Clinical Reasoning
Clinical reasoning is the cognitive process by which clinicians gather and interpret information, generate and refine hypotheses, and arrive at diagnostic and management decisions. It combines pattern recognition with analytical thinking and is a core competency that simulation and case-based practice are designed to develop.
Current accounts of clinical reasoning use a dual process model: a fast, largely automatic retrieval of likely diagnoses that experienced clinicians recognize from having seen the pattern before, and a slower, conscious, analytical working through of the evidence. Norman and colleagues set this out in Academic Medicine in 2017 and argued something that gets lost in the retelling, which is that errors arise from both processes and that the likelihood of error falls as knowledge and expertise grow.
That has a direct consequence for teaching. Handing learners a list of cognitive biases feels productive and does little on its own, while giving them many varied cases and specific feedback on where their thinking went wrong builds the knowledge the fast process draws on. Norman and colleagues took issue with the idea that bias instruction is the main lever on diagnostic error, and their argument is worth reading before a program builds a curriculum around it.
Reasoning is also case specific. A resident who works up chest pain well can handle an undifferentiated headache poorly, so a stable read on how someone reasons takes many short cases rather than one long one. A program that judges reasoning from a single case is mostly measuring how familiar that particular case happened to be.
Reasoning also depends on the history a learner actually elicits, so communication and reasoning are not separable skills. A resident who never asks what a family already understands, or who talks over a frightened parent, ends up reasoning from an incomplete picture no matter how good their differential looks on paper. That is the part conversation practice reaches and a written case does not.
What this looks like in a program
- Ask learners to commit to a differential out loud, with a reason for each item, before you give them any additional data.
- Score the history a learner obtained and not only the conclusion they reached, since a right answer from a thin history will not transfer to the next patient.
- Use several short cases across a rotation rather than one long case at the end, because reasoning does not travel from case to case as reliably as faculty assume.
Last updated August 2026