Taxonomic axes of epidemiologic study designs: a refutationist perspective.

Overlap among the axes of study design used for classifying epidemiologic research creates a taxonomic problem, the complexity of which may be illustrated by Venn diagrams. The diagrams also suggest a solution: to rank axes according to their bearing on study validity. This is consistent with the refutationist criterion for distinguishing strong from weak science--the potential to refute alternative explanations. In epidemiology, this means refuting confounding, reverse-causation bias, selection or allocation bias, and misclassification bias. Using susceptibility to bias as a criterion for ranking axes, a simple taxonomy emerges that is compatible with widespread usage of terminology.