Latent Variable Mixture Modeling of Ecological Momentary Assessment Data: Implications for Screening and Adolescent Mood Profiles

Abstract

Ecological momentary assessment (EMA) studies typically rely on arbitrary decision rules for identifying and excluding invalid responses from the data. In addition, most studies treat independent variables as separate from each other even if their combinations might have importance above the independent contribution of each. Our study aimed to conduct an exploratory latent profile analysis of EMA data to demonstrate an empirical method of identifying invalid responses, and to provide a preliminary investigation of mood profiles. We recruited 20 adolescents between the ages of 13–18 to complete 4 surveys about their internal states each day for 20 days. Participants provided responses on study smartphones using an Android app developed by the study team. Our profile analysis revealed 9 independent profiles. We determined that 3 of these profiles consisted of invalid responses because the integers provided by the participant were nearly invariant. The invalid responses comprised 24.9% of the sample. We also identified 6 valid profiles that were labeled: fatigued (8.7%), good mood/energetic (19.9%), angry/depressed (2.3%), good mood (37.1%), angry (5.7%), and depressed (1.4%). One important implication of the current study is that researchers and clinicians should screen electronic diary data, especially for invariant responding. In addition, it is important for clinicians to note that more than one internal state may drive the mood of an adolescent patient.