Epidemiology;UMCU;Cochrane;Health Innovation Netherlands (HiNL);Center for Side Effects Implants;Prediction Models;Medical tests&apps;Chief Editor Diag&Prog Res
We're seeing lots of reviews of prediction models, particularly as a consequence of lots of machine learning studies
So we've developed some reporting recommendations for systematic reviews / meta-analyses of prediction models
tinyurl.com/2p8vskxw#openaccess#statstwitter
ALT A new reporting guideline for systematic reviews and meta-analyses of clinical prediction model studies, TRIPOD-SRMA.
Are you a prediction modeling expert and not yet invited to participate in the Delphi of PROBAST AI, we would love to hear from you!
Send me a DM with your name and email, and I'll get you a personal invitation link
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.
We've heard on the grapevine, there are still people who are not using the TRIPOD statement (😲we know) to report their prediction model study.
Enlighten them by pointing them to
tinyurl.com/3z4fw4zx (checklist)
tinyurl.com/mrywn3mk (explanation/examples)
#statstwitter
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 samples
bmj.com/content/378/bmj-2022…@carlmoons
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 samples
bmj.com/content/378/bmj-2022…@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/bmj-2021…
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/bmj-2021…
We would like to thank all our participants today at the TRIPOD-#AI consensus meeting today.
Excellent discussion on a range of issues to report when developing or validating an #AI/#ML prediction model.
Now to refine the checklist. Watch this space!