Abstract: The older adults are suffering from declining cognitive function with the rapid population ageing. On the basis of 2011 and 2014 CLHLS database, this research explores the transition of cognitive function by age and gender, using the method of continuous time homogeneous Markov process. Risk factors affecting cognitive function are identified. The size of elderly population with cognitive impairment at four different levels is estimated and projected for the decade of 2015-2024. The probability of cognitive impairment decreases among older people, and the decline is most significant among the oldest old. Being male, having higher level education, doing persistent exercise, participating social activity, and better health status are associated with a relatively slow decline in the cognitive function. It is projected that there would be 27.85 million elderly population over 65 years old with cognitive impairment by 2024. In addition, this research provides the basic methodology for need assessment of long-term care services for older adults with dementia.

Keywords: Cognitive Impairment, Continuous Time Homogeneous Markov Process, Transition Probability Matrix, Long-Term Care

摘要： 伴随着人口老龄化和高龄化，大多数老年人面临不同程度的认知功能衰退。基于2011和2014年CLHLS数据，采用连续时间齐性马尔科夫过程刻画老年人分年龄、分性别的认知功能状态转移规律，识别了影响认知功能状态转移的风险因素，并估算了2015~2024年10年间不同认知功能状态老年人口需求规模。中国老年人整体认知功能缺损率呈现下降趋势，尤其在高龄老人中更加明显。男性、教育程度高、坚持锻炼、参与社会交往和拥有较好健康状况（自评健康好、所患慢性病数少、无功能受限和初始认知状态好）的老年人具有更好的抵御认知下降的能力。预测至2024年，中国65岁及以上认知缺损的老年人有2785万。为失智老年人长期照护需求规模的预测提供了方法论基础。

关键词: 认知功能缺损, 连续时间齐性马尔科夫过程, 转移概率矩阵, 长期照护