Skip to content

How missing answers are actually handled

Outcome measures handle missing items in three distinct ways — substituting the mean of the answered items within a subscale, prorating the score against a reduced denominator, or refusing to score the affected subscale entirely — and using the wrong rule silently produces a number that looks valid but is not.

Three different rules, not one universal one

The KOOS and HOOS family substitutes the mean of the answered items for up to two missing items within a subscale; beyond that, the subscale is not scored at all rather than calculated from too few items. That cut-off is deliberate — the developers judged that a subscale mean built from fewer than a certain fraction of its items no longer represents what the subscale is meant to measure.

The Oswestry Disability Index instead reduces its denominator: if a section is left blank, the percentage is calculated out of the remaining possible points rather than out of the full fifty, which is precisely why the sex life section — the one patients skip most often — does not sink an otherwise complete Oswestry score.

A single-item instrument like the Tegner activity scale, or a short branching classification, generally has no substitution rule at all, because there is no meaningful way to interpolate a missing answer from the others.

Why this matters more than it looks like it should

A calculator that silently treats a blank item as zero, or silently drops it without adjusting the denominator, will produce a plausible-looking number that is not the number the instrument’s developer intended. The error is invisible unless you know to look for it, which is what makes it worse than an obviously broken calculator.

Every instrument page on this site states its own missing-data rule explicitly, in the same place as the scoring formula, specifically so the number you see can be checked against what the developer actually published — rather than against a generic assumption that happens to be wrong for that particular instrument.

Where this applies

See also the full threshold database and all guides.