Causation, Correlation, Complexity, Confusion, and Oral Health
Donald M. Brunette, PhDFriday, February 7
9:30 am - 10:15 am
To establish causation, the cause and the effect should be connected through a biologically plausible mechanism. The definitive means of establishing causation is through experimentation, such as randomized controlled clinical trials. But such trials are often not possible, and criteria for judging the strength of the evidence for causation from less rigorous study designs demonstrating associations consequently have evolved. The standard causal inference criteria include strength of the association, dose-response relationship, time sequence, consistency, specificity, biologic plausibility, and independence from confounders. Besides frequentist-based statistical models, other ways of investigating relationships, such as Bayesian networks and complexity theory, have been developed. The output of sophisticated statistical analyses can be confusing even for scientists, so there is a need to consider how best to present such information. This presentation will discuss these concepts using examples from the dental literature and focusing on questions dentists can use in assessing causal relationships, as well as methods of presenting their analysis in understandable terms to their patients.