Results from well-conducted randomised controlled studies should ideally inform on the comparative merits of treatment choices for a health condition. In the absence of this, one attempts to use evidence from the impact of treatment when administered according to decisions of the physicians and the patients (observational evidence). Naïve comparisons between treatment options using observational evidence will lead to biased results. Under certain conditions, however, it is possible to obtain valid estimates of the comparative merits of different treatments from observational data. Causal inference can be conceptualised as a framework aiming to provide valid information about causal effects of treatments using observational evidence. It can be viewed as a missing data problem in which each patient has two outcomes: the observed outcome under the treatment actually received and a counterfactual (unobserved) outcome had the patient received a different treatment. Methodological developments over the last decades clarified the appropriate conditions and methods to obtain valid comparisons. This article provides an introduction to some of these methods.
Statistics from Altmetric.com
Competing interests None declared.
Provenance and peer review Not commissioned; externally peer reviewed.
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.