Striking a Nerve: Docs Flunk Diagnostics Math
Written by Editor   
Tuesday, April 29, 2014 10:18 PM

A study in JAMA Internal Medicine suggesting that most newly-minted MDs, and a sizeable chunk of more experienced physicians, don't understand how "positive predictive value" (PPV) works.  It's a commonplace that medical tests are over-prescribed -- many are diagnostic procedures said to be of limited value.  The study by Jain and colleagues points to another explanation -- many doctors may simply fail to understand how uninformative the test results really are, at least when it comes to low-prevalence conditions.

A group based at Harvard Medical School led by Sachin Jain, MD, rounded up 24 attending physicians, 26 house staff, 10 med students, and a retired physician at "a Boston area hospital" and asked them to estimate the PPV of a hypothetical diagnostic test.

Participants were told that the disease detected by the test had a prevalence of 0.1% and that the test's false positive rate was 5% while being 100% sensitive.

"We calculated that the correct [PPV] is 1.96%," Jain and colleagues wrote, and accepted answers of "2%" or "less than 2%" as close enough.

Guess how many answers met those standards? If you said "not many," then you were close enough.  

Another lesson from the study comes from its finding that students and recent medical graduates performed no better than attending physicians.  That suggests medical education is falling short when it comes to concepts surrounding diagnostic testing.