The main aim of this study is to explore the patterns, determinants and subsequent mortality prediction of change in self-rated health in the elderly American population. To achieve this purpose, we constructed logistic regression models and Cox proportional hazard regression models with the complex survey dataset from the National Second Longitudinal Study of Aging (LSOA II) to calculate the odds ratios (OR)/ hazard ratios (HR) and confidence intervals (CI) of risk factors. Our results show that chronic disease condition and difficulty in daily activities are the main reasons for change in self-rated health status. Furthermore, change in self-rated health has significant impact on survival function in the elderly populations. When change in self-rated health status was considered, self-rated health was a stronger and more flexible predictor of mortality for elderly populations. These findings will provide important information to establish effective strategies for prolonging lifespan by improving self-rated health status for elderly populations.