This ought come to no surprise to various proponents of Bayesian reasoning – in the expected sense that outputs of testing individuals with inappropriate pretest probabilities will be garbage.
This is a prospective evaluation of consecutive patients referred for office-based stress myocardial perfusion imaging with single-photon emission computed tomography (SPECT-MPI). These authors evaluated referring clinician notes and patient histories and stratified 10-year coronary risk, chest pain syndromes, and categorized studies as appropriate, uncertain, and inappropriate.
Of the 1,707 patients referred for SPECT-MPI, 1,511 had complete follow up and were classifiable. Pre-test appropriateness of referrals varied – but 47% of primary care referrals were inappropriate and 28% of cardiologist referrals were inappropriate, with inappropriate referrals for women patients exceeding men. The kicker – abnormal MPI in the appropriate-testing group was associated with poorer outcomes, while abnormal tests in the inappropriate-testing group had no such association.
One of the basic principles of diagnostic testing is choosing patients for whom we can have faith in the results. In our world of flawed tests, when the likelihood of a positive result is low, the balance between true positives and false positive tilts to favor misleading rather than valid diagnoses. In a minority of instances, these tests may yet be appropriate – but, as we can see, cardiac testing under questionable circumstances provides no true prognostic value. These are the population costs and harms of our zero-miss culture.
“The Impact of Appropriate Use on the Prognostic Value of SPECT Myocardial Perfusion Imaging”
http://www.ncbi.nlm.nih.gov/pubmed/24021779