Stuart McDonald’s contribution to the Discussion of ‘Some statistical aspects of the COVID-19 response’ by Wood et al.

It can be valuable, with the benefit of hindsight, to analyse the way that statistics were presented and used in a crisis such as the Covid-19 pandemic, so as to learn lessons for the future.However, there are several points I disagree with in this paper [1].In the interests of brevity, I will focus my comments on the analysis of excess deaths.The authors are right to highlight the importance of allowing for changes to the age structure and size of the population when calculating excess deaths.The iterated life table approach described in the paper has much in common with the approach taken by the Continuous Mortality Investigation (CMI) in their weekly Mortality Monitor publications [2].The CMI estimates 151,700 excess deaths [3] from the onset of the pandemic until the end of 2022.This is 60% higher than the authors' estimate of "around 95 thousand total excess deaths".This significant discrepancy (56,700 deaths) arises from two key differences.Firstly, and most materially, the authors calculate the number of expected deaths based on mortality rates in 2017-19, with no adjustment for subsequent improvements in mortality.Although the rate of improvement has been slower since 2011 compared to previous decades, mortality rates continued to fall up until the onset of the pandemic [4] and it is reasonable to assume that they would have continued to fall thereafter.Actual mortality rates were typically above the authors' baseline in 2017 and early 2018, but below the baseline thereafter until the onset of the pandemic.Close examination of the authors' Figure 4 bears this out.Figure A seeks to make this point more clearly, showing that deviations between observed mortality rates and the authors' baseline are not random.Mortality rates were significantly lower than the baseline in the two years before April 2020.