“The Foegen Effect: A Mechanism by Which Facemasks Contribute to the COVID-19 Case Fatality Rate”
- Thread starter WVU82
- Start date
Abstract
Extensive evidence in the literature supports the mandatory use of facemasks to reduce the infection rate of severe acute respiratory syndrome coronavirus 2, which causes the coronavirus disease (COVID-19). However, the effect of mask use on the disease course remains controversial. This study aimed to determine whether mandatory mask use influenced the case fatality rate in Kansas, USA between August 1st and October 15th 2020.This study applied secondary data on case updates, mask mandates, and demographic status related to Kansas State, USA. A parallelization analysis based on county-level data was conducted on these data. Results were controlled by performing multiple sensitivity analyses and a negative control.
A parallelization analysis based on county-level data showed that in Kansas, counties with mask mandate had significantly higher case fatality rates than counties without mask mandate, with a risk ratio of 1.85 (95% confidence interval [95% CI]: 1.51–2.10) for COVID-19-related deaths. Even after adjusting for the number of “protected persons,” that is, the number of persons who were not infected in the mask-mandated group compared to the no-mask group, the risk ratio remained significantly high at 1.52 (95% CI: 1.24–1.72). By analyzing the excess mortality in Kansas, this study determines that over 95% of this effect can solely be attributed to COVID-19.
These findings suggest that mask use might pose a yet unknown threat to the user instead of protecting them, making mask mandates a debatable epidemiologic intervention.
The cause of this trend is explained herein using the “Foegen effect” theory; that is, deep re-inhalation of hypercondensed droplets or pure virions caught in facemasks as droplets can worsen prognosis and might be linked to long-term effects of COVID-19 infection. While the “Foegen effect” is proven in vivo in an animal model, further research is needed to fully understand it.
1 Introduction
The coronavirus disease 2019 (COVID-19) pandemic struck the world with over 228 million confirmed cases and over 4.69 million confirmed deaths worldwide by September 18th, 2021,[1] resulting in a case fatality rate (CFR) of about 2.06%. The mortality rate of COVID-19 has been shown to increase with the overall mortality rate of the population.[2] Mortality rate is the most commonly expressed measure of the frequency of occurrence of deaths in a defined population during a specified interval. However, the crude death rate calculates the number of deaths in a geographical area during a given year, per 100,000 mid-year total population of the given geographical area during the same year. Therefore, it is a better parameter to assess death rates among different populations.Mandatory wearing of masks to cover the nose and mouth is a widely applied strategy in the management of the COVID-19 pandemic across many countries in the world. A lot of focus has been centered on the question whether mask mandates reduce infection rates. A study conducted in the Kansas state of USA showed a reduction in infection rates,[3] while a Danish study did not find any protective effect of wearing masks.[4]
However, a lot less focus has been centered on the course of the disease while using masks. This is a questionable approach, as the question “how many lives can be saved?” is more important than the question “how many infections can be prevented?”.
Therefore, the aim of this study was to assess the influence of mask mandates on CFR by comparing the CFR between 2 groups, 1 with and the other without mask mandates. The corresponding two-sided hypothesis is that mask mandates change the CFR. While an increase in CFR may look unintuitive at first glance, more intuitively, one would not exchance his facemask with another person out of fear to breath in the virus that is caught in the facemask and get infected. Thus, breathing in one's own virus might increase the CFR.
The state of Kansas, USA has over 2.8 million residents. During the summer of 2020, Kansas State issued a mask mandate, but it allowed its 105 counties to either opt out or issue their own mask mandate – which was a rarity in the USA and 1 reason for the choice of this state, the other being that the comparison of infection rates among these counties has already been done by Van Dyke et al,[3] showing a benefit of mask mandates.
Out of the 81 counties that had opted out and did not issue their own mask mandate, 8 large cities from 7 counties, had issued a mask mandate. This current study focused on the CFR, and whether mask mandates actually had an effect on the number of lives lost during the COVID-19 pandemic.
2 Method
This study applied secondary data on case updates, mask mandates, and demographic status related to the Kansas state, USA. As this is a secondary data analysis, ethical approval was not necessary.A 3 + 3 step model was applied for the analysis of these data.
2.1 Step 1: Categorizing the counties into two groups
Using the information on counties with facemask-related regulations from the study by Van Dyke et al,[3] which used data from the Kansas Health Institute and CDC, 105 counties were categorized into counties with mask mandate (MMC) and counties without mask mandate (noMMC). Further, the counties without mask mandate were evaluated to identify cities with mask mandates[5] in them. Then the percentage of the county population[6] that was represented by these cities[7] was assessed in order to eliminate counties in which about half of the population was under a mask mandate, as they would dilute the results.Thus, in order to guarantee that the cities with mask mandates constituted either more than twice of or more than half of the county's population not under a mask mandate, if more than 2/3 of these counties’ population was either under mask mandate or not, the county was included in the analysis and moved to the corresponding group. Correspondingly, if the city's population was within +/-17% of half of the county's population (that is, between 33% and 67%), the county was excluded.
2.2 Step 2: Parallelizing the groups
Since the assumption was close that counties with a more vulnerable population had issued a mask mandate (bias by selection), the specific COVID-19 risk of each group's population was assessed. The study by Vasishtha et. al[8] demonstrates that COVID-19 mortality is closely matched with overall mortality, which is represented by the crude death rate (CDR) of any given population. The CDR represents age, pre-existing illness and all other mortality-bound cofactors in the underlying population.Further, the CDR of each county for 2019[9] was modified by subtracting deaths from causes that are clearly not a risk factor for COVID-19 to prevent statistical anomalies when comparing CDR, like an unusual spike in deaths from external causes or perinatal mortality in single counties. The following categories of the Kansas Health Institute death data were thus excluded to calculate a covid-related death rate (crDR): “pregnancy complications,” “birth defects,” “conditions of the perinatal period (early infancy),” “sudden infant death syndrome,” “motor vehicle accidents,” “all other accidents and adverse effects,” “suicide,” “homicide,” and “other external causes”.[9]
This crDR of the counties was then population-weighted (multiplied with population of county divided by population of group) and added up to calculate the crDR (total number of expected deaths per 100,000 people per year) of both the MMC and noMMC groups.
The assessement showed that, after step 1, the crDR of the noMMC group was 1012.6 deaths per 100,000, while the MMC group had an crDR of 782.5 deaths per 100,000, clearly indicating a bias of noMMC group being a more vulnerable population, counterintuitively.
Due to the lack of normality and homoscedasticity (as demonstrated in the scatterplot, Fig. 1), a regression was not possible, thus, the counties were parallelized for comparison based on crDR.
Figure 1:
Scatterplot of COVID-19-related death rate (crDR) vs. case fatality rate (CFR). Orange triangles pointing upwards represent mask-mandated counties (MMC), blue triangles pointing downwards represent counties without mask mandate (noMMC).
In this process counties were excluded until both groups had a matching crDR, meaning both populations are equally vulnerably to COVID-19.
This process of parallelization is a customized modification of the usual process used in parallel studies. It is based on larger groups (county populations) instead of individuals while likewise aiming to eliminate the aforementioned confounder.
There were 2 ways in order to get almost the same crDR in both groups:
- A) Removing primarily counties with the highest crDR in the group with a higher crDR until both groups had the same crDR: Configuration A.
- B) Removing primarily counties with the lowest crDR in the group with a lower crDR until both groups had the same crDR: Configuration B.
2.3 Step 3: Analyzing the data
As the mask mandate was issued on July 3rd, August 1st was considered as the start date to allow for necessary adjustments to the mask mandate and prevent overlap with time before the mask mandate as the effect of mask mandates may not be visible immediately.Moreover, October 15th was fixed as the end date as proof of mask mandates was available up to that point, and the existent mask mandates were revised after that date. The number of infected cases[10] was calculated for this period.
The COVID-19 death count in Kansas[11] is not personalized, meaning for each death counted there is no information on the person's infection date. After referring to the study by Khalili et al,[12] the calculation of deaths was delayed to 14 days after the COVID-19 infection time period. In order to mitigate the influence of the start and end of the time interval, the number of deaths as the average of death differences between August 7th and October 22nd, August 14th and October 29th, as well as August 21st and November 5th was calculated. This way, both infection and death data were obtained for a span of 76 days. Based on these numbers, infection rates and CFR were calculated for both groups in both configurations.
A fourfold table was applied for the Chi-Squared test (α = 0.05) and risk ratio (RR; MMC to noMMC), and 95%CIs were calculated to determine whether the mask mandates significantly increased or decreased the CFR by COVID-19.
All statistical calculations were done using LibreOffice 7.1. (The Document Foundation, Berlin, Germany).
2.4 Step 4a: Infection rate correlated bias check (when applicable)
If the RR was significant, a sensitivity analysis is used to verify whether a difference in infection rate explains the difference in the CFR. For this, λlow-CFR was considered the infection rate of the group with a lower CFR, and λhigh-CFR was considered the infection rate of group with a higher CFR.The 2 possibilities were:
- 1. The group with low CFR also has a lower infection rate.
- If λlow-CFR < λhigh-CFR, there might be a testing bias.
- The hypothesis to this would be that if both groups had been tested equally and both had equal infection rates, the CFR would not be significant. In order to prove this hypothesis, the number of deaths in the group with a lower CFR was reduced by multiplying it with the factor (λlow / λhigh), the fourfold table from step 3 was revised, and a repeat calculation of the Chi-Squared, RR, and 95%CI was done.
- 2. The group with lower CFR has a higher infection rate.
- If λlow-CFR > λhigh-CFR, there might be a bias by protection.
- The hypothesis would be that if those protected by a reduced infection rate were counted as survivors (although they could still be infected later), the CFR would not be significant.
- In order to prove this hypothesis, the number of infected people in the group with a higher CFR was increased by multiplying it with the factor (λlow / λhigh), the fourfold table from step 3 was corrected, and calculation of Chi-Squared, RR, and 95%CI was revised.
2.5 Step 4b: Confounder check (when applicable)
If the RR was significant, further analysis was performed to find whether a confounder caused the RR (for MMC) to increase or decrease independently of severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) infection. This could be, for instance, the accumulation of fungal spores or bacteria in the mask or mask-induced hypoxia (increasing RR), or the prevention of other possibly lethal viral or bacterial infections (decreasing RR).The hypothesis would be that a confounder in MMC causes increase or decrease in the RR independently from SARS-CoV-2. If this were true, the effect of masks would occur not only in the infected population but also among the not infected population under mask mandate. This can be proven wrong if the potential effect does not align with overall excess mortality in Kansas.
Therefore, it was necessary to calculate the additional deaths by mask mandates or the reduced death by mask mandates (for RR and both ends of its 95%CI as in step 3).
These additional/reduced deaths were calculated as the absolute value of
(1/ϕ – 1) ∗ deathMMC
where ϕ is RR (or the values of both ends of its 95% CI), and deathMMC is the number of deaths in MMC. Further, the expected additional/reduced deaths (in all infected and non-infected) in all MMC counties were calculated by dividing by the number of infected persons in MMC (as obtained in step 3) and multiplying with the total population in all MMC (from step 1).
This result was compared to the (total) Kansas non-COVID-19 excess mortality during the corresponding weeks as already calculated by the CDC.[13] The process involves calculating and adding up the difference between nonCOVID-19 deaths and the average expected number of deaths for each given week. The resultant value indicates the nonCOVID-19 excess deaths.
By dividing this number with the expected additional/reduced deaths in all non-infected in all MMC countries, it is possible to estimate the proportion of the RR increase/decrease calculated in step 3 that is not related to COVID-19 and thus indicating the influence of possible confounders.
2.6 Step 4c: Negative control (when applicable)
In case there is a difference after Step 3, the same group of counties would be analyzed using data from February 1st as starting date and April 15th as the end date for cases. The number of deaths was calculated as the average differences of February 8th to April 22nd, February 15th to April 29th and February 22nd to May 6th. These dates were chosen because shortly after April 15th, Kansas was hit by the 1st wave of the COVID-19 pandemic.This resulted in multiple problems. First, case numbers increased rapidly and resulted in a strong undertesting, resulting in a test positivity rate[14] of 18% on April 21st and 22nd, which then dropped consecutively due to massively expanded testing to 3.7% on June 7th, which is problematic as the positivity rate influences CFR. Furthermore, hospital capacity during the first wave was limited which may have resulted in medical undersupply and increased CFR. As the first wave hit all counties neither simultaneously nor in same intensity, I did exclude this timespan as it would incur massive bias.
As a comparison, during the chosen time span from Step 3, positivity rate was constantly between 6.9% and 9.9%.
Last edited:
