Immortal Time Bias in Observational Studies With Dr Kabir Yadav
Jan 4, 2024
auto_awesome
Dr Kabir Yadav, Statistical Editor, discusses immortal time bias in observational studies with Dr Roger Lewis. They explore the misclassification of immortal time in observational studies compared to randomized trials, address the bias in the UK Biobank study, and examine the higher risk of cardiovascular events in women with premature menopause.
Immortal time bias can occur in observational studies when there is a gap of time between hospital discharge and the initiation of treatment, leading to misclassification of treatment time.
When comparing the time to readmission in observational studies, the time between discharge and when patients pick up their medication should be accounted for as unexposed time, not treatment time, to avoid bias toward overestimating the benefit associated with drugs.
Deep dives
Explanation of immortal time bias in randomized clinical trials
In a randomized controlled trial, the start time for both groups is the same, which eliminates unexplained or missing intervals. The treatment is initiated at the same time, creating a common starting point for comparing time to events. In observational studies, immortal time bias arises when there is a gap of time between hospital discharge and the initiation of treatment. During this time, multiple events could occur, leading to misclassification of treatment time. Observational studies should account for this immortal time as unexposed to the treatment.
Impact of immortal time bias in observational studies
When comparing the time to readmission in observational studies, the time between discharge and when patients pick up their medication should be accounted for as unexposed time, not treatment time. Failure to do so introduces a bias toward overestimating the benefit associated with the drugs, showing a greater benefit in the treatment group. Misclassification immortal time bias and excluded immortal time bias are the two types of immortal time bias that can occur.
Examples of immortal time bias and addressing potential sources of bias
One example of misclassification immortal time bias is seen in a study examining medications after discharge for COPD patients. The immortal time is accounted for under the treatment group when it should be accounted for under the unexposed group. Excluded immortal time bias arises when time periods are not counted in either group, creating an imbalance in time under exposure. Addressing potential sources of bias can involve shifting the time zero or considering what a randomized controlled trial design would have looked like to identify gaps in the data.
JAMA Statistical Editor Roger J. Lewis, MD, PhD, discusses Immortal Time Bias in Observational Studies with Kabir Yadav, MDCM, MS, MSHS. Related Content: