[Application of multi-state Markov model in studying transition of number of chronic complications and influencing factors in type 2 diabetes mellitus patients].

Objective: To establish a multi-state Markov model of type 2 diabetes mellitus (T2DM) patients and explore the transition rule between the cumulative number of different chronic complications, estimate the transition probability and intensity between status, and explore the possible factors affecting the transition between status. Methods: A retrospective cohort study of 33 575 patients with T2DM was conducted. According to the baseline and the cumulative number of chronic complications during the follow-up period, the patients were classified based on five status: T2DM, one complication, two complications, three complications, four and above complication, indicated by S0, S1, S2, S3 and S4, respectively. A time-continuous and state-discrete multi-state irreversible Markov model was used for statistical analysis. Results: The study included 33 575 T2DM patients, and their average age was 60 years old, the median of follow-up length was 8 years. In these patients, 32 653 had no baseline complications. At the end of follow-up, the transition probabilities of S0→S1, S1→S2, S2→S3 and S3→S4 were 16.4%, 32.4%, 45.6% and 25.9%, respectively. The results of multivariate analysis showed that being female (HR=0.919), less than 60 years old (HR=0.929), higher fasting plasma glucose (HR=1.601), lower high-density lipoprotein (HR=1.087), higher total cholesterol (HR=1.090),weekly exercise (HR=0.897), vegetarian diet (HR=0.852) and heavy diet (HR=1.887) were the risk factors for S0 to S1. And being female (HR=0.768), less than 60 years old (HR=0.859) and lower high-density lipoprotein (HR=1.160) were the risk factors for S1 to S2. Conclusions: The probability of multiple complications in T2DM patients increased over time, the transition intensity of S2→S3 was largest, followed by S1→S2. Therefore, we need to conduct both early and long-term indicators monitoring and disease prevention, strengthen the health education to improve patients' daily living habits at early stage of the illness, encourage patients to have moderate exercise and balanced diet, strengthen the monitoring of fasting blood- glucose, cholesterol and high-density lipoprotein levels to prevent the deterioration of the illness.目的： 建立2型糖尿病（T2DM）患者多状态马尔科夫模型，探讨不同慢性并发症累积数量状态间的转移规律，估计状态间的转移概率和转移强度，并探索影响状态改变的可能因素。 方法： 对33 575例T2DM患者进行回顾性队列研究，根据基线情况和随访观察期间累积慢性并发症的数量将状态划分为单纯T2DM、合并1、2、3、≥4类并发症共5个状态，分别用S0、S1、S2、S3和S4表示。采用时间连续、状态离散的多状态不可逆马尔科夫模型进行统计学分析。 结果： 共纳入33 575名研究对象，平均年龄60岁，随访时间M=8年，32 653例患者基线无并发症。随访期间，S0→S1、S1→S2、S2→S3和S3→S4的转移概率分别为16.4%、32.4%、45.6%和25.9%。多因素分析结果显示，女性（HR=0.919）、<60岁（HR=0.929）、FPG升高（HR=1.601）、HDL-C降低（HR=1.087）、TC升高（HR=1.090）、每周运动（HR=0.897）、素食为主饮食（HR=0.852）和重口味饮食（HR=1.887）为S0→S1转移的危险因素。女性（HR=0.768）、<60岁（HR=0.859）和HDL-C降低（HR=1.160）是S1→S2转移的危险因素。 结论： T2DM患者合并多种并发症的概率随着时间的推移而增高，其中S2→S3的转移强度最大，其次为 S1→S2的转移，故我们需要兼顾患病初期和长期的指标监测和病情防范工作。从患病初期（合并2~3类并发症之前）即重点加强对患者日常生活习惯的宣传教育工作，适当运动，均衡饮食。尤其要加强对FPG、TC和HDL-C等指标的监测，防止患者病情进一步恶化。.