Preprint Study: Omicron Is Just as Pathogenic But Not As Deadly Because of Other Factors

In an earlier blog, I summarized some information on why so many vaxxed people were getting infected and what immunity was doing to those at-home antigen texts.  In that post I said: "Omicron cousins are less likely to cause severe disease, even if they are more contagious."

In response to that @SuziSuperKitn on Twitter sent me this link to a Reuters article about a month ago, that reported on a preprint study looking at the question of whether Omicron in its various flavors is really less pathogenic.

Note that this is a preprint, which means it has still not been reviewed. It is not unusual for the peer review process to turn-up issues with the study methodology that invalidate the study or require modifications. But this one includes authors from Massachusetts General Hospital and Harvard Med School. Such researchers don't usually do shoddy studies so it's a good bet it does not have fatal flaws.

What they found, in a nutshell, is that if you just compare mortality rates of the Omicron wave with other waves, it indeed looks like it is inherently less deadly. But if you control for variables other than the variant that might also influence mortality the difference between waves disappears. 

Researchers call these "confounds," i.e. things the confound your ability to measure the thing you're really interested in, which is the pathogenicity of the variant. Confounds considered in this study included:

• Demographics: It may be that different kinds of people are being hospitalized now than those in earlier waves. For example is has been reported elsewhere that more younger people are being hospitalized in later waves than earlier ones. They are less likely overall to die even if hospitalized, because they are...well...not old.
• Comorbidities: It's been known from the start that you're more likely to die from COVID-19 if you have "comorbidities" (i.e. other conditions making death more likely), like diabetes and obesity. It may be that people being diagnosed with Omicron have fewer or different comorbidities that are making them less likely to die.
• Vaccinations: The vaccines have been shown to reduce the risk of death from an infection by as much as 90%. If more people were vaccinated during the Omicron wave, we would expect fewer deaths no matter how pathogenic Omicron is.
• Healthcare utilization: It may be that fewer people are accessing health care because of an infection than in previous waves.

The researchers had access to the records of 148,876 people who had positive PCR tests through the Mass General Brigham health care system from the start of the pandemic to present. These records included data on the four confounding variables just described.

They used this data to compare various waves for mortality risk while statistically controlling for the confounds. How you do that is too technical for this blog, but essentially there are statistical models that allow you to tease out separate effects of a set of variables (like variant and confounds) on an outcome (like risk of death).

The study found "that after accounting for confounders, the Omicron variant was as deadly as the previous SARS-CoV-2 waves." In other words, once you control for factors other than variant that can also affect death rates, Omicron is just as pathogenic as the other variants.

According to the authors this is important information because:

Any comparison between SARS-CoV-2 variants without adequately adjusting and controlling for important confounders that may change over time such as vaccination status and healthcare utilization, can mislead both the public and medical experts of the true danger of the variant. It could also lead to mistrust among the public and poor choices by health policy experts.

My only criticism of the paper is that the did not give effect sizes, which they could have done because they used a causal model.* Effect sizes are estimates of the importance of the different variables of the model in accounting for the outcome. If I were reviewing this article I'd require a revision to include them.

It would be good to know the relative importance of the four confounds. My guess is that the biggest effect was for vaccination—especially in Massachusetts, which has one of the highest vaccination rates in the country. This information could help public health officials better see the importance of vaccination campaigns.

*They do give odds ratios for mortality between variants and these can be converted to effect sizes; however, they do not give similar information for the confounding variables themselves.