Patterns of Missing Data with Ecological Momentary Assessment: Feasibility Report from a Pilot Study of People Who Use Drugs.
Kelly L. Markowski,G. Gauthier,Jeffrey A. Smith,Sela R. Harcey
DOI: https://doi.org/10.2196/31421
2021-08-24
JMIR Formative Research
Abstract:BACKGROUND
Ecological momentary assessment (EMA) is a set of research methods that capture events, feelings, and behaviors as they unfold in their "real world" setting. Capturing data "in the moment" reduces important sources of measurement error but also generates challenges for non-compliance (i.e., missing data). To date, past EMA research has only examined overall rates of non-compliance.
OBJECTIVE
We identify four types of non-compliance among people who use drugs (PWUD) and examine the factors associated with the most common types.
METHODS
Data are from a recent pilot study of 28 Nebraskan PWUD who answered EMA questions for two weeks. We examine questions that were not answered because they were skipped, they expired, the phone was off, or the phone died after receiving them.
RESULTS
We find that the phone being off and questions expiring comprise 93% of our missing data. Generalized structural equation model results show that participant-level factors, including age (RRR=0.93, P=.005), gender (RRR=0.08, P=.006), homelessness (RRR=3.80, P=.04), personal device ownership (RRR=0.14, P=.008), and network size (RRR=0.57, P=.001), are important for predicting off missingness, while only question-level factors, including time of day (i.e., morning compared to noon/afternoon, RRR=0.55, P<.001) and day of week (i.e., Tuesday-Saturday compared to Sunday, RRR=0.70, P=.02, RRR=0.64, P=.005, RRR=0.58, P=.001, RRR=0.55, P<.001, RRR=0.66, P=.008) are important for predicting expired missingness. Week of study is important for both (i.e., week two compared to week one, RRR=1.21, P=.03 for off missingness, RRR=1.98, P<.001 for expired missingness).
CONCLUSIONS
We suggest a three-pronged strategy to preempt missing EMA data with high-risk populations: 1) provide additional resources for participants likely to experience phone charging problems (e.g., people experiencing homelessness); 2) ask questions when participants are not likely to experience competing demands (e.g., morning); and 3) incentivize continued compliance as the study progresses. Attending to these issues can help researchers ensure maximal data quality.
CLINICALTRIAL
Medicine,Psychology