There also isn't really any trend here - a nonsignificant drop from 23% to 21% over two months, and then a sudden halving of the death rate in just one week, which corresponds to a sudden decrease in the average age
-
Show this thread
-
Replying to @GidMK
Why are you ignoring the previous week w/ very low adjusted mortality (10%) & older patients (mean age 61), or 18% 3 wks earlier? A simple sample size weighted regression supports their claim of decreasing mortality (P=0.02). I don't think findings can be dismissed so quickly.
1 reply 0 retweets 1 like -
Replying to @DiseaseEcology @GidMK
One can question their model for adjustment (details not provided in paper) but can't dismiss data out of hand when it's actually 3 low weeks suggesting lower mortality.
1 reply 0 retweets 0 likes -
Replying to @DiseaseEcology @GidMK
(FYI: an unweighted regression is even more significant; P =0.007). Not the way I'd actually analyze the data, but with what's provided that's what evidence suggests.
1 reply 0 retweets 0 likes -
Replying to @DiseaseEcology
Not trying to dismiss it out of hand, but I think the broader point that I made holds true - the death rate dropped precipitously from ~20% to ~10% in one week, coinciding with a massive drop in the number of patients and changes in their characteristics
2 replies 0 retweets 0 likes -
Replying to @GidMK @DiseaseEcology
The most likely explanation, to me, would seem to be that the demographic changes were the major driver of this difference, both factors that could be adjusted for and those that can't/are unknown
1 reply 0 retweets 1 like -
Replying to @GidMK
The adjusted mortality adjusts for all the major factors shown to be important in mortality. Which ones did they miss that would make more than a tiny difference?pic.twitter.com/izNMDYZGHT
3 replies 0 retweets 2 likes -
Replying to @DiseaseEcology
Duration of symptoms would be important. Some measure of severity of symptoms based on either clinician or patient report. To be fair, they might not have included more covariates because they were using a logistic model and they only had so many outcomes
2 replies 1 retweet 0 likes -
Replying to @GidMK
I'm not sure how you could use symptom severity as a predictor of death. Wouldn't that be circular? BTW, they included O2 saturation on admission & they included dozens of predictors in their model so I don't think # of predictors was issue.pic.twitter.com/PnlGWOx086
1 reply 0 retweets 0 likes -
Replying to @DiseaseEcology
Right, but it's not best practice to use a logistic model when you have relatively few outcomes and so many predictors. And symptoms would be important because they impact admissions
1 reply 1 retweet 0 likes
Indeed, while they haven't given the confidence intervals for the logistic model, I'd expect that the last few weeks would be impressively wide
-
-
Replying to @GidMK @DiseaseEcology
So the tl;dr answer is maybe it's less deadly, but it can't be known definitively with this data?
0 replies 0 retweets 0 likesThanks. Twitter will use this to make your timeline better. UndoUndo
-
Loading seems to be taking a while.
Twitter may be over capacity or experiencing a momentary hiccup. Try again or visit Twitter Status for more information.