Mortality Impact 1975 – 2021 (part 1)

England & Wales differential monthly excess mortality by quinary age band & sex: males 45+y

Ok, so this mouthful of a subtitle may need some explaining! Get the kettle on and grab that biscuit barrel…

In the last bunch of slides we squinted at 36 scatterplots that tracked monthly all cause excess mortality over the period 1975 – 2021 by quinary age band and sex. We saw cycles, trends and outliers these being a pictorial summary of the relative health of the nation when we bring in the notion of a ‘baseline’. Excess mortality enables us to appreciate deviations from a norm, that norm being mean values attained over the previous five years. Whilst the ‘pandemic’ stood out for older age groups it was by no means a unique event for the nations of England & Wales, and we observed many instances of historic excess mortality greater than that witnessed over the last two years. Pandemically-minded protagonists keen to ride the gravy train until it smashes into the buffers may well respond with statements such as, “ah well, yes, but the rapid onset of the outbreak is what was so frightening and unique”. I have a rude word to say to these people and that word is…

Differentials!

If we are now talking about the scarily rapid onset of the outbreak that scourged the globe in next to no time like no other pathogen before then we are talking acceleration, rates of change and differentials. Just how fast out of the blocks was this bug?

We can determine this several ways, one way being to take the monthly time series for excess mortality and derive the first order differential. This sounds fancy but all it means is that we calculate the month-to-month differences in the values of excess mortality. Thus, if excess mortality was 4.0 deaths per 100k pop in June and 6.0 deaths per 100k pop in July then the differential we would chalk up would be +2.0 deaths per 100k pop; that is to say, excess mortality rose by 2.0 deaths per 100k pop between June and July. Simples!

What I am going to do now is to derive the first order differential for all 36 data series and plot them all out once more. We should expect to see a load of little red dots bobbing about the zero value of the y-axis since, by definition, the first order differential removes any linear trend in a data series i.e. if something was steadily going up in value over time then its differential would be as flat as a pancake. If something was accelerating in value over time then its differential would be an incline. In time series world a series that bobs along the zero value of the y-axis is said to be ‘stationary’.

What Are We Looking For?

In this set of slides we are looking for five things:

• Are the data revealing an incline indicating acceleration in excess mortality?

• Are the data revealing a stationary process indicating a linear rise in excess mortality?

• Are the data revealing heteroscedasticity revealing changes in underlying mechanisms?

• Are the data throwing up outliers marking disease outbreaks?

• How does the recent pandemic compare to historic outbreaks of something nasty?

The first three tell us something about the dynamics of health and well-being of both nations in a very general way (impact of policies, laws, healthcare provision, inventions, economy, nutrition, immunity etc) whereas answers to the last two questions reveal what pathogens were getting up to.

Cooking With Control Charts

I’ve taken a leaf from the big book of control systems theory and plotted out two dashed grey lines that bind each horizontal series. These mark the 3 sigma boundaries (±3 standard deviations) and reveal the point when we might jump up and down and raise the alarm because our ‘process’ is going wildly out of control. For a perfectly classical Normal distribution the 3 sigma boundary delineates a p-value of p=0.00135; that is to say we can be roughly 99.865% confident that something is genuinely amiss. Another way of putting this is to say the odds of such an unusual process event happening by chance are around 741 to 1. Let us now pull up the first slide and get the feel of these new-fangled ‘disease control’ charts…

Here’s the first order differential series for 0 - 4y males. We don’t see any incline which indicates excess mortality was not accelerating over time. We observe significant heteroscedasticity whereby scatter for the first few years is rather wild compared to recent years. This suggests that excess mortality, and thus the mortality rate for this group is under tighter control of whatever it is that is controlling the health and well-being of this group (standardisation of healthcare is where I’d put my money). There are no truly significant outliers save for modest outbreaks of something during 2006, 1996, 1986, 1981 and 1976. We may note the years 1976 and 1981 exceed the 3 sigma threshold, thus sounding the alarm.

In these charts It is evident that you may also witness outlying negative excess, with the happy alarm being sounded during 1976 and 1986 for 0 - 4y males. There are many reasons why the series will swing into negative excess, starting with the response of the authorities (health and otherwise) and ending with the response of the individual (immunity, behavioural changes etc) along with the response of bugs and environmental toxins etc etc etc. There’s also the harsh fact of disease culling the unhealthy, then we have the mathematical matters of perturbations to a resilient system and whether that system is under acceleration or trundling along in a linear fashion (i.e. what goes up must come down).

Feb 1976 was not a good month for 10 -14y males. Note the negative rebound and total lack of the 2020 ‘pandemic’ in terms of excess mortality.

Still no sign of a 2020 ‘pandemic’ for 20 - 24y males, though May 1997 most certainly wasn’t a good month for this group.

Signs of the 2020 pandemic at last amongst 30 – 34y males, though this peak excess didn’t occur during the first wave of Apr 2020 or second wave of Oct-Nov 2020 but toward the end of the third wave of Jan 2021, which is rather curious given earlier strains were supposed to be more potent. Not just any old peak either but one that stands head and shoulders above all others during the last 47 years and which exceeds our 3 sigma alarm boundary. Some will no doubt point out that young frontline workers as well as clinically vulnerable young males will have likely received their first dose of vaccine during this time. Food for thought indeed!

Now this is intriguing for I was expecting another Jan 2021 peak excess. There’s certainly an elevated value at Jan 2021 but this is merely reflective of historic levels. How come 30 – 34y males show a clear and unprecedented peak when 35 – 39y males do not? Yet more evidence to support my notion that SARS-COV-2 can determine dates of birth - unless something else is going on!

It is with 40 – 44y males that we get to see a decent outlier aligning with the Apr 2020 ‘pandemic’ first wave for the first time in this series of slides, though this spike doesn’t quite trigger the 3 sigma control process alarm. In fact the alarm was truly triggered back in Aug 1981 (I don’t recall wearing masks back then or doing odd things with soap). We may note that Jan 2021 also features as a decent outlier but the question is which ‘v-word’ pulled the trigger? The substantial negative rebound of May 2020 is also worthy of cogitation.