We’ve noticed something that looks like a potentially important data-entry problem regarding comorbidities in the “big” China CDC report on the Epidemiological Characteristics of COVID19. This report uses data from 72,314 patient records. (1/6)
The China CDC report has the most widely used numbers I’ve seen for age and sex fatality rates. It was reported in brief form in JAMA, and fully in a China CDC weekly (link below, it loads slowly sometimes).
weekly.chinacdc.cn/en/article/id/…
(2/6)
weekly.chinacdc.cn/en/article/id/…
(2/6)
See numbers here.
From the report:
“† The comorbid condition variable, only includes a total of 20,812 patients and 504 deaths and these values were used to calculate percentages in the confirmed cases and deaths columns.”
(3/6)
From the report:
“† The comorbid condition variable, only includes a total of 20,812 patients and 504 deaths and these values were used to calculate percentages in the confirmed cases and deaths columns.”
(3/6)
Summing up the numbers for each of the 5 conditions plus “none” (which I take to mean none of these 5) adds up to exactly 20,812, whereas we should expect double counting.
Add 23,690 reported as missing = 44,502, which is 170 less than the 44,672 total confirmed cases
(4/6)
Add 23,690 reported as missing = 44,502, which is 170 less than the 44,672 total confirmed cases
(4/6)
Does anyone have any insight into this? Or access to the data behind the report? I am worried that this might compromise the per-comorbidity fatality rates. I hope there is some simple fix that we have missed.
(5/6)
(5/6)
And this is the JAMA paper, which is a condensed version of the full China CDC report. It doesn't have the full breakdown of comorbidities, but does report the per-comorbidity fatality rates.
(6/6)
jamanetwork.com/journals/jama/…
(6/6)
jamanetwork.com/journals/jama/…
@SanjatKanjilal @EpiEllie @EpidByDesign @statsepi @_MiguelHernan @GSCollins @EpidByDesign @jd_wilko
(tagging in case you have an idea or know someone who knows someone with better access to the numbers behind this report)
(tagging in case you have an idea or know someone who knows someone with better access to the numbers behind this report)
Just to be clear: the reason the number of patients with each comorbidity and with no comorbidities should *not* add up to the total number of patients is because a large proportion of patients are expected to have 2 or more comorbidities: e.g. CVD+diabetes, diabetes+hypertension
The least concerning solution to this I can think of would be if the reported sum of patient with known comorbid condition was wrongly calculated by summing up the categories for the five conditions + “none”. Then the fatality rates per condition are still correct.
