Tools for critical appraisal of evidence in studies of diagnostic accuracy

Studies of accuracy are often more complex to understand than clinical trials, since there can be more than one outcome and scope (screening, diagnosis, and prognosis) and because results have to be reported in more than one way, than in clinical trials (relative risk or odds ratio). Sensitivity and specificity are common terms for practitioners, but to remember that sensitivity is the "ratio between true positive rate and true positive rate plus false negative rate" may sometime cause some frustration. Moreover, likelihood ratio, predictive values, diagnostic odds ratio, and pre- and post-test probability complicate the framework. To summarize these indexes from multiple studies can be also a little more difficult. However, understanding diagnostic test accuracy from different study results and how to interpret systematic reviews and meta-analysis can help every practitioner improve critical appraisal of evidence about the best use of diagnostic tests. Avoiding complicated mathematical formulas, this paper attempts to explain the meaning of the most important diagnostic indexes and how to read a Forest plot and a summary Receiver Operative Characteristic curve.