Associations between wearing masks, washing hands, and social distancing practices, and risk of COVID-19 infection in public: a cohort-based case-control study in Thailand

Objective. To investigate whether wearing masks, washing hands and social distancing practices are associated with lower risk of COVID-19 infection. Design. A retrospective cohort-based case-control study. All participants were retrospectively interviewed by phone about their preventive measures against COVID-19 infection. Setting. Thailand, using the data from contact tracing of COVID-19 patients associated with nightclub, boxing stadium and state enterprise office clusters from the Surveillance Rapid Response Team, Department of Disease Control, Ministry of Public Health. Contacts were tested for COVID-19 using PCR assays per national contact tracing guidelines. Participants. A cohort of 1,050 asymptomatic contacts of COVID-19 patients between 1 and 31 March 2020. Main outcome measures. Diagnosis of COVID-19 by 21 April 2020. Odds ratios for COVID-19 infection and population attributable fraction were calculated. Exposure. The study team retrospectively asked about wearing masks, washing hands, and social distancing practices during the contact period through telephone interviews. Results. Overall, 211 (20%) were diagnosed with COVID-19 by 21 Apr 2020 (case group) while 839 (80%) were not (control group). Fourteen percent of cases (29/210) and 24% of controls (198/823) reported wearing either non-medical or medical masks all the time during the contact period. Wearing masks all the time (adjusted odds ratio [aOR] 0.23; 95%CI 0.09-0.60) was associated with lower risk of COVID-19 infections compared to not wearing masks, while wearing masks sometimes (aOR 0.87; 95%CI 0.41-1.84) was not. Shortest distance of contact >1 meter (aOR 0.15; 95%CI 0.04-0.63), duration of close contact [≤]15 minutes (aOR 0.24; 95%CI 0.07-0.90) and washing hands often (aOR 0.33; 95%CI 0.13-0.87) were significantly associated with lower risk of infection. Sharing a cigarette (aOR 3.47; 95%CI 1.09-11.02) was associated with higher risk of infection. Type of mask was not independently associated with risk of infection. Those who wore masks all the time were more likely to wash hands and practice social distancing. We estimated that if everyone wore a mask all the time, washed hands often, did not share a dish, cup or cigarette, had shortest distance of contact >1 meter and had duration of close contact [≤]15 minutes, cases would have been reduced by 84%. Conclusions. Our findings support consistently wearing non-medical masks, washing hands, and social distancing in public to prevent COVID-19 infections.


Introduction
high-risk contacts. All PCR tests were performed at laboratories certified for COVID-19 testing 133 by the National Institute of Health of Thailand. Data of risk factors associated with  infection, such as type of contact and use of mask, were recorded during the contact investigation, 135 but not complete.

137
The central SRRT team was tasked to perform contact investigations for any cluster with at least 138 five PCR-confirmed COVID-19 patients from the same location(s) within a one-week period. 21 139 We primarily used these data to identify asymptomatic contacts of COVID-19 patients between 1 140 and 31 March 2020. To reduce the bias of the selection of asymptomatic contacts, all contact 141 tracing records of the central SRRT team were used in the study.

143
We then conducted telephone calls and asked details about their contacts with COVID-19 patients 144 (e.g. date, location, duration and distance of contacts), whether they wore masks, washed their 145 hands and performed social distancing during the contact period, and whether the COVID-19 146 patient, if known, wore a mask. We also asked, and checked using records of the DDC, whether 147 and when they were sick and diagnosed with COVID-19. To include only asymptomatic contacts 148 in the study, we excluded people from the analysis who already had any symptoms of 149 including cough, fever, fatigue, diarrhoea, abdominal pain, loss of appetite, and loss of smell and 150 taste, 23 24 on the first day of contact. We also excluded contacts whose contact locations were 151 healthcare facilities because this study aimed to focus on infection in the public. CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted June 12, 2020. . https://doi.org/10.1101/2020 Asymptomatic contacts, cases, controls, index patients, primary index patients and COVID-19 154 patients were defined as described in Table 1. The reporting of this study follows the STROBE 155 guidelines.

157
Selection of cases and controls 158 We defined asymptomatic contacts who were later diagnosed as COVID-19 patients using PCR 159 assays by 21 Apr 2020 as cases (Table 1). All asymptomatic contacts who were not diagnosed as 160 COVID-19 patients using PCR assays by 21 Apr 2020 were controls. We arbitrarily used 21 days 161 after 31 March as the cutoff based on the evidence that most COVID-19 patients would likely 162 develop symptoms within 14 days 25 and it should take less than another 7 days for symptomatic 163 patients, under contact investigations, to present at healthcare facilities and be tested for COVID-164 19 with PCR assays.

166
Statistical analysis 167 Odds ratios and 95% confidence intervals were estimated for associations between development 168 of COVID-19 and baseline covariates, such as wearing masks, washing hands and social distancing 169 using logistic regression with a random effect for location and a random effect for index patient 170 nested within the same location. The interviewer identified the index patient, the symptomatic 171 COVID-19 patient who had the closet contact, if an asymptomatic contact contacted more than 172 one symptomatic COVID-19 patient. The percentage of missing values in the variable whether the 173 COVID-19 patients wore a mask was 27%, and the variable was not included in the analyses. We 174 assumed that missing values were missing at random and used imputation by chained equations. 175 We created 10 imputed datasets and the imputation model included all listed confounders and the 176 . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted June 12, 2020. . https://doi.org/10.1101/2020 case-control indicator. We developed the final multilevel mixed-effect logistic regression models 177 on the basis of previous knowledge and a purposeful selection method. 26

179
We also estimated odds ratios and 95% confidence intervals for associations between compliance 180 of mask wearing and other practices; including washing hands and social distancing using 181 multinomial logistic regression models and the imputed data set. Logistic regression was also used 182 to estimate p value for pairwise comparisons. Bonferroni correction was not performed. We 183 estimated secondary attack rate using definitions as described in Table 1 187 We conducted a sensitivity analysis by including type of mask in the multilevel mixed-effects 188 logistic regression model for COVID-19 infection. We also tested a pre-defined interaction 189 between type of mask and compliance of wearing masks. To respond to the national policy, we estimated population attributable fraction (PAF) using the 193 imputed dataset and a direct method based on logistic regression as described previously (details 194 in Supplementary Text). 27 28 In short, the final multivariable model was modified by considering 195 each risk factor dichotomously, and PAF was calculated by subtraction of the total number of 196 predicted cases from total number of observed cases, divided by the total number of observed cases.  CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)

Sensitivity analyses
The copyright holder for this preprint this version posted June 12, 2020. . https://doi.org/10.1101/2020 Participants and public involvement 201 No participants were involved in setting the research question or the outcome measures, nor were 202 they involved in developing plans for design or implementation of the study. However, the study,

212
Characteristics of the cohort data 213 The contact tracing of the central SRRT team consisted of 1,716 individuals who had contact with 214 or were in the same location as a COVID-19 patient who were associated with three large clusters 215 in nightclubs, boxing stadiums and a state enterprise office in Thailand ( Figure 1). Overall, we 216 considered 18 individuals as primary index patients because they were the first who had symptoms 217 at those places, had had symptoms since the first day of visiting those places, or were considered 218 to be the origin of infection of cases based on the contact investigations; 11 from the nightclub 219 cluster, 5 from the boxing stadium cluster and 2 from the state enterprise office cluster. Timelines 220 of primary index patients from nightclub, boxing stadium and state enterprise clusters are 221 . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted June 12, 2020. . https://doi.org/10.1101/2020  After retrospectively interviewing each contact by phone and applying the exclusion criteria 226 (Figure 1), we included 1,050 asymptomatic contacts who had contact with or were in the same 227 location as a symptomatic COVID-19 patient between 1 and 31 March 2020 in the analysis. The 228 median age of individuals was 38 years (IQR 28-51) and 55% were male (Table 1). Most 229 asymptomatic contacts included in the study were associated with the boxing stadium cluster (61%,  Overall, 211 (20%) asymptomatic contacts were later diagnosed with COVID-19 by 21 Apr 2020 234 (case group) and 839 (80%) were not (control group). Of the 211 cases, 150 (71%) had symptoms 235 prior to the diagnosis of COVID-19 using PCR assays. The last date that a COVID-19 case 236 diagnosed was 9 April 2020. Of 839 controls, 719 (86%) were tested with PCR assays at least once. CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted June 12, 2020. . https://doi.org/10.1101/2020.06.11.20128900 doi: medRxiv preprint boxing stadiums and the state enterprise office at workplaces (n=277), households (n=230) and 244 other places (n=151). Association between compliance of mask wearing and other social distancing practices. 257 Since wearing masks all the time was found to be negatively associated with COVID-19 infection, 258 we wanted to explore characteristics of those patients because of a potential false sense of security 259 caused by wearing masks. We found that those who wore masks all the time were more likely to 260 have shortest distance of contact >1 meter (25% vs. 18%, pairwise p=0.03), have duration of 261 contact within 1 meter ≤15 minutes (26% vs 13%, pairwise p<0.001) and wash their hands often 262 (79% vs. 26%, pairwise p<0.001) compared with those who did not wear masks (Table 3). We 263 found that those who wore masks sometimes were more likely to wash their hands often (43% vs.

264
26%, pairwise p<0.001) compared with those who did not wear masks. However, they were more 265 . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

282
Using the direct method to calculate PAF, we estimated that the proportional reduction in cases 283 that would occur if everyone wore a mask all the time during contact with index patients (PAF of 284 not wearing masks all the time) was 0.28 (Table 4). Among modifiable risk factors evaluated, PAF 285 of shortest distance of contact <1 meter was highest at 0.40. If everyone wore a mask all the time, 286 washed hands often, did not share a dish, cup or cigarette, had shortest distance of contact >1 meter 287 and had duration of close contact ≤15 min, cases would have been reduced by 84%.

288
. CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

Statement of principal findings 292
This cohort-based case-control study provides a supporting evidence that wearing masks, washing 293 hands and social distancing are independently associated with lower risk of COVID-19 infection 294 in the general public. We observed that wearing masks all the time when expose to someone with 295 COVID-19 was associated with lower risk of infection, while wearing masks sometimes was not.

296
This supports the recommendation that people should be wearing their masks correctly at all times 297 in public and at home when there is an increased risk. 2 4 9 10 298 299 We also quantified the effectiveness of different measures that could be implemented to prevent 300 transmission in nightclubs, stadiums, workplaces and other public gathering places. We found that 301 those who wore masks all the time were also more likely to wash hands and perform social 302 distancing. We estimated that adopting all recommendations (wear masks all the time, wash hands 303 often, not sharing dishes, cups or cigarettes, maintain a distance of <1 meter and, if needed, have 304 less than 15 minutes contact) could result in controlling 86% of the burden of COVID-19 infections 305 in our setting during the study period. We recommend that all public gathering places consider 306 multiple measures to prevent transmission of COVID-19 and new pandemic diseases in the future.

308
Public messaging on how to wear masks correctly needs to be consistently delivered, particularly 309 among those who wear masks sometimes or incorrectly (e.g. not covering both nose and mouth).

310
. CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted June 12, 2020. . https://doi.org/10. 1101/2020 This is because, based on our findings, those who wear masks intermittently could be a group that 311 did not practice social distancing adequately.

313
Comparison with other studies 314 The effectiveness of wearing masks observed in this study is consistent with previous studies;  This study found a negative association between risk of COVID-19 infection and social distancing 326 (i.e. distance and duration of contact), which is consistent with previous studies which found that 327 at least 1-meter physical distancing was strongly associated with a large protective effect, and 328 distances of 2 meters could be more effective. 32 Effectiveness of hand hygiene is consistent with 329 the previous studies. 34 Although sharing dishes or cups was not independently associated with the 330 infection in our study, based on previous studies, 35 we still recommend not sharing dishes or cups. CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted June 12, 2020. . https://doi.org/10.1101/2020.06.11.20128900 doi: medRxiv preprint The household secondary attack rate in our study (17%) is comparable with those reported ranging 333 from 11% to 19%, 35 36 and relatively high compared to workplaces and other places. While CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted June 12, 2020. has and some individuals may have been contacts to more than one COVID-19 patient. Hence, our 376 estimated secondary attack rates among contacts with high-risk exposure could be over or under-377 estimated. Fifth, findings were subject to common biases of retrospective case-control studies; 378 . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted June 12, 2020. . https://doi.org/10.1101/2020.06.11.20128900 doi: medRxiv preprint including memory bias, observer bias and information bias. Nonetheless, we used structured 379 interviews, whereby each participant was asked the same set of defined questions, to reduce 380 potential biases. it is important that we continue to expand our understanding about the effectiveness of each 398 measure. Wearing masks, washing hands and social distancing are strongly associated with lower 399 risk of COVID-19 infections. We strongly support wearing non-medical masks in public to prevent 400 COVID-19 infections. We also suggest that medical masks should be reserved for healthcare 401 . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted June 12, 2020. . https://doi.org/10.1101/2020.06.11.20128900 doi: medRxiv preprint workers. Everyone should also wash their hands frequently and comply with recommendations of The authors declare that they have no completing interests.

Ethical approval 427
As this study is part of the routine situation analysis and outbreak investigation of the DDC MoPH 428 Thailand, it was not required to obtain ethics approval and no written informed consent was 429 collected. However, the study team strictly followed ethical standards in research, that is, all   CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted June 12, 2020.

486
. CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted June 12, 2020. . https://doi.org/10.1101/2020.06.11.20128900 doi: medRxiv preprint  CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted June 12, 2020.

553
. CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted June 12, 2020. CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted June 12, 2020. . https://doi.org/10.1101/2020

Classification
Definition Asymptomatic contacts Individuals who had contact with or were in the same location as a symptomatic COVID-19 patient, and had no symptoms of COVID-19 on the first day of contact. Cases Asymptomatic contacts of COVID-19 patients who were later diagnosed and officially reported as COVID-19 patients by 21 Apr 2020.

Controls
Asymptomatic contacts of COVID-19 patients who were never diagnosed as COVID-19 patients by 21 Apr 2020.

Index patients
The COVID-19 patients identified from the contract tracing data as the potential source of infection. Cases (as defined above) could also be included as index patients. Primary index patients The earliest COVID-19 patients whose probable sources of infection were prior to the study period (1 to 31 March 2020), whom we were not able to identify the source of infection from, or whose probable sources of infection were outside the contract tracing data included in the study COVID-19 patients Individuals who had PCR positive for SARS-CoV-2, officially confirmed and reported by Department of Disease Control (DDC), Ministry of Public Health (MoPH), Thailand Secondary attack rate The percentage of new cases among asymptomatic contacts with high-risk exposure High-risk exposure Individuals who lived in the same household as a COVID-19 patient, had a direct physical contact with a COVID-19 case, had face-toface contact with a COVID-19 case within 1 meter and longer than 15 minutes, or were in a closed environment with a COVID-19 patient at a distance of within 1 meter and longer than 15 minutes. Household contact Individuals who lived in the same household as a COVID-19 patient 573 574 . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted June 12, 2020. . https://doi.org/10.1101/2020.06.11.20128900 doi: medRxiv preprint Footnote of Table 2. a Both crude and adjusted odds ratios were estimated using logistic regression with a random effect for location and a random effect for index patient nested within the same location. b The state enterprise office was considered and included as a workplace. Others included restaurants, markets, malls, religious places, households of index patients or other people but not living together, etc. c Location was included in the model as a random effect variable. d During the contact period. e Sharing dishes but using communal spoons all the time was considered as not sharing dishes. f Included sharing electronic cigarettes and any vaping devices. g Included washing with soap and water, and with alcohol-based solutions. h Wearing masks incorrectly (i.e. not covering both nose and mouth) was considered as not wearing. Footnote of Table 3. P values were estimated using univariable multinomial logistic regression models. Missing values were imputed using the imputation model. Wearing masks incorrectly (i.e. not covering both nose and mouth) was considered as not wearing. a The state enterprise office was considered and included as a workplace. b Included restaurants, markets, malls, religious places, public places, households of index patients or other people but not living together, etc. c During the contact period. d Sharing dishes but using communal spoons all the time was considered as not sharing dishes. e Included sharing electronic cigarettes and any vaping devices. f Included washing with soap and water, and with alcohol-based solutions.
. CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted June 12, 2020. . https://doi.org/10.1101/2020  Footnote of Table 4. a Prevalence (Prev) was estimated using the imputed data set. b PAF was estimated using the direct method (Supplementary Text). c During the contact period. d Sharing a dish but using communal spoons all the time was considered as not sharing a dish. e Included sharing an electronic cigarette and any vaping device. f Washing hands included washing with soap and water, and with alcohol-based solutions. g Wearing masks incorrectly (i.e. not covering both nose and mouth) was considered as not wearing. i Age and gender were considered as non-modifiable risk factors, while other risk factors were considered as modifiable. Total PAF was directly estimated using logistic regression in the form of natural logarithm; therefore, total PAF was not equal to the direct summation of PAF of each risk factor. . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted June 12, 2020. . https://doi.org/10.1101/2020.06.11.20128900 doi: medRxiv preprint CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted June 12, 2020. . https://doi.org/10.1101/2020.06.11.20128900 doi: medRxiv preprint Footnote of Figure 2. A, B and C represent the nightclub cluster, boxing stadium cluster and state enterprise office cluster, respectively. Black nodes represent primary index patients, red dots represent cases, and green dots represent controls. Orange dots represent index patients (confirmed COVID-19 patients) who could not be contacted by the study team. Black lines represent household contacts, purple lines represent contacts at workplaces and gray lines represent contacts at other locations. Definition of index patients, cases and controls are listed in Table 1. . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

Supplementary Methods
To respond to the national policy, we estimated direct population attributable fraction (PAF) using the imputed dataset and the direct method as previously described. 27 28 Direct PAF can be obtained by calculating PAFs directly from individuals' data using logistic regression. 27 28 First, we had to modify our final logistic regression model by considering each risk factor dichotomously. Then, irrespective of exposure to each risk factor for each individual, that factor was removed from the population by calculating probability based on all observations as unexposed. The predicted probability of developing COVID-19 infection for each asymptomatic contact, with the assumption that there was no exposure to a certain risk factor, is: Pki is representative of predicted probability of COVID-19 infection in individual asymptomatic contact k, assuming no exposure to a specific risk factor (xi); β j indicates the regression coefficient of risk factor (xj), except risk factor number i (xi). Subsequently, the sum of all predicted probabilities for all individuals in the study would be equal to adjusted estimate of total cases, which is anticipated in the absence of that specific risk factor (xi).
Then, PAF was estimated by subtraction of the total number of predicted cases from total number of observed cases, divided by the total number of observed cases: . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted June 12, 2020. . https://doi.org/10.1101/2020.06.11.20128900 doi: medRxiv preprint PAF = Total number of observed cases − Total number of predicted cases Total number of overserved cases

Supplementary Results
For the pub cluster, we identified 11 primary index patients who started having symptoms from 4 to 8 March and were diagnosed (and isolated) from 3 to 10 March (Supplementary Figure 1). Those primary index patients visited multiple nightclubs included in the analysis during the study period, and 35 of 228 (15%) asymptomatic contacts at nightclubs had PCR-confirmed COVID-19 infections after the contact (Figure 2, Cluster A).
For the boxing stadium cluster, we identified 5 primary index patients who started having symptoms from 6 to 12 March and were diagnosed (and isolated) from 11 to 21 March (Supplementary Figure 2). Those primary index patients visited multiple boxing stadiums included in the analysis during the study period, and 125 of 144 (87%) asymptomatic contacts at the boxing stadiums had PCR-confirmed COVID-19 infections after the contact (Figure 2, Cluster B).
Of the two primary index patients for the office cluster; one had had symptoms since 15 March 2020 (Primary index patient C1 in Supplementary Figure 3) and was considered as the source of infection to one new case in the office during the study period. The other primary index patient (Primary index patient C2 in Supplementary Figure 3) was a household member of a staff at the office, and was considered as the source of infection to that staff via household contact.
. CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted June 12, 2020. . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted June 12, 2020. . https://doi.org/10.1101/2020.06.11.20128900 doi: medRxiv preprint Footnote of Supplementary Table 1. a Both crude and adjusted odds ratios were estimated using logistic regression with a random effect for location and a random effect for index patient nested within the same location. Missing values were imputed using the imputation model. b The state enterprise office was considered and included as workplaces. Others included restaurants, markets, malls, religious places, households of index patients or other people but not living together, etc. c Location was included in the model as a random effect variable. d During the contact period. e Sharing dishes but using communal spoons all the time was considered as not sharing dishes. f Included sharing electronic cigarettes and any vaping devices. g Included washing with soap and water, and with alcohol-based solutions. h Wearing masks incorrectly (i.e. not covering both nose and mouth) was considered as not wearing.
. CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted June 12, 2020.