FROM COVID or WITH COVID? Can Excess Deaths Help to Answer the Question?
Abstract
In order to test the efficacy of the mRNA vaccines in a way that doesn't rely on COVID-19 diagnoses, a multiple linear regression is used to compare state-level vaccination rates and 11 additional variables to excess deaths. The result of this regression analysis is a statistically significant correlation between higher vaccination rates and lower excess death rates. The magnitude of this relationship is consistent with the prevailing opinion in the medical community that the vaccines in the U.S. greatly reduced the risk of death. The magnitude also suggests that most reported COVID deaths were FROM the disease, but the results don't rule out the possibility of some WITH COVID deaths. Because this study uses excess death counts instead of reported COVID-19 death counts, hopefully it can help to persuade people, those who doubt the reliability of the COVID-19 death counts and who believe that the vaccines are dangerous, that most likely, the vaccines have so far been very effective at reducing the risk of death in the average American.
Contents:
Disclaimer
I am not a doctor or a statistician and I did this report on my free time. Some of the math may be a little "fuzzy", but I am confident the gist of the report is fairly sound and has merit.
Introduction
The prevailing opinion within the medical community is that being vaccinated dramatically decreased your chance of death. Various sources say that vaccinated people were much less likely to die. Hospital numbers from around the country seemed to bear this out, routinely reporting 10 to 20 times as many unvaccinated people with COVID-19 in their ICUs. But not everyone is convinced that the vaccines helped and some think that the vaccines are doing more harm than good. The common refrain in some circles is "FROM COVID or WITH COVID?" The contention is that although many people with COVID are dying, the cause of death is most often something else, or that the person would have died anyhow. And a second assertion is that the vaccines are actually killing a lot of people. In both cases people don't trust the cause of death recorded on death certificates.
This presents the challenge of determining other ways to assess the effect of the vaccines without using the recorded cause of death. One such approach is to use regression analysis to determine the correlation between different variables, such as vaccination status, and the number of excess deaths. Excess deaths are deaths in excess of the expected number. In the five years prior to the pandemic, the number of deaths each year deviated by less than a couple percent from expected. During the pandemic, excess deaths have been well over 10%. Therefore, it is reasonable to attribute most excess deaths during the past two years to the pandemic, bearing in mind that not all of them were directly caused by COVID. And it is thus reasonable to think of excess deaths during the pandemic as "all-cause pandemic deaths". By using excess deaths, the cause-of-death controversy is no longer a factor, and the assertion that the people would have died anyhow is also handled because excess deaths are people who would not have ordinarily died.
Results
A multiple linear regression is used to assess the impact of 12 variables on excess deaths during 2021 when the vaccines became available. Data aggregated to the state level is
the input. The
Statistics Kingdom: Multiple Linear Regression Calculator is used to perform the analysis. Using a significance level of 0.99999 (i.e. using all variables, even those that are not statistically significant) yields the equation:
Equation 1: Y = 587.82 + 6.03 * X1 + 5.87 * X2 + 174.03 * X3 + 83.93 * X4 + 5.99 * X5 + 21.99 * X6 - 25.57 * X7 + 1.10 * X8 + 48.21 * X9 + 27.60 * X10 - 54.44 * X11 + 9.43 * X12
By contrast, using a significance level of 0.05, only four of the 12 independent variables remain:
Equation 2: Y = 3713.54 + 25.21 * X6 - 34.30 * X7 + 85.49 * X9 - 48.02 * X11
The dependent variable Y is
excess death rate (the number of excess deaths per 1,000,000 people). The 12 independent variables are:
Table 1. Independent Variables
* - Correlation with excess death rate is not statistically significant.
! - The metric of population density used here is the average number of people per mile, assuming all people within a county are evenly distributed.
# - For this metric, black people are considered 0% white and Latinxs and Native Americans are considered 50% white.
Of the 12 independent variables, when considered together, only four (
Equation 2) have a statistically significant correlation with excess death rate. (Statistical significance is achieved when there is a less than 5% chance that the result can be explained by chance.) The four variables are: % Biden Voters (X6), Median Income (X7), % Uninsured (X9), and % Vaccinated (X11).
Discussion
Vaccinated vs. Unvaccinated: Correlation with Excess DeathsSo the correlation between % Vaccination (X11) and excess death rate (Y) is statistically significant. In
Equation 1 the regression coefficient for X11 is -54.44 (95%
CI [-123.6, 14.71]). Therefore, a vaccination rate increase of one percenage point corresponds to an excess death rate decrease of 54.44 people per million. (Even though most of the 12 variables are not statistically significant, I use
Figure 1 because I suspect some of the variables would be significant given finer-resolution data and a better regression model.) This translates to a decrease of 5,444 people per million when vaccination rate increases from 0 to 100%. When applied to the U.S. population of ~330 million, it predicts a savings of ~1.8 million lives in 2021. 100% vaccination was unrealistic, but if we started the year at 0% and reached 90% by mid-year, that's an annual mean of ((0% + 90%) / 2 + 90%) / 2 = 67.5% or a savings of ~1.2 million. In other words, the prediction is that if no one had been vaccinated, there would have been ~1.2 million more deaths than if the maximum number of people had been vaccinated as soon as possible. Still a huge number, but perhaps not far off since the U.S. had ~600 thousand excess deaths and a mean annual vaccination rate of
~37.5%, not much more than half of 67.5%. However, this prediction makes use of extrapolation, i.e. calculations performed on vaccination rate values (0% and 67.5%) that are well outside the range of the input data (from 28.2% in Alabama to 48.5% in Vermont), and thus is not very reliable. Yet, because the correlation is statistically significant and the magnitude of the relationship is large, it strongly suggests that, for the average American, the benefit from being vaccinated far outweighed any risk.
As a hypochondriac, I was somewhat afraid when contemplating getting the mRNA shots. I'd go from a very low daily risk of COVID infection, to 100% chance of getting a shot of a novel vaccine. If I was going to have an adverse effect, it definitely was going to happen. But I felt like getting vaccinated was the responsible thing to do, both for my sake and the sake of the people I lived with. However, if I believed in the description that the mRNA vaccines are dangerous, it would justify a decision to not get the shots. It would justify my fear and justify doing what some people said was irresponsible. I suspect this is part of the reason some people believe in the description that the vaccines are dangerous. So I looked at the numbers and some of the reports and decided that getting vaccinated was in my best interest. A month or so after my second shot, my lower leg swelled up and I walked with a cane for weeks. I am prone to inflammation, but this was abnormal. Was this an adverse effect of the vaccine? I don't know. But now I'm back to being concerned, as I contemplate getting the booster. Analytically, I think the odds favor getting boosted, but in my case, I don't know.
Vaccinated vs. Unvaccinated: Relative Death RateSo it appears that vaccination results in a lot fewer deaths. Various reports quantify this benefit as the relative rate of death between unvaccinated and vaccinated people, for example the report:
How do death rates from COVID-19 differ between people who are vaccinated and those who are not? Graphs in this report, derived from data published by Switzerland and the United States (CDC), show relative COVID-19 death rates of about 7 to 1 for unvaccinated vs. fully vaccinated (no booster) individuals. That's a big difference, but some people question the validity of data provided by organizations such as the CDC, and people also contend that the vaccines are killing people and that this isn't being taken into account. However, relative death rate can be estimated using the regression analysis which does not rely on a COVID-19 diagnosis and does take into account any deaths caused by the vaccines. To model what would have happened if no one had been vaccinated, a value of 0% is used for variable X11 in
Equation 1 for each state. The sum of the projected Y values for all states is 1.27 million, or an increase of 670 thousand excess deaths if no one had been vaccinated. On the other hand, when a value of 100% is used for variable X11, the sum of the projected Y values is -520 thousand, a decrease of 1.12 million excess deaths if everyone had been vaccinated at the beginning of the year. To calculate relative excess death rate between unvaccinated and vaccinated individuals, 1.27 million (excess deaths, 0% vaccinated) is divided by -520 thousand (excess deaths, 100% vaccinated) which equals -2.4. Whoops! Because the model projects negative excess deaths if everyone were vaccinated, the resultant ratio is negative, which seems nonsensical. As the denominator approaches zero, the ratio approaches infinity, and when it is negative, the ratio goes negative, or as Buzz Lightyear would say: To infinity and beyond! In other words, the regression analysis suggests that the vaccines were so effective that the relative all-cause pandemic (a.k.a. excess) death rate would not only have exceeded 7 to 1 for unvaccinated vs. fully vaccinated individuals, but that there would have actually been fewer deaths than in a normal year, if everyone had been vaccinated.
Although it seems possible that there could have been fewer deaths than normal, especially if the projected number of expected deaths had not been updated to take into account the excess deaths in 2020, it seems implausible that it would have reached 520,000 fewer deaths than expected. An exaggerated result is presumably the result of using coarse state-level data and a simplistic linear model. The 95% Confidence Interval for the % Vaccination (X11) regression coefficient stretches from negative 123.6 to positive 14.71, so given this analysis, there's even a slim chance that the relationship is not negative, i.e. that an increase in vaccination rate actually correlates with a higher excess death rate. Moreover, as in the previous section, the projected Y values in this section rely on extrapolation and there is less confidence in them.
FROM COVID or WITH COVID?Because the correlation of lower excess death rates with higher vaccination rates was both statistically significant and of a magnitude that it could explain most or all of the excess deaths, it seems reasonable to believe that most of the reported COVID-19 deaths in 2021 (
~480 thousand) are FROM COVID not WITH COVID, and that the real number of COVID deaths might actually be higher, since there were even more excess deaths (~600 thousand). (One assumption being made here is that vaccination only helps prevent death from COVID and not from other causes of mortality.) However, one of the other statistically significant variables is % Biden Voters (X6) and the correlation is the opposite (i.e. being blue is associated with more excess deaths) and the magnitude is roughly equal to that of % Vaccination (X11). Since vaccination rates were higher in blue states, this suggests that the higher death rate was caused by something else. In this analysis many other variables are included in hopes of identifying any confounding variables. This includes variables related to weather, race, age, obesity, and population density. What isn't included are any variables related to COVID restrictions/mandates. Some people contend that lockdowns caused deaths. Therefore, three additional independent variables are considered: # Days Emergency Order in Effect, % Change in Personal Income, % Unemployment. The latter two are considered because lockdowns presumably slowed the economy. However, the correlations between these three variables and excess death are not statistically significant. So I don't know why bluer states correlate with more excess deaths and although the results suggest there were a lot of deaths from COVID-19, they don't rule out the possibility that a significant number of FROM COVID deaths were actually WITH COVID.
IvermectinAlthough there now are copious amounts of data with which to analyze the efficacy of the vaccines, there is much less data on the effectiveness of alternative medicines, such as ivermectin. For state-level data in the U.S., the best thought I had is to use Google Trends to see where people have been searching for the word "ivermectin", assuming that this would correlate with ivermectin use. The result is not statistically significant, but in the regression analysis, a higher percentage of google searches for "ivermectin" (X12) correlated with a higher excess death rate. And if no one in the U.S. had searched for "ivermectin", plug 0% for each state into X12 in
Equation 1 and the resultant projection is 190,000 fewer excess deaths. So this result does not support the claim that ivermectin is beneficial in the fight against COVID-19 and does suggest that people looking into ivermectin may have suffered worse outcomes. However, as previously stated, the correlation is not statistically significant and the projection uses extrapolation.
Blue States, Poor, and the UninsuredIn addition to % Vaccinated (X11), the three other statistically significant variables are % Biden Voters (X6), Median Income (X7), and % Uninsured (X9). The first variable is discussed above and the latter two are discussed in the previous report:
What Factors Lead to Mortality Rate in the U.S. in 2021 During the Pandemic?.
Conclusion
The multiple linear regression analysis performed in this study finds a statistically significant correlation between higher vaccination rates and lower excess death rates. The magnitude of this relationship is consistent with the prevailing opinion in the medical community that the vaccines in the U.S. (primarily mRNA vaccines) greatly reduced the risk of death. The magnitude also suggests that most reported COVID deaths were FROM the disease, but the results don't rule out the possibility of some WITH COVID deaths. Because this study uses excess death counts instead of reported COVID-19 death counts, hopefully it can help to persuade people, those who doubt the reliability of the COVID-19 death counts and who believe that the vaccines are dangerous, that most likely, the vaccines have so far been very effective at reducing the risk of death in the average American.
Go back to top for the
Disclaimer.