Causal Risk Ratio and Causal Risk Difference in Longitudinal Studies With Frequent Outcome Events
Marginal structural models (MSMs) are recognized as useful methods for addressing the issue of time-varying confounding in longitudinal studies. In the analyses of longitudinal data with binary outcomes, using the generalized estimating equation (GEE) logistic regression model within the MSM framewo...
Saved in:
| Main Authors: | , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Taylor & Francis Group
2025-12-01
|
| Series: | Data Science in Science |
| Subjects: | |
| Online Access: | https://www.tandfonline.com/doi/10.1080/26941899.2025.2527144 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | Marginal structural models (MSMs) are recognized as useful methods for addressing the issue of time-varying confounding in longitudinal studies. In the analyses of longitudinal data with binary outcomes, using the generalized estimating equation (GEE) logistic regression model within the MSM framework is a common approach to estimate the odds ratio. However, due to the interpretive issues with the odds ratio, recent statistical guidelines recommend the use of the risk ratio and the risk difference. Nevertheless, there are no applicable MSM methods using GEE that enable straightforward estimation for the causal risk ratio and risk difference. In this article, we provide two straightforward and effective methods for estimating the causal risk ratio and risk difference based on the MSM-GEE framework, using Poisson regression and normal linear regression models. We validated these methods through comprehensive simulation studies, confirming their unbiased estimation of the causal risk ratio and risk difference. Importantly, these methods remain effective even in situations where the MSM-GEE logistic regression model, which is the most widely used method for binary outcome data, yields biased estimates of the causal odds ratio. In addition, we applied the proposed methods to real-world longitudinal data and clearly demonstrated their practical effectiveness. |
|---|---|
| ISSN: | 2694-1899 |