Why a classification’s reliability figure is not an MCID

A patient-reported outcome measure like the KOOS or the Oswestry produces a number that moves as the patient’s condition changes over time. It makes sense to ask how much movement counts as meaningful — which is what MCID measures.

A classification system like the Garden grade for a femoral neck fracture, or the Weber grade for an ankle fracture, does neither of those things. It is assigned once, from a single radiograph, to describe a pattern. There is no "before and after" Garden grade for the same fracture to compare — the fracture either is or is not displaced. Asking for its MCID is asking a question the instrument was never built to answer.

The published evidence around a classification system is almost always one of two things: a cut-off definition (Garden III is defined by partial displacement; Pauwels type I is conventionally below 30 degrees) or an interobserver reliability figure — a kappa statistic describing how often two clinicians looking at the same radiograph agree on the type.

Neither of those is a change score, and reporting a kappa of 0.31 under a heading that defines MCID as "the smallest change a patient perceives as meaningful" would be actively misleading. That specific error is one an earlier clinical review pass on this site caught and corrected across every classification and decision rule on the site, which is why those instruments’ published evidence is labelled Criterion rather than MCID, PASS, SCB, or MDC.

A low interobserver kappa is not a footnote — it tells you how much to trust the grade itself. The Garden classification’s four-type system has only fair reliability (kappa roughly 0.27 to 0.56), which is exactly why most surgeons and most trials collapse it to the two-group non-displaced-versus-displaced distinction, where agreement is substantially better. Reading the reliability evidence, not just the grade definitions, is part of using a classification system correctly.