In recent years, with the development of the economy and the advancement of the national income, personal healthcare has been becoming an important issue. Since the health examination can help people understand their own health conditions clearly, many people make regularly health examinations to avoid missing the best treatment time. So the health examination is an important role for people's health statuses. However, people only get aware of the result of health examination report but they are not clear whether their health trends are high-risk or not. In this paper, we proposed a novel framework for discovering health risk patterns and the relationships between the patterns and a target disease by mining historic health examination data. Moreover, a prediction model for the target disease can be constructed effectively by the discovered information. Through the framework, the physicians can early provide health alerts and medical treatments for people to effectively increase the quality of disease prevention.