Emergency Department Admissions: Analysis of ECDS Dataset (part 2)

Further analysis of an ECDS data dump of 237k adult admissions records to the emergency departments of an undisclosed UK NHS Trust: the COVID question (continued)

I left off in part 1 of this series with the promise of adjusting my net. Since it is possible to debate the treatment matrix until the (taxed) cows come home I decided to narrow the shortlist down to those 7 procedures that were very much respiratory orientated, these being:

• Drainage of pleural cavity

• Insertion of endotracheal tube

• Insertion of pleural tube drain

• Nasopharyngeal airway insertion

• Nebuliser therapy

• Noninvasive ventilation

• Oxygen therapy

The resulting cross-tabulation against COVID symptomatic status now looks like this:

There’s an improvement for sure, with 10.9% of asymptomatic COVID cases receiving a respiratory procedure compared to 17.9% for symptomatic COVID but I’m still puzzled as to what might have been going on. Could it be that the tag of a positive test result took those 68 ED admissions down a different (and unnecessary) pathway? And how about that seemingly low critical treatment rate of 17.9% for those 2,874 symptomatic COVID cases? If we’re talking genuinely novel and deadly shouldn’t that figure be way, way up?

Getting To The Bottom

Not content with this I decided to get right to the bottom of matters by pulling down those 68 asymptomatic COVID admissions that had received one or more of those 7 shortlisted procedures then ran a quick multiple response frequency analysis to reveal what precisely had been done. Herewith the resulting table:

OK, so insertion of an endotracheal tube occurred in just 1.5% of cases, Nasopharyngeal airway insertion occurred in just 1.5 % of cases (likely the same cases) and Nebuliser therapy occurred in 14.7% of cases. What really takes the biscuit is Oxygen therapy, occurring in 86.8% of cases – this is what is throwing the curve ball!

A Little Word…

I guess I better have a little word about oxygen therapy (a.k.a. supplemental oxygen). Back in Catastrophic Health Collapse (part 5) I dropped this innocuous para:

Again we must note the real world wrinkles, for these procedures are not necessarily reserved for respiratory conditions only. Many will be used for those suffering a heart attack or suspected heart attack (precautionary), along with many and varied other conditions – traumatic chest injury, for example. Readers might spot use of supplemental oxygen and automatically think severe COVID. Not so! Oxygen is the most commonly used non-invasive therapy in use around the globe.

There’s a nice summary over on Wiki explaining the ins and outs of oxygen therapy but all we need note is that oxygen therapy is so widely used such that it doesn’t offer any discriminatory power for the purpose of my analyses, so I shall drop this uncommonly common treatment and start again with a clean slate:

That’s better! There are now just 11 asymptomatic COVID admissions representing 1.8% of this sub-population. I can live with that – no funny business needed. But just look at the pathetic rate of treatment for symptomatic COVID admissions, which is now down at 3.1% for this sub-population. This is a bit of a shocker, for out of 2,874 such admissions only 90 received a treatment commensurate with a seriously nasty respiratory condition. And this was a novel and deadly virus was it?

Non-COVID Admissions

What we can do at this stage is run the same cross-tabulation again but for acute and chronic non-COVID admissions and see how these compare:

Well this is something of a curve ball, is it not?! Whilst our expectation is for acute respiratory conditions to be treated within the ED in no uncertain terms, we discover that it those with chronic conditions that are more likely to receive treatment within the department, with the rate up at 24.2% compared to 4.2%.

Some reader is bound to ask if chronically ill patients could be presenting as acutely ill. Whilst this most certainly happens in the real and muddy world of healthcare, for the purpose of this particular analysis I excluded those presenting with both conditions so we might gawp at ‘pure’ results.

But it does make sense. Whilst the ED can do an awful lot for those struggling with asthma, COPD etc, they are not going to hang about with those coming in with pneumonia/ARDS etc, who will be shunted down the corridor as soon as they are stabilised, and into the hands of physicians over at the department of respiratory medicine. Since this would appear to be ‘normal’ practice for non-COVID respiratory cases then we can’t make much out of the low treatment rate for symptomatic COVID.

When & What?

At this juncture I’m more inclined to investigate the flow of asymptomatic and symptomatic cases over time in this post-vaccine period, so here is the weekly count of such cases:

There’s nothing particularly unusual going on here with the cessation of the 2020/21 winter peak and beginnings of the 2021/20 peak as the autumn commences and mould/fungus spores start to wreak havoc on a damp island. What feels slightly odd is that we’ve got asymptomatic COVID admissions matching the rise and fall of symptomatic COVID admissions, though I concede this may simply be a function of prevalence of whatever it is that was supposed to be prevalent across the nation. With this in mind let us take a look at a scatterplot of weekly asymptomatic against symptomatic COVID admissions:

Now that is a strikingly linear relationship! Normally that thin black trendline would be the result of an ordinary least squares regression based on the assumptions of normality within the data but what we have here are counts taking up discrete values and so I rolled out a generalised linear model (GLM) using the assumption of an underlying Poisson distribution. Herewith the parameter estimates for this model:

I am pleased to see the intercept of 0.098 fetch-up as totally insignificant (p=0.868) because it infers zero asymptomatic admissions when there are zero symptomatic admissions and this keeps things neat and tidy for our assumption of disease prevalence as the driver. That coefficient of 0.216 tells us that for every 46 symptomatic admissions there were 10 asymptomatic admissions, on average (p<0.001), but we could have guessed as much from applying a ruler to the cluster of points. Modellers must model else they count beans.

It occurs to me that the ratio of asymptomatic to symptomatic COVID admissions might tell us something useful about the prevalence of whatever COVID was supposed to be as well as functionality of the PCR test, as well as how EDs approached COVID case management over time; so how about we squizz at the ratio over time?

There we go. Those deviations about a grand mean of 0.219 look pretty random to me and this is confirmed by a Wald-Wolfowitz test for randomness (runs test), which coughed out p=0.813. I shall conclude that asymptomatic admissions do not have their own dynamic that is independent of symptomatic admissions, with both being tied to test protocols. More evidence for a testdemic that was, methinks; but I’ll tell you what is well weird and that is the concept of a symptom-less disease in the first instance. I’d call that a koan designed to keep even the experts talking tripe, and the public from cottoning-on to the global game play. Show me a study that unequivocally demonstrates symptom-less carriers capable of infecting others and I’ll show you this bridge I have to sell…

Kettle On!