中文摘要：

目的 利用多状态Markov模型对社区老年人轻度认知损害（MCI）转归进行研究.方法 通过MCI患者智商（IQ）变化反映随访人群认知功能的变化趋势,构建一个四状态模型,采用多状态Markov模型进行分析.结果 600例MCI患者中,男性174例（29.0%）,女性426例（71.0%）,年龄65～90岁,平均（69.7±6.6）岁.单因素分析显示,性别、年龄、文化程度、婚姻状况、吸烟、经济收入、脑溢血、高血压、高胆固醇、糖尿病、低密度脂蛋白胆固醇（LDL-C）水平、收缩压（SBP）、舒张压（DBP）等因素与认知功能有关.多因素分析显示,女性、高年龄、发生脑溢血及高SBP为认知稳定转化为认知重度恶化的危险因素（系数分别为0.762、0.366、0.885、0.069）;高年龄是认知稳定转化为认知轻度恶化的危险因素（系数为0.038）;高文化程度均为保护因素（系数分别为-0.219和-0.297）.转移强度结果显示认知功能稳定进一步恶化（轻度＋重度）的进度是好转的1.2倍;轻度恶化到认知功能稳定的进展强度是进一步发展成重度恶化的11.4倍多.结论 多状态Markov模型是处理纵向资料的有效工具.

英文摘要：

Objective To introduce the Multi-state Markov model in studying the outcome prediction of mild cognitive impairment （MCI）. Methods Based on the intelligence quotient （IQ）changes that reflecting the trends in cognitive function in the patients under follow-up program, we constructed a four states model and used Multi-state Markov model to analyze the patients. Results Among 600 MCI patients, there were 174（29.0%） males and 426（71.0%） females, with age range of 65-90 years-old （average 69.7±6.6）. For univariate analysis, gender, age, education level, marital status, smoking, household income, cerebral hemorrhage, hypertension, high cholesterol, diabetes,LDL-C, SBP and DBP were found to be associated with cognitive function. For multivariate analysis,female, older age, cerebral hemorrhage and higher SBP were shown to be the risk factors for transition from the state of cognitive stability to the state of severe deterioration, and their coefficients were 0.762,0.366,0.885, and 0.069, respectively. Age （0.038） could influence the transition from the state of cognitive stability to slight deterioration. Higher education level was shown to be the protective factor for these transitions （-0.219 and-0.297）. Transition intensity from the state of cognitive stability to the state of slight and severe deterioration was 1.2 times that of transition to the state of improving. Transition intensity from state of slight deterioration to cognitive stability was 11.4times that of transition to severe deterioration. Conclusion Multi-state Markov model was an effective tool in dealing with longitudinal data.