Population Health and Causal Inference

Abstract

This chapter is concerned with the theoretical nature of causal inferences and effect measurement in epidemiology. Following Bird, I outline Hill’s nine criteria for causal inference in medicine, and demonstrate why evidence-based medicine deems well-conducted experimental studies to be a more reliable means of identifying causation than observational studies. I then present two models of causal inference commonly employed by epidemiologists. First, the potential outcomes approach is put forward as a means of identifying practicable (Woodward 2004), manipulable causes of disease, and measuring the strength of the causal effects of exposures. We shall see that this approach does not accommodate all causal factors relevant to epidemiological practices. To resolve this issue, I suggest supplementing the potential outcomes approach with Rothmans (1976) sufficient-cause model.