Misclassification Bias in Estimating Clinical Severity of SARS-CoV-2 Variants – Authors’ Response

We thank Christina Yek and her colleagues for their correspondence regarding our article.

They note that people who test positive for SARS-CoV-2 generally have more severe disease than those who are infected but not tested. This result could lead to an overestimate of absolute risks, but relative risks are not necessarily biased unless the proportion of severe cases detected differs systematically between variants. Citing modeling results indicating a decline in the detection rate of infections in the United States during the transition period between the predominance of delta (B.1.617.2) and omicron (B.1.1.529) variants, possibly in Because of the increased proportions of undetected infections in people with non-severe disease, Yek and colleagues hypothesize a mechanism of differential detection rates: the omicron cases for which a result of positive test has been recorded might have included a relatively higher proportion of infected people prone to severe disease than the analogous delta cases – for example, because a higher proportion of people infected with the omicron variant who sought to to be tested had a comorbidity.

However, the available data does not suggest a change in the proportion of infections detected in England by community PCR testing over the study period (although the extent of community testing was reduced later

). We believe the UK is unique in that it has conducted large-scale population-based COVID-19 prevalence surveys,

alongside its mass testing programs. To assess the hypothesis of Yek and colleagues, we compared estimates of the prevalence of infection in the population with estimates of the corresponding prevalence of infections detected by community testing (appendix pp 1–4). Contrary to hypothesis, we found that community testing detected similar proportions of people infected with the virus during delta-dominant and omicron-dominant periods in England (Appendix pp 2–3).

Yek and colleagues further argue that the relative risks of all-cause outcomes may be closer to zero than those of COVID-19-specific outcomes. They suggest that in the absence of direct measurement of COVID-19-specific outcomes, other data could indirectly discriminate between probable COVID-19-related and non-COVID-19-related events. In principle, we agree that cause-specific event data is desirable. However, assuming a constant background rate of unrelated hospitalizations and deaths, differential misclassification of outcome events by variant is unlikely, and non-differential misclassification is more likely to lead to a bias towards that to move away from zero. Additionally, we note that, during the study period, all people admitted to hospital in England were tested for COVID-19 on admission, so missed hospitalization events in people whose COVID-19 has not been detected are unlikely. Several studies that have reported relative risks of hospitalization specific to COVID-19 have estimated relative risks consistent with those of our study.

^,

We acknowledge that our dataset did not include comorbidity data. However, recent studies in other European countries with available comorbidity data reported only minor differences in comorbidity between delta and omicron cases, and provided comorbidity-adjusted relative risks consistent with those of our study.

^,

^,

One of these studies explored the effect of adjustment versus no adjustment for comorbidity and found only marginal differences.

TN and NMF contributed equally. The NMF declares research funding from the Bill & Melinda Gates Foundation and Gavi, the Vaccine Alliance, for research conducted within the Vaccine Impact Modeling Consortium. The NMF received consultancy fees from the World Bank Group for consultancy work, which ceased in 2019, on infectious disease threats. All other authors declare no competing interests. For information on funding and contributors, see appendix p 6.