UK Personal Independence Payments & Accumulated Vaccine Doses (part 1)

A statistical rummage through the data held on UK GOV servers that reveals a strong correlation between COVID-19 vaccination dosing and cleared new claims for Personal Independence Payments

Prologue

SCENE: A Forest Of The Mind

[A horn sounds. Dee enters stage right]

Yes, I know that correlation doesn’t equal causation. That’s a basic premise I learned 37 years ago as I embarked on my career as a UK government scientist with specialism in statistical modelling. As a section head in charge of national research programmes I was also called upon to brief various government ministers and secretaries of state, and to inform departmental policy as well as supply statements to various press offices. Sloppy thinking was beaten out of me very early on.

In my crusted experience this old chestnut is usually quoted when somebody doesn’t like what they’re seeing but can’t come up with a good explanation. It’s also quoted by those who simply don’t want to think, for thinking can take us down some uncomfortable paths. In many ways straining out that well-worn phrase is akin to straining at the toilet: constipation can come in many guises.

I know this because I’ve had this chestnut thrown back at me more times than I’ve had hot dinners (and I like hot dinners). I can always tell the genuine responses because there is always some sort of thoughtful follow through such as: “could you be wrong?”, “what might this mean?” or “have you thought about trying such and such?” Dialogue ensues and the analysis evolves, sometimes melting into nothingness because it was a frippery and other times growing into a powerful proposition.

Willingness to address an issue is not always the way, for those same wise words can also be used as a thought terminating cliché, which sometimes comes sprinkled with a sour coating of ad hominem. Such responses are designed to destroy rather than engage, and these days it is worth noting just who is doing what and why (following their money is always a good start).

The crux of the matter is simple: if we spot a correlation that gives cause for concern we ought to follow the matter through to the best of our ability to ensure that concern has no solid basis in reality. This is especially true of population-level statistical analysis concerned with health and welfare, for that weak correlation might be the only signal you get to avert disaster.

The position analysts find themselves in today is rather surreal. Large numbers of people are dying, and dying suddenly, some with rare conditions (that are thus no longer rare). Authorities are blaming everything from crisps to goal-scoring for the relentless rise in excess deaths and yet we are not permitted to consider the most obvious potential cause of them all. Instead of opening the jaws of the monster wide to reveal it is toothless we’re not even permitted to acknowledge a monster that is crawling over nation after nation.

And so, dear readers, we come to yet another analysis that suggests something is very wrong. If fortune is smiling on us we may be looking at nothing more than a numerical mirage. We could also be staring down the barrel of a gun. With no authority caring to determine the situation in an independent and credible fashion then it is down to each and every one of us to determine the truth for ourselves.

[A horn sounds. Dee exits stage left]

ACT 1, Scene 1

In my article entitled US Civil Labor Force Disability & Accumulated Vaccine Doses: A Case Of Bent Regression? I addressed a valid criticism of a regression analysis performed by bods at Phinance Technologies. The criticism concerned the rather technical matter of serial correlation with cumulative data that can give rise to a false impression of correlation. In this instance a bit of delving revealed the criticism to be unfounded and the original conclusion sound. And what a conclusion it was, there being a rise in disability amongst the 16-64 year group that powerfully correlates with mRNA vaccine rollout sufficient to disturb all but the most entrenched minds!

What should be a laughing matter but is not are those experts dismissing the argument out of hand on the basis that correlation doesn’t equal causation. Yawn. Yes, we all know that (and can recite it in our sleep), but what if that correlation is arising out of causation? Until we undertake the appropriate course of research we cannot be sure (n.b. a limited clinical trial of handpicked healthy people isn’t a sufficient evidence base when you go slamming drugs into millions).

Out of this work came a suggestion from a couple of readers that I take a look at UK data to see if the same pattern emerges. In this regard Jonathan Engler of HART fame very kindly pointed me to an incredible online resource operated by the Department for Work and Pensions (DWP). He also furnished two interesting articles that are worth a fresh pot of tea. First we have a dry statement by the DWP on Personal Independence Payments (PIP), and second we have a juicy article in Yahoo News that goes under the title, The 'astonishing' rise in people claiming one key benefit.

ACT 1, Scene 2

It took me 1.5 pots of tea and no biscuits to get the data I needed, this being PIP new claims (normal procedure) for the nation of England that have been cleared (i.e. have been ratified and awarded). This is an important distinction because we need to separate out DLA reassessments and special claims made for end-of-life cases (SREL) such that we end up with fresh claims from everyday folk that have passed through to award; this ensuring the validity of the claim.

Claim counts were provided on a monthly basis from April 2013 through to October 2022 with download available in Excel format – marvellous stuff! I then scoured the NHS COVID-19 vaccinations online resource for the latest daily dosing information for England, which may be found in the weekly announced vaccinations file of all places, that sits here. Doses 1, 2, 3 and 4+ were summed on a monthly basis to provide total administered doses, for which the cumulative count was derived.

ACT 2, Scene 1

I guess we better start by taking a look at the incidence of new PIP claims over time. Try this:

We need to trawl through the history of PIP to see how it got going and what benefit payment system it replaced, for this sort of admin wrinkly could explain that initial steep climb from April 2013 to January 2015. The important feature for me is the steady pattern of payments made that set in from 2015 onward, for these give a flavour of what was happening all the way through the beginning of the pandemic right up until July 2021 when boosters were thrust at folk. You can see now why Nadine Batchelor-Hunt used the word ‘astonishing’ in her December 2022 piece for Yahoo News.

ACT 2, Scene 2

The data period December 2020 – October 2022 was selected for analysis. A Pearson bivariate correlation between monthly new claims and cumulative total administered doses fetched-up at r = 0.880 (p<0.001, n=23). Whilst this strong and highly statistically significant correlation should raise more than an eyebrow we must recall that we are using cumulative dose data and are thus wading through the land of serial correlation leading to phantoms and illusions. The ‘true’ correlation is likely to be less than this, and this matter will be addressed later.

A correlation like that is going to produce a tasty scatterplot and here is that plot:

That there is a clear, positive relationship there is no doubt. But what we need to do now is use a spanner to account for oodles of serial correlation embedded within a rigid monthly cumulative series for vaccine dosing. The particular spanner of choice in this instance was autoregression using Prais-Winsten estimation.

This proved to be an essential tool for the Durbin-Watson statistic using ordinary least squares regression started out at a scary 1.133, this being below the critical threshold value of dL = 1.26, thus putting us right in the guano. Four iterations were needed to settle on a Rho(AR1) value of ρ = 0.416 that was then used to adjust both time series. Key tables spewed out by the procedure are as follows:

The beady-eyed will note the much improved Durbin-Watson statistic of 1.842, this putting us back in the safe zone for serial correlation. We can thus have confidence in the final model that is displayed in the lowest table. In that table we see a rate of 246.920 new claims per million total administered doses, this rate being highly statistically significant (p<0.001). This model is plotted as that thick green line.

Whilst correlation doesn’t prove causation we must remember that correlation serves as a signal for causation. To ignore such a signal at this stage in the game, with such a wealth of anecdotal, clinical, microbiological, statistical and other forms of evidence is tantamount to a criminal act. Perhaps it’s time to start calling certain folk out as vaccine research deniers if not vaccine harm deniers.

Ground Control To Major Tom

I appreciate this section sounds like gobbledegook so what I want you to visualise is a knitting needle rammed up the backside of 23 marshmallows, each representing data for each month from December 2020 to October 2022. Because of that knitting needle (a physical representation of serial correlation) all the tasty treats line up whether they like it or not, and so we arrive at a forced alignment that will be partly illusory (i.e. derived by virtue of the knitting needle). If we whip the needle out (the Prais-Winsten trick) then the marshmallows can fall how they want to. Some will still line up and others will not, so we get to see the sugary truth.

ACT 3, Scene 1

All this is pretty crude stuff but at least it’s a start, and we’ve gone and found a result for the UK that is in alignment with Phinance Technologies US disability study. What I’ve presented is the equivalent of licking the lid on a pot of cream, there being a great deal more that could be done with the study. I’ll let this drama stew for a while and let ideas flow in. One tray bake I’ll be trying is… you’ve guessed it… some ARIMA time series analysis! With that comes the opportunity to throw in confounding factors as independent variables. Judging by expert commentary in the media this ought to include consumption of crisps, blows to the body, jogging, goal-scoring and a dramatic rise in coincidences.

Kettle On!