Causation and Correlation in Medical Science: Theoretical Problems

Abstract

Establishing causal relations is a core enterprise of the medical sciences. Understanding the etiology of diseases, and the treatments to reduce the burden of disease, is in fact an instantiation of the very many activities related to causal analysis and causal assessment in medical science. In medicine, correlations have a “Janus” character. On the one hand, we should beware of correlations as they do not imply causation—a well-established “mantra” in statistics and in the philosophy of causality. On the other hand, correlations are a very important and useful piece of evidence in order to establish causal relations—a line of argument that is currently debated in the philosophical and medical literature. Understanding the limits and potentialities of correlations in medicine is all the more important if we consider approaches in “data-intensive science,” where the search for correlations in big data sets is key in the medical sciences, and the rise of complexity approaches, in which quantitative methods are complemented with qualitative ones.