COVID-19 Mask Usage and Social Distancing in Social Media Images: Large-scale Deep Learning Analysis.

Area of Science:

• Public Health Surveillance
• Epidemiology
• Social Media Analysis

Background:

• Nonpharmaceutical interventions (NPIs) are crucial for controlling COVID-19 transmission.
• Public behavior during large gatherings, like protests, can impact NPI effectiveness.
• Examining mask use and fit during the pandemic provides insights into public health compliance.

Purpose of the Study:

• To assess face mask usage and fit, alongside social distancing, in the U.S. using social media images.
• To analyze public behavior during large gatherings, specifically Black Lives Matter (BLM) protests, in relation to COVID-19 preventive measures.
• To evaluate the correlation between online mask usage trends and COVID-19 cases.

Main Methods:

• Analyzed 2.04 million social media images from six U.S. cities (Feb-May 2020).
• Evaluated mask use trends, group posting behavior, and mask fit scores.
• Specifically analyzed 195,452 posts from New York City and Minneapolis during BLM protests (May-July 2020).

Main Results:

• Group picture prevalence decreased post-lockdown; mask use increased in cities with mandates.
• BLM protest posts showed a higher percentage of group pictures and lower mask fit scores compared to non-BLM posts.
• Approximately 35% of individuals in protest images had masks with a fit score above 80%.

Conclusions:

• Social media data indicated increased mask use following mandates but revealed lower adherence and mask fit during large protests.
• The study demonstrates social media's utility in indirectly monitoring public adherence to government health policies.
• Findings underscore the challenges in maintaining public health compliance during large-scale social events amidst a pandemic.