Predicted #COVID19 US deaths from now until the 12/1/2020. Uses the @COVID19Tracking repo to fit a Poisson GAM with time and day of the week as covariates. Fixed the model so that it will not update after 11/17/20 (vertical gray line) to see how far out it can predict
Can make it fancier i.e. fit a separate trend smoother to each state, and random effects for the interaction of state X day of the week to predict further out. However even the simple (green) model has a R2 of 84%, so not worth the effort since will not projecting past 14 days
Played a little bit .. fit the model to all data up to 14 days ago and still get a reasonable prediction
But if one goes back 30 days, then the model starts crapping out in a predictive fashion .
So yeap sticking with 14 day predictions

• • •

Missing some Tweet in this thread? You can try to force a refresh

Keep Current with ChristosArgyropoulos

ChristosArgyropoulos Profile picture

Stay in touch and get notified when new unrolls are available from this author!

Read all threads

This Thread may be Removed Anytime!


Twitter may remove this content at anytime! Save it as PDF for later use!

Try unrolling a thread yourself!

how to unroll video
  1. Follow @ThreadReaderApp to mention us!

  2. From a Twitter thread mention us with a keyword "unroll"
@threadreaderapp unroll

Practice here first or read more on our help page!

More from @ChristosArgyrop

20 Nov
#AskRenal about this scenario (answer at the end):
pt on dialysis shows up with upper respiratory symptoms c/w COVID19. Entire household has same symptoms and they all got tested. Everyone is +ve except the patient. How would you approach the pt for infection control purposes
Two days have passed after your decision. Pt is now visibly short of breath and is sent to hospital to be admitted (family members have also deteriorated so they all travel to ED). Pt sent to dialysis for urgent HD and PCR is sent. How do you handle their 1st inpatient dialysis:
Pt stabilizes after the first emergency dialysis and a CXR is obtained. It shows multilobar pneumonia; PCR results within 16hrs (yeah we can do that in #NewMexico) & is negative. IM labels the patient as HCAP & IV ABx are initiated; condition improves somewhat & ABx stepped down
Read 10 tweets
16 Nov
Since we have two RNA vaccines against SARSCOV2 now, perhaps you should unfollow covidiots , libertarian flubros , Hoover institute faculty, GBD acolytes and every single individual who have suggested that a RNA vaccine against a CoV has never been developed before because ...
a (RNA) vaccine against a CoV has actually been developed before and within the same timelines that the two human vaccines are being developed in front of our own eyes. Enter PED - porcine epidemic diarrhea, a disease that appeared in North America in 2013…
PED is highly contagious and devastated the swine industry:…
Similar to COVID19 it is a mucosal disease (it affects the GI tract) but turns guts (rather than lungs) into mush…
Read 9 tweets
9 Oct
For the last 10 days, we have added about 3 hospitalizations per month. The way this is going, we will hit our peak of hospitalizations in about 3 weeks, nearly coinciding with the election…
#COVID19 #NewMexico
This is the age distribution of cases in our state. Many kids as you can see from the graph.

#COVID19 #NewMexico
Since we don't have major testing bottlenecks (in fact nearly 45% of the state has been tested at least once), we can catch patterns that other locales can't.
So let's look at the age distro (after collapsing the last category) v.s. the population age pyramid

#COVID19 #NewMexico
Read 8 tweets
30 Sep
Alright then, time to visit this. We currently have two publications for two different SGLT2i from two different studies (CANVAS and EMPAREG-Outcome) that show that the *relative* benefits of SGLT2i don't differ according to the baseline (KDIGO) cardiorenal risk of participants.
First the links and the visual abstracts:… (canagliflozin/CANVAS)… (empagliflozin-EMPAREG)
Before we discuss, let us all remember what the KDIGO category is. It is a convenient classification of the risk for badness (deathworsening kidney function/acute kidney injury) for patients with Chronic Kidney Disease, based on two tests: eGFR/albuminuria
Red = bad
Green = Good
Read 16 tweets
28 Sep
How can you be reassured that a COVID19 vaccine will be safe for YOU ?

1. Pick the side effect which you fear the most
2. Find your age specific Rona death rate
(use CDC not FoxNews for this step)
3. Decide how large the prob of 1 over the prob of 2 should be to matter
3. Go to any online power calculator for one-sample proportions and input these values, the desired false positive and negative rates for the study, the difference in 3 and fire away
4. If the observation time of the study (Number of patients x Number of follow up for each participant), is smaller than the sample "size" determined at step 3, then you don't have a guarantee that the vaccine is safe for you
Read 5 tweets
24 Sep
This work suggests ~50% seroprev in Japan (Tokyo).…

One can search the web and find that the characteristics of the assay…
May use my code to see how this plays out……
Let's see some of the problems. Note:
1) Sensitivity of the IgG assay 0.967 and specificity 0.98
2) Rule employed by the authors for assigning positives (+ve) : any +ve test, even if subsequently became -ve labelled the participant as +ve for ever
3) They ran the test twice
Assume a cohort of 100 participants: 5 had #COVID19 and 95 didn't. Let's calculate the prevalence of +ve *tests*
Round 1: 5*0.967 + 95*0.02 ~6.7~7
Round 2: 5*0.967 + 95*0.02 ~ 6.7 ~7
The false +ve tests during rounds 1/2 are unlikely to have come from the same patients, so ...
Read 5 tweets

Did Thread Reader help you today?

Support us! We are indie developers!

This site is made by just two indie developers on a laptop doing marketing, support and development! Read more about the story.

Become a Premium Member ($3/month or $30/year) and get exclusive features!

Become Premium

Too expensive? Make a small donation by buying us coffee ($5) or help with server cost ($10)

Donate via Paypal Become our Patreon

Thank you for your support!

Follow Us on Twitter!