JMIR Publications

ABSTRACT

Background:

Ecological momentary assessment (EMA) has become a popular study design to harvest insights into how people interact with dynamic environments. Its potential to produce rich longitudinal data to inform interventions may be limited when burden on participants is not carefully limited.

Objective:

The current study explores design and momentary determinants of adherence that may inform the suitability of EMA implementations for examining dynamic public health environments.

Methods:

EMA from Phase I Family Matters study (n=150 adult Black, White, Hmong, Somali, Latino/a, and Native American caregivers; n=1,392 observation days) was examined to understand how participant self-reported survey burden was related to both design and momentary antecedents of adherence. Daily burden was measured by the question “Overall, how difficult was it for you to fill out the surveys today?” on a 5-item Likert scale (0, Not at all to 4, Extremely). Daily protocol adherence was defined as completing at least two signal-contingent surveys, one event-contingent survey, and one end-of-day survey each. Stress and mood were measured earlier in the day, sociodemographic and psychosocial characteristics were reported on a comprehensive cross-sectional survey, and EMA timestamps for weekends and weekdays were used to parameterize time-series models to evaluate prospective correlates of end-of-day study burden.

Results:

Burden was low at 1.2 (sd=1.14) indicating “a little” burden on average. Participants with elevated chronic stress levels (P=.043), immigrant status (P=.02), and the language primarily spoken in the home (P=.042; non-English speaking households) were found to be population characteristics of elevated moderate-high burden. Current and lagged non-adherence were correlated with elevated burden (P=.001), and the pattern decayed over subsequent days. Momentary antecedents including daily mood (P=.002) and across-day change in stress (P=.008) were independently associated with end-of-day burden after controlling for current-day adherence.

Conclusions:

The eight-day EMA implementation appeared to capture momentary sources of stress and depressed mood without substantial burden to a racially/ethnically diverse and immigrant/refugee sample of parents. Attention to sociodemographic attributes (e.g., EMA in the primary language of the caregiver) was important for minimizing participant burden and improving data quality. Momentary stress and depressed mood was a strong determinant of participant-experienced EMA burden and may affect adherence to mHealth study protocols. There were no strong indicators of EMA design attributes that created a persistent burden for caregivers. EMA stands to be an important observational design to address dynamic public health challenges related to human-environment interactions when the design is carefully tailored to the study population and study research objectives.