RR:C19 Evidence Scale rating by reviewer:
Reliable. The main study claims are generally justified by its methods and data. The results and conclusions are likely to be similar to the hypothetical ideal study. There are some minor caveats or limitations, but they would/do not change the major claims of the study. The study provides sufficient strength of evidence on its own that its main claims should be considered actionable, with some room for future revision.
Paredes et al. have analyzed the impact of SARS-CoV-2 variants of concern (VOC) on the risk of hospitalization, using a dataset of viral sequences linked to medical records in over 20,000 cases in Washington State. A significant strength of this study is that the main analysis used specimens randomly selected for sequencing through a sentinel surveillance program rather than those specifically chosen for sequencing for clinical reasons or for targeted outbreak investigations. This means that the results of the analysis are more likely to reflect the impact of the VOCs on hospitalisation risk in the general population.
The estimates in this article are wholly consistent with previous published values for the impact of the Alpha and Delta VOCs on hospitalization risk in unvaccinated individuals. The article also provides estimates of the elevated risk of hospitalization for other less common VOCs, for which there exists less high-quality evidence in the literature. The protective effect of vaccination is also estimated separately for each VOC, although this is limited due to a dichotomous comparison based on initial vaccine doses rather than giving a full breakdown by vaccine types and dose number and/or intervals for 2-dose regimens.
Establishing causal effects of factors of interest from retrospective observational data is always challenging, but the methodology used for statistical analysis is appropriate and broadly in line with previous peer-reviewed papers on the impact of VOCs. The authors have adjusted their analyses for age and sex, which are important predictors of clinical outcomes following SARS-CoV-2 infection. Ideally, the adjustment would also include pre-existing co-morbidities for each person, in case these were associated with the risk of infection with a VOC. Most similar analyses have presented results with adjustment for calendar time, to account for the potential impact of seasonal variations and healthcare pressures on outcomes. The authors make the argument that they prefer not to adjust for calendar time – which I think would be challenged by some other research groups – but they do also present such analyses, without a substantial impact on their findings. The discussion section appropriately acknowledges limitations of the dataset and the analysis such as the potential impact of vaccination on testing behaviour.
I am a statistician. After reading the current draft of the article, I am left with some technical queries that I think would ideally be addressed before final publication of the work. These include clarifying the exact structure used for the mixed effects models in analyzing the definitions of zero timepoints and any potential censoring conditions for time-to-event outcomes. However, any such clarifications are unlikely to have a major impact on the text or the conclusions of the article.
Claims are reliable by the data and methods used.