Mask Wearing and Control of SARS-CoV-2 Transmission in the United States

Abstract

Introduction

In December 2019, SARS-CoV-2 and the resulting COVID-19 disease were first identified in Wuhan, China¹. The disease has spread globally since its identification, causing widespread morbidity and mortality^2,3. In the absence of a vaccine or curative therapy, governments in the United States and worldwide have adopted numerous approaches to curb the virus’s continued transmission^4,5. Despite widespread implementation, the efficacy of various nonpharmaceutical interventions has been intensely debated⁶, resulting in significant heterogeneity in the acceptance of these interventions at the individual and community level⁷, including the use of face masks and respirators⁸.

Evidence strongly supports N95 respirators as an effective method to prevent viral respiratory transmission, but supply chain shortages necessitate their preferential allocation to high-risk, front-line medical personnel^9,10. Consequently, cloth face coverings and surgical masks (henceforth collectively, “face masks”) are recommended as an alternative for the general public¹¹. Following the initial spread of the virus in the United States, many local and state jurisdictions have mandated the use of face masks in public settings⁵. These masks are intended to serve as a mechanical barrier that absorbs virus-laden droplets expelled by the user¹². Therefore, their purpose is to reduce transmission events by the user, rather than protect the individual from infection. Accordingly, face masks are advocated as a source of collective benefit, akin to herd immunity with vaccination, which is most successful with high levels of adoption¹³.

Empirical, population level evidence on the effectiveness of face masks at preventing respiratory transmission of SARS-CoV-2 is limited but growing^11,14. While work on other respiratory infections¹⁵ and two recent case reports^16,17 suggest mask usage may be effective, a recent global analysis found a limited marginal effect of mask mandates in the presence of other interventions¹⁸. Here, we use an ecological approach to assess mask compliance directly, irrespective of mandates. We combine survey data on personal mask wearing habits across the United States with a time varying measure of transmission control, as quantified by the effective reproductive number (R_t) in each state. We then evaluate the association of a change in mask wearing with the timing of mask mandates to better understand their impact on transmission of COVID-19.

Materials and Methods

Results

Self-reported mask wearing was evaluated using N=378,207 survey responses recorded between June 3, 2020 and July 27, 2020. Most (84.6%) reported “very likely” to wearing a face mask to the grocery store, while just under half (40.2%) did so to visit friends and family. A similar fraction (39.8%) reported they were very likely to wear a mask to the grocery store and with family and friends. Very few (4.7%) reported they were “not likely [to wear a mask] at all” in either setting. The percentage of individuals in each county who reported “always” wearing a mask in the New York Times interviews was correlated with those who reported they were “very likely” to wear a mask to the grocery store (Spearman’s ρ = .80, p<0.0001), with family and friends (Spearman’s ρ = .57, p<0.0001), and the composite score (Spearman’s ρ = .57, p<0.0001).

Mask wearing was higher among women, elderly, non-white or Hispanic, and lower income respondents (p<0.0001, Table 1). There was substantial geographic heterogeneity in survey responses (Figure 1, A.), with the highest percentage of mask wearers along the coasts and southern border, as well as in large urban areas. There was a general trend of increased mask use across U.S. Census Divisions throughout the survey period (Figure 1, B.) and the West North Central Census Division reported the lowest mask usage across the entire period surveyed.

Open in a separate window

Figure 1.

Mask wearing across three-digit zip code and time

The percentage of individuals who responded they were very likely to wear a mask to the grocery store and with family and friends was averaged across the entire study period and each 3-digit ZCTA prefix (fuzzy matched with reported zip code) in the United States (Panel A). ZCTA clusters with less than 50 responses are shown in gray. Observations of mask wearing were also aggregated at the daily level across each Census division for the study period (Panel B).

We find a negative relationship between the mean percentage of people that report wearing a mask and the instantaneous reproductive number (Figure 2). In a multivariate logistic regression model adjusting for confounders, social distancing and a time trend, there is a significant association between percent reporting mask wearing and community transmission control (i.e. R_t<1) with an OR = 1.14 [95% CI: 1.07,1.20] (Table 2, Model 1). A 10% increase in mask wearing was associated with an over three-fold increase in odds of transmission control [OR = 3.53, 95% CI: 2.03, 6.43]. The association between mask wearing and community transmission control was robust to changes in definitions of mask wearing, R_t estimated from epiforecasts.io, control for peak R_t, and survey-weight standardization (Table 2, Models 2–6). When adjusting for community self-reported contacts instead of mobility, the association between one percent change in mask wearing and community transmission control was attenuated [OR = 1.09, 95% CI: 1.02, 1.16], but remained significant (Figure S1). Mask wearing was also significantly associated with reduced transmission across multiple R_t dichotomization thresholds (Figure S2) and when categorized as an ordinal variable (Figure S3). A mixed model with a random intercept for state found a stronger association of mask wearing on community transmission control [OR = 1.18, 95% CI: 1.07, 1.30].

Open in a separate window

Figure 2.

Mask Wearing and the Effective Reproductive Number

Box (median and IQR) and whiskers (minimum and maximum excluding outliers) of R_t estimates (rt.live) for each week and state (grey circles). Plot is stratified by quartiles (Q1: 28.8%, Q2: 36.6%, and Q3: 45.2%) of the percentage of individuals that report they are very likely to wear a mask with family/friends and to the grocery store.

Communities with high levels of mask wearing and social distancing are predicted by the model to have the highest probability of community transmission control (Figure 3). States with the highest percentage of reported mask wearing but lowest levels of social distancing have less than a 35% probability of community transmission control, though there are wide confidence intervals. The interaction between social distancing and mask wearing was not statistically significant (Table 2, Model 7).

Open in a separate window

Figure 3.

Mask Wearing, Social Distancing and the Predicted Probability of R_t below 1

Projected values from a logistic regression model measuring the association of community transmission control (R_t <1) with mask-wearing and social distancing in US states adjusting for population density, percent non-white and a time trend (Model 1). Values of social distancing are from the Google Community reports of relative residential time and represent no change from baseline, the national median during the study period (June 3 – July 31, 2020), and 97.5^th percentile observed during the study period. Observed mask wearing was between 8.1%−73.7%, so estimates outside this range are model-based extrapolations. Horizontal line placed at 80% probability of community transmission control, though the desired % may be higher.

We evaluated the change in mask wearing in the 2 weeks before and after statewide mask mandates for 12 states (Figure 4). While there was a general trend of increased mask usage across these states over time, the linear segmented regression models resulted in no significant change in slope in crude, β = .04 [95% CI: −.47, .54], or weighted, β = .31 [95% CI: −.28, .91], mask usage following the interventions. There was a non-significant 2.2% [95% CI: −2.1%,6.5%] change in average mask usage following the mandate in the unweighted model and a 2.31% [95% CI: −2.76%,7.38%] change in the weighted model.

Open in a separate window

Figure 4.

Mask Wearing in 14 Days Preceding and Following Statewide Mask Mandates

Daily measure of the percentage of individuals that report they are very likely to wear a mask with family/friends and to the grocery store in 12 states. Values are for the 14 days preceding and following each state’s institution of a statewide mask mandate. The mean daily value across all states is presented for the same timeframe.

Discussion

The effect of mask wearing on community transmission of SARS-CoV-2 has been the subject of substantial debate, despite evidence from detailed transmission studies and population-wide data from other respiratory pathogens. Here we present findings from over 300,000 serial cross-sectional surveys administered daily in June and July 2020 that confirm a high percentage of reported face mask wearing is associated with a higher probability of transmission control in U.S. states. Face mask wearing was more commonly reported among certain socio-economic groups (especially non-white and lower-income respondents). However, the impact of face mask wearing on lowering the instantaneous reproductive number below one – the threshold required for transmission control – remained despite adjusting for demographics, social distancing and prior peak transmissibility.

Mask wearing is shown to increase the odds of transmission control across all levels of social distancing, suggesting that any intervention to improve this community-based behavior may be worthwhile. The absence of a statistical change in mask wearing the two weeks following statewide mandates highlights the point that regulation alone may not drive increased masking behavior. However, we found that there is a general increase in mask wearing prior to the implementation of these policies and mask mandates may be important tools in maintaining this trend. Masking behavior assessed from anonymous surveys may provide insight into where education or other interventions should be directed. These results are consistent with case studies of mask wearing in the United States^16,17, and potentially provide some insight into why one report found no substantial effect of mask mandates in conjunction with other interventions¹⁸. The data presented here may highlight a gap between governmental policy and actual user behavior, but more research is needed.

Our evidence supports the role of mask wearing in controlling SARS-CoV-2 transmission; however, this ecological study cannot inform questions of causality. It is difficult to disentangle individuals’ engagement in mask wearing from their adoption of other preventative hygiene practices, and mask wearing may be serving as a proxy for other risk avoidance behaviors not queried (e.g. avoiding indoor spaces) – especially considering the magnitude of the reported effects. While our findings remained despite controlling for individual-level contacts from the Facebook survey, a proxy that likely includes other aspects of risk reduction, the possibility of residual confounding is likely. Additionally, observations from smaller states are overrepresented when results are aggregated at the state-level and further observations from survey respondents may not reflect the general population. However, our findings were consistent even with application of US census-based survey-weights.

A state that demonstrates social distancing may also be subject to additional non-modeled but impactful interventions including gathering size reductions, travel limitations, and the closing of businesses³⁸. While the social distancing proxy used here captures the broad activity level that would result from the implementation of these policies, future research should focus on incorporating data from disaggregated interventions with empirical assessments of mask wearing. Additionally, potential R_t and mask wearing within a state may be the result of prior transmission. While we showed the effect of mask wearing was robust to peak R_t in the first wave of the epidemic, our methods do not control for time-dependent confounding or variations in mask usage by susceptibility status.

The validity of epidemiologic parameters of transmission are only as accurate as the incidence data to which the models are fit. If states reporting low mask wearing also underreport incidence (e.g. limited testing), we may be underestimating the true effect of mask wearing. Conversely, instantaneous R_t estimations of transmission are subject to uncertainty, especially towards the end of the time-series before reports are complete²⁷. While our results were robust to different estimators, our model does not account for estimation error. Additionally, mask wearing measures, social distancing and R_t all exhibit significant temporal autocorrelation. To combat this, we dichotomized R_t and aggregated our exposures by week, but further analyses may consider complex time-series models, mechanistic and quasi-experimental methods to estimate the effect of face masks.

We find a community benefit for “face masks” which collectively include masks of various hypothesized efficacies³⁹. We did not query specific mask type or the use of face shields in conjunction with masks. The reported association may understate the maximum potential for face masks to curb respiratory transmission, which can only be realized through increasing the utilization of superior mask materials (i.e. choosing surgical masks over fleece gaiters). Additional study of effect modification by mask type is necessary to estimate true causal effects.

When considering the various challenges the US population has faced in slowing the spread of SARS-CoV-2, evidence on the impact of nonpharmaceutical interventions is paramount. Our data suggest widespread use of face masks by the general public may aid in limiting the SARS-CoV-2 epidemic as social distancing restrictions are rolled back around the United States. Given mixed evidence on the effect of mask mandates, but a strengthening body of evidence on the effect of masks, policy makers should consider innovative strategies for evaluating and increasing mask usage to help control the epidemic.

Supplementary Material

Acknowledgements

Footnotes

References