Theory 050/Medicine & Health /Diagnostics
Errors Scale With Test Volume
More information is assumed to improve diagnosis. Past a point it may dilute it.
Diagnostic error rates rise with the number of tests ordered rather than falling, because each additional test adds a chance of a false positive that can displace the correct diagnosis.
01The evidence
Diagnostic error remains common despite increasing test availability, and false positives are a recognised source of harm in screening.
02The tension
Diagnostic strategies assume that more data improves accuracy, on the model of signal detection. That model ignores the cost of false positives in a sequential search.
03The mechanism
Each test has a false positive rate. With many tests, the probability that at least one is falsely positive approaches one, and a false positive that is clinically plausible can anchor the diagnostic process away from the true cause.
That makes error a function of test count, and predicts that error should rise with volume at constant underlying disease.
04What this is not
Those models treat tests as signal. This treats them as noise sources, which predicts harm from volume.
05Predictions
- P1Error should rise with test count.
- P2Reducing low-value tests should reduce error.
- P3Anchoring on false positives should be demonstrable.
06Refutation
Error falling monotonically with test count would restore the information model.
07What it changes
It would support diagnostic stewardship.