Why epidemiology has not had a larger influence on health policy. As already noted, policy decisions are too often made more on the basis of political ideology, cost savings, pressure from interest groups and media attention than research evidence. Epidemiology is as much affected by this problem as any other scientific discipline. But to some extent, our limited influence is our own fault. North American epidemiologists have tended to emphasize aetiologic research, and to scorn the descriptive epidemiology that is so relevant to policy-making. European epidemiologists have done much better in this regard (as can be seen by comparing the American and International journals of epidemiology; I do not know the situation in Japanese epidemiology, but have the impression that it may be rather closely allied to the laboratory sciences–admirable in its own right, but usually fairly far from policy). Many epidemiologists have preferred to confine their role to “the science”, avoiding the grime of policy-making. There is no doubt that policy-oriented epidemiology is distinctly practical in nature, and that the machinations one must go through in a policy-relevant epidemiologic project can sometimes appear to lack rigour. Students who complete their MSc theses under my supervision are often asked by incredulous examiners, “So, you combined data from [several disparate sources], and you interpolated to make this estimate and extrapolated to make that one [when no one knows the shape of the association], and then you assumed [much more than one would like to assume]?” To which I encourage the students to answer, “Would you prefer to see the policy-maker pull her estimates out of the air?” Imperfect estimates that have the best available empirical basis are usually better than wild guesses (and so far, all of my students have passed!). Of course, there is a longstanding debate regarding the scope of epidemiology: the Dictionary of Epidemiology includes “the application of this study to the control of health problems” in its definition of the discipline, but this part of the definition is often not honoured, at least in North America (it appears to much more honoured in Europe and the Third World; I do not know about Japan). Policy-oriented epidemiology is simply one example of “the application of this study to the control of health problems”–and is the most appropriate approach in cases where problems call for policy solutions. The journal Epidemiology prohibits policy recommendations from scientific articles, arguing that these require different expertise from that required to do the research. Even if this is true (which is by no means clear), it does not question the conduct of policy-relevant research–it just criticizes one format for its presentation. On another front, one can argue that epidemiologists have looked too much at individual factors (strange, for a discipline that defines itself as working at the level of populations), neglecting the population-level determinants that are fundamental to the health of populations. One hopes for more evidence of the influence of Geoffrey Rose, with investigators seeking the reasons for sick populations, not just the reasons for sick individuals. This implies more ecological studies, and especially more multilevel studies, and in this respect things are looking better (as indicated by the American Journal of Public Health series in 1994 and a fair number of recent papers using multilevel analysis).

Our limited influence on policy is not all our fault: some of it is inevitable. Policy-makers tend to come from very different backgrounds from those of epidemiologists (in Canada, public servants come mainly from social science and administration, while politicians come mainly from law), making communication difficult. The few epidemiologists in government (at least in the Canadian government) tend to be in technical positions, well removed from the sectors where policy is made. Policy-makers want “the answer”–not a range of possibilities presented with a bunch of qualifications–and they want it immediately, while epidemiologists are trained to be sceptical (emphasizing possible sources of error rather than providing the unqualified advice that policy-makers want) and cautious (which tends to mean slow). Perhaps partly for this reason, politicians tend to listen to clinicians, the “real doctors”, even on epidemiologic topics: real doctors usually admit to no doubts. The fact that clinical epidemiologists are “real doctors” may explain why they seem to have had a larger influence on health policy than population epidemiologists have had: it certainly is one reason that clinical epidemiologists are so valuable, as real doctors who know epidemiology!