Temporal Stability of Repeated Self-Reported vs Device-Measured Behavior: Observational Study Using Ecological Momentary Assessment
Journal of Medical Internet Research ·
Background: A central concern in ecological momentary assessment (EMA) research is whether repeated self-report alters the reporting process itself. In most EMA domains, longitudinal stability of reporting is assumed but rarely tested. Objective: This study leveraged concurrent accelerometry as a temporal anchor to evaluate (1) whether self-reported physical activity (PA) and sedentary behavior (SB) exhibited temporal shifts across a 7-day monitoring period and (2) whether this temporal shift differed across near-momentary (5-min reporting period) and coverage-based (120-min recall period) EMA protocols. Methods: We conducted a secondary analysis of 315 US adults recruited from the Understanding America Study panel between January and July 2023 for the current study. Participants were monitored in their natural environments for 7 days via 2 measurement methods: continuous thigh-worn activPAL4 accelerometry and 5 daily smartphone EMA prompts assessing PA and SB. Reporting reactivity was defined as divergence in the temporal trajectories of EMA-reported behavior relative to accelerometry. Multilevel models with random intercepts and slopes tested whether linear and quadratic study day effects differed across the 2 methods (method × study day interactions). Protocol differences were examined using multivariable method × protocol × study day interactions, comparing coverage-based and near-momentary EMA designs. Results: In the coverage-based protocol, EMA-reported PA and SB exhibited temporal trajectories that diverged significantly from concurrent accelerometry (SB linear divergence: =1.40, 95% CI 0.20-2.61, =.02; PA linear divergence: =−1.91, 95% CI −2.93 to −0.90, <.001; and PA quadratic divergence: =0.28, 95% CI 0.12-0.44, <.001). No evidence of such divergence was found in the near-momentary protocol. The divergence in reporting trajectories differed significantly between the 2 protocols (SB linear 3-way interaction: =−2.58, =.02; PA linear 3-way interaction: =1.86, =.04). Conclusions: Reporting reactivity in EMA appears protocol-dependent. The coverage-based protocol was associated with modest reporting reactivity, which challenges the common assumption of reporting invariance in EMA and underscores the importance of considering protocol design when interpreting within-person change.