A Cox Proportional Hazard Model with Time-Varying Coefficients Using Revised Survival Time to Investigate Stroke Risk Factors

Abstract

According to estimates by World Health Organization (WHO), globally, cardiovascular disease is the first leading cause of death. Among various cardiovascular diseases, the risk of stroke is constantly increasing even though the mortality rate for it is decreasing. Including representative studies, Framingham study and Northern Manhattan stroke study (NOMAS), many of studies relating to the risk factors of stroke have been performed. In this study, using the dataset obtained from NOMAS, the Cox proportional hazard model is used in order to investigate stroke risk factors. In particular, the NOMAS used follow-up time to onset of stroke, whereas, in this study, the age of having a stroke is treated as survival time for fitting the models. The proportional hazard assumption is evaluated through graphs, and some risk factors whose risk effects vary according to the variation of age are detected. The pattern of change and the significance of these risk effects are evaluated by constructing a Coxs model with time-varying coefficients. In addition to the study concerning occurrence of all types of strokes, the study concentrating only on the ischemic stroke occurrence is also performed and these two are compared.