Study population and setting
This study estimated the effect of state-level mask mandates on daily county-level growth rates of confirmed COVID-19 cases from March 31 to May 22, 2020 using an event study design (also known as “difference-in-differences”). Fifteen states, including the District of Columbia, enacted mandates requiring individuals to wear face masks (non-medical, usually cloth face coverings) while in public places where maintaining six feet of distance from others may be impractical, while 36 states remained without a public mask mandate during the study period. Each of the 15 states with public mask mandates, and an additional 20 states, required employees in certain industries with high degrees of personal contact to wear masks at all times. Using the 5-day period before a mask mandate as the reference period, the authors estimated policy effects for the following periods after a mandate: 1-5 days, 6-10 days, 11-16 days, 16-20 days, and 21+ days. Effects of employee mask mandates were estimated in a supplementary analysis. The authors conducted several sensitivity analyses and robustness checks.
Summary of Main Findings
After state-wide mask mandates, estimated county-level growth rates of COVID-19 cases declined, with larger effect sizes seen at longer times following the mandates. Estimated average declines for each period following a mandate were -0.92% (1-5 days), -1.07% (6-10 days), -1.42% (11-15 days), -1.66% (16-20 days), and -1.97% (21+ days); p<0.05 for each estimate. The authors estimated the number of cases averted compared to a counterfactual scenario with no mask mandates: this estimate ranged from 230,000 to 450,000 cases averted during the time period, depending on how prediction error was accounted for. Effect estimates for time periods preceding mask mandates were not significantly different from zero, supporting the assumption that the trends seen in the non-mandate states was a reasonable proxy for what would have happened had the mandate states not implemented mask mandates. Estimated effects of employee-only mask mandates were small and statistically insignificant. Results of the primary analysis remained similar under a range of sensitivity analyses, including using states as the unit of analysis and excluding certain subsets of states and counties.
The authors took advantage of a temporally well-defined exposure proxy in state-level mask mandates. The analytic approach was reasonable, and authors performed a set of robustness checks using alternative assumptions and approaches.
Actual mask use was not measured, limiting inference about the effectiveness of mask wearing itself. Illustrating impacts on mask wearing would have also provided further evidence for the pathway of effects on case rates, and for the reasonableness of those effect estimates. The analysis assumes that case trends in non-mandate states provide a reasonable proxy for what would have happened in mandate states if mandates had not been implemented. The authors provide some justification for this assumption, but it is not directly testable. The models do not incorporate infectious disease dynamics– it is possible that growth rates in some counties may have declined independently of mask mandates (due to, for example, depletion of susceptible individuals in high-risk groups of people defined by institutional residence, type of employment, or socioeconomic status). County-level and municipal-level mask mandates were not considered. There was heterogeneity in the content and enforcement of mask mandates across states that is not accounted for. Raw data were not presented either visually or in tabular form.
This study adds to the evidence supporting mask requirements as an effective means of lowering transmission of SARS-CoV-2.
This review was posted on: 21 July 2020