Confirmation bias is not always bad

Jesus Rivas jesus.rivasce at GMAIL.COM
Wed May 25 19:50:19 UTC 2016

Hello friends,

First, excuse my poor

I’d like talking to you about a case I saw in my hospital a few weeks ago.
It’s about a 62 years´old woman with a 15 days history of cough and
intermittent hemoptisis. The first presumptive diagnosis was lung
Tuberculosis. Three tests of acid-fast bacilli in sputum and gastric
aspirate were negative. A Chest-X-ray and CT-scan showed a right lower lobe
cavitary lung lesion consistent with bronchiectasis. Ceftriaxone was
initiated. The physician  insisted on TB diagnosis and ordered a fourth
test in sputum and a fiberoptic bronchoscopy. That sputum smear was
positive (4 bacilli per 100 fields) before FOB be done.

In endemic regions like Peru, it´s not difficult to identify typical
pictures of TB (probabilistic and prototypical-based reasoning). However,
if we´re lacking a rapid positive mycobacterial test, diagnosis and
treatment of TB can be omitted o delayed, a serious situation which has
been considered as the main factor of nosocomial propagation of TB. Thanks
to doctor’s insistence on his first clinical suspicion in spite of repeated
negative tests (confirmation bias) cases like this (paucibacillary
tuberculosis) are not missed.

I agree with Monteiro’s opinion that we should call these situations as
“active pursuit of supportive data” and use the label “bias confirmation”
only in the event that the end result is an error1. In defining clinical
reasoning the experts give different weight to several aspects: cognitive,
memory, experience and narrative skill. I think that in some cases like
this there is also a little room for behaviors like “clinical

(1)Teach Learn Med. <> 2013;25
Suppl 1:S26-32. doi: 10.1080/10401334.2013.842911.

Jesús Rivas-Ceballos, MD

Departamento de Medicina. Hospital Grau, Essalud

Lima, Perú

Moderator: David Meyers, Board Member, Society to Improve Diagnosis in Medicine

HTML Version:
URL: <../attachments/20160525/a991e867/attachment.html>

More information about the Test mailing list