Measuring stability and change in response patterns to a set of hierarchical scales in a randomized intervention study: an innovative application of latent transition analysis.

The objective of this study was to measure stability and change in response patterns to a set of hierarchical Physical Activity Self-Efficacy (PASE) Scales with latent transition analysis (LTA). To accomplish this objective a multiple-group LTA modeled binary responses to six ordered items within each PASE scale. Data (N_baseline = 461 and N_{30 days post-baseline} = 428) from the Well-Being and Physical Activity (WBPA; ClinicalTrials.gov, identifier: NCT03194854) study were analyzed. A four-class solution with interpretable parameter restrictions explained response patterns to each PASE scale at baseline. Evidence for temporal measurement invariance of this four-class solution was provided. Stability of latent class membership from baseline (i.e., pre-intervention) to 30 days post-baseline (i.e., post-intervention) was modest, consistent with substantive theory. Desirable differences in LTA probabilities (e.g., transitioning from confidence to engage in 10 min of weekly physical activity at baseline to confidence to engage in 90 min of weekly physical activity at 30 days post-baseline) by intervention group were observed, consistent with objectives of the WBPA study.