Compliance With Mobile Ecological Momentary Assessment of Self-Reported Health-Related Behaviors and Psychological Constructs in Adults: Systematic Review and Meta-analysis

Abstract

Background Mobile ecological momentary assessment (mEMA) permits real-time capture of self-reported participant behaviors and perceptual experiences. Reporting of mEMA protocols and compliance has been identified as problematic within systematic reviews of children, youth, and specific clinical populations of adults. Objective This study aimed to describe the use of mEMA for self-reported behaviors and psychological constructs, mEMA protocol and compliance reporting, and associations between key components of mEMA protocols and compliance in studies of nonclinical and clinical samples of adults. Methods In total, 9 electronic databases were searched (2006-2016) for observational studies reporting compliance to mEMA for health-related data from adults (>18 years) in nonclinical and clinical settings. Screening and data extraction were undertaken by independent reviewers, with discrepancies resolved by consensus. Narrative synthesis described participants, mEMA target, protocol, and compliance. Random effects meta-analysis explored factors associated with cohort compliance (monitoring duration, daily prompt frequency or schedule, device type, training, incentives, and burden score). Random effects analysis of variance (P≤.05) assessed differences between nonclinical and clinical data sets. Results Of the 168 eligible studies, 97/105 (57.7%) reported compliance in unique data sets (nonclinical=64/105 [61%], clinical=41/105 [39%]). The most common self-reported mEMA target was affect (primary target: 31/105, 29.5% data sets; secondary target: 50/105, 47.6% data sets). The median duration of the mEMA protocol was 7 days (nonclinical=7, clinical=12). Most protocols used a single time-based (random or interval) prompt type (69/105, 65.7%); median prompt frequency was 5 per day. The median number of items per prompt was similar for nonclinical (8) and clinical data sets (10). More than half of the data sets reported mEMA training (84/105, 80%) and provision of participant incentives (66/105, 62.9%). Less than half of the data sets reported number of prompts delivered (22/105, 21%), answered (43/105, 41%), criterion for valid mEMA data (37/105, 35.2%), or response latency (38/105, 36.2%). Meta-analysis (nonclinical=41, clinical=27) estimated an overall compliance of 81.9% (95% CI 79.1-84.4), with no significant difference between nonclinical and clinical data sets or estimates before or after data exclusions. Compliance was associated with prompts per day and items per prompt for nonclinical data sets. Although widespread heterogeneity existed across analysis (I2>90%), no compelling relationship was identified between key features of mEMA protocols representing burden and mEMA compliance. Conclusions In this 10-year sample of studies using the mEMA of self-reported health-related behaviors and psychological constructs in adult nonclinical and clinical populations, mEMA was applied across contexts and health conditions and to collect a range of health-related data. There was inconsistent reporting of compliance and key features within protocols, which limited the ability to confidently identify components of mEMA schedules likely to have a specific impact on compliance.

Covariation of psychobiological stress regulation with valence and quantity of social interactions in everyday life: disentangling intra- and interindividual sources of variation

“…As alternatives, it would have been necessary to either allow user-initiated prompts (i.e., whenever there currently is a social interaction) or to assess social interactions much more frequently (e.g., every 30 min). However, both approaches did not seem viable because they would have come with increased participant burden and possibly a reduced compliance (Williams et al 2021 ). In addition, it would not be feasible for participants to report on their interactions during their everyday life routines in such a high frequency and, most importantly, the reporting itself could have interfered with social interactions (e.g., by interrupting conversations), thereby reducing ecological validity.…”

“…In case of cross-level interactions, the predictor on the higher level was interpreted as moderator (Andersson et al 2014). Interaction plots were built using the R packages 'effects' (Fox and Weisberg 2019) and 'ggplot2' (Wickham 2016).…”

Patient-Reported Outcome Measure for Real-time Symptom Assessment in Women With Endometriosis: Focus Group Study

“…Compared with other ESM studies, this compliance rate is relatively low, as meta-analyses have shown completion rates of 82% to 85% [ 38 , 39 ]. However, comparing data with other ESM studies is difficult because the absence of methodological guidelines related to the use of this method has resulted in a large heterogeneity of designs [ 39 ], and compliance rates have not been reported in approximately half of the studies [ 40 ]. For better compliance, fewer study days, less assessments per day, and fewer items per assessment are advised [ 38 , 40 ].…”

“…However, comparing data with other ESM studies is difficult because the absence of methodological guidelines related to the use of this method has resulted in a large heterogeneity of designs [ 39 ], and compliance rates have not been reported in approximately half of the studies [ 40 ]. For better compliance, fewer study days, less assessments per day, and fewer items per assessment are advised [ 38 , 40 ]. In addition, as previous ESM studies recommend at least 3 completed questionnaires per day for a reliable analysis, which occurred consistently only during the first week of this study, we recommend using the ESM for a maximum of 7 days [ 41 ].…”

Tracking Infant Development With a Smartphone: A Practical Guide to the Experience Sampling Method

“…Only if compliance is high, one is able to generalize the answers to daily life and data analysis are sufficiently powered. In most studies, participants answer on average 50–95% of all ESM surveys ( Wen et al, 2017 ; van Roekel et al, 2019 ; Williams et al, 2021 ). As a rule of thumb, 70–80% compliance indicates good data quality.…”