Effect of factors with time-dependent covariates on the survival status of the elderly in Taiwan

The elderly population has been increasing rapidly in recent years because of improvements in medical care and the progress in economics in Taiwan. Now, the population of old people age above 65 has approached 10%, and problems of the elderly have become a major concern for public health. In this study, we use five waves of Survey of Health and Living Status of the Elderly in Taiwan held from 1989 to 2003 to explore 15 variables related to demographic characteristic, health status, health behavior and home condition etc. affect the survival status of the elderly by employing Cox proportional hazard model. The results show that there are nine variables, e.g. age, gender, ethnic group, ADL, self-rated health, physical function, smoking, chewing betel nuts and marital status, were strongly related to the survival status of the elderly. In addition, Aalen's nonparametric additive model is not only used as an alternative to a proportional hazards to describe the effects of covariates on survival time but assist in detecting and describing the nature of time-dependent effects of covariates used for Cox proportional model. We find there are some time-dependent covariates (e.g. self-rate health, ADL function, and physical function), and Aalen's nonparametric additive model provides a flexible and nonparametric method for investigating the time-dependent variables through plots of the estimated cumulative regression coefficients, with confidence intervals.