The differential isn't a list. It's a probability cloud.

You learned the differential as a bulleted list: chest pain → MI, PE, dissection, pericarditis, GERD, costochondritis, anxiety. Then you tried to use the list on a real person and it fell apart. Here's why.

A list has no shape

A list says these things are possible. It says nothing about how likely each one is for the human in front of you, or how much you'd regret missing it. Those two axes — probability and cost of being wrong — are the entire game.

Re-draw it as a cloud

Picture every diagnosis as a dot on two axes:

• x-axis: how likely, given this patient (age, history, the actual story).
• y-axis: how bad if you miss it.

Costochondritis is likely but cheap to miss. Aortic dissection is unlikely but catastrophic to miss. They are nowhere near each other on the cloud, even though they sit one line apart on your flashcard.

You don't work down a list. You work toward the corner — high likelihood or high lethality first.

Why this scales

Every test you order is just an attempt to move a dot. A D-dimer doesn't "rule out PE" — it shoves the PE dot left, toward less likely. Pre-test probability is literally where the dot started. Bayes isn't an exam topic; it's the floor plan.

Once you see the cloud, you stop asking "what else could it be?" and start asking "what's the cheapest move that changes where the dangerous dots sit?" That's the whole job.