Emergency Department Admissions: Analysis of CDS Dataset (part 13)

I analyse an anonymised data dump of 1.9 million admissions records to the emergency departments of an undisclosed NHS Trust for the period June 2017 – September 2021

Continued from part 12 after stopping for a much-needed mini-break, some really real UK sunshine and a fair few beers…

Other Cardiac Conditions

Back in part 12 my model churned out an odds ratio of 1.26 (p=0.036) for other cardiac conditions, this being suggestive of a slight tendency toward these cases during the peculiar death ratio indicator period that was the three weeks immediately after vaccine rollout (2021/w2 - w4 covering the period 9 January – 29 January). I shall quickly remind readers that other cardiac conditions is dominated by chronic ischaemic/atherosclerotic heart disease. The bivariate crosstabulation above vaguely shows this result, with the percentage of deaths rising from 6.4% to 6.9% (factor increase of 1.08), but the effect is nowhere near as beefy: that’s multivariate messing about for you!

Cancer diagnosis

Cancer was another of the peculiar features of my model offered up in part 12 with OR = 0.77 (p<0.001), and here we have the crosstabulation confirming an inexplicable drop in the percentage of these cases from 30.3% to 22.2% (a factor of 0.73). So where did a sizeable chunk of in-hospital cancer deaths go during those first three weeks after vaccination rollout? All very peculiar if you ask me! Since cancer is relentless I can only presume the usual fun and games with clinical coding and/or patient management.

COVID-19 diagnosis

Well, here we are with that dodgy COVID-19 classification again that I am taking as a proxy measure for pretty much anything other than a novel and deadly virus. We ought to call this the mistreatment administrative indicator (MAD). In the model (that is now gathering dust) this offered up a whopping great odds ratio of 4.47 (p<0.001), and in this bivariate crosstabulation we can see the percentage of deaths rise from 12.8% to 40.8% (a factor of 3.19). A seeming really real result, then, but the question that needs asking once more is why the first three weeks of vaccine rollout during early 2021 were characterised by excess in-hospital COVID deaths?

Safe and effective my Aga!

Time, please!

That concludes my long-winded walk through a staged multivariate logistic tray bake aimed at identifying the characteristics of deaths during two very peculiar periods when the ratio of in-hospital death to emergency department death went through the roof.

I’m sure readers are going to groan when I declare that there is yet another method we can use to distinguish these peculiar periods (discriminant analysis) but I think we’ve had enough of this sort of thing; besides which, we really ought to be getting back to looking at time series data for emergency department admissions ‘coz that’s wot this series is all about.

So far I’ve plotted time series for all admissions, respiratory diagnoses, respiratory procedures, infectious diseases, probable COVID/ILI, (all cause) deaths in department, and all cause in-hospital deaths, so I guess we ought to look at historic ED admissions for cardiac cases, and I shall start with raw weekly counts:

There’s that great big lockdown dip as we may well expect. You’d think that folk would be bothered by this because that sudden and dramatic drop in cardiac admissions to the emergency department means a proportion of these people were dying at home. Try telling that to those who promoted lockdown and all you’ll get is a blank stare: they simply don’t want to know. Providing people die behind their own closed front door then all is well with the world, it would seem.

So that’s the raw numbers, then, and here’s them turned into a percentage of all admissions:

This is totally fascinating. We’ve got a wall of cardiac illness for a pandemic that was supposed to be respiratory in nature, though I do appreciate that the heart and lungs work in tandem, so elevated cardiac admissions is to be expected. However, we must be careful here in assuming we’re looking at elevated case rates within the population. What we are likely looking at is a funnelling effect whereby the most serious cases, by necessity, have to end up in A&E, whereas those with minor ailments will have chosen to stay at home and use the 111 service, their GP (by telephone) or even their local pharmacy. The case profile will thus have changed dramatically during 2020 and there are many ad hoc reports of this being so. We might say that the sharp end of medicine got sharper.

Despite the alleged pathogen fizzling out, and despite safe and effective vaccines allegedly saving millions of souls, and despite a return to normal work practices, normal case profiles, and normal workloads, we seem to have a nasty hump during early 2021 that requires explanation. Are we looking at the delayed effect of service closure and/or lockdown… or are we staring at vaccine harm? All is possible, but without a detailed audit of casenotes we are not going to get anywhere.

Cardiac Procedures

There’s no finesse to the coding of cardiac admissions within CDS: you get a choice of three options and that’s yer lot, these being: Cardiac conditions, Cardiac conditions - myocardial ischaemia & infarction, and Cardiac conditions - other non-ischaemia. Cases thus range from someone having palpitations and getting a bit worried to those trolleyed straight through to life-saving surgery. As a consequence I decided to take a look at procedures carried out within the emergency department that are uniquely cardiac. Herewith that shortlist:

• Defibrillation/pacing

• Defibrillation/pacing - defibrillation

• Defibrillation/pacing - external pacing

• Parenteral thrombolysis

• Parenteral thrombolysis - recombinant

• Parenteral thrombolysis - streptokinase

What we are going to do now is squint at a plot of the number of admissions requiring one or more of these specialised treatments:

Wow! Now there’s queer old result. That incredible mega spike took place during 2017/w43 and w44 and was largely driven by an extraordinary number of codes for Defibrillation/pacing – 230 cases to be exact. The individual records for this period look kosher so we can only guess as to what was going on back then in this single NHS Trust. Either a desk clerk new to the job made a right old pig’s ear or they opened up a pacing clinic for two weeks in a regional drive to improve health.

But it wasn’t only defibrillation and pacing for we observe a kick-up in use of Parenteral thrombolysis and Parenteral thrombolysis – streptokinase to the tune of 48 cases. You don’t mess around with that stuff for these poor souls would have been arriving as ‘category A’ calls with ST Elevation Myocardial Infarction (STEMI) – i.e. a heart attack owing to clogged arteries. Get thrombolysis wrong and you kill people through brain bleed so it’s a procedure that requires registrars and senior house officers who are fully clued. An uptick in appropriate staffing would open up the possibilities; then there’s investment in fancy telemetry enabling ambulance crews to broadcast critical ECG data for A&E medics to be ready with the needle when the back doors swing open.

Setting that wacko result aside my eyeballs are scanning that hump that marks the 2018/19 winter season, stretching into spring 2019, and we may wonder what was going on back then in this region. As well as trends within the catchment area we have to embrace the possibility of a new and enthusiastic consultant and/or registrar deciding that cardiac was THE top level priority for the emergency department, and/or they might have been trying out new facilities, drugs and protocols. Then there’s the wrinkle of the coronary care unit (CCU) and/or cardiac ward beds blocking back to push A&E medics into taking action while the patient was ‘hot’. Then there are suits like me who’d go and stir all of this up in order to maintain gold standards for door-to-needle time in order to secure funding for the Trust.

It’s hard to make any sense of this because of the scaling and because of the many confounding variables arising from lockdown and service closure, so let’s have a closer look at the percentage of cardiac procedures against all admissions to flatten the playing field a little:

Well there you go – this reveals a great deal! Let’s remove the spike and the hump from our minds for the moment and concentrate on the rest. That looks pretty samey to me and it’s the sort of thing I’d like to call ‘noise’ on account of my sound engineering hobby. That makes sense but the thing to note is that the noise didn’t budge from doing its thing during and after the pandemic. Under normal circumstances cardiac procedures trundle along at 0.02% - 0.04% of all admissions to the emergency department, so what’s with that hump straddling 2018/19? Curious. Most curious.

Another thing we can do at this point is look at the ratio of cardiac treatments to cardiac diagnoses. I like this approach because it irons out a fair few confounding factors and it gives us a proxy for seriousness of admitted cases:

Same again! If we assume an average of 0.005 for ‘normal’ life then we’re talking about a need to treat for 1 in 200 cardiac admissions. Before we go assuming that emergency department staff need to pull their socks up it’s worth noting that cardiac cases tend to get whizzed through to cardiac wards and facilities as soon as can be, so those 199 forgotten cases are not necessarily sent home with aspirin! The management of cardiac admissions can be looked at in more detail in a future article if there is sufficient interest but for now the main take away message is business as usual during and after the pandemic, with a spike and a bump that needs explanation – especially that bump!

Coffee & Cogitation

So there’s cardiac for you. Nothing is showing pandemic-wise and nothing is showing vaccine-wise though we must remember that folk need to stay alive to make it to the emergency department. My gut feeling is that something genuine happened amongst the population at the beginning of 2019 and this would be worth a slice of sleuthing.

Kettle On!