medical diagnosis: diagnostic accuracy and diagnostic research
This tag relates to medical diagnosis, including quantifying diagnostic accuracy, designing diagnostic studies, and optimizing diagnosis. It is a good place to discuss the pros and cons of various measures of diagnostic accuracy, including significant problems inherent in traditional measures such as sensitivity, specificity, positive predictive value, and negative predictive value. These measures assume that the diagnosis is all-or-nothing and likewise for diagnostic tests, and do not properly handle continuous diagnostic markers. Sensitivity and specificity have special problems in that they envision the problem in backwards time order / backwards information flow. This leads to all kinds of interpretation problems, made worse by non-constancy of sensitivity and specificity over different types of patients. Some background and alternative approaches may be found in Biostatistics for Biomedical Research, Chapter 18 under http://biostat.mc.vanderbilt.edu/ClinStat .