Hospitalisation Rate For Adult Respiratory Conditions (part 3)

Lessons from an undisclosed NHS Trust

We may well ask how the hospitalisation rate for all adult (18y – 99y) respiratory admissions (non-COVID + COVID) compares historically and so I have baked this colourful little slide covering the period Jan 2017 – Sep 2021. There are quite a few things to pick out here starting with the exceptionally lean year of 2018. We may raise an eyebrow or two to learn that pre-pandemic hospitalisation rates during 2017 and 2019 matched or exceeded those of 2020 and 2021.

There is a most surge in rates for 2019 around week 5 onward that lasts until week 39 when rates once again fall back to 2017/18 levels. Even more intriguing is the week 39 decline from ~0.55 (55%) to ~0.25 (25%) is matched by a similar pattern for 2017. This is a strange precision! Also strange is the steady climb upward from week 45 of 2019 to meet the rate at the beginning of 2020 – I suspect that something was already happening before the authorities declared anything.

Despite the Delta scares and Omicron scares in the press these last few months, the latter weeks of 2020 and the weeks of 2021 are not revealing anything catastrophic in terms of a sudden need for hospitalisation of respiratory cases: bed numbers are one thing but bed need is quite another!

If we asked a Martian which year represented a pandemic they’d probably put their money on 2019 owing to the step-shaped curve. If the Martian pondered on the issue they’d figure this was a trick question because 2017, 2018, 2019, 2020 and 2021 all yield hospitalisation rates for adult respiratory admissions that all fall very much in the same ballpark. We may conclude that the hospitalisation rate for adult respiratory admissions made during the pandemic hasn’t differed from bed need pre-pandemic; a bizarre conclusion indeed!

Just to double-check this, and ensure those colourful squiggly lines are not deceiving our eyes, we can under formal hypothesis testing using a repeated paired samples T-Test. The result is provided by this nerdy-looking table….

Over on the far right under the column called ‘Sig. (2-tailed) we get to see those all important p-values, which there are 3 non-significant results (p>0.05), 3 highly significant results (p<0.001) and one marginal result (p=0.047). These us that weekly hospitalisation rates for the 37 matching weeks of 2021 and 2020 are not significantly different (p=0.213). They also tell us that hospitalisation rates for the 37 matching weeks of 2021 and 2019 are not significantly different (p=0.795). They do, however, tell us that hospitalisation rates for the 37 matching weeks of 2021 and 2018 are highly significantly different (p<0.001), as are the 37 matching weeks of 2021 and 2017 (p<0.001). If I were being unscrupulous I could choose 2021 and 2017/18 as my comparative years and you’d not be any the wiser. I could also be equally unscrupulous and choose 2021 and 2019 as my comparative years and push the argument the other way!

We observed a marginal difference between 2020 and 2019 (p=0.047), with the grand mean for 2020 being slightly higher at 0.480 compared to 0.441 for 2019. I don’t know about you but I’d be expecting more than a marginal effect in exchange for destroying the economy and slaughtering healthcare for everybody. We find a highly statistically significant difference between the 52 matching weeks for 2020 and 2018 (p<0.001), as may be expected but an insignificant difference between the 52 matching weeks for 2020 and 2017 (p=0.748).

Sharp-eyed statisticians may grunt at this point and ask what happened to the Bonferroni correction that should have been made given I’d run five statistical tests; namely, my alpha should have been lowered to p=0.01 and not p=0.05. My answer is that I was being generous to the pandemic priesthood. Take away that generosity and we find that marginal difference for 2020 and 2019 (p=0.047) melting into insignificance.

Well, that’s the rigorous bit for you! In plain English this means we can choose years to prove the pandemic was a problem in terms of the hospitalisation rate of adult respiratory admissions and we can choose years to prove the pandemic wasn’t a problem. The overall picture indicates nothing unusual happened – certainly nothing we’d expect from a deadly novel virus and its variants – with variability well within that observed historically.