We briefly outline diagnostic strategies and introduce a measure of agreement, called kappa, between clinical diagnoses. This statistic is useful not only as a measure of diagnostic accuracy, but also as a means of quantifying and understanding disagreement between diagnosticians. It is disconcerting to many, clinicians included, that given a general deficit of data on sensitivity and specificity, the level of agreement between many clinical diagnoses is only moderate at best with kappa values of 0.3 to 0.6.
Sensitivity, specificity, pretest odds, and posttest probability of disease are defined and related to the interpretation of clinical findings and ancillary diagnostic test results. An understanding of these features and how they relate to ruling-in or ruling-out a diagnosis, or minimizzing diagnostic errors will greatly enhance the diagnostic accuracy of the practitioner, and reduce the frequency of clinical disagreement. The approach of running multiple tests on every patient is not only wasteful and expensive, it is unlikely to improve the ability of the clinician to establish the correct diagnosis.
We conclude with a discussion of how to decide on the best therapy, a discussion which centers on, and outlines the key features of, the well designed clinical trial. Like a diagnosis, the results from a clinical trial may not always be definitive, nonetheless it is the best available method of gleaning information about treatment efficacy.
相似文献