UKHSA Reporting on Vaccines & Pregnancy: 100% Data Crimes
Some of the work I’m most proud of on this stack is showing how survivorship bias, due to women getting vaccinated at different times in their pregnancy, implies we should not take the raw numbers on stillbirths contained in the UKHSA reports at face value. I’ve written a number of pieces on that, starting here:
Besides stillbirths, the other measures the UKSA reports on pregnancies & vaccines, in order to reassure us about the safety of the vaccines is information on births with low or very low birthweight & information on premature, very premature & extremely premature births. Screenshots of the charts from the most recent report (I have added the actual numbers for clarity):
Wow, this is great news, right? Rather than any sort of negative outcome with the vaccines, we see that those who were vaccinated had less premature, very premature, and extremely premature births, as well as less babies with low or very low birthweight. Heck, even if COVID were eliminated from the earth, it would see we should continue giving pregnant women the vaccine in order to achieve these positive results. Cool!
Hang on a second, though. If, like with stillbirth data, many of these women received vaccines late in their pregnancy, would this affect the above numbers? Good question! If, for example, a woman was vaccinated in her pregnancy after the 28th week, it might be hard for her to have an “Extremely Premature” birth, which is defined as before the 28th week. This would also seem to make her less likely to have a baby with very low birthweight.
Let’s run the following hypothetical to illustrate the principle and why the charts above should be lining a birdcage…
Hypothetical assumptions:
100 women, all with a due date of Jan. 1
All 100 are scheduled to have their 1st COVID vaccine shot on Dec 15, during week 38 of their pregnancy
All women face the following probabilities in pregnancy
1% chance of giving birth before week 28 (extremely premature), in which case the baby will have very low birthweight
2% chance of giving birth between week 28-32 (very premature), in which case the baby will have low birthweight
6% chance of giving birth between week 32-37 (premature), in which case the baby will have normal birthweight
Results
Amazing! If we would just move to a policy whereby all women received their vaccine in the 38th week, we could have perfect results for them!
I was reading another study (can’t remember the exact one) from the US (I think) that seemed to use this sort of time censoring.
I plotted out their methodology, enrollment and typical gestation period on a time line and it was obvious they’d show a benefit just by the study design.
Oh yes. This is a very important aspect. They show us what they want us to see but as an abstract from the whole picture. I always felt there is something awry with these figures, you have just put the last piece in for me. Makes so much sense, Thanks!