Older adults’ compliance with mobile ecological momentary assessments in behavioral nutrition and physical activity research: pooled results of four intensive longitudinal studies and recommendations for future research

Abstract

Abstract Background Mobile Ecological Momentary Assessment (EMA) is increasingly used to gather intensive, longitudinal data on behavioral nutrition, physical activity and sedentary behavior and their underlying determinants. However, a relevant concern is the risk of non-random non-compliance with mobile EMA protocols, especially in older adults. This study aimed to examine older adults’ compliance with mobile EMA in health behavior studies according to participant characteristics, and prompt timing, and to provide recommendations for future EMA research. Methods Data of four intensive longitudinal observational studies employing mobile EMA to understand health behavior, involving 271 community-dwelling older adults (M = 71.8 years, SD = 6.8; 52% female) in Flanders, were pooled. EMA questionnaires were prompted by a smartphone application during specific time slots or events. Data on compliance (i.e. information whether a participant answered at least one item following the prompt), time slot (morning, afternoon or evening) and day (week or weekend day) of each prompt were extracted from the EMA applications. Participant characteristics, including demographics, body mass index, and smartphone ownership, were collected via self-report. Descriptive statistics of compliance were computed, and logistic mixed models were run to examine inter- and intrapersonal variability in compliance. Results EMA compliance averaged 77.5%, varying from 70.0 to 86.1% across studies. Compliance differed among subgroups and throughout the day. Age was associated with lower compliance (OR = 0.96, 95%CI = 0.93–0.99), while marital/cohabiting status and smartphone ownership were associated with higher compliance (OR = 1.83, 95%CI = 1.21–2.77, and OR = 4.43, 95%CI = 2.22–8.83, respectively). Compliance was lower in the evening than in the morning (OR = 0.82, 95%CI = 0.69–0.97), indicating non-random patterns that could impact study validity. Conclusions The findings of this study shed light on the complexities surrounding compliance with mobile EMA protocols among older adults in health behavior studies. Our analysis revealed that non-compliance within our pooled dataset was not completely random. This non-randomness could introduce bias into study findings, potentially compromising the validity of research findings. To address these challenges, we recommend adopting tailored approaches that take into account individual characteristics and temporal dynamics. Additionally, the utilization of Directed Acyclic Graphs, and advanced statistical techniques can help mitigate the impact of non-compliance on study validity.