COVID-19, or the Coronavirus Disease 2019, is the disease that has plagued nations worldwide and led to over 1.3 million deaths. As many know, this disease stems from the SARS-CoV-2 virus, which has a mortality rate of approximately 2%, and causes acute respiratory distress syndrome (ARDS) and multiple-organ failure (MOF) to develop in those who become infected. To hinder the spread of this airborne disease, public health officials and medical experts pushed for the usage of masks across the globe, but the effectiveness of mask adherence in preventing the spread of COVID-19 stays debated. CMAP, or the COVID Mask Analyzer Program, is a tool that combines nationwide mask adherence data and simulates what the effects of different levels of mask adherence in counties across the United States would be on COVID-19 cases. Through this tool, researchers have been able to back the claim of mask usage being very effective by providing data that proves the efficacy of masks in the COVID-19 pandemic. In fact, the CMAP tool shows that earlier mask mandates have been the most effective in curbing COVID-19 spread and that the usage of masks, in general, has a positive impact in decreasing the spread of this disease.
The CMAP tool was developed by the University of South Carolina’s Artificial Intelligence Institute, under the guidance of Professor Biplav Srivastava in collaboration with Tantiv4- Sparsh, Chinmayi, Kartikaya and Lokesh. The goal of the tool is to look for any correlation between areas in the US with mask mandates and compare the number of COVID-19 cases in those areas to the number of cases in similar counties without mask mandates. The backbone of this tool is the New York Times survey from the summer of 2020. This survey assigned nearly every county in the United States with a mask adherence score from 0 to 5, with 5 being the most adherent, giving the team a clear idea of the extent to which masks were being worn in different areas across the United States. Combining this data with data from Johns Hopkins for case numbers, population size, etc. throughout counties in the United States, the tool returns to the user how many likely cases and possible deaths would have taken place in a county if the mask adherence had been lowered at a given intervention time. This intervention date is chosen by the user.
Utilizing a mathematical technique known as Robust Synthetic Control (mRSC), the CMAP tool is able to extrapolate results to answer the question of “what if masks had not been worn as much?” This analysis technique is often seen being used in drug research and involves creating a counterfactual from provided data sets in which covariate information is not available. The tool works in the following manner: The user must select a US county, a start date, an end date, and enter an intervention date, which is the date after which the mask adherence level in the selected county “decreases” in comparison to its actual mask adherence rate. The mortality rate is left up to the user to change but is pre-set in the tool to the currently known COVID-19 mortality rate of 2%. With this data, the tool then forms a control group and a treatment group.
A control group is a synthetic group created from a set of other counties with similar mask adherence levels in the state of the selected county. The treatment group returns values showing how the intervention, or decrease in mask adherence, would affect the control group. This counterfactual is calculated in the following manner: the tool calls back data points from times prior to the intervention date, and reads census data and combines it with land information. This data is then logarithmically scaled (except median age), and min-max scaling is applied to all variables as well. Closeness and mask-wearing strength are calculated between the selected county and each other county within the state, with a threshold of 3.0 set to distinguish between high and low adherence. Using the counties as donor pools, the mRSC is performed to automatically calculate the counterfactual. The CMAP tool then graphically displays the actual progression and counterfactual of the spread of COVID-19 to its user, along with printing both a calculated percent increase in cases and the numerical difference in COVID-19 deaths between the counterfactual and control groups.
With the mechanism of the tool established, let’s take a look at the results that this tool has provided researchers with. The output from CMAP trials shows most importantly the harmful effects of having a delayed mask mandate. Using different intervention dates, June 1st, July 1st, and August 1st, set for any specific county, the CMAP tool will show that mask adherence early on leads to lower numbers of cases and deaths than mask adherence later on. In addition, the CMAP tool proves not only the effectiveness of masks but shows that counties with mask mandates or high public mask adherence early on in June fared much better in mitigating COVID-19 cases than counties that put mask mandates in place in July or later. In essence, the tool effectively shows the efficacy of masks in preventing the spread of COVID-19, as well as shows diminishing returns, although still beneficial, in mask efficacy towards mitigating COVID-19 spread, as the date of mask mandate implementation comes at later points in time.
The CMAP tool, just as any other tool, has its limitations in use. As stated before, the core of this program is the New York Times survey from the summer of 2020. This data set is used cannot be altered or changed in the tool, but in reality, the level of mask adherence in each county across the United States is constantly changing, which is not taken into account by the CMAP tool. Additionally, the data being evaluated cannot take into account confounding factors, nor can it take into account the rise and fall of the pandemic’s waves. Since mask mandates are most effective at curbing the spread of COVID-19 or curtailing it, any preventative measures during the peak of cases will not be useful, nor will they be useful after a peak has occurred, as the goal is to prevent the peak from occurring in the first place. These factors cannot be taken into account by the tool, and it is up to the user to make sure intervention dates being used are not during or after a peak in the chosen location. In addition, the tool is limited in its use in that it only accounts for the United States, and other countries have not yet been accounted for. However, despite this limitation, it is important to note that the overall trend in data of mask efficacy in preventing widespread outbreaks of COVID-19 can, and should, be generalized to all populations in order to save lives in other nations fighting this pandemic as well.
Despite its limitations, the CMAP tool very effectively and strongly provides ample data that backs public health officials’ claims about the effectiveness of mask usage during this pandemic. This tool is not only a wonderful example of how the RSC technique for data analysis can be used in the future, but with the results provided from multiple CMAP trials, it becomes evident that mask adherence not only significantly aids in curbing the spread of COVID-19, but also proves that the earlier the mask adherence takes place, the more effective masks have been in preventing infections, and in turn, preventing COVID-19 related deaths. With the results provided from this tool in mind, public health officials and medical experts across the globe can firmly attest to the necessity of mask mandates in preventing the rise of COVID-19 cases, as well as any future airborne illnesses, making this a very useful tool.
Written by: Lokesh Johri