ABSTRACT
In evidence-based medicine, data from observational studies may be valuable to complement information from randomized controlled trials, especially when data are not available from the latter. The reliability of data from nonrandomized sources is, however, limited due to the potential for confounding bias and the inherent infrastructural impediments that should be overcome. We highlight the major statistical and infrastructural issues associated with such studies, and provide a summary of best practices developed for effective use of the data.
