JMIR Publications

ABSTRACT

Background:

Ecological Momentary Assessment (EMA) is useful for capturing the time-sensitive variability in mood. However, there is no widely used, gold-standard momentary measure of mood, making it difficult to compare results across studies. Measures often vary in number of items, as an attempt to resolve the trade-off between reliability and missingness.

Objective:

This study evaluated the performance of a brief EMA measure for mood in relation to reliability, missingness, and construct validity.

Methods:

We collected EMA data from a sample of young adults at a UK university (n=171). Participants received a notification from the PIEL Survey app on their mobile phones to respond to a short mood survey four times a day for two weeks. This survey was developed through co-production with a focus group and required participants to rate their mood (0–10) on five items: cheerful, frustrated, relaxed, sad, and worried. We assessed the within- and between-participant reliability using multilevel modelling methods. Construct validity was evaluated through comparison with scores on the Positive and Negative Affect Schedule. Principal Component Analysis was also used to explore the underlying dimensionality. We further considered compliance rates and the degree of ‘careless responding’ (defined by response times, standard deviations, and overall reliability).

Results:

Our data showed good internal consistency reliability and acceptable variance using both generalisability theory and multilevel modelling. There was high between-participant reliability (RkR≥0.98, RkRn≥0.98), and moderate within-participant reliability (Rc≥0.64, Rcn≥0.56) for both positive and negative items. High construct validity was also established: our measure showed moderate correlations with an existing, longer measure of mood. We found two Principal Components together explained 75.8% of the variance in our measure, which appear to represent distinct Positive and Negative Affect dimensions. We also found high compliance rates with only 23.3% of responses missing overall, showing some degree of variability across time window, weekday, and time in the study. Our sample contained no careless responders overall.

Conclusions:

In summary, the five-item measure was brief enough to maintain high compliance in a non-clinical sample, without sacrificing reliability or construct validity. This shows promise for the use of shorter-item surveys in Ecological Momentary Assessment, which may enhance compliance rates and reduce the level of careless responding. Future studies should consider whether this measure also maintains reliability for different populations.