Premium
Initial diagnostic hypotheses bias analytic information processing in non‐visual domains
Author(s) -
McLaughlin Kevin,
Heemskerk Laura,
Herman Robert,
Ainslie Martha,
Rikers Remy M,
Schmidt Henk G
Publication year - 2008
Publication title -
medical education
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 1.776
H-Index - 138
eISSN - 1365-2923
pISSN - 0308-0110
DOI - 10.1111/j.1365-2923.2007.02994.x
Subject(s) - medical diagnosis , confidence interval , selection (genetic algorithm) , debiasing , statistical hypothesis testing , information processing , alternative hypothesis , medicine , psychology , cognitive psychology , statistics , mathematics , social psychology , artificial intelligence , computer science , radiology , null hypothesis
Context Previous studies have shown that an initial diagnostic hypothesis biases automatic information processing. It is unclear if an initial hypothesis has a similar effect on analytic information processing. Our first objective was to study the effect of an initial diagnostic hypothesis on analytic processing. Our second objective was to assess the effect of clinical experience on analytic processing by evaluating the effect of clinical frequency and urgency of an alternative diagnosis on diagnosis selection. Methods During a 12‐minute objective structured clinical examination station, 19 subspecialty medical residents diagnosed the cause of 3 clinical presentations: dyspnoea; headache, and chest pain. Subjects were randomly allocated cases for which the suggested initial hypothesis was either correct or incorrect. For cases with an incorrect initial hypothesis, the alternative diagnoses varied in the frequency with which they are encountered in clinical practice, and their clinical urgency, relative to the initial diagnostic hypothesis. Results All correct initial hypotheses were retained, compared with 10.9% of incorrect hypotheses. All cases with a correct initial hypothesis were diagnosed correctly, compared with 65.2% of cases with an incorrect hypothesis (risk ratio 1.5 [95% confidence interval 1.2–1.9], P = 0.02). Clinical frequency and urgency were not associated with alternative diagnosis selection. Discussion Our results suggest that an initial diagnostic hypothesis biases analytic processing. The data used to reject an initial hypothesis appear to drive selection of an alternative hypothesis. Further studies aimed at finding strategies for increasing the likelihood of generating a correct initial hypothesis or debiasing an initial hypothesis are needed.