Impact of local mask mandates upon COVID-19 case rates in Oklahoma.

Use of face coverings has been shown to reduce transmission of SARS-CoV-2. Despite encouragements from the CDC and other public health entities, resistance to usage of masks remains, forcing government entities to create mandates to compel use. The state of Oklahoma did not create a state-wide mask mandate, but numerous municipalities within the state did. This study compares case rates in communities with mandates to those without mandates, at the same time and in the same state (thus keeping other mitigation approaches similar). Diagnosed cases of COVID-19 were extracted from the Oklahoma State Department of Health reportable disease database. Daily case rates were established based upon listed city of residence. The daily case rate difference between each locality with a mask mandate were compared to rates for the portions of the state without a mandate. All differences were then set to a d0 point of reference (date of mandate implementation). Piecewise linear regression analysis of the difference in SARS-CoV-2 infection rates between mandated and non-mandated populations before and after adoption of mask mandates was then done. Prior to adopting mask mandates, those municipalities that eventually adopted mandates had higher transmission rates than the rest of the state, with the mean case rate difference per 100,000 people increasing by 0.32 cases per day (slope of difference = 0.32; 95% CI 0.13 to 0.51). For the post-mandate time period, the differences are decreasing (slope of -0.24; 95% CI -0.32 to -0.15). The pre- and post- mandate slopes differed significantly (p<0.001). The change in slope direction (-0.59; 95% CI -0.80 to -0.37) shows a move toward reconvergence in new case diagnoses between the two populations. Compared to rates in communities without mask mandates, transmission rates of SARS-CoV-2 slowed notably in those communities that adopted a mask mandate. This study suggests that government mandates may play a role in reducing transmission of SARS-CoV-2, and other infectious respiratory conditions.

Introduction

Efforts for reducing community transmission of severe acute respiratory syndrome Coronavirus 2 (SARS-CoV-2), the causative agent for Coronavirus disease 2019 (COVID-19), initially focused on dramatic measures including federally-endorsed and state-mandated “shut-downs.” These included closure of schools (primary, secondary, and higher education), businesses, churches, and events, as well as restrictions on travel and nearly all commercial and non-commercial activities outside the home. These restrictions were effective in slowing the emergence of the pandemic in the US [1], but inflicted great costs, economic and otherwise [2, 3]. With the recognition that these measures could not be permanent, and as understanding of transmission increased, more targeted policy recommendations became common, including reduction of large gatherings, encouragement of social distancing, utilization of face coverings [4], and eventually vaccination. Once the shutdowns were eased transmission increased, particularly in areas of the US where cases had been low previously. This may be attributed to limited adoption of the targeted mitigation steps, which was likely driven by myriad of factors, including politicization of such measures as well as the economic burden of social distancing efforts. However, it also calls into question the effectiveness of the suggested mitigation strategies.

Of particular interest has been the effectiveness of use of masks or face coverings by the general public. Recommendations regarding use of masks or face coverings have varied since the beginning of the epidemic. Initially, wearing of face coverings was advised only for symptomatic individuals or in healthcare settings. Gradually, evidence suggested transmission occurred primarily via respiratory secretions rather than via fomites, leading to recommendations for face coverings to be more commonly employed [5]. Numerous studies have demonstrated the efficacy of face coverings in reducing transmission [6, 7], including a review which thoroughly examined the evidence on mask usage [8]. The authors recommended adoption of public cloth mask wearing, in conjunction with other measures. However, they also acknowledge challenges in evaluating efficacy of seeking population-wide mask usage, including deficiencies in compliance. Efficacy of masks were also called into question by Shah, et al. [9], who found that more commonly used cloth masks and even surgical masks offer relatively low apparent filtration efficiencies (<15%). Cheng, et al. [10] examined case incidence in populations with community-wide masking as compared to that of non-mask wearing communities, but this was across different countries where other societal, meteorological, or regulatory factors may have also played a role in reducing transmission. Importantly, this study compared the Hong Kong Special Administrative Region (HKSAR) to European and North American countries. The authors acknowledge that the HKSAR general population was on high alert after the previous SARS epidemic, and documented compliance of face mask usage of >95% on three consecutive days.

Compliance comparable to that reported by Cheng, et al. would not be expected in the US in the absence of government-imposed mask mandates, and perhaps not even then. Thus, one must be careful in considering what benefit could be derived from mask mandates in the US. To this question, incidence was assessed within a single US state before and after mitigation measures [11]. This examination included multiple mitigation steps, not just face covering mandates. Moreover, it was a simple longitudinal assessment, which can be confounded by changes over time due to other causes. A more recent project by Lyu and Wehby examined case occurrence data in light of statewide mask mandates or employment-related mask mandates, and in doing so attempted to account for shutdown restrictions in the states examined as well as week of the year [12]. A notable limitation of this work is the fact they only examined results over a seven week timeframe. Periodic “waves” of cases in various states or geographic regions has been a recurring phenomenon, but these waves have not necessarily followed seasonal or recurring intervals and are often difficult to predict or explain [13]. The impact of social norms, messaging, and now vaccination efforts vary enormously across states, and may account for some of this periodicity. Regardless of cause(s), it is unlikely that the complexities of this periodicity could be captured in any seven week period. A publication by Joo, et al., compared case data at the county level for states with mandates to those without mandates and described a reduction in hospitalization rates following state-wide mask mandates [14], over a roughly seven month period. Both Joo, et al. and Lyu and Wehby studies are limited by the fact that it is difficult or impossible to account for across-state variables, as well as changes in transmission over time due to other external factors. Specifically, the models included state as a variable in the regression. However, when no replicate exists for each state, such inclusion requires that the effect of the state is constant across the entire study period. This assumption cannot be evaluated and may be flawed. These factors complicate efforts to compare mitigation steps across states, and particularly over relatively short periods of time (as Lyu and Wehby examined).

Lyu and Wehby’s model included population density, socioeconomic status, and demographic information. Such information may strengthen the model in many ways, but it assumes that the effect of these considerations are consistent across multiple states. It is further likely that the prevailing political and social norms within a state are at least as important in influencing opinions and actions, as are the effects of socioeconomic or racial characteristics [15]. As such, nations, states, or even smaller municipalities that are considering mandating one or more mitigation strategies may prefer an examination of impacts of individual mitigation steps (as opposed to multiple policies concurrently), as well as comparisons within a state, to determine efficacy of mask mandates. Lyu and Wehby used an event study approach, and examined discreet time intervals following adoptions of mandates. This, or similar approaches (such as the simpler variation on a “difference in differences” approach reported here), may be best suited to account for variability over time.

Similar to most states in the US, Oklahoma instituted various state-wide restrictions early in the pandemic. These were initiated on March 24, with a comprehensive shut-down order issued on April 1. On April 22 the governor announced his “Open Up and Recovery Safely (OURS)” plan which involved various dates and stages for the state to reopen. By June 1, all state-wide restrictions were lifted and none were reimposed before late November 2020. Despite this absence of statewide restrictions, numerous municipalities instituted mask mandates. These municipalities are generally larger population centers in the state, including the two largest metropolitan areas- Oklahoma City and Tulsa- along with many of their surrounding suburbs. However, there are a number of municipalities with populations of 15,000 to 50,000 that are not a directly related suburb of Oklahoma City or Tulsa, as well as handful of non-suburban municipalities with populations of less than 10,000. These mandates were instituted in two general waves, with some variation in each of those. The first wave ranged from late March to early September, with most taking effect in July (Table 1). The second wave began in early November and continued through mid-December. The study described here analyzed the effect of mask mandates on case density rates. Rather than simply comparing within a location longitudinally, the primary goal was to compare populations with a mask mandate to those without a mandate at the same time point. Because mandates went into effect at different times, it was decided to use a modified difference-in-differences approach for comparing municipalities with mandates to those without. This entailed establishing a universal day 0 (d0) for mask mandate implementation and comparing all citizens residing in a given municipality with a mask mandate on that day to all citizens in the state in a municipality without a mask mandate. The hypothesis was that mask mandates would have no impact on case density rates.

Table 1

Oklahoma municipalities that adopted mask mandates, along with the population, mandate start and expiration dates.

City	Population	Mandate Start	Mandate End
Altus	18,338	3/23/2020	5/4/2021
Guthrie	11,376	4/7/2020	5/5/2020
Chickasha	16,337	4/10/2020	5/1/2020
Anadarko	6,504	4/18/2020	4/14/2021
Ada	17,235	4/20/2020	5/17/2021
Norman	124,880	7/7/2020	5/18/2021
Stillwater	50,299	7/11/2020	5/25/2021
Tulsa	401,190	7/12/2020	4/30/2021
Oklahoma City	655,057	7/17/2020	3/5/2021
Lawton/Ft. Sill	93,025	7/20/2020	3/23/2021
Warr Acres	10,118	7/21/2020	3/31/2021
The Village	9,564	7/22/2020	5/1/2021
Spencer	3,968	7/23/2020	12/31/2021
Edmond	9,4054	7/27/2020	3/23/2021
Nichols Hills	3,938	7/27/2020	4/30/2021
Shawnee	31,436	7/27/2020	4/30/2021
Midwest City	5,7407	7/28/2020	3/31/2021
Del City	21,822	8/3/2020	8/31/2020
Tahlequah	16,819	8/3/2020	3/31/2021
Choctaw	12,474	8/5/2020	4/20/2021
McAlester	17,814	8/20/2020	11/30/2020
Okmulgee	11,846	11/9/2020	4/14/2021
Jenks	23,767	11/11/2020	4/30/2021
Ardmore	24,698	11/12/2020	4/5/2021
Clinton	9,217	11/17/2020	3/19/2021
Sapulpa	21,278	11/18/2020	5/4/2021
Grove	6,957	11/20/2020	4/6/2021
Hominy	3,431	11/20/2020	4/14/2020
Okemah	3,178	11/23/2020	4/12/2021
Chouteau	2,066	11/24/2020	6/7/2021
Muskogee	37,113	11/25/2020	1/25/2021
Sand Springs	19,905	11/27/2020	4/27/2021
Enid	49,688	12/3/2020	3/17/2021
Ponca City	24,134	12/14/2020	4/12/2021
Seminole	7,219	12/14/2020	4/14/2021
Vinita	5,423	12/15/2020	2/17/2021
Glenpool	13,936	12/18/2020	4/5/2021
Claremore	18,753	12/20/2020	4/5/2021
Purcell	6,401	12/21/2020	3/1/2021

Materials and methods

Throughout the study period, SARS-CoV-2 testing was only available from healthcare settings (i.e., no tests were approved and marketed for at-home testing during the study period). Moreover, SARS-CoV-2 was a reportable disease in Oklahoma, with all healthcare providers and diagnostic laboratories required to report positive cases to the Oklahoma State Department of Health (OSDH), which there then recorded into the OSDH Public Health Investigation and Disease Detection of Oklahoma (PHIDDO) system. For this study, the OSDH PHIDDO records were queried for all diagnosed cases of COVID from March 17th, 2020 to March 1st, 2021. This included all cases diagnosed via polymerase chain reaction (PCR) testing (classified per criteria proposed by the Centers for Disease Control and Prevention as Confirmed Cases), antigen detection testing (classified as Probable Cases) and cases where a person had a known exposure to a COVID infected individual and subsequently developed symptoms consistent with COVID (designated as epidemiological links, and classified as Probable Cases) [16]. Cases were examined for city or town of residence and event date. Event date is the date of symptom onset if the person has symptoms, or date of sample collection if the person either did not demonstrate symptoms, or was sampled prior to symptom onset. Data were collected as part of routine reportable disease surveillance activities, and were deidentified prior to analysis. The study design was reviewed by the OSDH Institutional Review Board (IRB) administrator, who determined that the procedures are considered public health practice and further IRB review/oversight was not required.

Populations for towns of interest were derived from 2019 census estimates [17]. Date of implementation of mask mandates were obtained from websites compiling municipalities with mandates [18], through communication with public health officials throughout the state, and confirmed via search of city websites or other public records sources. Subsequently, city or town officials were contacted to confirm the adoption and effective dates of the mask mandate.

Daily case density rates (cases per 100,000 people) were calculated for each municipality that established a mandate, including calculation of rates for time points prior to implementation (designated d1, d2, etc.). Additionally, corresponding rates were determined for the rest of the state for the same date, excluding other municipalities that already had mandates in effect on that date. The difference in rates was then calculated by subtracting the rate for the “non-mandated regions” of the state from the rate for the municipality in question. A parallel trends analysis was done to confirm suitability of using a modified difference-in-differences approach. This entailed plotting case rates for municipalities that would adopt a mandate and those municipalities that did not, by date (without regard to date of implementation of mandates).

After the difference in rates were calculated for the 39 municipalities with a mask mandate, all were aligned to a “day 0” (d0) set point, and curated to include a range from d-45 to d90 (or as close to that range as possible, as some municipalities had instituted their mandate less than 90 days prior to analysis or the mandate was in effect for less than 90 days). For municipalities that had a mandate that expired within the period of analysis, effective at the time of expiration/termination the population for that municipality was removed from the state’s population estimate and the cases for that municipality were removed from all calculations. This prevents contamination of results of “mandated” communities by a municipality without a mandate in effect at the time, and also prevents counting cases and population as “non-mandated” in the period immediately following a previous mandate.

Comparisons were then done within a pre-mandate time period (ranging from d-45 to d0) and post-mandate (d14 to d90). No examination was done of d1 to d14, to account for the transition from pre- to post-mandate. Descriptive statistics were calculated for pre- and post-mandate periods. The difference in rates for each day was determined to be either less than or greater than zero (i.e., higher or lower in mandated municipalities vs. non-mandated municipalities [or municipalities that would eventually adopt a mandate, in the case of pre-mandate time period]). The collection, organization, and curation of data is summarized in Fig 1.

Fig 1

Summary of data collection, organization, and curation process for comparing case rates between municipalities with mask mandates vs. those without mandates.

To assess statistical significance of the differences found, a piecewise linear regression model was fit to the difference in case rates and a test was performed to determine if the slopes of the pre- and post-mandate periods were equal. A Durbin-Watson test was also performed to determine if serial correlation was an issue, and showed significant serial correlation. Consequently, a one-lag moving average term was added to the model, and the Durbin-Watson test for this adjusted model showed no evidence of serial correlation (p = 0.864). Additionally, residual plots from this modified model were obtained to examine possible departures from the required assumptions and 95% confidence intervals were calculated for slopes for pre-mandate rate differences, as well as post-mandate rate differences. Analysis was repeated with exclusion of apparent outlier observations to see if their inclusion altered conclusions.

Results

The search process identified 39 municipalities as issuing mandates. A listing of those municipalities is provided in Table 1.

The “non-mandated” population ranged from a maximum of approximately 3,957,000 (i.e., the full state population) to approximately 1,994,000. When assessed strictly by date, parallel trends were clearly demonstrated between mandated and non-mandated communities, meaning assessment of difference-in-differences was appropriate (S1 Fig). Residual plots of data after setting to d0 showed no evidence of assumption violations, and no outlier observations exerted disproportionate impact; therefore, all data points were retained in final results. Fig 2 shows a plot of the resulting model for the case rates. The pre-mandate time period showed rapidly increasing case rate differences between the locations that were to enact a mask mandate and the rest of the state (slope of 0.32; 95% CI 0.13 to 0.51). In other words, case rates in those areas were increasing faster than in other parts of the state, with the mean case rate in municipalities that will adopt a mandate increasing by 0.32 cases per 100,000 people per day (as compared to the rest of the state). For the post-mandate time period, the differences are decreasing (slope of-0.24; 95% CI -0.32 to -0.15, reflecting that changes in daily case rates were now 0.24 cases per 100,000 lower in communities that adopted mandates (as compared to the rest of the state). The pre- and post- mandate slopes differed significantly (p<0.001). The change in slope direction (-0.59; 95% CI -0.80 to -0.37) shows a move toward reconvergence in new case diagnoses between the two populations. Thus, while overall rates remained higher in mandate communities, growth in rates had slowed and were approaching comparability to non-mandated areas.

Fig 2

Difference in case rates per 100,000 population, between municipalities with mask mandates vs. those without mandates.

Discussion

Early state-wide restrictions coincided with very low counts of COVID-19 cases. Only a few, relatively small municipalities instituted mask-mandates during the period of state-wide restrictions. Case counts throughout the state increased after the state relaxed restrictions and permitted greater return to normal activities. However, restrictions were lifted in the summer, at a time where transmission was somewhat dampened by meteorological conditions and/or other factors (Fig 3). Nonetheless, the major metropolitan areas saw case count increases, and adopted mandates in mid-July. This was followed by mandates in many surrounding suburban communities. Despite these measures, case counts across the state increased notably as fall conditions emerged and schools returned to classes. Case counts (as well as hospitalizations and deaths) increased greatly in late fall and early winter (Fig 3), resulting in the second wave of mandates in November and December. The method used in the analysis reported here (a modification of the difference-in-differences approach, setting to a d0 implementation period) allows inclusion of this second wave of mask mandates in an analysis of efficacy of these measures, despite being implemented at notably different times than the first wave. While accurately reflecting statewide disease burden, these general descriptions of case rate dynamics obscure diversity of changes across the state. Our analysis shows that, prior to adoption of mask mandates, rates were increasing faster in the areas that would eventually adopt mandates than in those that would not (i.e., the difference in rates was greater than zero). After communities began adopting mask mandates, case rates became much less divergent in mandate and non-mandate communities, albeit they did not return to equal in the time period examined. Despite the difference in case rates remaining greater than zero, the change in slope of the difference is dramatic and clinically significant.

Fig 3

Daily case numbers and 7 day average of cases for the state of Oklahoma.

Source: OSDH Weekly Epidemiology and Surveillance Report.

Results reported here show that implementation of mask mandates coincided with a decrease in transmission of COVID-19 within those communities implementing the mandate. Prior to mandates being adopted, rates of transmission were much higher in communities that would eventually adopt mandates, as compared to the remainder of the state (i.e., the difference between rates was greater than zero, and had a clear discernible positive slope). It could be argued that the higher rates of disease prior to mandates point to inherent differences between those communities and others that did not adopt mandates. The most apparent of these differences would seem to be degree of urbanization. While relatively little research is available on COVID-19 transmission in rural vs. urban areas, increased population density has been shown to favor disease propagation [19]. The municipalities adopting mask mandates were generally larger than communities not adopting mandates, and include not only the two largest major metropolitan areas (Oklahoma City and Tulsa), but eight of the ten largest population centers of the state. Much of the remainder of the state is more rural. The fact that mandates appeared to largely mitigate this difference (which was manifest prior to d0) is supportive of the efficacy of mask mandates. There are several possible explanations for the persistently higher case rate difference between communities with mask mandates and those without, even after mandate implementation. The most dramatic increase in case rates initiated the introduction of many of the mandates, and continued for some time following them. This period coincided roughly with colder ambient conditions, and may reflect the limitations of mask mandates to overcome increased transmissibility in urban environments. Alternatively, it may reflect gradual non-compliance with the mandate, as people tired of mitigation efforts. Finally, it may reflect something of a spillover effect, where the benefits of mask mandates are not simply limited to those localities with the mandates but also impact transmission in other areas [20].

The study reported here employs a simpler analytical approach but may offer several advantages compared to previous work examining the effectiveness of mask mandates. The authors are unaware of any other studies that have compared communities with and without mandates within the same state. This approach avoids complications of varying social norms and attitudes, as well as other governmental restrictions or mitigation efforts. Moreover, the approach of calculating rate differences between municipalities with mandates versus the remainder of the state, and then setting all implementation dates to a standard d0 basis avoids any potential confounding effects of seasonal variation and changes over time that have been seen throughout the country and world with COVID-19. If utilization of masks is effective in preventing COVID-19 transmission, a compounding effect would be expected to occur over time, as secondary cases that would have developed are reduced by elimination of primary infections. However, without the d0 approach used here, this could not be appreciated unless all mandates were implemented on the same day. Our study also allows examination in one analysis of disease transmission dynamics between two general time periods of mandate adoption (mid- to late summer vs. late fall/early winter).

There are several limitations of the study reported here. The first is the observational nature of the study, which precludes determination of cause and effect. Most reported studies on mask usage impact on COVID-19 transmission have been observational, and our results are consistent with the majority of them. One randomized control trial has been reported, which found no benefit to wearing a mask in preventing COVID-19 infection [21]. Additional research is warranted to determine if this discrepancy is due to differences in prevalence at the time of study, compliance, or other factors; or, whether mask mandates or even voluntary use of masks corresponds with increased awareness and adoption of other mitigation strategies.

The current study is also ecological in nature. It was not possible to assess compliance with mask usage in communities with or without mandates, and there is no ability to determine whether cases were occurring disproportionately among individuals wearing masks. Enforcement of mandates varied notably across the various municipalities and even over time. Given that many of the mandates were limited to indoor gatherings, enforcement was done via business or venue staff, or at least was dependent upon staff notifying law enforcement of failure to comply [22]. Thus, it is also impossible to determine that mandates increased utilization of masks. Nonetheless, these limitations do not reduce the relevance of this study for policy makers who are most concerned with outcomes at the population level.

An additional limitation for the present study includes the need to rely on diagnosed cases. The d-45 period for early adopting cities aligned with the time of lifting of state-wide restrictions. However, this time also coincided with increasing access to widespread testing. As such, it is important to recognize that any noted increases may reflect increased detection as well as increased transmission. Specifically, if case rates increased in locales after mandate implementation, it should not be assumed that the mandates failed. We feel that this concern is limited by the approach used for our analysis, including comparison of communities at the same time (rather than comparing communities to themselves on a “before and after” basis). There would be no such concerns for municipalities in the second wave, as COVID-19 testing was not limited in availability during the fall and winter. Additionally, it is known that diagnosed cases rates do not account for all disease transmission. The relative efficacy of detection of a disease can be assessed through percent positivity, which refers to the percentage of all tests performed that are positive. Robust and effective surveillance efforts can be assumed to be present only if a relatively large proportion of people needing to be tested have access to the test. This would include anyone with symptoms consistent with COVID-19, as well as those with a history of exposure, those at high-risk of exposure, etc. When such widespread availability of testing is present, many of the results should be negative. In this instance, some of the people being tested have some other explanation for their symptoms, while others who were potentially exposed are not infected (or at least not shedding virus yet). When percent positivity is relatively high, it is likely that testing is only being performed on people who are highly symptomatic or most likely to be infected. In this case, subclinical and pre-clinical infections are likely being missed. Oklahoma maintained a relatively high percent positivity rate throughout the study period (always >5% and typically >10%) [23]. This indicates that a relatively large proportion of infections were not being detected in the state. We know of no reason why there would be a bias in percentage of infections diagnosed between communities with mask mandates and those without, but it cannot be ruled out as a potential bias as accurate and comprehensive percent positivity is not available at a city or county level in Oklahoma.

The study reported here includes a timeframe (late December 2020 to March 1, 2021) that coincided with introduction of SARS-CoV-2 vaccines. During this period, vaccination efforts were focused on high-risk situations (residents and employees at long term care facilities and healthcare providers), with only a small percentage of the state population becoming fully vaccinated before the termination of the study. It is therefore unlikely that findings of our study are biased in any way due to vaccination. It is unclear what impact widespread vaccination or immunity would have on the findings; thus, further study is warranted. Finally, more complicated methods exist that attempt to account for other variables, including population density, and socioeconomic or demographic characteristics. We chose to employ a simpler approach for a number of reasons. First, while population density is assuredly associated with increased risk of disease, inclusion of population density as a covariate in our study would be difficult. Specifically, the largest metropolitan areas adopted mask mandates, and almost no rural areas adopted mask mandates. Only in the middle range of population density do we have the ability to compare within roughly equal sized communities. Extrapolating this to the high and low ranges of population density would risk violating unverifiable assumptions. Moreover, inclusion of other variables, including socioeconomic status and demographic information (at the municipal level) adds complexity to the model and is only valuable if it is predictive of risk of infection. Applying additional variables, including socioeconomic status and demographic information at the ecological level risks assumptions of association that cannot be tested or verified. Without adequate empirical data to confirm associations of these variables with the outcome of interest, we were reluctant to unnecessarily complicate the model.

The CDC previously issued guidance that fully immunized people do not need to wear face coverings or practice social distancing [24]. However, this was later revised to state that masking may be indicated, even for vaccinated individuals, in circumstances of high community spread. Vaccine uptake slowed throughout the US after initial offerings, and vaccination efforts in other parts of the world also lag far behind what will be needed to contain the virus. In addition, the emergence of variants, uncertainty of duration of immunity, and vaccine hesitancy all suggest achieving “herd immunity” could be difficult, and impacts in the interim may be devastating. It remains to be seen whether seasonal patterns will become evident, with return of high numbers of cases in winter; seasonality of COVID-19 remains undetermined, as evidence is contradictory [25-27]. In such cases, policy makers may again need to examine what measures (voluntary or compulsory) are needed to protect public health. Indeed, recent work has suggested that improved compliance with mitigation strategies could be even more critical than vaccine efficacy, at least in terms of short term outcomes of hospitalizations and deaths [28]. As such, it is critical that states and municipalities be able to assess the impact of mask mandates on community transmission. The results reported here demonstrate that local mandates are indeed associated with reducing case densities. This evidence may also prove beneficial in considering mitigation strategies for future infectious disease outbreaks.

Conclusions

Our research identified notable change in disease dynamics associated with implementation of mask mandates. Many factors will affect the future impacts of COVID-19, with vaccination expected to be the most critical. However, it is likely that COVID-19 will remain an important public health threat for the foreseeable future. Just as importantly, infectious disease experts caution that future pandemics remain a very real risk, with respiratory diseases being the most likely. The notable economic and social impacts of early “shutdowns” make clear that more targeted interventions are needed for future control efforts. Policy makers should consider the possibility that less intrusive measures, including mandating use of face masks, may be effective at minimizing disease spread while avoiding disruptive effects of less focus strategies.

Graphic representation of parallel trends for municipalities that would adopt mandates and those without mandates.

(JPG)

Click here for additional data file.

(XLSX)

Click here for additional data file.

20 Jan 2022

A rebuttal letter that responds to each point raised by the academic editor and reviewer(s). You should upload this letter as a separate file labeled 'Response to Reviewers'.

A marked-up copy of your manuscript that highlights changes made to the original version. You should upload this as a separate file labeled 'Revised Manuscript with Track Changes'.

An unmarked version of your revised paper without tracked changes. You should upload this as a separate file labeled 'Manuscript'.

If you would like to make changes to your financial disclosure, please include your updated statement in your cover letter. Guidelines for resubmitting your figure files are available below the reviewer comments at the end of this letter.

If applicable, we recommend that you deposit your laboratory protocols in protocols.io to enhance the reproducibility of your results. Protocols.io assigns your protocol its own identifier (DOI) so that it can be cited independently in the future. For instructions see: https://journals.plos.org/plosone/s/submission-guidelines#loc-laboratory-protocols. Additionally, PLOS ONE offers an option for publishing peer-reviewed Lab Protocol articles, which describe protocols hosted on protocols.io. Read more information on sharing protocols at https://plos.org/protocols?utm_medium=editorial-email&utm_source=authorletters&utm_campaign=protocols.

We look forward to receiving your revised manuscript.

Kind regards,

Sinan Kardeş, M.D.

Academic Editor

PLOS ONE

Journal Requirements:

When submitting your revision, we need you to address these additional requirements.

1. Please ensure that your manuscript meets PLOS ONE's style requirements, including those for file naming. The PLOS ONE style templates can be found at

https://journals.plos.org/plosone/s/file?id=wjVg/PLOSOne_formatting_sample_main_body.pdf and

https://journals.plos.org/plosone/s/file?id=ba62/PLOSOne_formatting_sample_title_authors_affiliations.pdf

2. Thank you for stating the following financial disclosure:

"Unfunded study."

At this time, please address the following queries:

a) Please clarify the sources of funding (financial or material support) for your study. List the grants or organizations that supported your study, including funding received from your institution.

b) State what role the funders took in the study. If the funders had no role in your study, please state: “The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.”

c) If any authors received a salary from any of your funders, please state which authors and which funders.

d) If you did not receive any funding for this study, please state: “The authors received no specific funding for this work.”

Please include your amended statements within your cover letter; we will change the online submission form on your behalf.

3. We note that you have stated that you will provide repository information for your data at acceptance. Should your manuscript be accepted for publication, we will hold it until you provide the relevant accession numbers or DOIs necessary to access your data. If you wish to make changes to your Data Availability statement, please describe these changes in your cover letter and we will update your Data Availability statement to reflect the information you provide.

[Note: HTML markup is below. Please do not edit.]

Reviewers' comments:

Reviewer's Responses to Questions

Comments to the Author

1. Is the manuscript technically sound, and do the data support the conclusions?

The manuscript must describe a technically sound piece of scientific research with data that supports the conclusions. Experiments must have been conducted rigorously, with appropriate controls, replication, and sample sizes. The conclusions must be drawn appropriately based on the data presented.

Reviewer #1: Yes

Reviewer #2: Partly

Reviewer #3: Yes

**********

2. Has the statistical analysis been performed appropriately and rigorously?

Reviewer #1: No

Reviewer #2: Yes

Reviewer #3: Yes

**********

3. Have the authors made all data underlying the findings in their manuscript fully available?

The PLOS Data policy requires authors to make all data underlying the findings described in their manuscript fully available without restriction, with rare exception (please refer to the Data Availability Statement in the manuscript PDF file). The data should be provided as part of the manuscript or its supporting information, or deposited to a public repository. For example, in addition to summary statistics, the data points behind means, medians and variance measures should be available. If there are restrictions on publicly sharing data—e.g. participant privacy or use of data from a third party—those must be specified.

Reviewer #1: Yes

Reviewer #2: No

Reviewer #3: No

**********

4. Is the manuscript presented in an intelligible fashion and written in standard English?

PLOS ONE does not copyedit accepted manuscripts, so the language in submitted articles must be clear, correct, and unambiguous. Any typographical or grammatical errors should be corrected at revision, so please note any specific errors here.

Reviewer #1: Yes

Reviewer #2: Yes

Reviewer #3: Yes

**********

5. Review Comments to the Author

Please use the space provided to explain your answers to the questions above. You may also include additional comments for the author, including concerns about dual publication, research ethics, or publication ethics. (Please upload your review as an attachment if it exceeds 20,000 characters)

Reviewer #1: 1. In lines 55 and 56 you said "Numerous studies have demonstrated the efficacy of face coverings in reducing transmission, but these have generally been in a healthcare setting", while in reference 6 efficacy of face coverings and other restrictions are compared to healthcare and non-healthcare setting. Please correct this.

2. From lines 106 to 129 you have provided a description of Table 1 that has no specific reference. Please provide a reference for this content. If the information was collected from a specific center (such as your center) you should mention this.

3. Please tell us, have you done an analysis to check for the presence of outliers in the statistical analysis section? In Figure 1, there appears to be an outlier value that affects your analysis output.

4. In lines 208, 238, 249, and 308, you said that seasons can be a risk factor for increasing the number of cases of Covid-19. It is not true. Please check "Pan, J., et al., Warmer weather unlikely to reduce the COVID-19 transmission: An ecological study in 202 locations in 8 countries. Sci Total Environ, 2021. 753: p. 142272."

5. Please indicate whether during your study, vaccination was in progress or not. If yes, you should not ignore the effect of vaccination.

Reviewer #2: The authors seek to contribute to the fight against the COVID-19 pandemic. Use of face masks or coverings is one of the major recommendations for reducing transmission, hence the relevance of this manuscript. Comments to be addressed to improve on the quality of the manuscript are as follows:

1. The methodology is not clear. A flow chart on the sample analyzed will bring some clarity. Also case counts for the period of analysis is needed.

2. Authors should provide literature/references on the enforcement of the mask mandate in Oklahoma. This will strengthen argument on the impact of the mandate.

3. Authors should provide the number of municipalities in Oklahoma and the exact number that implemented the mask mandate (39 or 40: refer to lines 110 and 160)

4. Why the selection of d-45 to d90 and not d-45 to d45 (providing equal intervals pre-and post-)?

5. Authors should provide the actual dates for period of analysis. Which date was d0?

6. Authors should check table 1:

- There are two columns for "Mandate start".

- Should also check dates for Midwest city.

- Implementation of state-wide restrictions (March-June 2020) is likely to dilute the impact of mask mandates. Authors should consider excluding the first five municipalities.

7. Authors should provide reference for the CDC classifications (lines 134-139)

8. There is no data or/and references in the manuscript to support lines 200-210)

9. Difficult to appreciate higher rate of increase in municipalities to adopt the mask mandate. What was peculiar about these municipalities? (lines 215 - 217). Authors need to describe the study sites in the methodology

10. One would have expected cases to increase faster in municipalities not implementing the mask mandate. This needs explanation (lines 218 -2020)

11. Authors should provide references for categorical statements (lines 280 - 282)

Reviewer #3: Please see the attached document for my review.

**********

6. PLOS authors have the option to publish the peer review history of their article (what does this mean?). If published, this will include your full peer review and any attached files.

If you choose “no”, your identity will remain anonymous but your review may still be made public.

Do you want your identity to be public for this peer review? For information about this choice, including consent withdrawal, please see our Privacy Policy.

Reviewer #1: Yes: Mehrdad Bagherpour Kalo

Reviewer #2: No

Reviewer #3: No

[NOTE: If reviewer comments were submitted as an attachment file, they will be attached to this email and accessible via the submission site. Please log into your account, locate the manuscript record, and check for the action link "View Attachments". If this link does not appear, there are no attachment files.]

While revising your submission, please upload your figure files to the Preflight Analysis and Conversion Engine (PACE) digital diagnostic tool, https://pacev2.apexcovantage.com/. PACE helps ensure that figures meet PLOS requirements. To use PACE, you must first register as a user. Registration is free. Then, login and navigate to the UPLOAD tab, where you will find detailed instructions on how to use the tool. If you encounter any issues or have any questions when using PACE, please email PLOS at figures@plos.org. Please note that Supporting Information files do not need this step.

Submitted filename: PONE-D-21-39472 Reviewer Comments.docx

Click here for additional data file.

11 Mar 2022

As always, we are appreciative of the time and effort of the reviewers. We believe that the revised manuscript is notably improved by consideration of the suggestions and issues raised. We have endeavored to address all inquiries below, and have accepted/addressed nearly all suggestions or requests. The one suggestion that we are reluctant to incorporate is the graphic reporting of direct rates for day counts (as opposed to the difference in rate counts). We have concerns in doing so, which we ask the reviewers and editor to consider, but we would further evaluate such a figure if reviewers and editor deem it appropriate and beneficial.

Thank you for your continued investment of time. We look forward to continued improvement and publication of this timely work.

Review Comments to the Author

Please use the space provided to explain your answers to the questions above. You may also include additional comments for the author, including concerns about dual publication, research ethics, or publication ethics. (Please upload your review as an attachment if it exceeds 20,000 characters)

Reviewer #1: 1. In lines 55 and 56 you said "Numerous studies have demonstrated the efficacy of face coverings in reducing transmission, but these have generally been in a healthcare setting", while in reference 6 efficacy of face coverings and other restrictions are compared to healthcare and non-healthcare setting. Please correct this.

Done

2. From lines 106 to 129 you have provided a description of Table 1 that has no specific reference. Please provide a reference for this content. If the information was collected from a specific center (such as your center) you should mention this.

The process of identifying these municipalities was described in the methods. Table 1 has been moved from the introduction section into the results section to clarify the listing was created as part of the research process, and no reference for outside sources can be provided.

3. Please tell us, have you done an analysis to check for the presence of outliers in the statistical analysis section? In Figure 1, there appears to be an outlier value that affects your analysis output.

Yes, impact of outliers was assessed. It was found that none exerted disproportionate influence, and therefore all data points were retained in final results. Statements were added in the methods and results section to reflect this.

4. In lines 208, 238, 249, and 308, you said that seasons can be a risk factor for increasing the number of cases of Covid-19. It is not true. Please check "Pan, J., et al., Warmer weather unlikely to reduce the COVID-19 transmission: An ecological study in 202 locations in 8 countries. Sci Total Environ, 2021. 753: p. 142272."

Line 87 states “these waves have not necessarily followed seasonal [patterns]…,” reflecting uncertainty of the role of season. Line 213 (formerly line 208) does not assert season as a risk factor; it merely notes that cases increased in late fall and early winter. Figure 3 has been added to show the timeframe of case dynamics within the state.Similarly, line 242 (formerly 238) states “coincided roughly,” conveying both the factual occurrence but conceding uncertainty of true seasonality. Line 253 (formerly 249) was modified to reflect seasonality (or general changes over time) as a potential confounder. Line 320 (formerly 308) was expanded to acknowledge the conflicting information related to seasonality (the provided citation was included, as well as two additional references).

5. Please indicate whether during your study, vaccination was in progress or not. If yes, you should not ignore the effect of vaccination.

A paragraph was added in the discussion (line 291) addressing vaccine availability.

Reviewer #2: The authors seek to contribute to the fight against the COVID-19 pandemic. Use of face masks or coverings is one of the major recommendations for reducing transmission, hence the relevance of this manuscript. Comments to be addressed to improve on the quality of the manuscript are as follows:

1. The methodology is not clear. A flow chart on the sample analyzed will bring some clarity. Also case counts for the period of analysis is needed.

An attempt at a flow chart of the methodology has been made and provided as figure 1. Case rate differences for each day count (which is what the actual analysis was based upon) will be provided in the supplemental materials. Raw case counts for each municipality cannot be provided due to data reporting restrictions.

2. Authors should provide literature/references on the enforcement of the mask mandate in Oklahoma. This will strengthen argument on the impact of the mandate.

Enforcement of mandates varied immensely geographically and over time. An explanation to this effect has been added (lines 271-275).

3. Authors should provide the number of municipalities in Oklahoma and the exact number that implemented the mask mandate (39 or 40: refer to lines 110 and 160)

The correct number is 39. Line 160 has been corrected.

4. Why the selection of d-45 to d90 and not d-45 to d45 (providing equal intervals pre-and post-)?

Any interval was going to be artificial. We felt that -45 to 90 days gave a fairly comprehensive assessment for several reasons.

a. 45 days would seem adequate to create a trend for pre-mandate implementation, and it would be unlikely that longer periods would bring more clarity. When policy makers were considering mandates, it is unlikely they would examine data from 60 days prior.

b. 45 days also minimizes the loss of data. Stretching farther back than 45 days would necessitate excluding municipalities that implemented mandates early in the pandemic.

c. 90 days would seem adequate to establish any trend post-implementation. If a mandate works, the benefit should be amplifying (i.e., the resulting reduced transmission rate at day 21 should further feed a reduction in transmission on days 28, 35, etc.). Because of this, we felt the post-mandate interval should be longer than the pre-mandate.

d. That said, extending beyond 90 days would have again resulted in a loss of data, as many municipalities that implemented later mandates did not have data for even 90 days post-mandate.

5. Authors should provide the actual dates for period of analysis. Which date was d0?

The specific date for D0 varied by municipality, depending upon when the mandate was implemented for that location. See the flow chart for further clarification.

6. Authors should check table 1:

- There are two columns for "Mandate start".

- Should also check dates for Midwest city.

Thank you! Those errors were addressed

- Implementation of state-wide restrictions (March-June 2020) is likely to dilute the impact of mask mandates. Authors should consider excluding the first five municipalities.

It is possible that state-wide restrictions would alter case rates. However, any effect should be largely non-differential between communities with mandates and those without. Because the study analysis only examines differences in case rates, we believe that the effect of restrictions on our study should be minimal. Our approach chose to remove from the calculations any locations from all calculations after a mandate expired. Two of the early adopters expired rapidly, meaning they contributed only to the shortest of observed intervals and were not included in considerations beyond that point. The other three had much longer lived mandates, and they therefore contributed to the full length of examination. Removal of them from all analysis would reduce our sample size. Adding them in as D0 at a time after state-wide restrictions were lifted would misrepresent the effect of mandate over time. While your comment is appreciated, after consideration, we’ve decided it is preferable to not remove those early municipalities.

7. Authors should provide reference for the CDC classifications (lines 134-139)

Added.

8. There is no data or/and references in the manuscript to support lines 200-210)

Figure 2 has been added to show the dynamics of case rates throughout the state.

9. Difficult to appreciate higher rate of increase in municipalities to adopt the mask mandate. What was peculiar about these municipalities? (lines 215 - 217). Authors need to describe the study sites in the methodology

Our best efforts at explanation are offered in lines 233 through 240- the fact that more urban areas were more likely to adopt mask mandates. Given that there are 39 municipalities included in the study, and they are notably diverse in geography, demographics, and other features, it would be impossible to describe the sites in any satisfactory detail.

10. One would have expected cases to increase faster in municipalities not implementing the mask mandate. This needs explanation (lines 218 -2020)

Our best explanation for this is again offered in lines 233-240. The municipalities that implemented mask mandates were more urban, and thus subject to generally higher rates of transmission. Mask mandates reduced much of the higher risk, but cannot necessarily be expected to completely reverse those dynamics.

The study design was essentially a difference in differences method, whereby a difference was calculated (rate in mandated municipality minus rate in non-mandated population) and then changes in that difference was assessed over time (and specifically, before vs. after implementation). Thus, the comparison does not necessitate a faster increase in rates in non-mandate municipalities in order to detect an effect. Indeed, increases aren’t even necessary (as cases dropped at different times throughout the study). The real assessment is the difference between rate dynamics in mandate vs. non-mandate communities.

11. Authors should provide references for categorical statements (lines 280 - 282)

Done

Overall

This article makes a valid contribution to the scientific record by evaluating the efficacy of local mask mandates in reducing COVID-19 case rates. The present work highlights an important limitation of previous mask mandate efficacy research and employs an analytic strategy that minimizes this limitation. Previous research has focused on comparing individual municipalities with themselves at two or more different points in time (i.e., prior to and following mask mandate implementation). These strategies are limited by their inability to largely account for temporal changes in other factors besides mask mandates that may influence COVID-19 transmission. To address this longitudinal limitation, the present manuscript instead compares mandated and non-mandated municipalities in Oklahoma at the same points in time. Specifically, a modified difference-in-difference analytic approach was used to compare the difference in case rates between 40 Oklahoma municipalities that eventually implemented a mask mandate and the rest of the non-mandated municipalities at the same points in time. The finding that locally-mandated mask mandates can mitigate differences between these two groups even when COVID-19 case rates are initially higher in the “eventual mandate” population than the “never mandate” population prior to implementation is very encouraging.

While this work makes a substantive contribution to knowledge, some minor revisions are first required to improve readability, resolve conflicting or confusing statements, add a little bit of information to the Methods and Materials, fix in-text and Table errors, and explain/elaborate on certain statements.

Data Availability

I know the authors can’t share raw data owing to the protection of confidential medical information. However, the authors could provide the case rate for each municipality and non-mandate population at each d(-45) to d(90) time point.

The case rate differences of rates in mask-mandate municipalities minus rates in non-mandated population of the state is provided in supplementary materials

Abstract

Lines 28-31:

In the abstract, the slope of difference estimates given in brackets may be difficult for the reader to understand since they have not yet read the paper. Word count permitting, I’d like to suggest the authors explain the first slope of difference in words. For example:

“Prior to adopting mask mandates, those municipalities that eventually adopted mandates had higher transmission rates than the rest of the state, with the mean case rate difference per 100,000 people increasing by 0.32 per unit increase in time (slope of difference= 0.32; 95% CI 0.13 to 29 0.51).”

If the authors cannot do this in the abstract, I’d recommend doing something similar in the results the first time an estimate for the slope of mean case rate difference is presented.

Done

Introduction

The authors provide an excellent summary of the COVID-19 timeline and mitigation efforts, a thorough background of existing research, a clear identification of research gaps, and a sound justification for the present work. However, the authors should consider reorganizing the summary of existing research to enhance reader clarity.

Attempts were made to improve the clarity and transitions between topics. If reorganizing is still justified, we would welcome specific suggestions (we may be too close to the subject to appreciate the issue!).

Line 48-49:

Could the authors briefly list some of the other “number of factors” besides “limited adoption” in this sentence (e.g., politicization of masks, economic burden, etc.)?

Done

Lines 51-104 (Paragraph 2):

This is a very long paragraph-- 2.5 pages! To help guide the reader, this text should be divided into multiple paragraphs that follow a clear, logical framework.

Done

Suggested breaks:

- End of line 71 seems like a good cut-off for a new paragraph. After this point, you transition into specific US studies and then focus on the “Joo, et al.” and “Lyu and Wehby” publications.

- Line 92 after “short periods of time” seems like a natural break because you stop talking about temporal factors and switch to other factors.

Line 73:

In this sentence it is unclear why “longitudinally” is not ideal. Perhaps the authors could consider adding a few words to help out the reader. For example:

“…the assessment was don’t longitudinally, which cannot account for xyz, and included…”

Done

Lines 74-76:

You reference two papers for the first time and then refer to them by “Joo, et al.” and “Lyu and Wehby” only later in the paragraph. It would be clearer if you explicitly wrote “Joo, et al.” and “Lyu and Wehby” in this first sentence that introduces their work.

Done

Lines 84-85 and 89-90:

The sentence “And while Joo…” is confusing with where it is located. The authors just finished discussing how including US states as a variable in a regression model requires the state effect to remain constant over time, and then the present sentence implies that having a seven-week study observation is a bad thing. This is confusing because the reader may presume (like I did) that having a shorter observation period = less opportunity for the effect to vary over time. Consider appending this sentence to the end of the “Regardless of causes(s)…”. For example:

“…in any seven-week period, such as the seven-week period observed by Lyu and Wehby.”

For further resolution, the authors could consider explicitly contrasting the trade-offs of using longer observational periods in order to include the complexity of periodicity versus using a shorter observational period where the effect is more likely to be constant but periodicity is not captured.

Lines 101-104:

These sentences seem out of place because they go back to discussing variability over time. Consider moving to the end of line 92 after “…short periods of time”.

Extensive changes were made to these sections to improve clarity and flow of content.

Lines 121-123:

The authors saying that they used a modified difference-in-difference approach because of mandates going into effect at different times will likely not be self-explanatory to non-epidemiologists/statisticians. Perhaps a quick explanation as to why difference-in-differences is beneficial could be added at the end of the sentence to help out the reader. For example, a layman’s explanation of how the difference-in-differences approach does not assume exchangeability between the mandate and non-mandate groups. This could also be used to briefly justify not adjusting for the other factors discussed on the previous page (population density, socioeconomic status, political factors, etc.).

Materials and Methods

Lines 132-134:

Since the information captured by state health departments likely varies by state, it may be beneficial to specify which events qualify for capture/what data sources the OK registry uses (i.e., specify the source population). Does the registry count COVID-19 cases in pharmacies, universities, assisted-care living facilities, etc., or just hospitals?

A couple sentences were added to clarify this. The authors are uncertain whether this information is more appropriately placed in the materials and methods, or in the discussion. We welcome comment on this matter.

Lines 150-153:

I think “…were calculated for each municipality to establish a mandate” is a typo. It reads as if the presence of a mandate was determined by municipality case density rates, which doesn’t make sense. I’m not sure what this means.

We apologize for the confusion. The sentence wasn’t intended to imply “… for each municipality IN ORDER TO establish a mandate…” Rather, it is meant as “…rates were calculated for municipalities that adopted mandates.” The sentence was modified to attempt to clarify.

Additionally:

Can the authors comment on the number/size of the aggregated non-mandate population? Just so we know there is an adequate non-mandate population size for comparisons with the mandated municipalities (which already have the sizes provided).

This was added on line 220

Results:

Line 186:

To support this sentence (i.e., the parallel assumption), perhaps the authors could supplement with a graph.

This has been added as a supplemental figure.

Line 189:

I was confused by this sentence because the preceding paragraph in the Methods and Materials said that there was significant autocorrelation. Specifying that this is for the one-lag moving average term model would improve clarity.

The p-value was moved to follow the statement that there was no remaining autocorrelation after addition of the one-lag moving average term, and the sentence was stricken from the results.

Lines 190-192:

- Can the authors comment on the parallel assumption not being violated despite the case rate difference rapidly increasing in the pre-mandate period (i.e., the case rate lines for mandated and non-mandated populations not being parallel since the difference between the two lines increases by 0.32 per unit time)? I had a hard time resolving these two pieces of information, which seem to contradict each other. Is it because the parallel assumption was done prior to date alignment, but the increasing case rate difference in on the aligned time scale?

- Can the authors explain the slope of 0.32 in words since this is the first time they present the slope of mean case rate difference (see comments for Lines 28-31)?

Additional explanation was added at the end of the results section.

Discussion:

Lines 234-236:

- This sentence is out of place. This concluding sentence (“The fact that…”) is sandwiched by two other sentences (rural versus urban areas and additional possible explanations) that aren’t directly related to it or support it. Instead, this sentence seems to interrupt the discussion about possible reasons for the difference. The authors should consider relocating it.

The sentence was moved to earlier in the same paragraph, to speak more generally about rural vs. urban, and to not interrupt the specific comments about the state of Oklahoma.

Lines 236-243:

- These sentences are confusing. I think “… subsequent increase in case rate difference between communities with mask mandates and those without” indicates there is an increase in the case rate difference after the mandate. This isn’t communicated in the results (which say there is a decrease), so at first I thought I was misinterpreting this sentence. However, the following sentences hypothesize why the mask mandates don’t work as well over time following the mandate (temperature, gradual non-compliance), which makes me think that the “… subsequent increase” sentence is indeed about post-mandate periods of time. Then, the last sentence talks about a spillover effect with mandates benefitting nearby non-mandate regions by reducing transmission to them. I don’t understand how this could explain an increase in the case rate difference—wouldn’t this transmission reduction decrease the case rate difference?

- I’m not sure if something is missing from the results or this part of the discussion needs work or shuffling to improve clarity. Presently, it is difficult read and understand.

The authors agree that the statement was poorly constructed and difficult to understand. Lines 279-282 have been rewritten in attempt to more effectively communicate the findings. Critically, the phrase “subsequent increase in” was replaced by “persistently higher.” The statement “increase in” was factually incorrect. However, it is important to note that, despite the fact that the slope reversed course, the mean difference remained above zero, even 90 days after mandate implementation. This means that case rates in mandated communities still exceeded those in non-mandated communities. However, the difference was shrinking, and may have eventually reached equality.

Line 246:

Change “compare” to “compared”.

Done

Line 247:

Is aligning time to d0 solely responsible for avoiding confounding? Wouldn’t calculating and comparing non-mandate rates with mandate rates on the same calendar date be extremely important for avoiding confounding by temporal factors? I think it is important to mention this, otherwise it reads as if d0 is solely responsible for preventing confounding. In fact, I’m not entirely clear on how d0 avoids temporal confounding when the rate differences for a particular city and the non-mandated population are always calculated on the same day—would it not be this technique of using the same day for the mandated city and non-mandated population comparison that removes confounding and not aligning to d0?

The sentence has been changed to state that it is both the use of differences in rates, and the use of D0, that avoids the risk of confounding. Utilizing the difference of rates largely eliminates risks of confounding by season, etc., because this essentially creates a comparison by date. However, given that implementation dates varied immensely, setting to D0 permits consolidation of municipalities that adopted mandates at radically different times into a single assessment. Because it is reasonable to think that effects of a mandate compound, there should be a linear relationship of declining rates with longer period of mandate implementation. This could not be assessed without setting to d0.

Lines 270-277

It is very important for us to remember that observed COVID rates rely on the availability of widespread testing. The authors explain this issue very well and then succinctly describe how/why this bias is limited in their study.

Thank you!

Lines 280-283:

It is not explicitly clear why high positivity rate implicate a large number of infections not being detected. Could the authors briefly elaborate?

Additional discussion was added

Lines 287-300:

This is an excellently written and technically sound justification for why adjustment for other rate-influencing covariates was not performed.

Thank you!

Line 313:

Change “by able” to “be able”.

Done

Tables and Figures:

Table 1:

I think there is a typo in the rightmost column of Table 1. I think this is supposed to be “Mandate Expiration” and not “Mandate Start”.

Fixed

Addition of Figure:

To improve the reader’s understanding of the results, the authors could consider supplementing an additional figure with:

- Y-axis: mean case rate per 100,000

- X-axis: alignment (d(-45) to d(90))

- One line for case rates in the non-mandated population

- A second line for case rates in the mandated population

- Perhaps mean case rate data points for each population (mandated and non-mandated) at each time point

This will allow the reader to observe the case rates over time for the mandated vs non-mandated groups. Changes in case rates in two populations over time are less abstract and more intuitively understood than changes in case rate differences over time. The possible inclusion of the suggested case rate graph would help non-statistician/epidemiologist readers understand the results.

From a statistical perspective, we are reluctant to provide such a graphic because the population for the non-mandate denominator would be undefined (i.e., each person is not represented once and only once for each calculation). We are also hesitant to provide it from a non-statistical perspective because it may create confusion as to the actual method of analysis. We would consider the possibility further if the reviewer considers it an imperative point.

Figure 1:

- Both the regression lines and the data points in Figure 1 are plotted with circles of the same shape. When I first reviewed the figure, I thought the regression lines were also data points. Since the graph is in gray-scale, different shapes/symbols should be used for the regression lines and data points.

- It is not immediately clear what the data points (not lines) are in Figure 1. I think, but am not sure, that each point is the average of the 40 separate differences in case rates between mandated and non-mandated municipalities for each time point. This should be explicitly specified in the Figure title. Better yet, add a legend explaining what the individual data points and regression lines are.

- The methods describe calculating the difference in rates from d(-45) to d(90), but Figure 1 only shows up until d(75) instead of d(90). Is there an error in the axis? If not, perhaps the authors could briefly mention why only up until d(75) is shown.

All of the described issues have been addressed.

Submitted filename: Response to reviewers.docx

Click here for additional data file.

18 Apr 2022

22 Apr 2022

Response to editor:

Comment was made in the review regarding citation of retracted articles. I am unclear if this is a standard request or a specific citation raised concern. All citations were verified as having no retractions. One article had an erratum issued in a subsequent publication. Information on this has been added to the references section. This addition is the only change made to the manuscript itself. The only other changes requested were in relation to figures 1 and 3.

Response to reviewers:

The authors are again appreciative of the time and expertise of the reviewers. No comments were provided from reviewer #1 for the resubmission, and thus no actions were taken. Reviewer #2 provided suggestions for figure #1, many of which were adopted. It is the authors’ opinion that other recommendations (specifically, changing the reference of dates to days) would misrepresent the data analysis process. For Step 2, the reviewer asked: “Is this number of cases for days or dates? If days, then consider revising to ‘…..each day (d-45 to d90)…..” The number at that step is for specific dates. The conversion to days does not occur until after the difference has been calculated for each municipality (i.e., rate for municipality - rate in remainder of the state), which is shown as step 8. The data analysis had to be done in this order was because there is no single “d0” rate for the non-mandated portion of the state. Conceptually, the figure could be altered to state that the differences were calculated for each day count without necessarily changing the interpretation. This would indeed simplify the figure. However, since this is not consistent with the actual method of data analysis, the authors are reluctant to adopt such changes.

Figure 3 was updated with truncation of data to be consistent with intervals of the study period.

Submitted filename: response to reviewers Apr.docx

Click here for additional data file.

19 May 2022

Impact of local mask mandates upon COVID-19 case rates in Oklahoma

PONE-D-21-39472R2

Dear Dr. Taylor,

We’re pleased to inform you that your manuscript has been judged scientifically suitable for publication and will be formally accepted for publication once it meets all outstanding technical requirements.

Within one week, you’ll receive an e-mail detailing the required amendments. When these have been addressed, you’ll receive a formal acceptance letter and your manuscript will be scheduled for publication.

An invoice for payment will follow shortly after the formal acceptance. To ensure an efficient process, please log into Editorial Manager at http://www.editorialmanager.com/pone/, click the 'Update My Information' link at the top of the page, and double check that your user information is up-to-date. If you have any billing related questions, please contact our Author Billing department directly at authorbilling@plos.org.

If your institution or institutions have a press office, please notify them about your upcoming paper to help maximize its impact. If they’ll be preparing press materials, please inform our press team as soon as possible -- no later than 48 hours after receiving the formal acceptance. Your manuscript will remain under strict press embargo until 2 pm Eastern Time on the date of publication. For more information, please contact onepress@plos.org.

Kind regards,

Sinan Kardeş, M.D.

Academic Editor

PLOS ONE

Additional Editor Comments (optional):

Reviewers' comments:

Reviewer's Responses to Questions

Comments to the Author

1. If the authors have adequately addressed your comments raised in a previous round of review and you feel that this manuscript is now acceptable for publication, you may indicate that here to bypass the “Comments to the Author” section, enter your conflict of interest statement in the “Confidential to Editor” section, and submit your "Accept" recommendation.

Reviewer #1: All comments have been addressed

Reviewer #2: All comments have been addressed

**********

2. Is the manuscript technically sound, and do the data support the conclusions?

The manuscript must describe a technically sound piece of scientific research with data that supports the conclusions. Experiments must have been conducted rigorously, with appropriate controls, replication, and sample sizes. The conclusions must be drawn appropriately based on the data presented.

Reviewer #1: Yes

Reviewer #2: Yes

**********

3. Has the statistical analysis been performed appropriately and rigorously?

Reviewer #1: Yes

Reviewer #2: Yes

**********

4. Have the authors made all data underlying the findings in their manuscript fully available?

The PLOS Data policy requires authors to make all data underlying the findings described in their manuscript fully available without restriction, with rare exception (please refer to the Data Availability Statement in the manuscript PDF file). The data should be provided as part of the manuscript or its supporting information, or deposited to a public repository. For example, in addition to summary statistics, the data points behind means, medians and variance measures should be available. If there are restrictions on publicly sharing data—e.g. participant privacy or use of data from a third party—those must be specified.

Reviewer #1: Yes

Reviewer #2: Yes

**********

5. Is the manuscript presented in an intelligible fashion and written in standard English?

PLOS ONE does not copyedit accepted manuscripts, so the language in submitted articles must be clear, correct, and unambiguous. Any typographical or grammatical errors should be corrected at revision, so please note any specific errors here.

Reviewer #1: Yes

Reviewer #2: Yes

**********

6. Review Comments to the Author

Please use the space provided to explain your answers to the questions above. You may also include additional comments for the author, including concerns about dual publication, research ethics, or publication ethics. (Please upload your review as an attachment if it exceeds 20,000 characters)

Reviewer #1: (No Response)

Reviewer #2: (No Response)

**********

7. PLOS authors have the option to publish the peer review history of their article (what does this mean?). If published, this will include your full peer review and any attached files.

If you choose “no”, your identity will remain anonymous but your review may still be made public.

Do you want your identity to be public for this peer review? For information about this choice, including consent withdrawal, please see our Privacy Policy.

Reviewer #1: No

Reviewer #2: No

8 Jun 2022

PONE-D-21-39472R2

Impact of local mask mandates upon COVID-19 case rates in Oklahoma

Dear Dr. Taylor:

I'm pleased to inform you that your manuscript has been deemed suitable for publication in PLOS ONE. Congratulations! Your manuscript is now with our production department.

If your institution or institutions have a press office, please let them know about your upcoming paper now to help maximize its impact. If they'll be preparing press materials, please inform our press team within the next 48 hours. Your manuscript will remain under strict press embargo until 2 pm Eastern Time on the date of publication. For more information please contact onepress@plos.org.

If we can help with anything else, please email us at plosone@plos.org.

Thank you for submitting your work to PLOS ONE and supporting open access.

Kind regards,

PLOS ONE Editorial Office Staff

on behalf of

Dr. Sinan Kardeş

Academic Editor

PLOS ONE