Street-View Greenspace Exposure and Objective Daily Rest-Activity Patterns

Abstract

BACKGROUND AND AIM: Rest-activity patterns (sleep and physical activity) are modifiable risk factors for obesity. The sparse research on greenspace and sleep or physical activity in children is dominated by satellite-derived measures of greenspace, which do not capture ground-level exposures. Google Street View (GSV) images measure greenspace exposures as participants experience them and may be most relevant for health. We aimed to examine GSV-based greenspace and objective rest-activity patterns in adolescents in Project Viva, a cohort in eastern Massachusetts participating in the Environmental influences on Child Health Outcomes (ECHO) consortium. METHODS: We applied deep learning algorithms to GSV images from 2012-2016 to derive metrics of greenspace (e.g., % trees, %grass within each image) within 250m of participant’s residential addresses. We derived rest-activity metrics from the early adolescence visit (2012-2016; median age 12.7), when participants completed 5 days of wrist actigraphy for 10 hours/day. We used linear regression to examine associations between GSV-based greenspace and rest-activity patterns, adjusting for child’s sex, race/ethnicity, and age; mother’s education and marital status; father’s education; household income; neighborhood median income and population density. RESULTS:In unadjusted cross-sectional analyses (N=505), higher %grass, %trees, %plants, %fields, and %flowers combined (%total greenspace) was associated with lower activity in the least active 5-hour period (L5) in the 24-hour cycle, suggesting more consolidated sleep (L5 difference per IQR [24%] increase in %greenspace: -10.3 [95%CI -16.4,-4.2]), and slightly higher relative amplitude (RA), reflecting both higher activity during wakefulness and more restful sleep at night (RA difference per IQR increase in %greenspace: 0.01 (95%CI 0.01, 0.02)). These associations were consistent for both %grass and %trees. However, in multivariable models, these associations were no longer present. No associations were observed between GSV-based %greenspace exposure and moderate-vigorous nor light physical activity. CONCLUSIONS:GSV-based greenspace exposure was not associated with objective rest-activity patterns among participants in early adolescence. KEYWORDS: Greenspace, Google Street View, Deep Learning, Accelerometry, Rest Activity, Children