10 minute read
Problem representation
A problem representation compresses a case into one sentence built from semantic qualifiers: who (age, sex, key context), tempo (acute, subacute, chronic; constant or episodic), and syndrome (the abstracted clinical pattern). 'A 58-year-old man with diabetic CKD on RAAS blockade presents with subacute weakness and oliguria after GI losses' is a search key the mind can use; a shapeless retelling is not.
Illness scripts and hypothesis testing
Experienced clinicians match representations against compact disease scripts: who gets it, how fast it moves, what it shows, what finds it, and what separates it from its neighbors. Learning explicitly in script form (who / time course / symptoms / findings / investigations / differentiators) builds the same machinery on purpose.
Each new datum should do work: raise a hypothesis, lower one, or split two. Data that cannot change any ranking is not worth collecting yet.
Failure modes
Premature closure (stopping at the first fit), anchoring (staying with the triage label), and confirmation bias (collecting only agreeable data) cause more ward harm than ignorance does. The antidote is procedural, not moral: force one 'what else could this be?' before every disposition, and schedule the reassessment that gives reality a chance to disagree.
Pearls
- Semantic qualifiers (acute versus chronic, unilateral versus bilateral, exertional versus positional) are the vocabulary of expertise.
- The differential is a ranked, living list, re-sorted by each result, not a recital.
Connected