Now that we have LR+, we multiply it by the pretest probability to get the post-test probability:

Post-Test Probability = 0.05 x 2.85 = 14.3%.

This approach for diagnosis and management of the patient seems very appealing; being able to calculate an exact probability of disease seemingly eliminates many issues in diagnosis! Unfortunately, Bayes theorem breaks down in clinical practice for many reasons. First, a large amount of data is required at every step to update the probability. No physician or database has access to all the contingency tables required to update the Bayes theorem with every historical element or lab test result discovered about the patient. Second, this method of probabilistic reasoning is unnatural for humans to perform. The other techniques discussed are much more conducive to a performance by the human brain. Third, while the model may work for single diseases, it doesn’t work well when there are multiple diseases and comorbidities. Finally, and most importantly, the assumptions of conditional independence and exhaustiveness and exclusiveness that are fundamental to the Bayes theorem don’t hold in the clinical world. The reality is that symptoms and findings are not completely independent of each other; the presence or absence of one finding can influence that of many others. Together, these facts render the probability calculated by the Bayes theorem to be inexact and even misleading in most cases, even when one succeeds in calculating it. Nevertheless, Bayes theorem is important in medicine for many subproblems when ample evidence is available (for example, using chest pain characteristics to calculate the probability of MI during the patient history).