No. 014 / Original reader worksheet
Trust the evidence,
not the headline.
A four-box field card for checking a dramatic health-study number. Type on this page or print and write by hand. Entries stay on your device.
01. Who is the result about?
Population, selection, sample size, follow-up and exclusions. Who might be different?
02. What actually changed?
Association, experiment or prediction statistic? Write the actual unit and comparison.
03. What does it NOT prove?
Separate prediction from causation and group averages from a personal forecast.
04. What would change my view?
Look for methods, outside validation, calibration, limits and follow-up studies.
Red-flag test: If a headline promises an individual benefit, was that benefit actually tested in an intervention? If it claims a model improved, which metric and validation support that claim?
ETHOS example: 562,234 adults in clinical treadmill testing; model C-statistic 0.662 → 0.682. These compare discrimination of two models, not individual probabilities.
Primary source: Myers et al., 2026, DOI 10.1249/MSS.0000000000004175. Read the original indexed record, not a quote alone.