Abstract:

Clinical reasoning is an iterative process in which physicians transform patient information to diagnose patients and create treatment plans by synthesizing information (Cluter, 1985; Kuiper, 2013). Although diagnostic skills are fundamental to medical practice it is a basic skill for reaching a diagnosis, pedagogical approaches to reproduce this skill remain scarce (Chamberland, et al., 2011). For creating effective pedagogical techniques of medical diagnosis, we first need to understand clinical reasoning.

Therefore, this study aims to increase the understanding of clinical reasoning by examining the differences in the clinical reasoning behaviors between high and low performing medical students. We analyzed the amount of clinical reasoning behaviors they performed and the patterns of behavior transition. We examined five discrete steps to final diagnosis (Artino, et al., 2014; Kuiper, 2013):
1. selecting relevant symptoms of the patient’s history (referred to as ‘adding symptoms’);
2. searching for supporting evidence from laboratory tests and literature (referred as ‘additional studies’);
3. selecting differential diagnosis ( referred as ‘adding hypotheses’);
4. doing connections between the evidence and the potential diagnosis (referred as ‘linking’);
5. categorizing and prioritizing evidence that support the selected diagnosis (referred as ‘prioritizing’).

Medical students from a large North American University volunteered to solve a clinical case in BioWorld (Lajoie, 2009), a technology-rich learning environment that simulates a hospital environment. Their performance was measured by percentage of overlap with an expert’s solution. Based on their score, they were categorized as high (n = 14) or low performing (n = 14). BioWorld logfiles were used to classify students’ clinical reasoning behaviors.
To identify group differences, a series of Chi2 analyses were conducted. Low performers performed more ‘adding symptoms’ (p=.002), and ‘additional studies’ (p<.001). High performers performed more ‘adding hypotheses’ (p=.007), and ‘prioritizing’ (p<.001).

Analysis of transitions of clinical reasoning behaviors showed statistically significant differences (p<.001). Low performers had more transitions across: 1) ‘adding symptoms’ followed by ‘additional studies’ (p<.001) and ‘linking’ (p<.001); 2) ‘additional studies’ followed by ‘adding symptoms’ (p<.001), ‘additional studies’ (p<.001), and ‘linking’ (p=.001); 3) ‘linking’ followed by ‘prioritizing’ (p=.002); 4) ‘adding hypotheses’ followed by ‘adding symptoms’ (p=.001), ‘additional studies’ (p<.001), and ‘prioritizing’ (p<.001). High performers had more consecutive ‘adding hypotheses’ behaviors (p<.001); and ‘prioritizing’ behaviors (p<.001).

We found that low performing students tend to focus more on adding evidence. In contrasts, high performers focused more in selecting potential diagnosis, and organizing the supporting evidence. Similarly, in patters of behaviors transitions, low performers had more iterations across different behaviors, whereas the high performers were more likely to repeat the same kind of behaviors.

Our findings could be used to identify cognitive caveats experienced during low performance, and adaptive behavioral traits in medical diagnosis. Thus, the results presented in this paper could provide insight to create new pedagogical approaches for learning clinical reasoning skills.

Keywords:

Clinical reasoning, learning performance, medical education, technology-rich learning environments.