Test-Retest Reliability (Outcome Measurement & Instrument Science)

The degree to which an instrument produces stable and consistent scores when administered repeatedly under conditions where no true change in the domain being measured is expected. Test-retest reliability evaluates the extent to which variation in scores over time reflects measurement error rather than actual change in the construct of interest. Within OMERACT, this property is used to determine whether an instrument can reliably discriminate between stable and changing states.A high-quality test-retest reliability study includes repeated administrations of the instrument during a stable period, while keeping factors such as observer, timing, and mode of administration as consistent as possible. The interval between assessments should be long enough to reduce recall effects but short enough to minimize the possibility of true clinical change.Agreement statistics such as the Intraclass Correlation Coefficient (ICC) and Weighted Kappa (Kw) are recommended because they evaluate concordance between repeated scores. Simple correlations alone are insufficient because they may fail to identify systematic differences between assessments.Test-retest reliability also supports calculation of the Minimum Detectable Change (MDC), which represents the amount of change that must be exceeded before observed differences can be interpreted as true change rather than measurement error.