An Enigma: Transmission Of Epidemic Influenza (part 12)

I attempt to shed light on the riddle that is seasonal influenza using my bag of spanners. Today I check out the percentage of influenza to all cause death and find more oddities

And so we continue with the last three slides…

A well-behaved plot, as promised, with less wayward outliers: just the ticket! I’d like readers to notice the upper cluster of points and how they seem to be biased toward females for we shall see this bias develop bigly over the next two slides.

---

There it goes! A nice, tight cluster of points - yes indeed - but that systematic bias of a bigly nature is darn intriguing. Let’s clock the last slide then have a think…

---

The systematic bias is well evident, though we’ve gained a couple of rogue years by way of 2014 and 2018 that suggest a higher and somewhat transient rate for elderly males compared to females that doesn’t really make sense when you sit and think through things. I’m not into sex-specific pathogens, especially those that are also age selective, so my money is on something else going on.

That something else may well be how physicians go about certifying deaths in the first instance. If this is indeed the case then I’m not that confident of any and all cause of death analysis, which pretty much leaves the field high and dry (and prone to misinterpretation). H’mmm…

What Now?

We’ve got two weird and wonderful things going on:

1. Isolated wacko years that indicate a serious but transient imbalance in the counts of ‘causal’ influenza that largely favours influenza as a cause of death for females;

2. A general trend toward ascribing causal influenza death to females that may be viewed as systematic bias.

It doesn’t look to me like we’ve got a reliable, consistent and coherent reporting system for cause of death despite the best efforts of the ONS. I know this for an unpleasant tasting factoid; for I know it from experience as the suit who chased cause of death for a living in a busy teaching hospital, being the same bloke who organised the monthly mortality and morbidity meetings when consultant cardiac firms chewed over the fine detail of each and every case they failed to save.

Death is never straightforward, and even a gunshot victim can die from liver failure if liver function is not closely monitored as drugs are pumped down lines. Couple that with a senior house officer or registrar filling out certificates late into the night, or a GP doing the same well after the fact (and at distance from the facts) and you have a recipe for rubble.

That rubble then gets fed into some fancy ONS software (MUSE) that forces a multi-causal situation into a unicausal outcome. Now couple this with a condition that isn’t that well-defined and comes embedded in a truck load of complication such as bacterial opportunism and it is no wonder we get problems when we come to analyse the data.

But analyse it I shall.

Readers better run and hide now because I’m talking-up more blasted ARIMA. Eighteen more models to be precise, with as many indicator variables as I can fit on a (large) side plate.

Run! Run for the hills before part 13 is penned!

Kettle On!