Handling time-dependent exposures and confounders when estimating attributable fractions – bridging the gap between multistate and counterfactual modeling
share
The population-attributable fraction (PAF) expresses the percentage of events that could have been prevented by eradicating a certain exposure in a certain population. It can be strongly time-dependent because either exposure incidence or excess risk may change over time. Competing events may moreover hinder the outcome of interest from being observed. Occurrence of either of these events may, in turn, prevent the exposure of interest. Estimation approaches thus need to carefully account for the timing of potential events in such highly dynamic settings. The use of multistate models (MSMs) has been widely encouraged to meet this need so as to eliminate preventable yet common types of bias, such as immortal time bias. However, certain MSM based proposals for PAF estimation fail to fully eliminate such biases. In addition, assessing whether patients die from rather than with a certain exposure, not only requires adequate modeling of the timing of events, but also of the confounding factors affecting these events and their timing. While proposed MSM approaches for confounding adjustment may be sufficient to accommodate imbalances between infected and uninfected patients present at baseline, these proposals generally fail to adequately tackle time-dependent confounding. For this, a class of generalized methods (g-methods) which includes inverse probability (IP) weighting can be used. Because the connection between MSMs and g-methods is not readily apparent, we here provide a detailed mapping between MSM and IP of censoring weighting approaches for estimating PAFs. In particular, we illustrate that the connection between these two approaches can be made more apparent by means of a weighting-based characterization of MSM approaches that aids to both pinpoint current shortcomings of MSM based proposals and to enhance intuition into simple modifications to overcome these limitations.
READ FULL TEXT
