Ethnic disparities in COVID-19: increased risk of infection or severe disease? - Authors' reply.

We thank Daniel Pan and colleagues for raising an important point regarding our cohort study. We agree that separating the risk of infection from the risk of severe disease among those infected is a key issue in understanding and tackling ethnic disparities in COVID-19.

Separating these risks in routine electronic health data sources is challenging. During the first wave of the pandemic in the UK, testing for SARS-CoV-2 was not random, with prioritisation of health-care workers and those admitted to hospital. Although testing was more common during the second wave, those receiving a test were still not entirely representative of the population infected with SARS-CoV-2.

Indeed, in our own study, we found that minority ethnic groups were less likely to be tested for SARS-CoV-2 than White groups were in the second wave. Therefore restricting our analyses of severe COVID-19 outcomes to those who tested positive for SARS-CoV-2 could have introduced bias and limited interpretation of the resulting estimates.

To generate unbiased estimates of ethnic differences in COVID-19 severity among those infected, we would have needed to be confident that uptake of testing in relation to infection prevalence was equivalent between ethnic groups at the point of analysis. Alternatively, we would have needed to adjust for all factors associated with testing positive for SARS-CoV-2, including occupation, which is not captured in routine electronic health data.

Our study provides other clues that ethnic differences in severe COVID-19 outcomes might be driven by exposure risk, including the attenuation of risk on adjustment for household size and the rapid change in ethnic patterning of outcomes from the first wave to the second wave, which would be unlikely if ethnic differences were more related to the effects of underlying susceptibility than to increased exposure.

There are now a number of population-based cohorts in the UK, within which it might be possible to reliably estimate COVID-19 severity in a representative sample of people with known infection. These include the REACT-2 study, which has already reported higher levels of SARS-CoV-2 antibodies in minority ethnic groups but no ethnic differences in infection-to-mortality ratios. Studies from the Office for National Statistics and the UK Biobank highlighting the role of occupational exposure in COVID-19 mortality further support the hypothesis that ethnic differences in severe COVID-19 outcomes might be related to differences in exposure risk rather than other explanations, such as biological differences or health-care related factors.6, 7

We look forward to findings from ongoing prospective cohort studies, which will build upon existing insights from sociological and community engagement work to disentangle the complex interactions between ethnicity, social disadvantage, occupational and household factors, and access to health care in explaining health inequalities related to COVID-19 and beyond.

RM reports research funding from the Wellcome Trust and personal fees from AMGEN, unrelated to this Correspondence. All other authors declare no competing interests.