Clinical reasoning: how experienced doctors think, and how to learn it

Watch an experienced clinician with a new patient and the diagnosis often seems to arrive before the history is finished. It looks like intuition. It is not magic, and it can be learned — but only if you understand what is actually happening in the expert's head.

Two systems, one diagnosis

Clinical reasoning runs on two modes that cognitive psychologists call System 1 and System 2 ^[1]. System 1 is fast pattern recognition: the breathless young woman on the pill with a swollen calf is a pulmonary embolism until proven otherwise, and the expert knows it in seconds. System 2 is slow, analytic, hypothesis-testing reasoning: listing the differential, weighing probabilities, ordering the discriminating test.

Experts use both, and the skill is knowing when to switch. System 1 is efficient and usually right in typical presentations; it is also where most diagnostic errors are born when the pattern is subtly wrong. System 2 is the safety net — slow, effortful and underused when the ward is busy.

Illness scripts: what experts actually store

The difference between a student and an expert is not more facts but a different organisation of them. Through repeated encounters, clinicians compress textbook knowledge into illness scripts — compact mental packages that hold, for each disease, the typical patient (who gets it), the time course, the key findings and the things that must be excluded ^[2]. When a new patient walks in, the expert is not reasoning from physiology; they are matching the patient against hundreds of stored scripts.

This is why reading alone does not create clinicians. Scripts are built by seeing patients — real or simulated — and getting feedback on what you missed.

The problem representation

The single most useful habit you can learn early is the one-sentence problem representation: a summary of the case in abstract, clinical terms using semantic qualifiers — acute versus chronic, unilateral versus bilateral, sharp versus dull, progressive versus intermittent ^[3].

"A 24-year-old nulliparous woman with acute-onset, unilateral, severe lower abdominal pain, six weeks of amenorrhoea and a positive pregnancy test."

That sentence already contains the differential: ectopic pregnancy until proven otherwise, then ovarian torsion, then a complicated cyst. Students who cannot produce this sentence usually cannot produce the differential either, because they are still holding the case as a pile of unsorted details.

Where reasoning fails

In a classic study of diagnostic errors in internal medicine, the most frequent cognitive cause was premature closure — settling on a diagnosis and stopping the search once it felt plausible ^[4]. Its relatives are well known:

• Anchoring — locking onto the first impression (the referral letter says "gastritis") and under-adjusting as new information arrives.
• Availability — judging the likely diagnosis by what you saw last week.
• Confirmation bias — noticing the findings that fit and explaining away the ones that do not.
• Framing — "the anxious patient" becomes a label that hides a thyroid storm.

The defence is not "be more careful". It is a routine: before committing, run a deliberate diagnostic time-out — What else could this be? What does not fit? What is the worst thing this could be, and have I excluded it?

How to train it

Deliberate practice with cases and feedback is the only method with consistent evidence. Work through many short cases rather than a few long ones; commit to a diagnosis before you see the answer; and after each case, write down what you would do differently. MedIQ's virtual patients are built for exactly this loop — you take the history, choose investigations, commit to a diagnosis and then see what the expert pathway was and which red flags you missed.