We agree and work hard to have the handbook out at the end of 2024
View quoted post on X
Carl Moons
@CarlMoons
Epidemiology;UMCU;Cochrane;Health Innovation Netherlands (HiNL);Center for Side Effects Implants;Prediction Models;Medical tests&apps;Chief Editor Diag&Prog Res
315 Following 1.4K Followers 435 Posts
Latest posts 5
It is indeed - and the paper that shows that is almost out as well.
@pauladhiman @GSCollins @DrGSBullock @Richard_D_Riley @BenVanCalster @BenjaminSpeich @CarlMoons @CSMOxford Great work. I have my suspicion this is not only in oncology but generalized across medicine and health, unfortunatelyView quoted post on X
A great overview of methodological aspects that may still need to be improved when developing machine learning or AI based prediction models for diagnosis, prognosis, prevention or monitoring. Thanks to many colleagues who made this tremendous amount of work possible.
NEW PAPER in @JClinEpi -> "focus is required on handling of missing values, methods for internal validation, & reporting of calibration to improve [...] studies on #machinelearning-based prediction models" tinyurl.com/3zfjs5wu
#OpenAccess #StatsTwitter #DataScience #MLtwitterView quoted post on X
If one is interested in the accuracy of Covid-19 self tests - when used at home (a real world validatie study) - you may want to read this paper. A comprehensive self test validation study, also investigating the added value of throat to nose sampling.
The performance of some home testing kits for covid-19 appears to have declined as the omicron variant emerged, suggests study based on three widely used rapid antigen tests. Performance improved when tests used both nose and throat samplesView quoted post on X
bmj.com/content/378/bm… @carlmoons
A unique endeavour this has been. My great Thanks to everyone across the globe, who contributed their data and/or expertise to this challenging but hugely important project that we started already in 2020! You all did it!
NEW PAPER
Large external validation of covid-19 prediction models for mortality @bmj_latest
TL;DR: performance highly heterogenous -- often pretty bad --, but some positive exceptions too that require some recalibration
bmj.com/content/378/bm…View quoted post on X
