Is Evolving Face Mask Guidance During a Pandemic Harmful to Public Perception?: An Analysis of Sentiment and Emotion on Twitter

Abstract

Throughout the COVID-19 pandemic, U.S. Centers for Disease Control and Prevention policies on face mask usage fluctuated. This study assessed how mask sentiment evolved surrounding two policy announcements: (1) mask recommendation on April 3, 2020 and (2) relaxation on May 13, 2021. Interrupted time series methods were applied to U.S. Twitter data surrounding each policy change. Outcomes were changes in the (1) proportion of positive, negative, and neutral tweets and (2) number of words within a tweet tagged with a given emotion (e.g., trust). Results were compared to COVID-19 Twitter data without mask keywords for the same period. There were fewer neutral mask-related tweets in 2020 (β = -3.94 percentage points, 95% CI: -4.68, -3.21, P < .001) and 2021 (β = -8.74, 95% CI: -9.31, -8.17, P < .001). Following the April 3rd recommendation (β = 0.51, 95% CI: 0.43, 0.59, P < .001) and May 13th relaxation (β = 3.43, 95% CI: 1.61, 5.26, P < .001), the percent of negative mask-related tweets increased. The amount of trust-related terms decreased following the policy change on April 3 (β = -0.004, 95% CI: -0.004, -0.003, P < .001) and May 13 (β = -0.001, 95% CI: -0.002, 0, P = .008). The U.S.-Twitter population responded negatively and with less trust following policy shifts, regardless of whether the policy recommended or relaxed mask usage. Federal agencies should ensure that changes in public health recommendations are communicated concisely and rapidly.

Keywords: Face Masks, COVID-19, Twitter, Science Communication, Political Communication, Public Policy, Public Health, Sentiment Analysis, Emotion Analysis