Vaccines & Death (part 7)

Further investigation into delayed death following vaccination

In part 6 of this series I investigated use of NHS staff absence records as a predictor of weekly non-COVID excess deaths on the assumption that staff sickness could be a handy proxy for long COVID as well as the stresses and strains of lockdown despite this being a highly unrepresentative sample that would also be contaminated with vaccine harm. You gotta start somewhere!

We observed a noteworthy initial result in that levels of staff absence were far higher for the first wave of COVID back in spring 2020 than they were for the third wave case mega-peak of winter 2020/21, this being indicative of many things but especially of an overly zealous DIY testing regime for folk that were completely healthy. This got the nation nowhere, though it did make some people very rich indeed and I shall analyse this matter in a little more detail in a future article

ARIMA time series modelling revealed the weekly staff absence count to be a statistically significant predictor of excess non-COVID death over the period 2020/w50 - 2022/w30 (p=0.018) and thus a useful independent variable to be considered in future modelling work, though we’ll need to exercise caution in interpreting results. Long COVID will be in there but so will the kitchen sink!

I ended the article with an unpleasant plot twist, this being the revelation of a negative relationship between excess non-COVID death and the time series for combined weekly doses when lagged by 23 weeks precisely. In just one paragraph I went from seemingly providing evidence of vaccine harm to seemingly providing evidence of vaccine benefit. The key word here is seemingly, for it is possible to use the very same data in support of these two opposing outcomes: all depends on the fine detail of the methodology employed.

This brought the ONS method of derivation of excess death into question, this being an estimate based on subtraction of historic 5-year means from observed counts on a week-by-week basis. If disease was a clockwork mechanism, ticking along with digital precision, this would be an eminently suitable and splendidly simple method. Since disease of any kind doesn’t carry a diary or recognise the numbered weeks of the calendar year we may arrive at estimates that are sufficiently meaningless. For example, if the flu season arrives two weeks later than ‘usual’ then excess death will rocket during those two historically naïve weeks, leaving a gaping hole in the preceding two weeks. We thus find ourselves trying to model data that is bouncing up and down on a trampoline and doesn’t want to come in for supper. Imagine your carefully crafted results are a beverage that you are holding in your hand on that trampoline and you’ll get some idea of the mess we can get into!

In this article I am going to explore alternatives to excess death as it is currently derived, and shall start out by taking a closer look at the counts that we use as raw ingredients. I’ve no idea what I’ll find so I suggest we carefully suck any and all results up through a straw. With that said, let’s have a fiddle with T4253H smoothing to iron the crumpled bed sheets of death certificate processing…