Immortal time bias (survivor treatment selection bias) can make a treatment seem better than it really is. There is an element of bait and switch, with different exposures for the two groups over time.
Features of a study that shows immortal time bias:
(1) An observational study of the effectiveness of an intervention using "matched" controls.
(2) The intervention group does not undergo the intervention for a period of time after the start of the study. To get the intervention a participant must survive this interval (the period of "immortality").
(3) The control group is evaluated from the start of the study, and some members may die before their match in the intervention group completes the "immortality period".
(4) The intervention appears on first evaluation to have a benefit that disappears when the effect of the "immortal period" is accounted for in the data analysis.
Ways to avoid the immortal time bias:
(1) matching subjects at the time of the intervention (avoiding any "immortal period") and comparing within pairs only
(2) analyze the data using time-dependent covariates
To read more or access our algorithms and calculators, please log in or register.
