Information distortion in physicians' diagnostic judgments

Information distortion suggests that people change the evaluation of new information to support an emerging belief. The present study was designed to measure the extent to which physicians distort incoming medical information to support an emerging diagnosis.

Data were collected via an anonymous questionnaire. The experimental group (102 physicians) read 3 patient scenarios, each with 2 competing diagnoses. Physicians first read information that favored 1 of the 2 diagnoses (the "steer"). They then rated a series of neutral cues that favored neither diagnosis. At each cue presentation, respondents rated the extent to which cues favored either diagnosis and updated the strength of their diagnostic belief. After the neutral cues in the third scenario, respondents rated cues that opposed the initial steer. A control group (36 physicians) rated all the cues in random order and not within scenarios, thus providing unbiased baseline ratings for calculating distortion in the experimental group.

Distortion was statistically significant (P < 0.001) and was associated with the strength of belief in the leading diagnosis. Physicians with over 10 years in practice distorted less than their less experienced counterparts ( $$\overline{X}$$ = 1.04 v. $$\overline{X}$$ = 1.78, P < 0.05). Having developed an initial diagnostic leaning consistent with the steer, 56% of physicians remained committed to it after receiving the conflicting cues. Distortion was strongly associated with commitment to the steer (odds ratio, 1.4; 95% confidence interval, 1.03-1.79; P = 0.03).

Physicians did not elicit information; therefore, the authors cannot estimate the size of distortion in tasks involving information search.

Distortion could partly explain commitment of physicians to an early diagnosis. Both distortion and strength of initial diagnostic belief seem to decline after 10 years in family medicine.