Abstract
Face mask usage is one of the most effective ways to limit SARS-CoV-2 transmission, but a mask is only useful if user compliance is high. Through anonymous surveys (n = 679), it was shown that mask discomfort is the primary source of noncompliance in mask wearing. Further, through these surveys, three critical predicting variables that dictate mask comfort were identified: air resistance, water vapor permeability, and face temperature change. To validate these predicting variables in a physiological context, experiments (n = 9) were performed to measure the respiratory rate and change in face temperature while wearing different types of three commonly used masks. Finally, using values of these predicting variables from experiments and the literature, and surveys asking users to rate the comfort of various masks, three machine learning algorithms were trained and tested to generate overall comfort scores for those masks. Although all three models performed with an accuracy of approximately 70%, the multiple linear regression model provides a simple analytical expression to predict the comfort scores for common face masks provided the input predicting variables. As face mask usage is crucial during the COVID-19 pandemic, the goal of this quantitative framework to predict mask comfort is hoped to improve user experience and prevent discomfort-induced noncompliance.
Acknowledgments
We would like to acknowledge Nicholas Jeffreys, the Harvard Face Mask Committee, and the Harvard Active Learning Labs for providing excellent guidance throughout our investigation process. We would especially like to acknowledge Harvard Face Mask Committee members Stephen Blacklow, John Doyle, Willy Shih, Mary Corrigan, Sarah Fortune, and Sara Malconian for their invaluable support and advice. We would also like to recognize Evan Hunsicker and Katia Osei for their initial data analysis, and Paola Carrillo Gonzalez, Jordan Daigle, Taisa Kulyk, Kazi Tasnim, Meghan Turner and James Young for their contributions during the Harvard SEAS engineering design course. Last but not least, we would like to thank Juan C., Mason D., Benjamin F., Rosie P., Daniel R., Miguel S., and Yann T. for volunteering as additional study participants.
Disclosure statement
The authors have declared no financial support. The authors have declared no conflicts of interest.
Data availability statement
The data that support the findings of this study are available from the corresponding author upon reasonable request. All code used for the creation, training, and testing of the ML models is available on Github at https://github.com/mythriambatipudi/Comfort-Model.