Abstract
Background: Although interest in objective screen time measurement is growing, questions remain regarding data processing. The aim of this study was to investigate how different image capture intervals and processing assumptions influence screen time estimates from wearable camera images.
Methods: Screen time was measured using chest-worn Brinno TLC130 cameras which took static images every two seconds over four days in children participating in a crossover trial manipulating sleep. Images were coded for screen usage and only children with data from the same time block (before school, after school, weekends) during both intervention conditions were included. Analyses compared estimates of screen time using different intervals of image capture (2, 4, 6, 8, 10, 20, 30, 60 s) and processing rules (images with screens only, Rules 1 and 2 which allowed for blocked [device being used not visible in photos] images as long as the surrounding images were coded as screen time).
Results: 51 children (51% female, 14% indigenous Māori) had sufficient data (187 observations, 1.8 million images). Intervals of up to 60 s between images did not meaningfully influence total screen time estimates compared to 2 s, at the group level. At the individual level, a 10-s interval provided the optimal balance between reducing the number of images to code and accurate screen time estimates. Allowing blocked images between successive screen images to be coded as screen time increased screen time by a median of 8.8 (25th,75th percentiles: 4.4,14.9, Rule 1) to 59.8 (28.6,87.4, Rule 2) minutes.
Conclusion: Researchers can confidently use up to 60 s intervals between images to measure total screen use in children at the group level, but shorter intervals are required for individual level data. Processing rules which allow blocked images to be coded as screens may influence average screen time estimates by up to 32%.
Trial registration: Australian New Zealand Clinical Trials Registry ANZCTR ACTRN12618001671257, 10th Oct 2018.