Analyzing deleted VAERS death reports

Why were they deleted?

I recently came across an interesting dataset - apparently, VAERS records for COVID19 vaccines get periodically deleted.

So I downloaded the Excel file provided at the end of that post and started analyzing the reports.

Is anyone else looking into this?

My first question, of course, was to see if anyone has already discussed this.

The only peer reviewed study I found was by Jessica Rose, as part of a larger analysis of VAERS data.

In the paper, Jessica writes:

Nowhere in the VAERS handbook or on the website published by the CDC/FDA is there mention of deleted data or transparent description of the processes and criteria used for record deletion. The only reference I could find to legitimate removal of data, from WONDER’s ‘Reporting Issues’ section, claims that ‘Duplicate event reports and/or reports determined to be false are removed from VAERS’.

I visited that link to see if there is any additional information about deleted reports on the official VAERS website and I did not find any.

So I decided to dig a little deeper into this.

As you can see, there are two possibilities according to the official source: either the report is a duplicate (meaning we should be able to find the same information elsewhere) or it is a false report.

Initial dataset

I took the Excel file and loaded it into Zoho Analytics (which is actually very handy for these kinds of tasks).

I first filtered the dataset down to rows where DIED = ‘Y’ (meaning it was a death report). This provided 594 rows (out of a total of nearly 25000 rows).

Then I further filtered only the rows where the following fields where non-empty:

• age (AGE_YRS)

• sex (SEX)

• state (STATE)

• vaccination date (VAX_DATE)

After I filtered it, I had a total of 318 rows left in the table.

Missing entries

From this list, I looked for reports which were not found in any of the three datasets for 2020, 2021 and 2022. (By dataset, I am referring to the CSV files provided by the CDC. My most recent download was on 31 Oct 2022)

This was a two step algorithm:

I first check to see if there are any rows where DIED=’Y’ in one of the three years databases which also match on age, sex and state (no check for vaccination date yet).

Sometimes this already returns no results.

Then I implement a second step, where if there are any results returned by step 1, I check to see if the vaccination date matches the one in the deleted report. If there is no vaccination date mentioned in the CSV file, or if the vaccination date matches, then I include it in the result set.

The NUMRESULTS column is a count of the number of rows which match the deleted VAERS report.

There are 183 rows where NUMRESULTS is zero.

In other words, these are VAERS death reports which

a) have been deleted but

b) which are not found if you search the database using the CDC Wonder tool - meaning the duplicate was not added into the system

Here is the table with the list of reports

How to verify this data

First, visit the MEDALERTS_URL. You will see information about the report which was deleted (e.g. age, sex, state and vax_date)

Then, visit the SEARCH_URL. This will do a search and show you the results from the OpenVAERS website for the given parameters.

Note that there is no way to search on OpenVAERS using the vaccination date, but you can use the other parameters.

For the rows where NUMRESULTS = 0 in the table, you will notice one of two things:

a) either there are 0 results in OpenVAERS

b) even if there are a handful of results, you can click into them and verify that the vaccination date does not match

On the other hand, if you see NUMRESULTS > 0, you can also visit the OpenVAERS search result and confirm that there are indeed results which match all the parameters we are searching for, including the vaccination date. This also proves that I have not accidentally added bad links into the system which somehow always fail.

How this analysis could be wrong

It is also possible that this analysis is wrong. But I hope this is a good starting point for future investigations.

First of all, it is possible that the report itself is false. However, you can see at least a handful of these reports which have some narrative text with words like “reviewed” and “regulatory authority” etc which indicates that someone has reviewed these cases before entering them into the report. This makes it unlikely to be a “prank” report.

Second, it is not a false report, but the search is failing because some parameters are missing in the report which is currently used in the CDC database. I cannot rule this out because this happens quite a lot (more on that in the next article). If this is true, however, it does mean that some crucial information is missing from a death report in the VAERS database, even if the same information is available in a duplicate report.

If you can think of any other explanation for these missing VAERS reports, please do leave a comment below.