Pacific D
Time-varying Exposure Effect for Adjusting Time-varying Confounding on Competing Risks (307810)
*Hyunsun Lim, National Health Insurance Service Ilsan HospitalJongHun Kim, National Health Insurance Service Ilsan Hospital
Youn Nam Kim, Severance Hospital, Yonsei University Health System
Keywords: competing risk, marginal structure model, blood thinner, cerebral infarction , time-varying covariate
On competing risks data, Cox’s regression model can be naturally extended for time-varying covariates, while Fine and Gray’s regression model has a limitation to work with time-varying covariates (Latouche, 2005). Fortunately, Beyersmann (2007) have induced the subdistribution hazard using relationship between the cause-specific hazard and the subdistribution hazard. However, there has been no study of regression model for time-varying exposure which controlled time-varying confounding under competing risk structures. In Korean nationwide data (N=112,000), we aim to estimate the effect of blood thinner to delay the reinfarction in cerebral infarction patients. The usage of blood thinner usually accelerates the occurrence of the hemorrhage, the hemorrhage usually reduces the usage of blood thinner. Therefore, we suggested the marginal structure model for the cause-specific hazard and the subdistribution hazard of the time-varying usage of blood thinner, adjusting by time-varying occurrence of the hemorrhage. The simulation presented to estimate the effect of time-varying exposure from marginal structure model on competing risks.