Psychological Assessment, Vol 38(6-7), Jun-Jul 2026, 441-450; doi:10.1037/pas0001454
Missing observations in intensive longitudinal data may be attributable to between-person differences and within-person states, either of which may systematically bias assessment of associations between predictors and outcomes. We examined this issue in intensive longitudinal data from people receiving medication for opioid use disorder (n = 306), each of whom was asked to provide an end-of-day (EOD) report via smartphone every night for 8 weeks, along with thrice-daily randomly prompted reports (person-moments = 39,321). We examined whether missing EOD reports were associated with mood, stress, craving, and substance use, estimated both as within-person changes (in random-prompt entries) and between-person characteristics (means across random-prompt entries). Between-person characteristics were the strongest correlates of missing EOD reports, though the magnitudes of the associations were modest. Greater likelihood of missingness was associated with greater mean low-arousal mood (i.e., fatigue; OR = 1.13, 95% CI [1.01, 1.27]), stress (OR = 1.14, 95% CI [1.02, 1.28]) and proportions of random-prompt entries with heroin craving (OR = 1.14, 95% CI [1.02, 1.27]). Within participants, the likelihood of a missing EOD report decreased when positive mood was higher than typical (OR = 0.82, 95% CI [0.68, 0.99]). We conclude that people who tend to have more difficult day-to-day experiences (i.e., higher stress, craving), and moments when people’s moods are more positive, may contribute to under- or overestimation of effects tested in EOD data. The adaptive delivery of surveys based on person- and prompt-level characteristics (i.e., following elevated positive moods) could be one approach to maximizing survey completion. (PsycInfo Database Record (c) 2026 APA, all rights reserved)