The cdc published

Severity of Disease Among Adults Hospitalized with Laboratory-Confirmed COVID-19 Before and During the Period of SARS-CoV-2 B.1.617.2 (Delta) Predominance — COVID-NET, 14 States, January–August 2021

i would clear up some data from the table

TABLE. Demographic characteristics and clinical interventions and outcomes among 7,615 nonpregnant adults aged ≥18 years hospitalized with COVID-19,* by vaccination status† and period relative to SARS-CoV-2 B.1.617.2 (Delta) variant predominance

| Weighted % of COVID-19 hospitalizations (95% CI)
|                | Unvaccinated                                     |
| TOTAL          | 4,896            | 1,145                         |
| ICU admission  | Pre Delta period | Delta period       | p-value  |
| ≥18            | 20.1 (18.3–21.9) | 22.6 (19.1–26.3)   | > 0.99   |
|                | Fully vaccinated                                 |
| TOTAL          | 389              | 393                           |
| ICU admission  | Pre Delta period | Delta period       | p-value  |
| ≥18            | 19.9 (14.2–26.6) | 24.6 (18.2–32.0)   | > 0.99   |

My understanding is that

  • based on the wide (95% CI) the sample varies and is probably small
  • based on the high p-value the findings are not signifikant, could be chance, are of low confidence

Regarding these footnotes

* Data are from a weighted sample of hospitalized nonpregnant adults with completed medical record abstractions and a discharge disposition. Sample sizes presented are unweighted with weighted percentages.

** Total hospitalizations include data from selected counties in all 14 COVID-NET states with vaccination status, including fully vaccinated, partially vaccinated, and unvaccinated adults. As a result, the number of total hospitalizations exceeds the sum of fully vaccinated and unvaccinated adults.

The second footnote ** is clear. what is unclear the meaning of the first *

  1. Is my understanding above regarding sample variation, ci and p-value correct?

  2. What numbers are weighted and which are not?

  3. How are they weighted?

  4. Does 20.1 of unvaccinated ≥18 mean that 20.1% of all unvaccinated hospitalizations did require ICU admission? Would this be n ~ 230 ~ (1,145 * 20.1%)

  5. solved i hope: Does the p-value refer to a line? So pre-delta and delta? It seems that this is the case. My understanding of the footnote †† is that they compared delta with pre-delta and found no significant difference within a group (vaccinated pre vs delta, and unvaccinated pre vs delta)

  6. solved: Which values are compared for the p-value? (see above)

  7. What gives 100%?

Update 1 - regarding p-value

I overlooked a footnote regarding the meaning of p-value††

†† Proportions between the pre-Delta and Delta period were compared with chi-square tests; p-values <0.05 were considered statistically significant, adjusted for multiple comparisons using the Bonferroni correction method

Update 2 - wrong numbers corrected

After the first reply i corrected the number of hospitalized for fully vaccinated

1 Answer 1


You have the data a little wrong. Here's a screenshot of the figures:


From this you can see that your figures show the total hospitalizations over the period, rather than the fully vaccinated. So the total hospitalizations for the fully vaccinated are 389 and 393 respectively for pre-delta and delta (that's interesting of itself, but not relevant here).

You have the data correct for the ICU admission, ≥ 18 category:


  1. Correct - you can't say that the populations are different, however, incorrect on sample size. The general way to phrase this is:

If, in the population from which this sample was drawn, there was no effect, how likely is it that we would get, in a sample of this size, get a test statistic this large or larger

Based on a p-value of >0.99, there is no difference in these populations, however, the sample sizes are probably large enough to see some variation if it existed. Generally (at least when I was taught) you need at least 30 samples for statistical validity - they have many more than this. However, this depends on what sort of effect they are trying to find. It could still be that the sample sizes are too small, particularly in the fully vaccinated when broken down into age groups.

  1. The sample sizes are not weighted - i.e. they presented total numbers (as you might expect). They did weight the percentage in each group (age, sex, etc) and the CI correspondingly.

  2. I don't know for sure, but based on this line in the text:

Unadjusted age-specific monthly population-based hospitalization rates (hospitalizations per 100,000 persons) among all adults aged ≥18 years irrespective of pregnancy status during January–August 2021 were calculated by dividing the total number of hospitalized COVID-19 patients by population estimates within each age group in the surveillance catchment area.¶

It sounds like weighted relative to the estimated age proportion in the general population.

  1. Yes, but you can't calculate the numbers like that, you would have to adjust for the age weighting in the calculation, but we aren't given the age weighting in the data provided.

  2. Yes, comparing rates pre- and post-delta. They used a χ-square test for all groups in the corresponding cells, hence the single value for some of the groups (e.g. ethnicity). Where they have looked closer at the differences they have provided tests for each sub-group (e.g. comparing age 18-49 with other 18-49, but not to 50-64).

  3. see 5.

  4. Nothing. Because of the age weighting in applying the percentages you can't simply sum the percentages for each age group to reach 100% of the total population. This is why the calculation in 4. doesn't work.

  • Regarding <18 versus ≥18 i believe that they do mean all above 18 years of age based on a population-based surveillance system for COVID-19–associated hospitalizations, were used to examine trends in severe outcomes in adults aged ≥18 years hospitalized with laboratory-confirmed COVID-19
    – surfmuggle
    Commented Nov 23, 2021 at 22:24
  • Thanks for your reply. I fixed the wrong numbers and reverted the order of questions so that they match to the state they were in at the time you were answering. Regarding: #4, total number of icu admissions and #7 100%. Thanks Why do you think that hospitalization fully vaccinated pre- and post delta 389 and 393 is interesting of itself? Are they to close together to be of chance? May be a typo?
    – surfmuggle
    Commented Nov 23, 2021 at 22:33
  • 1
    @surfmuggle Heh - reading failure on my part, should have looked at the title. The closeness of the numbers pre and post delta are interesting because of the disparity in the numbers for the unvaccinated group. Implies that the vaccine is less effective against delta, however it could also be an effect of shorter time for delta, so the true numbers aren't known yet, and also that the vaccines haven't been in place for all that long, so the pre- numbers are lower than otherwise might be.
    – bob1
    Commented Nov 23, 2021 at 23:20
  • Since ≥18 means all participants and total delta are 1,145 Person. Does that mean that the ICU admission value 22.6 (19.1–26.3) can be used to calculate the number of people that went to ICU? Does 22.6 mean 22.6% of 1145 Persons went to ICU?
    – surfmuggle
    Commented Nov 24, 2021 at 12:08
  • 1
    @surfmuggle that percentage looks about right for the statistics on those admitted to hospital with COVID, somewhere between 20-30% enter ICU. There are plenty of papers with similar statistics out there. I would say, yes you could probably use that number for a raw percentage conversion, but can't say for sure.
    – bob1
    Commented Nov 24, 2021 at 20:24

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.