NEW PREPRINT!
Enjoyed working with @tom-parks.bsky.social and @petedodd24.bsky.social on immortal time bias
In its simplest form, ITB results from misallocation of person time, i.e. pre treatment time in intervention group being counted as time on treatment
https://www.medrxiv.org/content/10.1101/2025.01.09.25320251v1
[1/n]
Enjoyed working with @tom-parks.bsky.social and @petedodd24.bsky.social on immortal time bias
In its simplest form, ITB results from misallocation of person time, i.e. pre treatment time in intervention group being counted as time on treatment
https://www.medrxiv.org/content/10.1101/2025.01.09.25320251v1
[1/n]
Comments
In lay terms, people receive the intervention because they survive, rather than survive because they receive the intervention
[2/n]
[3/n]
It is a common issue in observational analyses of interventions in acute severe infection
[4/n]
IMMORTOOL will estimate the likely extent of the spurious protective association in studies that have been analysed naively, or which have used common but problematic workarounds, e.g. landmark analysis
[5/n]
Inputs include reported or assumed distributions of time to intervention and time to event
These are fit as Weibull distributions
The tool will help you fit distributions to your data
It provides a graph to illustrate how well the distributions approximate the data
[6/n]
First, we assume the intervention has no effect (often true!)
Second, we assume no other biases are in play (usually not true!)
The tool outputs the rate ratio we expect purely as a result of misallocation of person time
[7/n]
https://authors.elsevier.com/a/1kP4e,RGPgW0JO
So. Good.