Epidemiology 801 Course Syllabus

Department of Epidemiology

Professor James S. Koopman MD MPH

1. Stimulate students to develop a new vision of how their thesis work contributes to the science of epidemiology.
2. Examine how the methods used by epidemiology, its traditions for developing theory, and its traditions for causal inference fit into a broader philosophy of science.
3. Explore issues of discovery, explanation, and inference in epidemiology.
4. Examine the role of Bayesian inference in scientific reasoning.
5. Examine a new paradigm in epidemiology that focuses on the population systems causing disease and clarify how this new paradigm relates to the old one of examining individual risks.
6. Introduce approaches to epidemiology that examine how interactions between individuals lead to the emergence of disease causes and preventives at the population level in a non-linear fashion where individual effects do not sum up to population effects.

Students lead discussions of papers and books selected by the professor. Students may decide to supplement or even subvert the professor's selection of papers. The professor often provides questions to which the discussion can be directed. The students, however, are not obligated to discuss these questions. The readings often do not deal directly to epidemiology. To provide a focus for relating them to epidemiology, we will discuss their relevance to the work of Lynch and Kaplan on the health effects of income disparity. A web page to which students and professors contribute discussion will be maintained regarding the nature of inquiry into whether and how income disparity causes mortality. When students or professors find some relevance of course material to the issue of income disparity effects, they are to type up comments and put them on the course web page.

Doctoral students should not care about grades. They should only care about developing collegial practices contributing to their profession and to writing a good thesis. Therefore, I reserve the right to give everyone an A- unless exceptional circumstances dictate otherwise. 25% of grade is on student leading of discussions, 25% is on participation in other sessions and on contributions to the web page discussion of income inequality effects, 50% is on the final paper.

The final paper should focus on the student's thesis topic. It should deal with the philosophical issues of discovery, explanation, and inference in relation to how the student is pursuing their thesis or what impact they want their thesis to have. It should critically evaluate the base of theory on which the thesis is developed and outline how the thesis will advance that theory.

There is a call for change in the way epidemiology conceptualizes cause and conducts its inquiries. We will discuss some visions of the future of epidemiology and provide time for each student to discuss their research and how it fits into the future of epidemiology.

1. Krieger N. Epidemiology and the web of causation: Has anyone seen the spider? Soc. Sci Med. 1994;39:887-903
2. Wing S. Limits of epidemiology. Medicine and Global Survival 1994;1:74-86
3. Savitz DA. In defense of black box epidemiology. Epidemiology 1994;5:550-2
4. Skrabanek P. The emptiness of the black box. Epidemiology 1994;5:553-5
5. Susser M, Susser E. Choosing a future for epidemiology: I. Eras and paradigms. Amer J Pub Hlth 1996;86:668-73
6. Susser M, Susser E. Choosing a future for epidemiology: I. From black box to chinese boxes and eco-epidemiology. Amer J Pub Hlth 1996;86:674-7
7. Koopman JS. Emerging objectives and methods in epidemiology. Amer J Pub Hlth 1996;86:630-2
8. Winkelstein W. Eras, paradigms, and the future of epidemiology. Amer J Pub Hlth 1996;86:621-2
9. Weed DL. Beyond black box epidemiology. Amer J Pub Hlth. 1998;88:12-14

We want to relate the topics discussed in this course to your thesis work. But instead of focusing on each student's thesis, we will focus on the work of John Lynch and George Kaplan from our department. John will attend the class and contribute to the web page discussion of class material relevant to his work with George. This work raises issues of relevance to the class in terms of the difference between individual and population level effects and the necessity of explanations and theories regarding the findings so that interventions can be appropriately instituted on the basis of the findings.

1. Kaplan GA, Pamuk ER, Lynch JW, Cohen RD, Balfour JL. Inequality in income and mortality in the United States: analysis of mortality and potential pathways.
2. Lynch JW, Kaplan GA. Understanding how inequality in the distribution of income affects health. J Health Psychology 1997;2:297-314
3. Lynch JW, Kaplan GA, Pamuk ER, Cohen RD, Heck KE, Balfour JL, Yen IH. Income inequality and mortality in metropolitan areas of the United States. Amer J Publ. Hlth. 1998;88:1074-80
4. Wilkinson RG. Health inequalities: relative or absolute material standards? British Med J 1997;314:591-4
5. Gravelle H, Wildman J, Sutton M. Income, income inequality and health: what can we learn from aggregate data?

The philosophy of science provides a context for examining how epidemiology relates to other sciences in its objectives, methods, and traditions. In the decade before the present decade, epidemiologists even explicitly discussed philosophy of science in their journals and meetings. These discussions largely focused on inferences about causality. We will examine those discussions. Viewing those discussions in the light of changing objectives for epidemiology provides another focus as to where we have been and where we are going in our discipline. A broader view of philosophy of science, however, is needed to see the place epidemiology occupies among the sciences.

1. Robert Klee. Introduction to the philosophy of science: Cutting nature at its seems Oxford University press 1997
2. Readings from Causal Inference edited by Ken Rothman. Epidemiology Resources Inc. 1988
1. Rothman K. Inferring causal Connections-habit, faith or logic
2. Weed DL. Causal criteria and Popperian refutation
3. Susser M. Falsification, verification and causal inference in epidemiology: Reconsideration in the light of Sir Karl Popper's philosophy
4. Greenland S. Probability vs. Popper: An elaboration of the insufficiency of current Popperian approaches for epidemiological analysis.
5. Petitti DB. The implications of alternative views about causal inference for the work of the practicing epidemiologist.
6. Poole C. Induction does not exist in epidemiology either.
7. Susser M. Rational sciences vs. a system of logic
8. Weed DL. Criticism and its constraints: a self-appraisal and rejoinder
3. Kuhn TS. The structure of scientific revolutions

Bayesian inference now underlies much of modern statistical development. This approach builds on what we know to extend knowledge rather than building inference on the base of rejecting models that are inconsistent with data. Most epidemiology students are being left behind the statistical community because our quantitative training in inference methods has been inadequate. The text by Howson and Urbach can at least help us make our discourse relevant to these new trends.

1. Colin Howson and Peter Urbach. Scientific reasoning: the Bayesian approach Open Court Press 1993
1. Chapters 1, 5, 7, 8, & 9

Our profession has productively founded its logic on rather radical abstractions. First, we have conceptualized cause in terms of relationships between discrete variables that either ignore time or build time into the categorization procedures for variable measurement. Second, we have largely ignored system phenomena that generate important and controllable causal effects. Finally, we have assumed that the structure of interactions between individuals can be ignored in the causal models used to analyze data. We have made this assumption even when the data we are examining deals with interactions between humans. This approach avoids many complexities that just could not be feasibly managed in epidemiological theory before the modern computer age. The fundamental problem in causal inference with this approach is confounding. Confounding can be addressed using the causal abstraction of discrete variables capturing time relationships. Our understanding of the underlying logic of confounding and how to address this problem have advanced considerably in the past decade. Thus, the dominant paradigm in epidemiology seems to be advancing despite the fact that we will argue later that it needs to be superseded. Our first three readings review recent advances in the logical structure of the dominant paradigm. The final three readings begin to show how this logical structure is insufficient for some of our most pressing problems, including the income disparity issues in the research of Lynch and Kaplan.

1. Rothman KJ, Greenland S. Chapter 2: Causation and Causal Inference. 7-28 In Modern Epidemiology edited by Kenneth J. Rothman and Sander Greenland Lippincott-Raven Press 1998
2. Holland PW. Statistics and Causal Inference. Journal of the American Statistical Association 1986;81:945-70
3. Greenland S, Pearl J, Robins JM. Causal diagrams for epidemiologic research. Epidemiology 1998 (Publication pending)
4. Koopman JS. Causal models and sources of interaction. Amer J Epidemiol 1977; 106:439-443.
5. Halloran ME, Struchiner CJ. Causal inference in infectious diseases. Epidemiology 1995;6:142-51
6. Kaufman JS, Cooper RS. Seeking causal explanations in social epidemiology. Amer J. Epidemiol. 1998 (publication pending)

The abstractions of the sufficient-component cause model are too limiting for many areas of needed theory development in epidemiology. Continuous compartmental models of dynamic systems offer an alternative for the construction of epidemiological theory. This tradition has productively addressed important problems in the epidemiology of both chronic and infectious diseases. It provides a systems view as to where causation lies and how prevention should be promoted. It thus expands causal concepts from those of individual risk. This type of modeling is the subject matter of Epidemiology 802 for which an introduction will be provided. Another modern development in the construction of causal models comes out of operations research. The development of discrete event models offers many different ways to advance epidemiology. I will discuss how we are using such models for the development of new information systems for the control of HIV and STDs.

Material for both Epidemiology 802 and 606 will be updated in a manner appropriate for this course. I will try to adapt this material so that it addresses questions that arise in the earlier part of the course.

A new paradigm for scientific investigation and the logic of discovery has arisen in recent years. It comes out of chaos and complexity theory and is based largely in discrete causal models. It addresses not only systems dynamics, but evolutionary and adaptive dynamics as well. We introduce this approach through two works of a major contributor to its development who is an inspiring member of our University community. We hope this will stimulate epidemiology doctoral students to pursue a certificate from the Program for the Study of Complex Systems (PSCS) which you can investigate at http://pscs.physics.lsa.umich.edu/pscs.html

1. John H Holland, Hidden Order: How Adaptation Builds Complexity, Addison Wesley 1995
2. John H Holland, Emergence: From Chaos to Order, Addison Wesley Helix Books, 1997