Hi Suhail. In my experience, experienced medical practitioners do not interpret diagnostic tests by dichotomising them first but interpret each value individually. This process can be modelled by creating curves displaying the probabilities of an outcome conditional on medical test results by using logistic regression etc (for example, see Figures 1 and 2 in this recent post Risk based treatment and the validity of scales of effect ). This means that sensitivity and specificity, ROC curves etc are not very applicable to clinical practice and cause confusion; they are tools for use public health as far as I can understand so I hardly mention them in the Oxford Handbook of Clinical Diagnosis.
Bayes rule is also used rarely in the differential diagnostic thought process, which is based on lists of diagnosis associated with individual symptoms, signs or test results. Each diagnosis has a ‘posterior’ probability conditional on the test result based on the directly observed frequency of each diagnosis in those with the finding. If a diagnosis occurs in more than one list of the findings of a patient then this increases the probability of that diagnosis. I explain the probability theory of this process too in the Oxford Handbook of Clinical Diagnosis.