Using 2002 cross-sectional data and 1998, 2000, 2002 three waves of panel data from the Chinese Longitudinal Healthy Longevity Survey, we study health in oldest old population. We measure health using the Katz Index of Activities of Daily Living (ADL) and in term of mortality. Our results suggest that besides aging, there are other important factors contributing to bad health both in term of the Index of ADL and mortality. Effects of gender on the Index of ADL and on mortality are different. Female tends to be more dependent in daily living, but has higher probability to survive. Oldest-olds living in urban area are also more dependent, but are less likely to die. Socioeconomic status such as financial resources, education level of the oldest-old and of his/her spouse and etc. plays insignificant role in the health of oldest-old. Oldest-olds still in marriage are more independent in daily living and are more likely to survive. There exists reverse causality between health and risky behaviors. After controlled for simultaneity bias by instrumental variable method, the coefficients of risk behaviors are consistent with theory and common wisdom.