Observation Effect in Ecological Momentary Assessments: A Study of Sun Protection Practices

Abstract

Daily diaries and ecological momentary assessments are plagued by the assessment itself becoming an intervention, known as the observation effect. Bayesian hierarchical level modeling is a technique to analyze repeated measures or multiple outcomes. In a study of twice-daily self-reporting of sun protection behavior among high-risk individuals, we investigate observation effects, agreement between retrospectively self-reported reminder effect and observation effect, differential observation effects, and consistency of behaviors. Participants who retrospectively reported no reminder effect showed a decrease in protective behaviors over time, whereas those who reported they were reminded showed sustained use. Advantages of the Bayesian methodology are demonstrated for assessing consistency of behaviors. Although we cannot observe prior behavior, we theorize that individuals experience an initial elevation at the onset of observation, though this unobserved increase is only sustained for a subset who later attribute this sustained behavior to a reminder effect. Implications for study designs with repeated observations are discussed.