Psychometric properties and reference values for the instrumented timed-up and go test in adults

The instrumented Timed Up and Go (iTUG) test is widely used in assessing mobility and predicting falls. However, the psychometric properties of a smartphone-based iTUG remain unclear, particularly with respect to the factors that may affect reliability and measurement error. In addition, the lack of established reference values limits its clinical applicability and interpretability. Therefore, this study aimed to evaluate the test-retest reliability and measurement error of iTUG using a smartphone application in healthy adults and to establish reference values for selected outcome measures.

This psychometric study involved fifty healthy adults who completed two iTUG test sessions 1 week apart. Each session included one practice trial and six test trials, comprising three trials in participant's preferred turning direction and three in nonpreferred direction. Outcome measures included time-based and angular velocity-based parameters, extracted using a custom MATLAB algorithm from raw data recorded by a smartphone application. These outcome measures were categorized into three subgroups based on turning directions: overall, preferred direction, and nonpreferred direction. Test-retest reliability was assessed using the two-way random effects model, with variance components for all sources reported. Measurement error was evaluated with standard error of measurement for agreement (SEM_agreement), smallest detectable change for agreement (SDC_agreement), and Bland-Altman plots. The influence of performance approach (average vs. best) and turning directions on reliability was also analyzed. Reference values were calculated based on data distribution characteristics.

Most outcome measures demonstrated moderate to good reliability. Angular velocity-based measures showed consistently higher reliability than time-based measures. Across all outcome measures, residual variance was the main source of nonsubject variability (range: 13.58%-69.44%), whereas interoccasion variance was consistently small or negligible (range: 0.00%-5.11%), suggesting limited systematic bias between sessions. The average performance approach yielded a higher intraclass correlation coefficient (ICC) and a smaller SEM_agreement than the best performance. Turning directions also affected reliability and measurement error. Specifically, outcome measures in nonpreferred turning direction generally showed lower ICC and higher SEM_agreement and SDC_agreement, indicating greater variability. Bland-Altman plots showed no proportional bias or heteroscedasticity. Reference values were successfully established for 15 outcome measures.

The iTUG demonstrates moderate to good test-retest reliability in healthy adults. Average performance approach generally shows higher reliability and lower measurement error, and turning direction should be considered in practice. The reference values offer a benchmark for interpreting performance and support broader clinical use of iTUG.