Fact checker Repustar, represented on Twitter as “FactSparrow,” published a fact brief that (we believed) confused hospital admissions for coronavirus with coronavirus hospitalizations, though the CDC reports very different statistics for each.
After seeing no movement on the issue based on Twitter outreach, Zebra Fact Check emailed Repustar to encourage a correction to the misinformation (we believed we detected).
Update Sept. 16, 2021:
On Sept. 14, 2021, Repustar Editorial Director John Marcom responded to our request with an email showing we based our correction request on a wrong interpretation of CDC data. Find that email, with our response, in the document embedded after the original correction request. here.
In response [to] Marcom’s email, we have updated this item, added parenthetical corrections to the original post in bold above, and published a correction announcement with the details of the correction.
Of note, our response suggests a different reason why the Repustar fact check may mislead its audience. CDC appears to count any hospitalized patient as a patient hospitalized with coronavirus if that patient tests positive for the virus. That method of defining hospitalized coronavirus patients would appear to count cases where the presence of the positive test was incidental to the hospitalization.
If Repustar responds to that point, we will continue to update this page.
Dear Repustar/FactSparrow,
Your Sept. 11, 2021 fact brief purporting to answer a question about covid-related hospitalizations for children 18 and under overlooked the CDC’s distinction between hospital admissions and hospitalizations.
The fact brief answers a question about hospitalizations in terms of hospital admissions. So the brief’s information on hospital admissions, taken more-or-less directly from the CDC website, was accurate. But it wasn’t responsive to the question.
Here’s how the fact brief presented the question:
Have about 2,500 children been hospitalized with COVID-19 in each recent week in the US?
The fact brief asserts that the answer to the question is “yes.” Here’s the opening paragraph, exemplifying the brief’s focus on hospital admissions as a measure of hospitalizations:
According to Centers for Disease Control data, in the week ending Sept. 9, 2021, new hospital admissions of COVID-
19 patients under 18 years of age averaged 350 a day, close to a pace of 2,500 weekly or 10,000 a month.
But the CDC website keeps statistics for hospitalizations separate from statistics for hospital admissions, apparently distinguishing between the two in a way that remains unacknowledged in the Repustar fact brief.
Your fact brief relied on information from this URL:
https://covid.cdc.gov/covid-data-tracker/#new-hospital-admissions
Here’s the URL that contradicts FactSparrow’s answer to the question:
https://covid.cdc.gov/covid-data-tracker/#covidnet-hospitalization-network
It’s the same cdc.gov website. The second URL puts a number on the cumulative hospitalizations of children aged 0-17.
It’s 3,899. That’s the total number since the pandemic began, 2020 through 2021. That number cannot support the Repustar claim of 2,500 hospitalizations of children in each recent week. Two weeks at that rate would eclipse the cumulative total since the start of the pandemic.
The peak number of hospitalizations for one week? One hundred twenty-one. That’s well short of 2,500.
Everyone makes mistakes. It is the hallmark of conscientious journalism (and fact-checking) to correct promptly, completely and transparently. Transparency is supposed to encompass the attempt to make sure that those who read the incorrect information see the correction (that can look like this).
Hi Mr. White,
Thanks so much for writing and your close attention to our work.
The URL you shared that appears to contradict the other CDC page is measuring a subset of total data, with a view of establishing a rate of hospitalizations. The data on this is in general clearly imperfect, but we opted to use the CDC data based on the full reporting set.
As the top line on the page you shared with us says, the raw data is “laboratory-confirmed COVID-19-associated hospitalizations among children and adults through a network of over 250 acute-care hospitals in 14 states” so in our understanding and that of our researcher it is not complete as a total national number of hospitalizations.
The CDC further explains this on a linked page:
“COVID-NET differs from hospitalizations reported in national and state case counts in two ways. First, state and national COVID-19 case reporting are based on all people who test positive for COVID-19 in the United States. COVID-NET is
limited to COVID-19-associated hospitalizations captured in the COVID-NET surveillance area. Second, COVID-NET reports rates and not just counts. These rates show how many people are hospitalized with COVID-19 in the surveillance area, compared to the entire number of people residing in that area.”
For the Brief, it seems to be appropriate to draw on the other data, which is based on reporting from across the US.
As corroboration we considered a further source, cited in the Brief, the American Academy of Pediatrics. They present some base estimates based on data from 24 states + New York City. Noting that about 243,000 new cases of positive tests for children were added in a week would be consistent with a rate of hospitalization of just over 1%. If you consult Table2B on this page, you’ll see that they estimate a hospitalization rate of 0.9%, which would be about 2,200 hospitalizations. Not quite 2,500 but close.
Finally - which we did not have space to consider in the brief - you can consider the rate cited in the CDC report you are focusing on, which peaked at 1.7/100,000 for the under 18 age group. Per census data 73.1 million people are under 18.
That does indeed provide a different estimate of 1,243. But this is by its own admission a limited data set, so amid imperfect choices it seems like the CDC topline reports, being the most complete nationwide, remain for now the best first place to consider.
Thanks again for your thoughtful attention,
Dear John Marcom,
I greatly appreciate you taking the time to address my concern. I agree that the fine-print caveats in the COVID-NET data allay that concern.
While verifying the information in your reply, however, I have developed a new concern. I’m looking for the methodologythe site uses to identify covid-related hospitalizations, and the best description I’ve found so far is “Hospitalization rates are calculated by the number of residents of a defined area who are hospitalized with a positive SARS-CoV-2 laboratorytest divided by the total population within that defined area.” If, as it appears, “covid-related” hospitalization represents any hospitalized patient who happened to test positive for covid, the Repustar answer would still count as misleading.
The natural inference from “Hospitalized with COVID-19” is that covid infection served as the primary reason for hospital admission. So far I’ve found no solid evidence that would support that interpretation of the CDC data.
If the CDC information does not address that aspect of the question, the fact brief should mention that weakness in the data.
Again, thanks for taking the time to defend Repustar’s fact brief. That’s an action that too few take in the fact-checking business.