|
Abstract:
|
In studies with long-term follow-up, exposures are often measured overtime and have a protracted effect on the risk of developing an event. Also, the intensity, duration, and timing of the exposure vary among individuals, which create challenges in modeling the exposure-time-response association. Meanwhile, an increasing number of clinical studies involve data with competing risks, where subjects may fail from one of multiple events. In this study, we proposed a subdistribution hazards regression model incorporating weighted cumulative effect of time-dependent exposures. This model simultaneously takes into consideration the intensity, duration, and timing of an exposure in estimating the cumulative incidence rate of an event when data involve competing risks. When the exposure is drug use, the model is able to distinguish different usage patterns even though these patterns have the same cumulative doses over a period of time. Performance of the proposed model was evaluated through a simulation study. We applied this model to investigate the effect of opioid use patterns on the risk of overdose among Medicare beneficiaries, treating mortality before overdose as a competing risk.
|
