 1–2. The rise and perhaps fall of U.S. breast cancer incidence rates. (Krieger, 2008)

Figure 1–3 is the graph of changing trends in mortality among women and men ages 55 to 64 years in England and Wales from 1850 to 1950 that Jerry Morris (1910–2009) used in his classic 1955 article on “Uses of Epidemiology” (Morris, 1955) and with which he opened his pathbreaking 1957 textbook by the same name (Morris, 1957). During this time period, mortality rates fell in both groups, but not evenly so: whereas the male:female mortality ratio was approximately 1.1 in 1850, it was 1.3 in 1920, and 1.9 in 1950. The growing divergence, Morris noted, resulted chiefly from the “emergence of three diseases from obscurity to become exceedingly common, disease which particularly affect men and are very frequent in middle-age: duodenal ulcer, cancer of the bronchus and coronary thrombosis” (Morris, 1957, pp. 1–2). As Morris also wondered: Why?

Figure 1–4 shows maps from the “Worldmapper” project, in which the size of countries is scaled to the size of the outcome depicted: population size, economic resources, and health status (Worldmapper, 2008). Figure 1–4a provides the conventional map of the world, with countries scaled to land size; in Figure 1–4b, the countries are scaled to the size of their population. Figure 1–4c shows the data for “absolute poverty,” defined by the World Bank as living on an income of at most $2 per day; Figure 1–4d displays the data for wealth, as measured by the gross domestic product (GDP). In the former, the African continent and Asian subcontinent loom large; in the latter, the United States, Europe, and Japan are bloated, and the Asian subcontinent shrinks and the African continent dwindles to the merest strand. Figure 1–4e presents data on infant mortality; Figure 1–4f provides data on lung cancer deaths; Figure 1–4g shows data on “often preventable deaths,” defined (p.9)

                   Does Epidemiologic Theory Exist?                   On Science, Data, and Explaining Disease Distribution

Figure 1–3. Trends in mortality by gender and cause of death in England and Wales, 1850–1950, as presented in Morris’s 1955 article on “Uses of Epidemiology” (Morris, 1955) and incorporated into his 1957 pathbreaking textbook Uses of Epidemiology (Morris, 1957, pp. 1–2).

in relation to communicable infections and maternal, perinatal, and nutritional conditions and accounting for one-third of the world’s deaths in 2002; and Figure 1–4h depicts data on sewerage sanitation. In Figures 1–4e and 1–4g, the African continent and Asian subcontinent again loom large, whereas the United States, Europe, and Japan are massively shrunk. In Figures 1–4f and Figures 1–4h, the reverse occurs. Why? (p.10)
                   Does Epidemiologic Theory Exist?                   On Science, Data, and Explaining Disease Distribution

Figure 1–4. Maps from the “Worldmapper” Project (Worldmapper, 2008), in which country size is scaled in relation to the outcome depicted. © Copyright 2006 SASI Group (University of Sheffield) and Mark Newman (University of Michigan).

Figure 1–4a. Countries scaled to land size.

                   Does Epidemiologic Theory Exist?                   On Science, Data, and Explaining Disease Distribution

Figure 1–4b. Countries scaled to population size (2002).

                   Does Epidemiologic Theory Exist?                   On Science, Data, and Explaining Disease Distribution

Figure 1–4c. Absolute poverty (up to $2 per day) (2002).

(p.11)
                   Does Epidemiologic Theory Exist?                   On Science, Data, and Explaining Disease Distribution

Figure 1–4d. Wealth (gross domestic product) (2002).

                   Does Epidemiologic Theory Exist?                   On Science, Data, and Explaining Disease Distribution

Figure 1–4e. Infant mortality (2002).

                   Does Epidemiologic Theory Exist?                   On Science, Data, and Explaining Disease Distribution

Figure 1–4f. Lung cancer deaths (2002).

(p.12)
                   Does Epidemiologic Theory Exist?                   On Science, Data, and Explaining Disease Distribution

Figure 1–4g. Often preventable deaths (communicable infections, maternal, perinatal, and nutritional conditions) (2002)

                   Does Epidemiologic Theory Exist?                   On Science, Data, and Explaining Disease Distribution

Figure 1–4h. Sewerage sanitation (1999).

Finally, Figures 1–5, 1–6, and 1–7 display data from the “Gapminder” project regarding associations between child survival (children dying before age 5 years per 1000 live births) and Gross National Income per capita (Gapminder, 2008). Figure 1–5 depicts these country-level associations for 2006, with the size of each country’s data point scaled to population size, and countries within the same global region shaded the same color. Although it shows an overall robust direct association between child survival and income (the lower the income, the poorer the survival), as underscored in Figure 1–6, at any given level of per capita income, countries vary considerably in their rates of child survival (e.g., South Africa fares worse than Malaysia, despite similar per capita income), and at any given level of child survival, countries vary considerably in their per capita income (e.g., Malaysia fares as well as the United States, despite its lower per capita income). Figure 1–7 in turn presents data on within-region distributions of income in 2000, along with data on within-countries inequities in child survival and income in 2003, for India, Bangladesh, Peru, Guatemala, Yemen, South Africa, and Vietnam. Illuminating (p.13)
                   Does Epidemiologic Theory Exist?                   On Science, Data, and Explaining Disease Distribution

Figure 1–5. Gapminder World Chart 2006: Child survival (children dying before age 5 per 1000 live births) in relation to Gross National Income per capita.

(p.14)
                   Does Epidemiologic Theory Exist?                   On Science, Data, and Explaining Disease Distribution

Figure 1–6. Between-country comparisons of child survival and per capita income, by level of child survival and by per capita income, excerpted from the Gapminder Human Development 2005 presentation (Gapminder, 2008).

Figure 1–6a. Income and child survival inequities: South Africa and Vietnam (2003)

                   Does Epidemiologic Theory Exist?                   On Science, Data, and Explaining Disease Distribution

Figure 1–6b. Income and child survival inequities: South Africa and Malaysia (2003)

                   Does Epidemiologic Theory Exist?                   On Science, Data, and Explaining Disease Distribution

Figure 1–6c. Income and child survival inequities: Malaysia and the United States (2003)

(p.15)
                   Does Epidemiologic Theory Exist?                   On Science, Data, and Explaining Disease Distribution

Figure 1–7. Within-country inequities in child survival and per capita income, excerpted from the Gapminder Human Development 2005 presentation (Gapminder, 2008)

Figure 1–7a. Income distribution by global region

                   Does Epidemiologic Theory Exist?                   On Science, Data, and Explaining Disease Distribution

Figure 1–7b. Income and child survival inequities: within Bangladesh and India (2003)

                   Does Epidemiologic Theory Exist?                   On Science, Data, and Explaining Disease Distribution

Figure 1–7c. Income and child survival inequities: within Peru and Guatemala (2003)

(p.16)
                   Does Epidemiologic Theory Exist?                   On Science, Data, and Explaining Disease Distribution

Figure 1–7d. Income and child survival inequities: Yemen (2003)

                   Does Epidemiologic Theory Exist?                   On Science, Data, and Explaining Disease Distribution

Figure 1–7e. Income and child survival inequities: South Africa (2003)

                   Does Epidemiologic Theory Exist?                   On Science, Data, and Explaining Disease Distribution

Figure 1–7f. Income and child survival inequities: Vietnam (2003)

(p.17) the variability behind the on-average values, these figures show that within-country differences in income and child survival can dwarf between-country differences. Once again: Why?

Before even considering the role of theory in answering these “whys,” it is important to step back and ask: What is the thinking that leads to data allowing these questions even to be posed? And where does theory fit into this process?

One place to begin is to realize that Figures 1–1 through 1–7 are premised on a host of assumptions. What ideas are built into these figures? To start: population rates of disease—a phrase that requires understanding population, rate, and disease. Other ideas at play include: changing incidence rates over time; geographic variation in disease occurrence; and differences in disease rates by social group. None of these ideas are intuitively obvious. They make sense only if one already has a theoretical orientation that finds it compelling and reasonable to think abstractly about populations, about individuals in numerators and denominators, about averages and distributions, about disease occurrence in space and time, and about disease as a definable entity apart from (as opposed to uniquely residing in) the individual persons in whom it is experienced—and hence diseased persons as countable cases.

Prelude, then, to Figures 1–1 through 1–7 are the ideas that would compel someone to collect and display their data. Also notable is who and what is omitted, not simply who and what is included—for example, if the data are or are not separately shown by such social categories as social class, gender, race/ethnicity, sexuality, or by subtypes of disease.

In other words, data are not simply “observed”: there is active thinking behind the act of data acquisition. Not to mention the active thinking that guides data analysis, display, and interpretation.

And this active thinking is the stuff of theory.

Meaning: contrary to its etymologic origins, data are not a “given” (“datum” is the past participle of the Latin verb “dare,” “to give” [OED, 2008; Krieger, 1992]). Nor do data tell stories. People do. An important caveat, however, is that the stories that people who are scientists tell are not simply or simple “stories”: they are (or are supposed to be) transparent accounts, informed by theory, and premised on the public testing of ideas and explanations, using explicitly defined concepts and methods.

So What Is a Scientific Theory?

To appreciate what an epidemiologic theory is (or ought to be), it helps first to have a sense of what counts as a scientific theory—and also: what counts as science. The literature on these topics is vast, contentious, and complex (Mendelsohn et al., 1997; Archer et al., 1998; Ziman, 2000; Collins, 2001; Gould, 2002; Grene & Depew, 2004; Daston & Gallison, 2007; Sober, 2008). That said, some common contemporary criteria for science and scientific theories do exist (see Textboxes 1–1 and 1–2).

To begin, most current scholarship would agree that scientific theories, in contemporary terms, are coherent and presumptively testable sets of inter-related ideas that enable scientists to describe, explain, and predict features of a commonly shared biophysical reality in which cause-and-effect exists (Mendelsohn et al., 1997; Ziman, 2000; Krieger, 2001a). Science, in turn, is both a human activity and a body of knowledge premised on the thinking and action of people to describe and to test their explanations and predictions about features of their commonly shared reality. Particular fields of scientific inquiry are, in turn, distinguished by the domains they seek to understand, the substantive and explanatory (p.18) (p.19) (p.20) (p.21) concepts they use, and the metaphors and mechanisms they employ for their causal explanations (see Textbox 1–2) (Martin & Harré, 1982; Ziman, 2000; Krieger, 2001a). Additionally, those sciences whose domains encompass non-deterministic phenomena (e.g., excluding what are held to be invariant “natural laws,” such as the law of thermodynamics) can further be characterized by historical contingency (meaning what occurs depends on context, hence is not universally invariant)—and among these are the subset of reflexive sciences, which are focused on phenomena that can be influenced by human action (e.g., societal characteristics), such that the explanation adduced can be used to transform that which is being explained (Lieberson, 1992; Archer, 1998; Gannett, 1999; Ziman, 2000; Gadenne, 2002; Krieger, 2001a).

Core to the theorizing and conduct of science are a host of assumptions (Lieberson, 1992; Mendelsohn et al., 1997; Archer et al., 1998; Ziman, 2000; Collins, 2001; Gould, 2002; Grene & Depew, 2004; Daston & Gallison, 2007; Sober, 2008). One such assumption is that we humans live in a commonly shared biophysical (including social) world—and, more broadly, universe—which provides the referent for what we term reality. Another is that this commonly shared biophysical world encompasses diverse processes, structures, and events that are in principal knowable by humans and amenable to scientific investigation. A third is that the existence of this commonly shared knowable biophysical world can be investigated by—and is independent of—any particular human individual. A fourth is that independent humans (in solo and in groups) can independently formulate and test their ideas about “how the world works” and collectively compare ideas, methods, and results. All four of these assumptions are preconditions for the existence and evaluation of scientific theories. More bluntly, no postulated referent reality shared by and accessible to independent humans, no science.

Equally essential is the assumption that causal processes exist. Whether these processes are “deterministic” or “probabilistic” is another question entirely. I note only in passing that (p.22) debates have raged for millennia over the meaning of causality—and, more recently, within a variety of scientific disciplines, over connections between “chance” and “necessity,” and whether “randomness” is “real” or simply a reflection of ignorance of otherwise deterministic causes (Moyal, 1949; Monod, 1972; Stigler, 1986; Desrosières, 1988; Hacking, 1990; Daston, 1994; Gannett, 1999; Weber, 2001; Gadenne, 2002; Russo & Williamson, 2007; Machamer & Wolters, 2007; Groff, 2008). Regardless of the positions argued in these debates, however, the basic point remains that the scientific work of causal inference necessarily presumes that some sort of underlying causal relationship exists, either of the inevitable or contingent variety. Hence, one key corollary to the assumption about a referent reality: no causal processes, no science—and no scientific explanations.

This is all very abstract. It is supposed to be. Science and scientific theories require abstract thinking: to imagine and discern the causal processes behind the observed and postulated specifics, to derive meaning from pattern, and, as the poet William Blake (1757–1827) put it so well, “[t]o see a world in a grain of sand/And a heaven in a wild flower/Hold infinity in the palm of your hand/And eternity in an hour” (Blake, 1977, p. 506). Or, as stated more prosaically by Stanley Lieberson (b. 1933) in a 1991 presidential address to the American Sociological Association: “[T]heory involves generating principles that explain existing information; but it also goes beyond those observations to integrate and account for a variety of other phenomena in ways that would not otherwise be apparent” and would further “‘predict’ all sorts of observations not yet made” (Lieberson, 1992, p. 4).

Why bother with these abstract assertions? Because to understand and evaluate epidemiologic theories, it is important to know what science and scientific theories presume—and what they do not.

First, scientific theories are, by definition, conceptual. But they are not about just any set of ideas. They are instead sets of inter-related ideas intended to explain phenomena in specified domains of the commonly shared biophysical world. Additionally, both the ideas and what they refer to are capable of being independently evaluated and employed by different individuals. Accordingly, some of the concepts in scientific theories pertain to the phenomena that are being described and explained. Others pertain to the causal processes that are theorized to explain the selected phenomena. And both kinds of concepts—substantive and explanatory—are essential for scientific theory; neither alone suffices. What is being explained and how it is being explained are constituent and complementary—and often contested—aspects of scientific theory. Within any given discipline, different theories can exist, simultaneously or successively, offering different and debated explanatory accounts; across disciplines, theories additionally differ because of their respective focus on different aspects of what nevertheless is presumed to be a shared referent reality—whether physical, chemical, biological, or social. A theory of biological evolution, for example, needs not only the concepts of organism, environment, reproduction, and heredity (all of which presumably can in some way be studied by independent investigators) but also the causal ideas (which may be convergent, competing, or complementary) that tie these concepts together to explain the occurrence of evolution (Mayr, 1982; Eldredge, 1999; Gould, 2002; Grene & Depew, 2004; Sober, 2008).

Moreover, to express the ideas at play, scientific theories inevitably employ a combination of metaphor and mechanisms—metaphor to convey concepts describing both phenomena and causal processes and mechanisms to explain the pathways between cause and effect (Lakoff, 1980; Osherson et al., 1981; Martin & Harré, 1982; MacCormac, 1985; Young, 1985; Holton, 1988; Krieger, 1994; Keller, 1995; Krieger, 2001a; Keller, 2002). As I have noted in prior essays, this use of metaphor in scientific theories—essential for enabling the “unknown” to be comprehended in terms of the “known”—can simultaneously free and constrain thought (Krieger, 1994). A salient example, relevant to (p.23) epidemiology, concerns the widespread—and now increasingly contested—metaphor of DNA as the “blueprint” or “master program” for the organism (Watson, 1968). This conceit, as pointed out by the biologist Richard Lewontin (b. 1929) (Lewontin, 2000, pp. 10–11), has dominated the genetics research agenda since the mid-twentieth century. Attesting to its widespread acceptance are the statements of prominent scientists, such as Sydney Brenner (b. 1927; Brenner, 2002), who in 1968 asserted, “The goal of molecular biology is to be able to compute an organism from a knowledge of its genes” (Melnechuk, 1968), and Walter Gilbert (b. 1932; Gilbert, 1980), who in 1992 declared that the complete sequencing of the human genome will enable us to know “what makes us human” (Gilbert, 1992, p. 84). Explicit articulation of the “blueprint” metaphor, moreover, was likewise provided in 1992 by James D. Watson (b. 1928), one of the co-discoverers of the double-helical form of DNA, who declared that the human genome constitutes “the complete genetic blueprint of man (sic),” arguing, “if you can study life from the level of DNA, you have a real explanation for its processes” (Watson, 1992, p. 164), a statement echoed in one newspaper account of the first full sequencing of the human genome on June 26, 2000: “The blueprint of humanity, the book of life, the software for existence—whatever you call it, decoding the entire three billion letters of human DNA is a monumental achievement.” (Carrington, 2000). Although this architectural/computer programming conceit may initially have fruitfully guided genetic research (with the idea of DNA being “in command”), it is increasingly understood to disregard how DNA—and biological development—is dependent on and subject to myriad exogenous influences on gene regulation and expression (Keller, 1992; Keller, 1995; Gilbert, 2000; Lewontin, 2000; Keller, 2002; Van Speybroeck et al., 2002). The key point is that the concepts employed by scientific theories—whether to describe phenomena or causal processes—are not simply self-evident terms. Instead, they are usually rife with connections to other concepts—which is only to be expected, as theories, by definition, must employ interrelated ideas, and the people who use and develop these theories must employ words and symbols that convey these ideas to others interested in understanding them.

Second, the scientific assumption that there is a commonly shared biophysical world is a precondition for science—even as this assumption does not presume this referent reality is commonly perceived or understood by all individuals. Depending on people’s specific characteristics and worldviews, individuals within and across different societies and time periods may vary in their perceptions and interpretations of any given biophysical phenomenon. At a fairly trivial level, color-blindness in particular individuals does not mean the absence of reflected light at the frequency at which these individuals are color-blind (Gibson, 1979). At a more profound level, different individuals may agree on the existence of the same set of associations—for example, when the sun passes below the horizon, it gets dark—and yet may have completely different interpretations of why these associations exist (e.g., because the sun is passing through the underworld; because the sun revolves around the Earth and has moved to location where it is not observable by the person on Earth; or because the Earth revolves around the sun and has rotated to a point where the sun is no longer observable by a person on that point of the Earth’s surface; Hanson, 1958). Or, more epidemiologically, the shared observation of an association between two variables—say, race/ethnicity and disease—does not mean the variables or their association are comprehended in the same way. Whereas some might deem “race” a biological characteristic that explains the observed association (Burchard et al., 2003), others might argue instead that racism, and its associated socially-constructed categories of race/ethnicity, is what has causal relevance (Krieger, 2005). That said, disputes over the causal ideas at issue—and the substantive phenomena under study—nevertheless presume that there is a common reality to which they refer; otherwise, (p.24) attempting to elucidate the reasons for disagreement—and testing competing hypotheses—would be impossible.

Third, scientific observation is not a passive phenomenon: what we “see” and apprehend depends on the ideas we have about we expect—and do not expect—to “see” and our technical capacity to do so (Fleck, 1935 [1979]; Hanson, 1958; Daston & Gallison, 2007). In one sense, this means meaningful observation is, at some level, theory-laden: what we “see” depends in part on what our ideas are about what we expect to see and what assumptions underlie the methods used to “observe” the data. If our theoretical ideas do not include micro-organisms, we would not devise methods to see them—and if offered a microscope, we would not know what we are seeing, regardless of the magnification employed. Similarly, if we do not have the idea of birth cohort effects, we will not “see” their impact on a population’s age-specific disease incidence rates. For example, whereas Johannes Clemmesen (b. 1908) in the late 1940s (Clemmesen, 1948) saw the slight dip in the period’s breast cancer incidence rates after age 50 as evidence that the risk of the disease was lower in women just older than 50 compared to those just younger than 50 and those in their late 50s and older (Figure 1–8a), Brian MacMahon (1923–2007) in the late 1950s saw this same pattern as evidence of a change in risk among women who reached age 50 before rather than after the mid-twentieth century (Figures 1–8b to 1–8d; MacMahon, 1957)—and others since have explored the impact of age–period–cohort effects on the observed yearly incidence of breast cancer (Krieger et al., 2003; Chia et al., 2005).

                   Does Epidemiologic Theory Exist?                   On Science, Data, and Explaining Disease Distribution

Figure 1–8. Data on breast cancer incidence: differing interpretations by Clemmesen and MacMahon (MacMahon, 1957)

Figure 1–8a. Clemmesen’s age-specific breast cancer incidence data for Denmark (1943–1947)

(p.25)
                   Does Epidemiologic Theory Exist?                   On Science, Data, and Explaining Disease Distribution

Figure 1–8b. MacMahon’s analogous age-specific breast cancer incidence data for Connecticut (1935–1951)

                   Does Epidemiologic Theory Exist?                   On Science, Data, and Explaining Disease Distribution

Figure 1–8c. MacMahon’s re-expression of the Connecticut data for specific age groups, by birth cohort

(p.26)

                   Does Epidemiologic Theory Exist?                   On Science, Data, and Explaining Disease Distribution

Figure 1–8d. MacMahon’s re-expression of the Connecticut age-specific incidence data, by birth cohort

In another sense, meaningful observation is experience-laden: we need familiarity not only with the concepts at issue but also the experience of looking at the data themselves and working with the methods to do so. In other words, trained judgment (Daston & Gallison, 2007). Or, as Ludwig Fleck (1896–1961) wrote in the early twentieth century, even with the expectation that when we look through a microscope we will see cells and micro-organisms, we need to learn to prepare the sample with appropriate methods (e.g., stains) and likewise need to learn to “see,” to decipher what is “signal” and what is “noise” (based on theory-laden ideas about what is being observed) (Fleck, 1929; Daston, 2008); the same holds for when we look at epidemiologic data. These statements do not mean that when we do science, we can “see” just anything we please. What counts as scientific evidence is not idiosyncratic; it is instead bound to the assumption of a shared biophysical world and the replicable, contestable, and debatable work of scientists, conducted in the public domain and collectively interpreted and argued.

Fourth, science is by definition fallible—in part because the testing of evidence and ideas, with or without new technologies, can result in the refinement and at times partial (and occasionally wholesale) replacement of explanatory theories, leading to new insights and new predictions as well as new interpretations (or dismissals) of prior observed associations (Fleck, 1935 [1979]; Cohen, 1985; Mendelsohn et al., 1997; Ziman, 2000; Sober, 2008). The recognition that science yields provisional and fallible knowledge, however, does not render all scientific knowledge equally tentative: some theories and their diverse (p.27) predictions have withstood repeated tests; some hypotheses have been tested only a handful of times. For example, the scientific evidence that biological evolution occurs is rich and robust to the point where scientists concur its existence is a fact—even as lively scientific controversies exist over the causal processes at play (Mayr, 1982; Eldredge, 1999; Gould, 2002; Grene & Depew, 2004; Eldredge, 2005; Sober, 2008).

The testing and evaluation of scientific theories, however, as recognized by an enormous literature, is multifaceted and complex and involves debates over methods as well as substance (Fleck, 1935 [1979]; Lieberson, 1992; Mendelsohn et al., 1997; Ziman, 2000; Gadenne, 2002; Grene & Depew, 2004; Daston & Gallison, 2007; Archer et al., 1998; Sober, 2008). Rarely, if ever, does it simply follow the pristine hypothetico-deductive logic of particular observations refuting entire theories—a stance famously postulated by influential philosopher of science, Sir Karl Popper (1902–1994; Popper, 1959, 1985), and one that has been subjected to serious critique in contemporary philosophy of science (Hacking, 2001; Collins, 2001; Mjøset, 2002; Sober, 2008) (even as it has had its share of adherents in epidemiology [Rothman, 1986; Rothman, 1988; MacClure, 1995]) as well as some epidemiologic critics [Susser, 1986; Pearce & Crawford-Brown, 1989; Krieger, 1994; Greenland, 1998]). The theory of general relativity, for example, does not mean Newtonian mechanics are wrong, but rather that the latter is a subset of the former, applicable only at certain spatiotemporal scales (Hanson, 1958; Holton & Brush, 2001). Moreover, an important asymmetry exists between evaluating results from a particular study to (1) decide if they are compatible with a particular theory versus (2) determine how much they strengthen or weaken confidence in a theory (Lieberson, 1992). In part, this is because even if the study results are accurate and valid, it is highly implausible a given data set contains enough elements to test all competing hypotheses (especially under alternative sets of conditions). Thus, as noted by Lieberson, in the case of probabilistic theories, “a theory may be correct even if there is negative evidence” (Lieberson, 1992, p. 1)—and understanding why this can occur requires in-depth consideration of the conditions under which certain associations would or would not be expected.

More deeply, however, science is fallible because as historians and other analysts of science have extensively documented (Fleck, 1935 [1979]; Rose & Rose, 1980; Desrosières, 1988; Holton, 1988; Hubbard, 1990; Rosenberg & Golden, 1992; Keller, 1995; Massen et al., 1995; Mendelsohn et al., 1997; Lock & Gordon, 1988; Ziman, 2000; Keller, 2002; Harraway, 2004; Longino, 2006), scientists are part of the societies in which they are raised and work and, consequently, both think with—and sometimes challenge—the ideas and beliefs of their times. The eighteenth to nineteenth century scientific shift from a constrained biblical time-scale to expansive notions of “deep time” not only reflected fundamental changes in theories of geology, cosmology, physics, and biology but also constituted a profound rupture with dominant and deeply held religious views (Mayr, 1982; Gould, 1987; Holton, 1988; Eldredge, 2005). Closer to home for epidemiology are the powerful and painful connected examples of scientific racism and eugenics and their views of innately biologically inferior and superior “races”—which, far from being “crackpot” theories, were widely accepted and promoted by leading scientists in the nineteenth and the first half of the twentieth centuries (Chase, 1977; Harraway, 1989; Harding, 1993; Kevles, 1995; Gould, 1996; Banton, 1998; Harris & Ernst, 1999; Allen, 2001; Proctor, 2003; Lewontin et al., 1984; Jackson & Weidman, 2004; Stern, 2005a). Their lingering influence on how epidemiologists and others analyze racial/ethnic—and also socioeconomic—health inequities remains a topic of considerable concern (Krieger, 1987; Muntaner et al., 1996; Stern, 2005b; Krieger, 2005; Duster, 2006; Braun et al., 2007).

Fifth and finally, science is not the sole arbiter of knowledge, and scientific theories are not the only path to wisdom. It would be hubris to think otherwise (and not just because of (p.28) myriad changes in what is “scientifically known”). Knowledge and insights generated and obtained from the arts, the humanities, healing practices, and religious and spiritual beliefs can be of profound importance—and, like scientific knowledge, can be profoundly destructive as well. Although these alternative approaches to different kinds of knowledge are not, by definition, premised on the scientific approach of the testing ideas by independent individuals using methods and data that in principle are public, they can—on other grounds—question whether particular scientific studies are immoral, unethical, or in violation of human rights and hence should not be done (Chase, 1977; Kevles, 1995; Proctor, 2003; Gould, 2003; Lavery et al., 2007). Beyond this, they can challenge underlying assumptions that scientific theories posit about “how the world is” or “works”—and, hence, potentially raise critical questions that can be addressed empirically. Challengers engaged in the repeated waves of effort to discredit scientific racism, for example, by and large came from outside the ranks of science, even as some of these critics worked with scientists to make the scientific case (Chase, 1977; Krieger, 1987; Harding, 1993; Kevles, 1995; Allen, 2001; Jackson, & Weidman, 2004). In this example and many others (Longino, 2006), the “non-science” criticisms of scientific theories and purported evidence were brought into the public domain of debate about publicly testable ideas. Recent contentious debates over scientific evidence (e.g., in the case of so-called “creationist science” or “abstinence-only” sex education) additionally underscore there is a world of difference in testing hypotheses about and debating the evidence and its implications—versus ignoring, distorting, or fudging the evidence or contriving invalid “tests” of ideas (Mooney, 2005; Schulman, 2006; Sober, 2008). Unsubstantiated opinion is insufficient to counter the empirical findings produced by science; valid counter-evidence matters.

A useful metaphor, employed by John Ziman (1925–2005; Ziman, 2000), accordingly posits scientific theory is a map: one can never map reality per se, but one can construct and test different representations of this reality. Defining aspects of any given scientific discipline thus include: the domain of phenomena it seeks to explain, the theories it uses to explain and predict the phenomena within the specified domain, and the methods employed to test competing and potentially refutable hypotheses suggested by these theories.

What Is Epidemiologic Theory?

Enough about science in general. What about the science of epidemiology? What would considerations about the nature of science and of scientific theories (as listed in 1–2) lead us expect to be key features of epidemiologic theory?

First and foremost: that epidemiologic theory would exist. (This would seem to be an obvious statement—yet, as I discuss below, it is, in fact, a contentious one.)

Second: that the content of epidemiologic theory, like that of any scientific theory, would be premised on the domain it seeks to explain. In the case of epidemiology, this domain concerns “population distributions of disease, disability, death, and health and their determinants and deterrents, across time and space” (Krieger, 2001b; see Textbox 1–2). A corollary is that if the types, rates, or distributions of diseases and causes of death change over time, it would follow that epidemiologic theories would need account for these changes in their explanations of the population patterning of health.

Third: that epidemiologic theory would necessarily employ domain-specific substantive concepts and explanatory concepts relevant to describing and analyzing extant and changing population distributions of health, disease, and well-being (see Textbox 1–2). (p.29) (p.30) An additional expectation is that both types of concepts—substantive and explanatory—would be informed by metaphors, which in turn would influence the kinds of causal mechanisms proposed.

Fourth: that there would not be “one” epidemiologic theory but, rather, many—even as they all would share the common domain-specific focus of explaining the population occurrence of disease, disability, death and health. Given diverse and changing societal and technological contexts, it would be expected that different epidemiologic theories, employing different substantive and explanatory concepts, would exist, within and across different time periods and societies.

Fifth: that the process of developing, testing, refining, and at times replacing epidemiologic theories would involve data (“observations”) that are theory-laden and whose use and interpretation is experience-laden. And it would likewise involve methods that are themselves theory-laden and whose use and interpretation is likewise experience-laden. That said, the conceptual ideas of epidemiologic theory—both substantive and explanatory (i.e., causal)—would be distinguishable from the methods (and their underlying theories) used to test epidemiologic theories and hypotheses.

Sixth: that epidemiologic theories would be influenced by—even as they might contest—the ideas and beliefs of the societies in which they are formulated. The implication is that epidemiologic theory would reflect not only the theorizing conducted within the field of epidemiology but also likely the responses to and engagement with epidemiologic theory by the diverse sectors of the populations and societies whose health is being described and analyzed.

One implication of these statements for the scope and mandate of epidemiologic theory is that they clarify that epidemiologic theories are, in essence, what I have termed theories of disease distribution—a shorthand phrase meant to be inclusive of the population occurrence of not only disease but also of disability, death, health, and well-being (Krieger, 2001c). They accordingly cannot be reduced to—even as they must incorporate—explanations of disease mechanisms (Krieger, 1994; Krieger, 2000; Krieger, 2001a; Krieger, 2001b). Consider, for example, the epidemiology of diseases related to tobacco use. It is unlikely that the mechanisms by which cigarette smoking causes lung cancer have notably changed between the early and late twentieth century (plus or minus alterations in cigarette additives) (Brandt, 2007, pp. 360, 393; Rabinoff et al., 2007); by contrast, the social patterning of cigarette smoking in the United States and many Western European countries changed dramatically, from initially being more common among professionals and affluent populations to becoming increasingly concentrated among working class and impoverished populations, a pattern emerging in other regions of the world as well (Barbeau et al., 2004; Graham, 2007; Davis et al., 2007). The implication is that explanations of the changing epidemiology of lung cancer and other smoking-related disease requires considering not only specific disease mechanisms but also factors leading to the changing and differential distribution of the exposure.

Consequently, explaining disease distribution is not the same as explaining disease mechanism. By the same token, theories of disease distribution are not the same as theories of disease causation. Nevertheless—and this is key—epidemiologic theories of disease distribution require appraising whether postulated disease mechanisms are compatible or not with observed spatiotemporal and social patterns of disease distribution. In other words, can the hypothesized mechanisms account for increases, declines, or stagnation of rates over time, space, and social group? If not, is this because still other mechanisms contribute to the observed disease distribution? Is it because the wrong time-scale was used to evaluate the exposure–outcome association? Or, alternatively, is the posited mechanism (p.31) simply (or not so simply) wrong? The thinking enabling the triangulation of disease distribution data with hypothesized disease mechanisms is yet another reason why epidemiologic theories of disease distribution are essential (Davey Smith & Egger, 1996; Krieger, 2001a).

The mandate of epidemiology, however, imposes yet one more requirement for epidemiologic theory, one not necessarily shared by sciences unconcerned with human (or other living) populations. As articulated by Morris, writing a half-century ago in his now classic text, Uses of Epidemiology (Morris, 1957), the promise—and responsibility—of epidemiology was clear: to generate scientific knowledge about the “presence, nature, and distribution of health and disease among the population” (Morris, 1957, p. 96), ultimately to “abolish the clinical picture” (Morris, 1957, p. 98). Stated another way, the objective of epidemiology, as long-argued by many leading epidemiologists (Morris, 1957; Terris, 1979; Lilienfeld & Lilienfeld, 1982; Susser, 1989), and as underscored in the “Ethics Guidelines” issued by the American College of Epidemiology in 2000 (American College of Epidemiology, 2000), is to create knowledge relevant to improving population health and preventing unnecessary suffering, including eliminating health inequities (Krieger, 1994; Krieger, 2001a; Krieger, 2000; Krieger, 2007a; Krieger, 2007b).

Hence, an additional reflexive feature of epidemiologic theory is that it seeks to generate valid knowledge that people can use to change the distribution of disease, the very phenomenon that the theory seeks to understand (an intellectual challenge and tension also evident in many of the social sciences; Lieberson, 1992). Not that all epidemiologists would agree: in 1998, some prominent epidemiologists felt impelled to assert, in the face of rising epidemiologic discussions about investigating societal determinants of health, that although “moral purpose of epidemiology is to alleviate the human burden of disease,” epidemiologists should nevertheless be free “to pursue knowledge for its own sake without fear of being badgered about the practical relevance of their work” (Rothman et al., 1998). Granted, the biological fact that we are mortal creatures, who are born and who die, means it is not in the scope of epidemiology—or any science—to eliminate the world of all morbidity and mortality. Yet, to the extent there is spatiotemporal and/or social variation in the age-specific patterns of any particular health outcome, it suggests modifiable causes are at play, whose mechanisms could presumably be altered by informed action.

In summary, key features of any epidemiologic theory, as one type of scientific theory and as summarized in Textbox 1–2, necessarily include interrelated sets of ideas—including both substantive and explanatory concepts—for describing, explaining, and ultimately transforming population distributions of health, disease, and well-being. It can likewise be expected that the ideas of epidemiologic theories, and the metaphors through which they are expressed and the mechanisms they propose, are influenced by the historical and societal context in which the epidemiologic theories are formulated, debated, and bolstered, modified, or rejected. A robust analysis of epidemiologic theory accordingly requires attention to each key aspect, in context, as the next few chapters will show.

And yet this listing of expected characteristics of epidemiologic theory, derived from analysis of critical aspects of science and scientific theories, rests on one very big “if.” It is premised on the logic that if epidemiology is a science, then it must have a scientific theory, hence epidemiologic theory must exist. But this holds ONLY if epidemiology is a science. Conversely, if epidemiology is not a science, but is instead something else, then there need not be any epidemiologic theory. The standard hypothetico-deductive approach would accordingly posit that if evidence of epidemiologic theory cannot be found, then epidemiology is not a science. The next section considers whether this is a reasonable approach.

(p.32) Epidemiologic Theory in Practice: “Lost-and-Found” or Something Else?

A logical place to look for discussion of epidemiologic theory—that is, theories of disease distribution—is in epidemiologic textbooks. This is because textbooks, a key sign of disciplinary institutionalization (Altbach et al., 1991; Apple & Christian-Smith, 1991; Keith & Ender, 2004; Topham, 2000; Morning, 2008), are designed by one generation of scholars to train the next in the fundamental issues of their field: the accumulated body of knowledge, the relevant theories, important controversies, and, in the case of science, the diverse methods used to generate evidence and test hypotheses (Krieger, 1994). Also shaping the content of textbooks is the context of their times, referring not only to the extant state of knowledge but also societal attitudes about what is being taught—especially regarding such controversial topics as the origins and evolution of life, sexuality, and the structure of human societies (Altbach et al., 1991; Apple & Christian-Smith, 1991; Keith & Ender, 2004; Mooney, 2005; Roughgarden, 2004; Morning, 2008). In the case of textbooks for health professionals, for example, recent content analyses have explored ways in which content has been affected by implicit and explicit assumptions about gender, race/ethnicity, sexuality, disability, and aging (Lawrence & Bendixen, 1992; Mendelsohn et al., 1994; Rabow et al., 2000; Byrne, 2001; Tompkins et al., 2006; Macgillivray & Jennings, 2008). A related but different question is how textbooks portray the theories of their fields.

In 1994, I conducted the first systematic evaluation of the coverage of epidemiologic theory in epidemiologic textbooks (Krieger, 1994). My search began with the first generation of books published as epidemiologic textbooks, which appeared in the late 1950s (all in the English language) and extended up to the early 1990s (with my search limited to books published in English, still the dominant scientific language for epidemiology). My strategy was to see how much text, if any, each textbook devoted to epidemiologic concepts pertaining to explaining disease distribution and also the history of ideas in the field. Table 1–1 shows both the original results (Table 1–1a), to which I have now added my findings for additional textbooks, including those published through 2007 (Table 1–1b), and I supplement both sets of results by newly adding, in the last column, the definition of epidemiology offered by each text, if any. All textbooks appearing in Tables 1–1a and 1–1b were published in English, with several of them translated into multiple languages and serving as foundational texts for epidemiology courses worldwide.

A curious pattern emerges. It would seem that from the late 1950s up to about 1980, epidemiologic textbooks, although not featuring overt discussion of epidemiologic theory as such, nevertheless typically did include sections on the history of epidemiologic thinking about disease in populations and the kinds of concepts needed to generate epidemiologic hypotheses. Exemplifying this approach was the stance Morris took in his 1957 text, Uses of Epidemiology, in which he wrote (Morris, 1957, p. 3):

In this book I am concerned mainly with epidemiology as a way of learning, of asking questions, and getting answers that raise further questions: that is, as a method.

From 1980 until the mid-to-late 1990s, this type of discussion virtually disappeared, with the emphasis instead shifting to a different sort of epidemiologic methods: technical methods, understood in relation to study design, data analysis, and causal inference. Since the latter part of the 1990s, however, several of the newer epidemiologic textbooks have again begun to include text on ideas germane to epidemiologic theories of disease distribution. Nevertheless, considering the last half-century of influential and mainstream epidemiologic textbooks, it is striking to note that none of these texts has a section explicitly focused on epidemiologic theories of disease distribution.

(p.33) One interpretation of the evidence presented in Table 1–1 is that the lack of serious attention to epidemiologic theory means epidemiology is actually not a science and hence has no need of domain-specific explanatory theories. The most usual form that this argument takes is that epidemiology is about methods, meaning it offers a “toolkit” of methodological approaches for obtaining and analyzing data on disease in populations, with epidemiological concepts equated with concepts referring to epidemiologic methods (Rothman, 1988; Mawson, 2002; Morabia, 2004). In keeping with the Popperian tradition of treating the origins of scientific hypotheses and theories as outside of the bounds of scientific inquiry (Popper, 1959, 1985), this orientation holds that asking whence epidemiologic questions arise is not in epidemiology’s domain. The focus is on applying methods, rather than on the source(s) of the questions being asked to which these methods are being applied. Less philosophically, another plausible reason that the origins of epidemiologic questions get little special attention is of the “it’s obvious” variety: epidemiologic questions simply build off the extant evidence—including its contradictions, gaps, and need for replication.

An alternative interpretation might start with a “lost-and-found” approach, recognizing the typically less-than-linear development of scientific thought (Mayr, 1982; Ziman, 2000; Krieger, 2000; Keller, 2002; Gould, 2002). That is, when the field of epidemiology was first producing textbooks, those epidemiologists who wrote their groundbreaking texts had inklings of theories of disease distribution—if not in whole, then in part. Subsequent textbooks then somehow either took for granted or lost these theoretical bearings in the 1980s, only to start to find them again in the latter part of the 1990s.

Offering some evidence in support of this latter interpretation are two different bursts of articles in the epidemiologic literature. Thus, around the time that epidemiologic textbooks began shifting in the late 1970s and early 1980s to their more technical methodological orientation, several articles written by leading epidemiologists trained in an earlier generation began to raise alarm at the increasingly technical bent of the field, which they felt was losing sight of the public health import of the questions being asked (Terris, 1979; Stallones, 1980; Najman, 1980; Lilienfeld & Lilienfeld, 1982; Susser, 1985; Susser, 1989). Also during the late 1970s, a brief flurry of articles and letters debated definitions of epidemiology, including its status as a science (Lilienfeld, 1978; Frerichs & Neutra, 1978; Abramson, 1979; Evans, 1979). Subsequently, starting in the mid-1990s, a new round of epidemiologists, many trained by and reacting to their heavily methodological textbooks, began to publish articles calling for the development of explicit epidemiologic theory (Krieger, 1994; McMichael, 1995; Link & Phelan, 1995; Susser, 1996; Pearce, 1996; Davey Smith & Egger, 1996; Victora et al., 1997; Berkman & Kawachi, 2000; Ben-Shlomo & Kuh, 2002; Carpiano & Daley, 2006; Popay, 2006; Dunn, 2006; Vågerö, 2006), with their arguments reflected in the newer textbooks published in the early part of the twenty-first century (see Table 1–1b).

If correct, however, this alternative “lost-and-found” interpretation raises more questions than it answers. First, how can a science misplace its domain-specific theories? Second, given that epidemiologists were nevertheless busily—and fruitfully—conducting studies to describe population patterns of disease and to generate evidence to test etiologic hypotheses about these patterns throughout this entire period, from whence did their hypotheses come? And on what sorts of ideas and theories were epidemiologists drawing before 1950?

To answer these questions, and those I posed at the outset about the ideas animating and spurred by Figures 1 to 7, a more nuanced approach to epidemiologic theory is warranted. In the following chapters, I will accordingly examine the diverse array of ideas that people have elaborated, in different places and different times, to explain the population patterning (p.34) of health—including during the latter half of the twentieth century, when epidemiology seemingly was “atheoretical.” And the three-pronged argument I will make is that:

  1. 1. theories of disease distribution are vital to the conduct of epidemiology;

  2. 2. these theories all too often inform epidemiologic research implicitly, rather than explicitly; and

  3. 3. analysis of epidemiologic theories of disease distribution can improve the intellectual rigor of the field.

In other words, theory is a practical necessity, not an obscure luxury.

Table 1–1. Analysis of English-Language Epidemiologic Textbooks for Content on Epidemiologic Theory: 1922 to 2007

Table 1–1a. Initial Survey of U.S. Epidemiologic Textbooks and Anthologies Published Since 1960: Content on Epidemiologic History and Theory*, and Diagram of “Web of Causation” (Krieger, 1994)

Text

Total Pages

Percent of pages

New addition: definition of epidemiology

History

Theory

Diagram of web

MacMahon B, Pugh TF, Ipsen J. Epidemiologic Methods. Boston: Little, Brown, & Co. 1960.**

302

0

11.6

+

p. 3: “Epidemiology is the study of the distribution and determinants of disease prevalence in man.”

Fox JP, Hall CE, Elveback. Epidemiology: Man and Disease. New York: Macmillan, 1970.

339

3.5

44.8

p. 1: “Epidemiologic curiosity centers about the causation of disease in human populations.”

Susser M. Causal Thinking in the Health Sciences: Concepts and Strategies of Epidemiology. New York: Oxford University Press, 1973.

181

12.2

8.3

p. 1: “In a current definition, epidemiology is the study of distribution and determinants of states of health in human populations. This definition has room for most present-day activities of epidemiologists. Some prefer to add that these activities are for the purpose of prevention, surveillance, and control of health disorders in populations. This addition emphasizes a determinant of health that weighs heavily in public health and medicine, namely, such conscious intervention in health matters as societies elect to undertake.”

Mausner JS, Bahn AK.Epidemiology: An Introductory Text.Phelapdelphia:Saunders, 1974.

377

0.0

4.0

+

p. 3: “Epidemiology may be defined as the study of the distribution and determinant of diseases and injuries in human populations.” (bold in original)

Friedman G. Primer of Epidemiology. New York: McGraw-Hill, 1974.

230

0.0

0.9

+

p. 1: “Epidemiology is the study of disease occurrence in human populations.”

White KL, Henderson M (eds). Epidemiology as a Fundamental Science: Its Uses In Health Services Planning, Administration, and Evaluation. New York: Oxford University Press, 1976.

235

0.9

0.9

p. 19: “However defined, epidemiology implies methods and strategies used to identify and study that which determines the level and distribution of health and disease in the community.”

Lilienfeld A, Lilienfeld D. Foundations of Epidemiology. New York: Oxford University Press, 1980.

375

6.1

5.1

p. 3: “Epidemiology is concerned with the patterns of disease occurrence in human populations and of the factors that influence these patterns.”

Kleinbaum DG, Kupper LL, Morgenstern H (eds). Epidemiologic Research: Principles and Quantitative Methods. Belmont, CA: Lifetime Learning Publications, 1982.

529

0.0

1.1

p. 2: “As exemplified by John Snow’s famous work on cholera, epidemiology was initially concerned with providing a methodological basis for the study and control of population epidemics. Currently, however, epidemiology (italics in the original) has a much broader scope—namely, the study of health and illness in human populations.”

Schlesselman J. Case–Control Studies: Design, Conduct, Analysis. New York: Oxford University Press, 1982.

354

0.6

0.0

None provided.

Kahn HA. An Introduction to Epidemiologic Methods. New York: Oxford University Press, 1983.

166

0.0

0.0

None provided.

Miettinen OS. Theoretical Epidemiology: Principles of Occurrence Research in Medicine. New York: Wiley, 1985.

359

0.0

1.4

p. vii: “This text treats theoretical epidemiology as the discipline of how to study the occurrence of phenomena of interest in the health field.” (italics in the original)

Feinstein AR. Clinical Epidemiology: The architecture of Clinical Research. Philadelphia: W B Saunders Co., 1985.

812

1.1

1.2

p. 1: “Clinical epidemology is concerned with studying groups of people to achieve the background evidence needed for clinical decisions in patient care.”

Weiss N. Clinical Epidemiology: The Study of the Outcome of Illnesses. New York: Oxford University Press, 1986.

144

0.0

0.0

pp. 3–4: “Epidemiology per se is the study of variation in the occurrence of disease, and the reasons for that variation … Clinical epidemiology is defined here in a parallel way: It is the study of variation in the outcome (italics in the original) of illness and of reasons for that variation.”

Rothman K. Modern Epidemiology. Boston: Little, Brown, 1986.

358

1.7

0.0

p. 23: “The clearest of many definitions of epidemiology that has been proposed has been attributed to Gaylord Anderson. His definition is: ‘Epidemiology: the study of the occurrence of illness.’ Other sciences are also directed toward the study of illness, but in epidemiology the focus is on the occurrence (italics in the original) of illness.”

Kelsey J, Thompson WD, Evans AS. Methods in Observational Epidemiology. New York: Oxford University Press, 1986.

366

0.0

7.4

p. 3: “Epidemiology, the study of the occurrence and distribution of disease and other health-related conditions in populations, is used for many purposes.”

Hennekens CH, Buring JE. Epidemiology in Medicine. Boston: Little, Brown, 1987.

383

2.6

3.9

p. 3: “… a useful and comprehensive definition of epidemiology: ‘the study of the distribution and determinants of disease frequency’ in human populations.”

Abramson JH. Making Sense of Data: a Self-Instruction Manual on the Interpretation of Epidemiologic Data. New York: Oxford University Press, 1988.

326

0.0

0.6

None provided.

Anthologies

Winklestein W Jr, French FE, Lane JM (eds.). Basic Readings in Epidemiology. New York: MSS Educational Pub. Co., 1970.

193

13.9

27.8

“Epidemiology may be defined as the study of disease distributions and the factors that influence them. Epidemiology shares with other disciplines an interest in the natural history of disease and the utilization of the scientific method. Its distinctiveness is more related to the design and execution of studies than to content and conclusions.”

Greenland S (ed). Evolution of Epidemiologic Ideas: Annotated Readings on Concepts and Methods. Chestnut Hill, MA: Epidemiology Resources, Inc., 1987.

190

7.9

0.0

None provided.

Buck C, Llopis A, Najera E, Terris M (eds). The challenge of Epidemiology: Issues and Selected Readings. Washington, DC: Pan American Health Organization, 1988.

989

14.8

24.9

p. x: “Besides its importance and usefulness in disease surveillance and prevention, epidemiology has an even more critical function to carry out—the gathering of knowledge for understanding the health-disease process. It can anticipate needs, identify risk conditions, and orient the definition of priorities and the use of available resources for planning and administering health systems. In short, by analyzing and evaluating health problems and health services, and their contexts, epidemiology can go beyond considering just specific health problems: it can help bring us closer to considering society as the source for explaining health problems and their solutions.”

Rothman K (ed). Causal Inference. Chestnut Hill, MA: Epidemiology Resources, Inc., 1988.

207

0.0

0.0

None provided.

(*) Epidemiologic theory: defined as explicit discussion of theories of disease causation and/or epidemiologic concepts (e.g., “time, place, person”) (NB: This footnote is per the 1994 text; I would now instead refer to “theories of disease distribution” [rather than “theories of disease causation”], and I would further clarify that the text pertaining to epidemiologic theory is that which provides guidance on theories and concepts required to generate epidemiologic hypotheses, to develop substantive explanations for patterns of disease distribution [as distinct from the methods to test the ideas])

(**) In the original 1994 table, I cited the 1970 version of MacMahon et al.; I have changed it to the 1960 version in this table.

(p.35) (p.36) (p.37) (p.38)

Table 1–1b. Additional Selected Introductory and Advanced Textbooks Not Included in the Original Review (1922–2007)

Text

Total Pages

Percent of pages

New addition: definition of epidemiology

History

Theory

Diagram of web

Vaughan VC. Epidemiology and Public Health: A Text and Reference Book for Physicians, Medical Students and Health Workers. Vol. 1: Respiratory Infections. St Louis, MO: CV Mosby, 1922.

683

34.0

7.8

p. 23: “We may be asked for a definition of epidemiology. It is the science of epidemic diseases, and these may appear in any given community at any given time singly or by the hundreds.”

Greenwood M. Epidemics and Crowd-Diseases: An Introduction to the Study of Epidemiology. New York: Macmillan, 1937.

378

32.5

5.0

p. 10: “Epidemiology came to mean the study of disease, any disease, as a mass phenomenon … the epidemiologist’s unit is not a single human being but an aggregation of human beings, and since it is impossible to hold in mind distinctly a separate mass of the particulars he (sic) forms a general picture, on average of what is happening, and what works upon that.”

Taylor I, Knowelden J. Principles of Epidemiology. Boston: Little Brown & Co., 1957.

292

1.4

7.9

p. 1: “One of the most fundamental tasks of the epidemiologist is to describe the pattern of disease in communities, whether national or some smaller groups … (in relation to ‘which diseases,’ ‘what persons are most affected., ‘when the disease occurs,’ and ‘where the disease is found’) … Armed with answers to these questions, which together describe the pattern of disease in a community, the epidemiologist may postulate theories of the mode of spread of the disease he (sic) finds, and these theories may be put to the test by clinical, field, or laboratory studies … Finally, having made his (sic) epidemiological diagnosis, he (sic) may be able to put forward logical ideas for the control of those diseases he (sic) describes” (italics in the original).

Morris JN. Uses of Epidemiology. Edinburgh: Churchill Livingston, 1957.

131

6.9

15.3

p. 5: “Epidemiology may be further defined as the study of health and disease of population and groups in relation to their environment and ways of living” (italics in the original).

Kark SL. Epidemiology and Community Medicine. New York: Appleton-Century Crofts, 1974.

463

2.6

14.3

p. 1: “The function of epidemiology is to study health in population groups.”

Barker DJP, Rose G. Epidemiology and Medical Practice. 2nd ed. Edinburgh: Churchill Livingston, 1979.

148

0.0

4.7

p. v: “Epidemiology, the study of the distribution and determinants of disease in human populations, has always been an integral part of medical practice.”

Ahlbom A, Norell S. Introduction to Modern Epidemiology. Chestnut Hill, MA: Epidemiology Resources, 1990.

100

0.0

2.0

+

p. 1: “Epidemiology is the science of occurrence of diseases in human populations.”

Walker AM. Observation and Inference: An Introduction to the Methods of Epidemiology. Chestnut Hill, MA: Epidemiology Resources, 1991.

165

0.0

0.0

None provided.

Beaglehole R, Bonita R, Kjellstrom T. Basic Epidemiology. Geneva: World Health Organization, 1993.

153

1.3

4.6

p. 3: “Epidemiology has been defined as ‘the study of the distribution of health-related states or event in specified populations, and the application of this study to the control of health problems’” (citing definition from Last J (ed). A Dictionary of Epidemiology. 2nd ed. New York: Oxford, 1988).

Friis RH, Sellers TA. Epidemiology for Public Health Practice. Gaithersburg, MD: Aspen Publishers, 1996.

406

2.5

4.2

p. 4: “Epidemiology is concerned with the distribution and determinants of health and disease, morbidity, injuries, disabilities, and mortality in populations.”

Young TK. Population Health: Concepts and Methods. New York: Oxford University Press, 1998.

306

3.6

4.7

p. 7: “The Dictionary of Epidemiology defines epidemiology as ‘The study of the distribution and determinants of health-related states or events in specified populations, and the application of this study to control health problems.’

Brownson RC, Petitti DB (eds). Applied Epidemiology: Theory to Practice. New York: Oxford University Press, 1998.

387

1.0

1.3

p. ix: “In our view, applied epidemiology synthesizes and applies the results of etiologic studies to set priorities for interventions; it evaluates public health interventions and policies; it measures the quality and outcome of medical care; and it effectively communicates epidemiologic findings to health professionals and the public.”

Berkman LF, Kawachi I (eds). Social Epidemiology. New York: Oxford University Press, 2000.

382

6.0

16.2

p. 3: “Epidemiology is the study of the distribution and determinants of states of health in populations.”

Rothman K. Epidemiology: An Introduction. New York: Oxford University Press, 2002.

217

0.9

1.8

p. 1: “Often considered the core science of public health, epidemiology involves ‘the study of the distribution and determinants of disease frequency,’ or, put even more simply, ‘the study of the occurrence of illness.’”

Bhopal RS. Concepts of Epidemiology: An Integrated Introduction to the Ideas, Theories, Principles, and Methods of Epidemiology. Oxford: Oxford University Press, 2002.

296

1.0

16.6

+

p. xxii: “… in short it (epidemiology) is the science and craft that studies the patterns of disease (and health, though usually indirectly) in populations to help understand both their causes and the burden they impose. This information is applied to prevent, control or manage the problems under study.”

Aschengrau, A, Seage GR. Essentials of Epidemiology in Public Health. Sudbury, MA: Jones and Bartlett, 2003.

447

5.1

3.1

+

p. 6: “The study of the distribution and determinants of disease frequency in human populations and the application of this study to control health problems.” (italics in the original)

Gordis L. Epidemiology, 3rd ed. Philadelphia, PA: WB Saunders, 2004.

323

1.2

1.5

p. 3: “Epidemiology is the study of how disease is distributed in populations and the factors that influence or determine this distribution.”

Webb P. Essential Epidemiology: An Introduction for Students and Health Professionals. New York: Cambridge University Press, 2005.

323

2.5

1.8

pp. 1–2: “Epidemiology … is about measuring health, identifying the causes of ill-health, and intervening to improve health … Perhaps epidemiology’s most fundamental role is to provide a logic and structure for the analyses of health problems both great and small.”

Fletcher RH, Fletcher SW. Clinical Epidemiology: The Essentials. 4th ed. Baltimore, MD: Lippincott Williams & Wilkins, 2005.

243

0.0

1.1

p. 3: “Epidemiology is the ‘study of disease occurrence in human populations.’” (bold in the original)

Oakes JM, Kaufman JS (eds). Methods in Social Epidemiology. San Francisco, CA: Jossey-Bass, 2006.

460

5.0

20.4

p. 3: “Epidemiology is the study of the distribution and determinants of states of health in populations.”

Yarnell J (ed). Epidemiology and Prevention: A Systems-Based Approach. Oxford: Oxford University Press, 2007.

275

1.1

2.5

p. 5: “At the beginning of the twenty-first century, epidemiology is a broad-based population science, drawing on many disciplines from biology and sociology to biostatistics and philosophy of science, which investigates the causes of human disease and methods for their control.” (bold in the original)

Szklo M, Nieto FJ. Epidemiology: Beyond the Basics. 2nd ed. Sudbury, MA: Jones and Bartlett, 2007.

482

0.6

0.6

p. 3: “Epidemiology is traditionally defined as the study of the distribution and determinants of health-related states or events in specified populations and the application of this study to the control of health problems.” (italics in the original)

(p.39) (p.40) (p.41)