Ondřej Vencálek's contribution to the Discussion of ‘Some statistical aspects of the Covid-19 response’ by Wood et al.
本文讨论了在评估新冠疫苗有效性时，因接种者健康状况差异导致的“健康疫苗效应”和“指征偏倚”如何使直接比较不同接种状态人群的死亡率产生误导，并引用英国国家统计局数据说明问题。
The current paper by Wood et al. is a must-read for anyone who has ever spoken publicly about COVID-19.It is written with the utmost effort to be objective.The procedure suggested in Section 2.4 could provide insight into an issue not discussed in the article, namely the evaluation of vaccine effectiveness.Let us look at ONS "Deaths by vaccination status in England" [1].The data file includes this warning: "the characteristics of people in the different vaccination status groups, such as health, may differ" An important pattern can be observed in the non-COVID death rates: Once each wave of the vaccination campaign begins, the group of already vaccinated individuals splits into those who opt for the new dose and those who stay with the previous dose.While non-COVID mortality rate decreases in the former group, it increases in the latter.For example, in the 90+ age group, age-standardized non-COVID death rate in the (very small) group "first dose, at least 21 days ago" is about 5 times higher than in the group "second dose, at least 21 days ago", and it is about 3.5 times higher than in the group of unvaccinated in June 2021.Does it mean that vaccination not followed by the next dose increases the risk of dying?No, the compared groups differ substantially in their susceptibility to die of non-COVID causes.This fact makes any direct comparison of their COVID-related mortality rates not accounting for this baseline difference highly misleading.This pattern could result from the decision (or inability) of a part of the frail cohort not to be vaccinated, see [2].Different non-COVID death rates indicate differences in the health status of the compared groups, known also as healthy vaccinee effect (HVE), see [3][4][5].HVE does not mean that the part of population opting for vaccination is "healthier on average" but it has certain characteristics that make them less likely to die.The presence of HVE can be demonstrated by showing significantly lower non-COVID mortality in a given subgroup than expected in the general population.The opposite (indication bias, when the frailest are preferentially vaccinated at the beginning of vaccination with the new dose) also plays a role, as can be seen in the 18-39 age group in September 2022.The effect of the interplay of these two biases on the interpretation of the comparison of death rates in groups defined by vaccination status is surprisingly large.