Emergency Department Admissions: Analysis of CDS Dataset (part 6)

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

Death is a big deal, especially if folk die in the emergency department, so I’m going to take a closer look at this outcome for the pre-pandemic period of January 2017 – February 2020 and then extend the analysis to consider all deaths through to September 2021.

Here’s what the multilayer perceptron neural network model for the prediction of deaths in A&E churned out in terms of variable importance:

This makes a lot of sense and matches my own experience of nine years of working between cardiology, cardiac surgery and A&E in a busy teaching hospital. Death-wise we’re talking about an elderly intake with multiple illnesses and likely suffering a cardiac condition requiring treatment, and arriving by ambulance (no doubt under a category 1 call). The cardiac procedure in question is almost certainly going to be thrombolysis, though defibrillation will also feature if paramedics were unable to secure some sort of stability.

So let’s have a look at deaths in the department each week for the p…