Read on Twitter
THREAD: Lots of alarm over the 
88% Mortalityamong #Covid19 patients on ventilators. The paper *should* say "Among pts on vents, mortality will range from 25-97%". This @JAMA_current study is a great for a JournalClub on "Beware Your Denominator"
(1/7)
https://jamanetwork.com/journals/jama/fullarticle/2765184 https://twitter.com/DrEricDing/status/1253232915158310912
88% Mortalityamong #Covid19 patients on ventilators. The paper *should* say "Among pts on vents, mortality will range from 25-97%". This @JAMA_current study is a great for a JournalClub on "Beware Your Denominator"(1/7)https://jamanetwork.com/journals/jama/fullarticle/2765184 https://twitter.com/DrEricDing/status/1253232915158310912
Now, I know that the abstract say "Mortality for those requiring mechanical ventilation was 88.1%." And I know it's @JAMA_current, so it has to be perfect, right? But its very, VERY misleading in the headlines on @CNN @business @medpagetoday & tweets by MDs, here's why...(2/7)
Here's the key: Median follow-up time to death or discharge = 4.4d (IQR 2.2-9.3), median follow-up at time of censoring = 4.5d (2.4-8.1). That's simply not enough time to say something useful about ventilated pt w/ #Covid19. More than 72% of their sample was excluded. Look(3/x)
(3/x)
Look at Table 5. First it's amazing to see a paper w/ >1,100 vented #Covid19 pts. But of these 1151 vented pts, *at a median follow-up time of 4.5d* only 38 went home, 282 died, while 831 were still hospitalized. 282/(38+282) = 88% mortality. But thats *quite* misleading(4/x)
(4/x)
A better conclusion?
- "Among the 1,151 pts who were mechanically ventilated, at a median follow up of 4.5 days, 3% had gone home, 25% had died, & 72% were still in the hospital".
- "The eventual mortaily rate will thus range between 25% (282/1151) and 97% (1113/1151)."(5/7)
- "Among the 1,151 pts who were mechanically ventilated, at a median follow up of 4.5 days, 3% had gone home, 25% had died, & 72% were still in the hospital".
- "The eventual mortaily rate will thus range between 25% (282/1151) and 97% (1113/1151)."
(5/7)
Still alarming, sure. But to put those numbers in context, let's remember that good ole fashioned ARDS has pretty frightening mortality. For mild, mod, severe ARDS the mortality is around 25%, 35%, 45%. Its likely *much* higher in ARDS from #Covid19, but we don't know yet.(6/7)
(6/7)
The rapid publication cycle of #Covid19 = lots of data misinterpretation. The @JAMA_current paper put all of the data out there. But that eye-popping 88% # in the abstract is too easy to misinterpret. The #s are scary, but they should be presented & reported more clearly (7/7)
@threadreaderapp unroll
Link to actual paper fell off: https://jamanetwork.com/journals/jama/fullarticle/2765184?guestAccessKey=6ad3f237-3962-483e-9a30-81ef5ad2c40f&utm_source=silverchair&utm_medium=email&utm_campaign=article_alert-jama&utm_content=olf&utm_term=042220
A clarification that others have pointed out (thx peers). Not sure # specific to vents but:
- LOS for pts dc'ed alive home: 3.9d (2.4-6.7)
- LOS for pts who died: 4.8d (2.3-7.4)
- LOS for pts inhospital at study endpoint: 4.5d (2.4-8.1)
- median post-dc follow-up: 4.4d (2.2-9.3)
- LOS for pts dc'ed alive home: 3.9d (2.4-6.7)
- LOS for pts who died: 4.8d (2.3-7.4)
- LOS for pts inhospital at study endpoint: 4.5d (2.4-8.1)
- median post-dc follow-up: 4.4d (2.2-9.3)
Also, to be very clear, the authors are forthwright about the risk of bias in the limitations. But writers of newspaper headlines dont always read the limitations. "3% discharged, 25% dead, & 72% still inpatient" is more representative. Most ARDS studies report 28-60d mortality.
