the following thread is purely my personal opinion as a @CCDD_HSPH @HarvardEpi epidemiologist. I would welcome feedback though I am trying to take some breaks so don't promise to reply. Thanks in advance to anyone who provides critique or further info
Impact of #Omicron obviously depends on severity. Number of cases is growing exceptionally fast and several countries have seen such growth continue for quite a while (weeks, which is a lot of time with doubling times of small number of days
AFAIK there are two publicly available assessments of severity. Kudos to @Imperial_IDE and Discovery Health for rapidly putting out reports… and… (site down as I write, but previously had a link to a detailed webinar
Below queries about what was done are not criticisms of these groups who got info out fast, but hopes for more clarity about how to interpret their findings.
SA Discovery finds in adults a 29% lower case-hospitalization ratio and a 77% lower hospitalization-ICU ratio, for (I think) an 82% lower case-ICU ratio for Omicron vs Delta.
Slightly higher case-hospital ratio for children w Omicron vs Delta. Not sure if the CIR can be calculated from the hospitalization and ICU/Hospitalization ratio because I don't understand the risk adjustment algorithm. Could be very important to understand how that was done.
But ~80% reduction in ICU risk per case would be very good news, if correct (see caveats below for issues that may contribute to bias)
2. Imperial finds in logistic regression for that hospitalization has an OR of 0.95 (0.61-1.47) for predicting SGTF, a proxy for Omicron. I cant tell if this is multivariable regression, adjusting for everything, but almost any single approach would be hard to interpret.
Would love to understand better what they did. Notably the point estimate of 0.71 relative risk of hospitalization from SA is within their 95% CI for the OR in England. Both analyses taken at face value suggest at most a modest reduction in hospitalization risk.
(though unfortunately no uncertainty estimates in the South Africa analysis). A modest reduction in hospitalization risk could be swamped by severalfold more cases as several models predict.
Complicating analysis could be
1) hospitalization may not have occurred yet, preferentially in the Omicrons, making the results too optimistic (applies to both analyses)
2) If SA is correct, the real reduction in severity is at the highest-severity end, which is not considered in the Imperial report (i.e. ICU)
3) Both may (I suspect do, at least for England) include those admitted for non-COVID causes and thus make hospitalization rates look higher (because Omicron is contagious enough that many people in hospital have it unrelated to their admission)
Including nonCOVID admissions who test positive for SARS-CoV-2 would make severity among hospitalized look lower as suggested by this account from SA… .
Bottom line we are far from having evidence that severity is low enough to make this not a worry at the societal level, especially with already-stretched hospitals. Evidence that vaccines, especially when boosted, offer protection against severe disease is more consistent.
Already learning from the responses, thank you. Two sets of data suggesting modest contribution of "incidental" SARS-CoV-2 diagnoses among hospitalizations, one in Tshwane, Gauteng, SA
which shares this . A constant proportion of hospital admissions over time that are incidental SARS-2 findings vs. admitted for COVID is indeed consistent with similar severity per infection to the past variants, as @chrischirp notes.
This presentation from yesterday is illuminating ht to a responder whom I can't find any more, twitter fail...
Several lines of evidence here suggesting a lower hospitalization proportion than in previous waves, limited to first 25 days. Caveats 1) this is growing faster, so higher proportion of recent cases makes bias (missing not-yet-hospitalized) worse even when compare first 25 days
2) the proportion vaccinated is going up, as is the proportion previously infected, so a more immune population. So hard to compare but declining in-hospital severity measures here as in Discovery Health are hopeful signs.

• • •

Missing some Tweet in this thread? You can try to force a refresh

Keep Current with Marc Lipsitch

Marc Lipsitch Profile picture

Stay in touch and get notified when new unrolls are available from this author!

Read all threads

This Thread may be Removed Anytime!


Twitter may remove this content at anytime! Save it as PDF for later use!

Try unrolling a thread yourself!

how to unroll video
  1. Follow @ThreadReaderApp to mention us!

  2. From a Twitter thread mention us with a keyword "unroll"
@threadreaderapp unroll

Practice here first or read more on our help page!

More from @mlipsitch

6 Dec
Clear explainer of a very carefully nuanced report. While the main thing to say is "small numbers," it does seem that the average severity among those in the hospital is comparatively low. Two thoughts about interpretation:
1/ Omicron growing very fast. Patients hospitalized for COVID are typically (not always) well into their infection. In a fast-growing epidemic, the proportion infected 10d ago, say, is lower on any given day than in a slow-growing one. This alone could cause unusually lo ...
proportion of hospitalizations with primary diagnosis. Fast-growing epidemic = most infections are new is basic demographic theory (just like in a fast-growing population of people, most people are young). The unusually high growth rate can at least partly explain observation.
Read 7 tweets
18 Aug
@CT_Bergstrom @jsm2334 Certainly this kind of bias merits consideration. I think the particular figure cited in that table is an example of Simpson's paradox, which is a special type of confounding.
@CT_Bergstrom @jsm2334 For those new to these terms, confounding is just the problem that (in this case) vaccine is not randomly distributed in the population, so the vaccinated have different risks from the unvaccinated for reasons other than their vaccine: in this case, age.
@CT_Bergstrom @jsm2334 Simpson's paradox is an extreme form of confounding in which a combined analysis for two groups of people gives an unusually misleading estimate, relative to the (more) accurate estimate for each group individually.
Read 17 tweets
13 Aug
As often happens, the headline is too simple for the more subtle message of the article…
The article's main point is unarguable: "we cannot control the delta variant by maximizing the immunity of only a segment of the population."
But some (incl me) have thought the main goal of vaccination all along should be to defang (make less harmful) not defeat (eliminate) the virus.
Read 7 tweets
20 Jul
At the risk of boiling down too much and certainly losing some detail, one way to summarize this wonderful thread is that when we think about vaccine effectiveness, we should think of 4 key variables: 1 which vaccine, 2 age of the person, 3 how long after vax, 4 vs what outcome.
We've been using the simple view that the major vaccines in use in the US/Europe are possibly less effective against infection/symptoms when a variant is involved, but remain highly effective against severe outcomes. Published data so far support this view.
To be more precise, we would say "so far in the general population, up to about 6 months after vaccination, the vaccines have held up against severe outcomes even from Delta, though there is some evidence from Israel, UK, and Canada of declines in effectiveness vs infection."
Read 14 tweets
7 Jul
Different approach from many other VE studies, following HCW vaccinated vs unvaccinated, tested when exposed to a case, to assess VE against infection given exposure, consistent with our recommendations in…
Also looked at infectiousness (proxied by Ct). Take home messages: fully vaccinated 65% (45-79) protected against infection given exposure. This is lower than other estimates of symptomatic or arbitrary mix of symptomatic and other cases, as expected.
Read 8 tweets
13 Jun
Now out with @rebeccajk13 Interpreting vaccine efficacy trial results for infection and transmission…
funded in part by #SeroNet
In which we show that earlier work by Rinta-Kokko et al on interpreting prevalence measures for vaccine efficacy generalizes to the COVID-19 case and that the odds ratio for PCR+ in vax vs unvax persons swabbed at random
is under reasonable assumptions a lower bound on the vaccine's effect against transmission, the critical quantity for herd immunity that combines reduced risk of acquiring and shorter duration.
Read 11 tweets

Did Thread Reader help you today?

Support us! We are indie developers!

This site is made by just two indie developers on a laptop doing marketing, support and development! Read more about the story.

Become a Premium Member ($3/month or $30/year) and get exclusive features!

Become Premium

Too expensive? Make a small donation by buying us coffee ($5) or help with server cost ($10)

Donate via Paypal

Or Donate anonymously using crypto!


0xfe58350B80634f60Fa6Dc149a72b4DFbc17D341E copy


3ATGMxNzCUFzxpMCHL5sWSt4DVtS8UqXpi copy

Thank you for your support!

Follow Us on Twitter!