A patient who feels substantially better after an intervention is the most persuasive evidence a clinician ever encounters, and the least reliable. Regression to the mean brings people who presented at their worst back towards their average. Natural history resolves a great deal. The attention, explanation and reassurance around a procedure have measurable effects of their own. And the period around an intervention is when medication is reviewed, alcohol is cut, sleep improves and a patient finally takes the diagnosis seriously.
None of that means the procedure did not work. It means that the improvement in front of you does not, on its own, establish that it did - which is precisely why sham-controlled trials keep returning smaller effects than open series.
The clinic version of this is a habit, not a calculation: when someone reports a marked improvement, ask what else changed. New drug, stopped drug, changed job, stopped drinking, started sleeping. Sometimes the answer changes what you do next.
- Ask what else changed in the same period before attributing improvement to the intervention.
- Patients who present at their worst tend to improve regardless - expect regression to the mean.
- Record the baseline severity score, so later improvement can be judged rather than recalled.
- An uncontrolled series and a sham-controlled trial routinely disagree; trust the latter.
- Say this to patients plainly - most accept it, and it protects against over-treatment later.
Why it matters
The single most convincing clinical experience any of us has is also the one most likely to be wrong.
Read the rest in the app
You have read your two free briefings this month. The app carries all 27 specialties, every morning, free — and this finding is waiting in it.
Scan to keep reading on your phone. No account needed to start.
Tomorrow morning, before your first patient
One edition a day for top clinical updates, written by the desk, every claim tied to its paper. Six minutes.
Get the app — free