The Real Story behind the 'Bad Batches'
but not an all-clear signal – the consequences are even worse
This is the 3rd and last article of three. Please read the other articles first, as an introduction:
The real story behind the bad batches (this one)
Abstract
It is a well-known fact that 95% of all adverse events were reported for only a 5% fraction of all Covid-19 vaccine batch codes in the VAERS system. While others assume that the number of reported adverse is a valid representation of the number of adverse events that actually occurred and the vaccine batches must be different, I found evidence that these 5% were the only batches that had an adequate level of adverse event reporting and the other batches had almost none at all.
In the following sections, I will search and identify factors that resulted in better or worse reporting rates for different batches and explain the reasons, why a small number of batches has much better reporting quality than the others. I boiled down the factors to three ones that explain the number of adverse event counts of a batch with a fit (r^2) of 70%.
Observations and conclusions are coherent and indicate, that the batches with the largest number of adverse events reported provide us with the most realistic estimate for the number of adverse events that actually occurred for each and every vaccine batch! If you had nightmares about receiving a bad batch, you can stop worrying: Each batch is as good or bad as the reputedly “Bad Batches”.
Could batches have significantly different reporting rates?
After finding inconsistencies in the 'Bad Batches Theory', I wrote an article about it and stated alternative explanations. This work led to a suspicion: What if all batches were created equal and caused an equal number of adverse events? Then, different report counts for different batches could only be explained by different reporting levels for different batches. In the beginning, this seemed implausible to me, but this changed with my findings.
Definitely plausible is the assumption, that any individual adverse event has an individual probability of being reported, depending on the vaccinated person's gender, treating doctor, knowledge, social and professional environment and other factors. I have described and explained these in article 2 of this series: The Rocky Road to an Adverse Event Report.
However, Covid-19 batches of most manufacturers provide for more than 1 million doses which are distributed over dozens of different geographic locations and, presumably, hundreds or thousands of different socio-economic groups. Different reporting rates for individuals are expected to statistically even out over sample sets of such a big size. But VAERS data shows that this was not the case, especially in the first months of Covid-19 vaccine availability.
As I will explain in this article, VAERS data alone reveals several factors which differ strongly between batches and exhibit a significant impact on the report count:
Profession
Vaccination date
Gender
Vaccination Site
In the following sections, I will continue the enumeration of observations that I started in article 1 of this series.
Observation I: Gender Reporting Levels
According to CDC statistics, about 53% of all vaccination doses have been administered to women. Supposed that women do not experience more side effects than men, the percentage of reports related to women should be 53%. But in VAERS, it is 71%. This means that an adverse event of a woman has twice the probability of being reported than an adverse event of a man.
Observation J: Gender Disparity in Batches
Let us count the number of batches with a certain quota of women and create a frequency histogram that displays the number of batches for each quota of women. I smoothed it a bit (3% ranges) to make it more intuitive. It expresses observations like this: „There were 31 batches of which 66% of reported adverse events related to women“. The result should be a symmetric bell-shaped distribution centered at 71% (as according to Observation I – Gender Reporting Levels).
The surprising result is a curve that is neither symmetric nor centered at 71%. There is a big cluster of batches with 60–73% of reports related to females and a smaller one with 74-83%. Obviously, some batches had a group of recipients with a significant majority of women.
The quantitative influence of the quota of women is confirmed when plotting each batch as a point with its report count (bottom to top) over the percentage of women (left to right): The higher the percentage of women vaccinated with a batch, the more reports were filed for it.
Observation K: Pharmacy Underreporting
One of the strangest observations is a strong underreporting of vaccinations that have been administered at pharmacies: According to the CDC, as of March 2022, 233.4 million doses have been administered at retail pharmacies in the US. According to Our World in data, the total number of doses administered in the USA until that date was 555.0 million. This means, 42% of all Covid-19 Vaccine doses in the USA have been administered at pharmacies.
In consequence, 42% of all adverse event reports should have had their dose administered at a pharmacy – in fact, it is only 21%, showing us, that for an adverse event related to vaccination at another site than a pharmacy, the probability of being reported is about three times larger than if the dose would have been administered at a pharmacy.
For confirmation, let us check the effect of vaccination at a pharmacy on the report count of the batch in the same way as in the previous diagram:
This shows that as the percentage of vaccinations at a pharmacy goes to 100%, the number of reports goes to zero. A strange correlation which is not supposed to be causation, but supposedly correlated outcomes of the same reason that caused the decision to get vaccinated at a pharmacy – and which I have not detected yet. But as a German, I have little knowledge about the US health system and consumer behavior whilst an American may have a simple explanation. Any feedback on this is welcome. For example, I would love to know if, maybe, aspiration is the standard vaccination procedure of US pharmacists and a possible cause for less adverse events occurred.
By the way, a plot like the one above would be expected if there were no reports related to pharmacy-administered doses at all – another confirmation that reporting rates for pharmacy-administered doses are much lower than the average.
Multilinear regression tests over VAERS data confirm that this is one of the three most important factors for the reporting quality of a batch.
First Conclusions
Observations J (Gender Disparity in Batches) and K (Pharmacy Underreporting) prove, that the doses of batches were distributed unevenly over groups of society, resulting in very different levels of reporting for different batches. So far, I am motivated to find out more, are you too?
Examining the most reported batches
For a systematic search of factors that influence the probability of an adverse event to be reported, I compared the most reported batches with the complete set of US domestic VAERS data, including only reports
not marked as invalid by VAERS
with a valid batch code
which are a real adverse event or a breakthrough case (thus excluding administration errors)
These sum up to 431,407 reports. I compared the reports for the 10% most reported batches with the averages of overall VAERS data. The most-reported batches were 12: 039K20A, 026L20A, 011J20A, 025L20A, 013L20A, 012L20A, 037K20A, 029L20A, 011L20A of Moderna and EK5730, EK9231, EH9899 of Pfizer/BioNTech.
Observation L: Gender Disparity in most reported Batches
Of the most reported batches, 77.7% of the patients were female, versus an average of 70.5% of all VAERS reports. This confirms the importance of women for reporting, as observed in Observation J – Gender Disparity in Batches.
Observation M: Severity Disparity
The quota of severe adverse events is small, but in direct comparison, the total set of VAERS reports has a significantly higher percentage of severe real adverse events than the most frequently reported batches. It seems that for the average vaccine recipient, as compared to one of the most reported batches, an adverse event needed to be more severe in order to get reported: For life-threatening symptoms, there were 72% more reports (1.6% vs. 0.9%), for permanent disabilities 70% (1.5% vs. 0.9%) and for hospitalization 69% more (6.2% vs 3.7%). Recipients of the most reported batches seem to tend to give feedback to their doctor also for milder symptoms or are more prone to suspect a symptom to be an adverse event of a vaccination.
Observation N: Vaccination Date Distribution
The vaccination dates of the most reported batches are concentrated to a few weeks after Covid-19 vaccination was made publicly available.
Observation O: Vaccination Site Disparity
Of the most reported batches, about 38.4% have been administered at a doctor's office or hospital, only 2.7% in pharmacies. On average, only 24.2% were administered in a hospital or a doctor's office (37% less), but 21% in pharmacies (677% more). One more confirmation of the special influence factor of vaccination at a pharmacy.
Observation P: Profession Disparity
One channel for VAERS report submission delivers reports to the VAERS system marked as 'spontaneous report' and including the reporter's profession. 2.7% of the most reported batches were self-reports of healthcare professionals (physicians, healthcare professionals, nurses and pharmacists), their quota in the whole VAERS data set is just 1.1%. Probably, an above-average quota of healthcare professionals was vaccinated with these batches.
Putting the Pieces together
Ramp-up of vaccine production lines takes time, so, in the beginning, there was a small number of doses available for a big demand. It became necessary to plan distribution in a manner that vaccines were available first, where they were the most valuable. Thus, COVID-19 Vaccine Priority Groups have been defined to be the first to receive vaccination:
Phase 1A: Health Care Workers and Long-Term Care Residents
Phase 1B: Older individuals, people with underlying health conditions, other congregate settings, child care workers, and employees of preschools and kindergarten through 12th grade
Phase 1C: Essential workers and people of any age at increased risk for COVID-19
More than that, vaccine distribution was centralized: In August 2020, The company McKesson was selected to be the sole distributor for all Covid-19 vaccines in the United States. They had to make sure that two doses were available for each person to receive vaccination in each of the early phases.
I assume that this resulted in a large concentration of vaccine doses of the first batches to very similar professional groups at the same time, though being widely distributed geographically.
Now, let us search for clues of relationships between our observations from above and the stated vaccine priority groups:
Observation L: Gender Disparity in Deciles: 71% percent of reports were filed for women: In healthcare and education, the percentage of women is about 75% and among the users of long-term care services more than 70%.
Observation M: Severity Disparity: When a mild adverse event occurs, healthcare professionals are more likely to acknowledge them as such, that it is important to report it and to encounter a doctor in their professional activities and consult him or her. The average American would not necessarily draw this conclusion and go to a doctor.
Observation N: Vaccination Date distribution: Since their administration was concentrated on the first weeks of Covid-19 vaccine availability, the most reported batches are likely to have been priority group shipments.
Observation O: Vaccination Site Disparity: Healthcare workers are likely to receive their vaccination at the workplace – doctor's offices, hospitals and clinics. This is the case for the recipients of the most reported batches. The vaccination site of the average VAERS report was much more likely to have been a pharmacy.
Observation P: Profession Disparity: The most reported batches had about 2.5 times the quota of self-reports of healthcare professionals than the complete VAERS data set.
It is a perfect match. And it makes sense that these are the most reported batches: For many people working in health care, it is more of a vocation than a profession. They likely know that adverse event reporting exists and that it is necessary to report adverse events in order to support the detection of possible side effects. They are likely to get vaccinated by a colleague which they meet again often, to whom they are likely to report an adverse event and ask him or her to report it. On the other side, a person who has vaccinated a colleague probably feels a stronger obligation to verify the impact and report adverse events.
Now, let us make some cross-checks: About 20 million US-Americans work in healthcare. This is 7.8% of the USA's adult population of 258.3 million. If Covid batches contain an average of 1.3 million doses and each batch was exclusively administered to healthcare professionals, only 30 batches would have been necessary to completely vaccinate each one. These numbers show, that it has not been necessary to produce exceptionally large batches for the priority groups and is a clue against the 'Giga Batches Theory' described in article 1.
The remaining batches mostly were administered to everybody else, outside the priority groups, and have caused comparably few reports. For the 555 million vaccine doses administered so far, about 427 batches were necessary.
Observation Q: Plummeting Report Quality
With VAERS data, we can calculate the total number of reports related to vaccinations on any given day, and with the vaccination statistics of Our world in data, we can calculate the evolution of reporting quality over time. By using a 7-day moving average the data curves are smoothed and weekend effects are factored out.
Summing up the counts of all VAERS reports for a vaccination date and comparing it with the number of vaccinations for this day, one can calculate the quota of reported adverse events per vaccination, or, for easier reading, adverse events per 100,000 vaccinations. Be aware, that this does not relate to batches; the numbers are calculated using the total number of reports and the total number of vaccinations.
With constant reporter motivation, the quota should roughly be constant over time and the diagram would show a horizontal line. Let us verify this:
Well, this is deluding: While in the first weeks of vaccination up to 400 adverse events were reported per 100,000 vaccinations, the quota went down to 50 within a few months, stagnated for a few months, and then even went down to a recent value of 25. That is a reduction by a factor of 16. In February 2022, the number sank to 12, but this may be the result of delayed reporting.
Maybe reporting dropped only for mild events? Doctors should at least document deaths, life-threatening symptoms, permanent disabilities and hospitalizations to a large degree and in a continuous manner. Let us prove that:
The next delusion: Also the quota of severe adverse events goes down from 16 per 100,000 vaccinations to 1 recently and the number of reported deaths per 100,000 vaccinations went down from 2.5 to 0.15, both at the same factor of 16.
I have no explanation for the reduction in reporting quality.
I observed some „Super Reporters“ who filed dozens of reports. Since it takes 30 minutes to file a report, they were committed strongly enough to sacrifice many hours of their spare time for (AFAIK unpaid) VAERS reporting. Obviously, no one does this forever and after a few days or weeks, each one stopped reporting. Since VAERS does not publicly provide any means to identify a reporter, it is a complex task to identify super reporters by report similarity analysis and this evidence is just anecdotal.
Other factors that may have influenced reporting behavior, were strong efforts to reduce vaccine hesitancy, allegations that people who report adverse events were anti-vaxxers, and hospital managers asking doctors to stop adverse event reporting.
Whatever the reason, there was an impressively sharp drop in adverse event reporting rates (not counts) of 50% within 18 days, from Jan 1 to Jan 18, 2021, and another 50% within further 35 days, until Feb 23, 2021. I emphasize that this number is independent of the number of vaccinations and I do not see an explanation other than a strong reduction in the motivation to report an adverse event at all.
It is worth checking, how plummeting reporting rates influenced the report counts of individual batches. In the following diagram, I visualized the number of adverse event reports for a batch (bottom to top) over the date at which most doses of this batch were administered. Pfizer/BioNTech batches were displayed as blue squares, Moderna batches as orange diamonds, and Janssen batches as yellow triangles:
This exhibits several important observations:
After gender, vaccination site and profession, we have identified the fourth and by far most important factor for the number of adverse event reports of a batch: Date. This effect is so strong that you will hardly be able to find any more batches with large adverse event counts in 2022.
Moderna batches are significantly smaller than batches of Pfizer/BioNTech. As far as I know, Moderna batches have an average of 1.1 million, Pfizer/BioNTech of 1.7 million, and Janssen of 500,000 doses.
An extremely large variety of report counts per batch is only observed in the first three months in which priority groups dominated,
Conclusions
Suspected variations in „toxicity“ of batches were primarily the consequence of plummeting report quality over time and secondarily of the percentage of women vaccinated with and the percentage of pharmacy-administered doses of a batch. A multilinear regression over female and pharmacy percentage explains the number of adverse events with a fit (r^2) of 50%, with date added, it is 68%. For severe adverse events, only date and age are important factors, with a fit of 57%. I will explain more about these analyses in another article.
Early batch report counts which look erratic at first glance can be explained by a high concentration of these early batches to a few priority groups of which one – by profession – had especially good health monitoring: Healthcare professionals.
The observations are coherent and plausible. Thus, having started with the observation that more than 95% of all adverse event reports relate to less than 5% of the batch codes, we can safely abandon the 'Bad Batches Theory'and embrace the idea, that it is simply the consequence of most batches having very little reporting. (By the way, even the 5% number is wrong, but that is a topic for another article).
It was pure maths and logic which have led us to this insight and dictate the conclusions: If all batches have similar size and properties, they must have a similar rate of side effects, hidden only by the absence of reporting. Thus, for estimating the number of actual adverse events based on the number of reported events, we must take the most reported batches as the measure.
The 12 most frequently reported batches, which constituted the first decile of reports, had 584 non-Covid-related deaths, providing us with an estimate of 49 deaths per batch. Furthermore, there were 186 severe adverse events and 3,157 real adverse events per batch.
With an average of 1.3 million doses, 427 lots would have been needed for the 555 million doses administered so far, allowing us to extrapolate the numbers. But we need to take underreporting into account. When assuming that 50% of the 12 most reported batches were administered to healthcare professionals and that for them there was no underreporting at all, we can neglect the reporting contribution of the other priority groups in comparison, like many studies have shown (see article 2). Thus, we need to double the numbers per batch calculated above for extrapolation to all vaccinated US Americans.
Thus, as a rough and conservative estimate, it is likely, that 2.7 million adverse events occurred, of which 160,000 were severe, of which 42,000 resulted in death. This would mean that one out of 6,000 vaccinated persons (of a total of were 255 million) died. The underreporting factor would be 7 which is moderate in comparison to other VAERS underreporting studies.
My Motivation and Approach
Some explanatory words about me and my motivation: Having heard and read quite a lot of contradictory information about Covid-19 vaccines and considering the novel approach of mRNA, I initially concluded that I do not have enough reliable information to make a qualified decision for or against Covid-19 vaccination. I decided to wait for proof of vaccine safety. This wait time was necessary to end when the German government declared the intent of mandating Covid-19 vaccination. Finding out that statements about vaccine safety were more polarized than ever, I realized that I have to find out myself and chose VAERS as a source of neutral and unbiased data.
After witnessing the limitations of the VAERS web interface soon, I decided to program an analysis software, which grew more and more powerful and provides me with new interesting findings each day, as still on the day of this writing. The data set used for this article was downloaded from the VAERS website on February 22, 2022, and I analyzed US cases only. Finally, assessing my results, the observed dramatic drop in report levels even raises doubt about the unbiasedness of VAERS reports.
All of this research was self-motivated and there is no party sponsoring this activity. I have an academic degree in physics and analyzing data and numbers is my everyday business.
Since the discussion of vaccine-safety information is so polarized and prone to personal defamation and sabotage rather than consideration and discussion of arguments, I prefer to stay anonymous and use the pseudonym Leonard Frey.
Feedback welcome
As you surely have noticed, I am not a native speaker. I did my best in articulating these articles, but expect some phrases to sound strange and cumbersome to a native speaker. Feedback for linguistic improvement is strongly welcome. I processed the data to my best knowledge and for each of my findings, I tried to make plausibility checks, cross-check with other data, and ask the devil's advocate. However, if you find mistakes, weaknesses or contradictions, your feedback is welcome. You can contact me via email: leonard_frey@yahoo.com.
Stay healthy in mind, keep thinking for yourself,
Leonard Frey
(Pseudonym)
I ran an analysis the same way I run my report cohort analysis on pervaers.com.
It turns out there were a couple of what looked like bad batches in there (just a few really). However those weren't the ones you'd expect by looking at the number of reports before age adjustment etc.
I think it's a dead end.
Observation M: Severity Disparity - Being that ONLY INITIAL REPORTS are made public, even though CDC/VAERS continues to collect data, it would stand to reason that perm disability would take longer to report/diagnose, and there for have lower reporting rates? https://imgur.com/yACUYpu