Abstract
Introduction: Screen time is predominantly measured using questionnaires assessing a limited range of activities. This project aimed to develop a coding protocol that reliably identified screen time, including device type and specific screen behaviors, from video-camera footage. Methods: Screen use was captured from wearable and stationary PatrolEyes video cameras in 43 participants (aged 10-14 years) within the home environment (May-December 2021, coding in 2022, statistical analysis in 2023). After extensive piloting, the inter-rater reliability of the final pro-tocol was determined in 4 coders using 600 minutes of footage from 18 participants who spent unstructured time on digital devices. Coders independently annotated all footage to determine 8 device types (e.g., phone, TV) and 9 screen activities (e.g., social media, video gaming) using Observer XT (behavioral coding software). Reliability was calculated using weighted Cohen's K for duration per sequence (meets criteria for total time in each category) and frequency per sequence (meets criteria for total time in each category and order of use) for every coder pair on a per-partici-pant and footage type basis. Results: Overall reliability of the full protocol was excellent (>= 0.8) for both duration per sequence (K=0.89-0.93) and the more conservative frequency per sequence (K=0.83-0.86) analyses. This protocol reliably differentiates between different device types (K=0.92-0.94) and screen behaviors (K=0.81-0.87). Coder agreement ranged from 91.7% to 98.8% across 28.6-107.3 different instances of screen use. Conclusions: This protocol reliably codes screen activities in adolescents, offering promise for improving the understanding of the impact of different screen activities on health. Am J Prev Med 2023;65(5):923-931. (c) 2023 American Journal of Preventive Medicine. Published by Elsevier Inc. All rights reserved.