Clinical reasoning and differential diagnosis
Introduced| English | Meaning |
|---|---|
| differential diagnosis/ˌdɪfəˈrenʃl ˌdaɪəɡˈnəʊsɪs/ | The list of possible conditions that could explain the findings. |
| premature closure | Accepting a diagnosis too early, before it is fully checked. |
The first idea is not the last word
- A 58-year-old is breathless. The referral letter says "probable chest infection", and the first doctor agrees within a minute. Three days later the real cause is found: a clot in the lung.
- Clinical reasoning is a method for keeping the list open long enough, and for spotting the thinking traps that close it too early.
From complaint to a broad list
- A differential diagnosis 鉴别诊断 is the list of conditions that could explain the findings. Build it broadly first, often from anatomy: what structures could produce this symptom?
- For breathlessness: heart (failure), lungs (infection, asthma), blood vessels (clot), blood (anaemia), and anxiety.
A broad differential for sudden breathlessness should include which of these? Select all that apply.
Difficulty 2/5 · 140 coins + a small bonus on first completion today
Heart, vessels, lungs and blood can all cause breathlessness. A broad first list costs little; a missing item can cost a life.
Rank by probability and danger
- Rank by how likely each cause is and how harmful it would be to miss. Probability alone misses rare killers; danger alone sends everyone for every test.
- The rule: common diagnoses head the working list, and "must-not-miss" diagnoses stay on it until they have been excluded, however unlikely.
A diagnosis with only a 2% probability can be crossed off the list straight away.
Difficulty 1/5 · 100 coins + a small bonus on first completion today
If it is dangerous (a must-not-miss), it stays until it has been actively excluded, however unlikely.
Tests move the odds
- A test does not give an answer by itself; it shifts the probability. The likelihood ratio (LR) from course 12 does this: post-test odds = pre-test odds × LR.
- (Synthetic) Pre-test probability 20% is odds of 0.25. A positive result with LR+ 9 gives odds 2.25, a probability of about 69%: much higher, but not certain.
- (Fictional) The breathless 58-year-old. Breathless when lying flat: heart failure moves up. A sudden fall in oxygen level with a clear chest X-ray: the clot moves up.
- Each finding reorders the list. The clot stayed on the list as a must-not-miss, so it was still there to move up.
(Synthetic) Pre-test odds are 0.25 and a positive test has LR+ = 9. What are the post-test odds?
Difficulty 2/5 · 140 coins + a small bonus on first completion today
Post-test odds = 0.25 × 9 = 2.25.
Convert post-test odds of 2.25 to a probability, in % to the nearest whole number.
Difficulty 2/5 · 140 coins + a small bonus on first completion today
p = odds ÷ (1 + odds) = 2.25 ÷ 3.25 ≈ 0.69, so about 69%.
Five biases and their antidotes
- Anchoring: sticking to the first impression. Availability: favouring yesterday's memorable case. Premature closure 过早闭合: stopping as soon as one diagnosis fits.
- Confirmation bias: seeking only agreeing evidence. Framing: adopting someone else's label. Antidotes: ask "what else could this be?", look for evidence against, and read the data yourself.
Match each thinking trap to its description.
Difficulty 2/5 · 140 coins + a small bonus on first completion today
Each has a habit that guards against it; the most general one is to ask "what else could this be?"
The working diagnosis
- The result is a working diagnosis: the current best explanation, held until evidence disproves it, together with a plan for what is still uncertain.
- Safety-netting is part of that plan: tell the patient which changes mean they should come back at once.
A working diagnosis is best described as —
Difficulty 1/5 · 100 coins + a small bonus on first completion today
It is provisional, comes with a plan for what is still uncertain, and includes safety-netting advice.
Common slips
- Treating the working diagnosis as final. It is the current best, not the answer.
- Chasing rare diagnoses. The base rate still counts.
- Letting one test decide. Tests change odds; they rarely settle a question alone.
Recap
- Broad list → rank by probability and danger → tests update the odds → working diagnosis with a safety net.
- Must-not-miss diagnoses stay until excluded; post-test odds = pre-test odds × LR.
- Biases: anchoring, availability, premature closure, confirmation, framing; ask "what else?"