Digging Deeper Into EudraVigilance (part 1)

Duplication and the Top Twenty Reactions

I’m just about getting the hang of handling data obtained via the EudraVigilance line report system. Actually, it’s not so much understanding the data as understanding how it hangs together and how to handle foibles. Foibles such as the keyword ‘death’ not producing a listing of all fatal reports, and foibles such as somebody receiving Pfizer, Moderna and AstraZeneca shots and being triplicated into the bargain. Then there’s the knack of handling 3,952 different reactions recorded across 60 entry fields. But, the data fog is lifting!

Duplication and triplication has been a big issue to the extent that the 41,328 deaths we may count using the slides generated for public consumption in the EudraVigilance website actually boil down to 23,890 unique individuals when we de-duplicate the raw data, and I attach a simple summary slide of actual fatalities by product by reporting year¹ as best I can estimate…

As for those 3,952 different reactions across 60 entry fields neither is it plain sailing there. Those 23,890 unique individuals were tagged with 82,296 reaction events (3.4 tags per fatality), many of which are not strictly adverse reactions. For example, there are 1,118 instances recorded as ‘vaccination failure’, 517 instances recorded as ‘drug ineffective’, 459 instances recorded as ‘off label use’ and 92 instances of ‘interchange of vaccine products’.  I’ve produced a listing of the top 20 reactions in descending order of frequency so we may get a feel as to what was happening just prior to death. This top 20 listing accounts for 34.7% of all reactions recorded…

Death is the number one reported reaction, occurring in 25.4% of fatalities (yes, the mind boggles). It should, of course, occur in 100% of fatalities but all depends on how healthcare professionals set about entering the data. What isn’t helpful is when we encounter records that tell us that person died because of ‘death’. If we flag all forms of inexplicable death as a reported reaction we find 7,156 individuals designated as such out of a total of 23,890 (30.0%). In terms of record completeness some 3,035 inexplicable deaths were recorded for which no other data entry had been made (12.7%): this seems like an awful lot of death for which no reason has been given.

Join me again soon for another rustle in the biscuit tin, meanwhile…

Kettle on!

1. Data to 28th February 2022