Raj Mehta
@raj_mehta
Family Doc, Clinical Informaticist, EBM & Bioethics enthusiast | Faculty @AdventHealth Family Medicine Residency | @UFMEDICINE alum
ID:32283096
https://tinyurl.com/3bbh8j7y 17-04-2009 03:17:38
19,7K Tweets
2,3K Followers
4,4K Following
Follow People
A new use for paroxetine (Paxil): Erythema of rosacea. The SSRI #antidepressant improved multiple outcomes in this well designed, 12 week randomized controlled trial of 97:
pubmed.ncbi.nlm.nih.gov/36806645
#psychiatry #clinicaltrials #dermatology
Interesting example for the 'is there ever heterogeneity of treatment effect?' debate. Congrats to my colleagues Berry Consultants Ben Saville Roger Lewis, Giorgio Paulon, Mark Fitzgerald, and Anna McGlothlin! Also an example for the 'how to analyze an enrichment trial' debate.
I knew UNOS and OPTN and certain OPOs were bad from Sumit Mohan and others work
But this….is still wild reading
statecraft.pub/p/how-to-stop-…
From Santi Ruiz h/t Raj Mehta
28000 organs are tragically wasted every year - including 17,500 kidneys
Anil Makam Manni Mohyuddin Just make sure that mgus patients have had FLC checked/tracked. About 10-15% of myeloma is light chain only, so someone can go from “it’s just mgus with a small m spike” to “full blown myeloma” if you check flc.
But also the reference ranges are wider now, which is good 😎
Anil Makam The values of what a normal light chain is were revised. Many people previously considered to have abnormal light chains (or light chain only MGUS) now have normal labs. Saves many a hematology referral!
ashpublications.org/blood/article/…
Students sometimes ask 'should I do a regression or calculate a correlation?'
This is a useful explanation about why it's the same thing: lindeloev.github.io/tests-as-linea…
(originally via Sam Abbott (@[email protected]))
New on our site: The best augmentation strategy varies by patient. Learn which features help personalize the plan. #mentalhealth thecarlatreport.com/articles/4662-…
#Statistics throught of the day: If sponsors knew how much money was wasted with fixed sample size designs, and how much earlier Bayesian sequential designs would have bailed out on ineffective treatments, they'd be shocked. hbiostat.org/bayes/bet/desi…
Our new study in JAMA Internal Medicine looking at the reasoning abilities of GPT-4 compared with human physicians just came out.
Big picture: AI displays (much) better reasoning than humans, makes diagnoses similarly, but hallucinates considerably more.
A 🧵to put in context ⬇️
John Mandrola, MD I think the most important steps are to separate 4 things:
1) priors based on theory
2) methodological stgth/amt of evidence
3) bounding uncertainty
4) attaching values LAST
I find even smart/knowledgeable/confident ppl are often unable to even understand such distinctions /1