Applicability of diagnostic studies--statistics, bias and estimates of diagnostic accuracy

Reviewing and using diagnostic research for decision making involves complex issues about what are the exact diagnostic questions, how they should be studied, and to whom the results of such studies apply in real life. In this paper we aim to address some of the main issues concerning applicability of diagnostic research by looking at different diagnostic questions, what different study designs can be used and how bias and variability may impact on applicability. Users of diagnostic research should be aware of these issues in order to avoid confusion and misunderstandings about why modern diagnostic research addresses particular patient groups and uses certain study designs, whilst choosing to ignore others that at first glance seem relevant. We conclude that there are 4 main points to be addressed in doing and using diagnostic research, and these are: "Get the question right"; "Get the study design right"; "Include patients for whom the test will also be used in practice"; and "Educate users of research". Simple as they may seem, these points cover extremely complex issues in practice, and these need to be addressed by more communication between methodologists, practitioners and decision makers.