THE PAST MAY DISAMBIGUATE THE PRESENT: PRIOR EXPERIENCE LEADS TO UNAMBIGUOUS INTERPRETATIONS OF AMBIGUOUS MEDICAL SYMPTOMS IN NOVICES

Purpose: Many medical areas have features with high variability of appearance. Prior experience with unambiguous cases may shape the interpretation of individually ambiguous features. Under certain circumstances participants may be able to view a previously interpreted feature as bi-stable (have two interpretations).

Method: Novices were trained to competence in simple diagnostic categories (n=40 per study). Features constructed to have two possible interpretations (ambiguous features) were shown in isolation (Study 1) to establish baseline rates of bi-stability. In Study 2, ambiguous features were included within overall unambiguous written vignettes during training. Features were later included in test cases with multiple possible diagnoses, and tested again in isolation to determine participants’ ability to identify bi-stability. In Study 3, midway through test cases, participants were shown a statement indicating that features could be interpreted in multiple ways. Similar methods were used in Study 4 & 5, with family history and peer-based feedback used to encourage re-interpretation. 

Result: In Study 1 features were identified as bi-stable 46% of the time. With a cued interpretation during training (Study 2), participants assigned more diagnostic probability to the disorder supported by the cued interpretation (54% to 40%, F(38)=9.9, p<0.05), and when features were presented in isolation, reported features as bi-stable 26% of the time. In Study 3, participants showed a decreased reliance on familiarity, and reported bi-stability was 40%. In Study 4 & 5, participants’ reported bi-stability was 31% and 26%, respectively. In all studies, participants reported higher diagnostic probability for the diagnosis supported by the familiar interpretation.

Conclusion: Participants assigned significantly more diagnostic probability to the diagnosis supported by the cued interpretation of the ambiguous feature, and were less likely to identify multiple interpretations for those symptoms than were control subjects. The likelihood of reporting bi-stability appeared to be related to the strength of the underlying causal relationship presented mid-way through the test phase. That is to say, the studies in which a clear causal mechanism was used to encourage re-interpretation (e.g. family history) showed higher rates of reported bi-stability, where those with little causal reasoning (e.g. peer feedback) showed little improvement over baseline rates of reported bi-stability. This suggests a strong role for familiarity, however some conditions can encourage some re-interpretation of ambiguous features.