Now online @bmj_latest! #COVID19 outcomes & risks @nyulangone among 5,279 patients (2,741 hospitalized). Thread with details and some special twitter bonus content. @petrillimd @jones_prof @WhiteCoatLuke @Cerf_MD @Francois1Fritz bmj.com/content/369/bm…
Those of you who saw our previous preprint: this version is substantially altered – updated data, comprehensive follow up, new analyses. E.g., 94% of the hospitalized patients have been discharged or died, giving us robust estimates of final outcomes. Some highlights follow.
Context: we include all patients testing positive for #COVID19 between March 1 and April 8, with follow up (in the paper) through May 5 (and here in this thread, through this morning!). We restrict to those positive before April 9 to ensure adequate follow up for outcomes.
1st, outcomes for hospitalized patients: 24% died or went to hospice, 6% still hospitalized. (Bonus Twitter update: now at 25% died/hospice, 3% still hospitalized.) 647 (24%) were ventilated, of whom 60% died/hospice, 16% discharged, 23% still hospitalized (now 63%, 24%, 13%).
Now for factors associated with adverse outcomes. First, hospitalization. Biggest risk is still age. We now show average marginal effects to make ORs more interpretable: e.g. risk of hospitalization for >=75 years is 58 pct points higher than those 19-44. h/t @kit_delgadoMD 
Other key risks heart failure (22pct points increased risk), male (16pp), chronic kidney disease (14pp), any increase in BMI (14pp for BMI 40+).
As previously shown current/former smoker “protective” BUT unknown smoking status 5pp increased risk. *I do not think smoking is good for you.* Other possible explanations: 1) smokers more likely to be tested b/c perceived higher risk; 2) smokers don’t want to admit it = unknown.
What about critical illness (ICU, ventilator, death) & death? As shown in preprint, additional risk age/comorbidity smaller once hospitalized though still present for age, unknown smoking status, BMI 40+, heart failure. Here's cumulative hazard of death by age group.
Special bonus Twitter-only graphic: unadjusted outcomes among the subset of ventilated patients, by age.
And still true: hypoxia on presentation and inflammatory labs bigger risks for critical illness than comorbidity among hospitalized group.
*New results* But this result is my favorite: adjusted risk of critical illness decreases every week. Are we getting better at taking care of these patients even without blockbuster drug, definitive treatments? I think we may be.
*New analyses* we have added a competing risk survival model for mortality to this paper (competing risk=discharge, since we don’t know if die post-discharge). Here you can see the dramatic differences in mortality by O2 sat and D-dimer on admission.
And here is risk of mortality and discharge by CRP and lymphocyte count
Finally, we do a number of sensitivity analyses (i.e. restricting non-hospitalized comparison group to those tested in ED; adding in ED patients with suspected covid but not tested to non-hospitalized groups) – similar results.
Thanks as always to extraordinary team here @nyulangone, to #medtwitter for feedback on preprint & to the thoughtful reviewers who inspired many improvements. BMJ has open peer review so you will be able to see their comments and names posted online with the paper.
And to @davidludwigmd, awesome shepherding editor!
