Validation of within-person associations with coverage and momentary EMA self-reports using an objective standard...

Ecological momentary assessment (EMA) is a useful tool for assessing peoples' experiences, behaviors, and environments in everyday life. EMA is considered a standard for self-reporting, because its short reporting periods and multiple assessments in everyday life offers reduced recall bias and increased ecological validity. In this study we tested how well within-person EMA measurements of levels of physical activity (prompt-level data) match objective reports of activity from accelerometry. We looked at two kinds of validity: duration of EMA recall (5 m; 120 m) and type of physical activity response scale (verbal response scales [VRS]; numerical duration scales). Concurrent criterion validity was tested by examining level differences and with correlations of EMA with accelerometry. Construct validity was assessed by examining whether the correlations between EMA and non-physical activity variables aligned with correlations between accelerometry and those same variables. Participants (N = 315) wore a thigh accelerometer and completed EMA five times per day for 7-days. We observed that EMA estimates of activity levels were very different compared with accelerometry levels, especially for active behavior, where EMA yielded much higher levels of active behavior than accelerometry (the first type of concurrent criterion validity). We found moderate within-person correlations between EMA and accelerometry (second type of concurrent criterion validity). However, some study factors moderated the concordance: coverage EMA had slightly stronger associations than near-momentary EMA and sedentary behavior was more accurately reported than high physical activity levels. Construct validity analyses showed that EMA self-reports could be used in place of accelerometry data. With the caveat that physical activity was greatly overestimated by EMA, findings largely support the utility of EMA for analyzing within-person associations with other coverage and near-momentary variables, although they suggest some forms of EMA may be somewhat less optimal.