COVID uncovered (part 7)

Case severity and NHS staffing levels: a conundrum.

In part 6 of this series I explored a freshly-minted case severity index score (CSI) using freely available bed occupancy data for NHS England. The thinking behind this was very simple: if COVID infection within the population is getting vicious then we’d expect to see mechanical ventilation (MV) beds used in preference to general beds. Number-wise all we do is calculate MV bed occupancy for COVID cases, then calculate general bed occupancy for COVID cases, then derive the ratio of MV bed occupancy to general bed occupancy. This ratio is what I’m calling the case severity index score (CSI) because it should act as a reasonable proxy.

This morning I thought I’d take CSI out for a spin and see what it may or may not explain. I decided to start nice and easy by calibrating the index against something that would make sense: NHS staff absence levels, for instance¹. When staff absence due to COVID is high we may reasonably expect the CSI to be running high and so the two series should run together hand-in-glove. Except they don’t.

The first sign of trouble came with a quick Pearson correlation that yielded r = -0.399, p<0.001, n=670. I put my cuppa down and asked how on Earth it was possible for staff absence due to COVID to be negatively correlated to case severity! One slide later and I was scratching my head….

At the beginning of the pandemic we see the CSI (ratio of MV to general beds) for COVID cases running high at 0.15 to 0.20, together with the percentage of staff absence due to COVID running very high at 70% or thereabouts. This makes total sense. Now look what happens during the second CSI bump during Sep - Oct ‘20!

We have a well-defined inverse whereby COVID patients are getting sicker but staff absence due to COVID declines markedly. As if that wasn’t strange enough the situation then reverses for the winter season of 2020/21, before big-time reversing again for most of 2021. Then we have the final flourish for 2022 when staff absences rise at al all-time CSI low. These two series are virtually mirrors of each other!

If staff absence is an indicator of viral outbreak then how come bed use swings from MV beds to general beds? If a swing toward MV bed use over that of general bed use is an indicator of viral outbreak then how come staff absence due to COVID declines?

Here we have two potentially useful indicators for disease severity and they are telling us not just a completely different story but an opposite story. This calls for some decent cake and a fresh brew.

Kettle on!

1. See ‘What bed is that, then?’ (part 3)