Editorial methodology

An instrument is included when it is validated in a peer-reviewed publication, in current clinical or research use, and when its scoring procedure is published in enough detail to be implemented exactly. Instruments that exist only as a proprietary algorithm are not included, because we cannot show you the formula.

Coverage is organised by body region, then by what the instrument does: patient-reported outcome measures, radiographic and clinical calculators, classification systems, and clinical decision rules. A clinic reaches for all four, so a site that only carries outcome measures is only a quarter of the job.

Every scoring formula is taken from the original development paper or the developer’s own published scoring manual, never from a secondary summary. The formula is then restated in plain English on the instrument page so that a clinician can reproduce the calculation by hand and confirm our implementation agrees.

Each implementation is independently reviewed against the published method by a second pass whose explicit brief is to find a discrepancy, including any mismatch between what the page says the formula is and what the code actually does. Where a formula cannot be established with confidence, the instrument is published as a profile without a working calculator rather than with a calculator we are not sure of.

Minimal clinically important difference, patient acceptable symptom state, substantial clinical benefit and minimal detectable change are reported with the population and follow-up interval each was derived in, and with a citation. A threshold quoted without its population is not usable — an MCID from an arthroplasty cohort does not transfer to an ACL reconstruction cohort, and presenting one figure as if it were universal is the most common error in this field.

Where published values conflict, we report the range rather than picking a favourite. Where no threshold has been published for an instrument, we say so instead of borrowing one from a related measure.

Every instrument follows its own developer’s missing-data rule, which is stated on the page. Where a developer specifies that a subscale is invalid beyond a certain number of missing items, the calculator declines to score that subscale rather than silently returning a number computed from fewer items.

Partial completion always produces a visible count of answered items, so an incomplete score is never mistaken for a final one.

Pages carry the name and credentials of the reviewing clinician and the date of review once that review has taken place. Pages that have not yet been reviewed say so plainly rather than implying an endorsement that does not exist.

If you find an error in a formula, a threshold or a citation, tell us and we will correct it and note the correction. Scoring errors are treated as urgent, because a wrong calculator is worse than no calculator.

These calculators report and interpret published scores. They do not recommend a treatment, direct a clinical decision, or substitute for examining a patient. Output is phrased so that the basis of every number is visible and independently checkable, which is both a clinical necessity and the boundary that keeps a scoring tool outside the definition of regulated medical device software.