The Application Of Multi-state Markov Model In The Prognosis Of Type 2 Diabetes Mellitus And Its Complications

ObjectiveThe multi-state “disease-progression” Markov model is used for laws of state transition exploration,transition probability,transition intensity and sojourn time estimation of patients with type 2 diabetes mellitus and its complications.And together with the multi-state Markov model,the classical Logistic regression model is fitted to screen the influence factors of state transition,so as to provide the basis for prevention and treatment of type 2 diabetes mellitus.MethodsPatients with type 2 diabetes mellitus and its complications were retrospectively collected from top three hospitals during January 2011 and May 2015.According to the combined number of chronic complications,patients with type 2 diabetes mellitus were discrete into five different states.And a homogeneous Markov process with continuous time and discrete state was fitted by the msm package in R software,so as to estimate the transition probability,transition intensity and sojourn time among states.Meanwhile,the logistic regressions were fitted too,and its analysis was carried out in the SAS9.2 software.ResultsMSM plot and Pearson type goodness of fit test show that the multi-state Markov model is basically qualified.And results of the multi-state Markov model suggest that the total retention time of patients in state 1,state 2,state 3 and state 4 is 3.19 months,14.11 months 23.87 months and 15.21 months.The transfer intensity matrix and transfer probability matrix show that patients in state1 have risk of transfer to state2,patients in state2 have higher risk of transfer to state3 and patients in state3 have higher risk of transfer to state4.Those variables which are significant in the Univariate logistic regression are chosen to the multi logistic regression analysis.And results show that hypertension,fasting plasma glucose,urea,urine microalbumin and high density lipoprotein are statistically significant for transition from state 1 to state 2.Age,fasting blood glucose,fasting insulin,low density lipoprotein,total cholesterol,apolipoprotein A1 and urea are statistically significant for transition from state 2 to state 1.Fasting blood glucose,fasting insulin,high density lipoprotein,triglycerides and free fatty acids are statistically significant for transition from state 2 to state 3.Fasting blood glucose,fasting insulin,high density lipoprotein,triglycerides and free fatty acids are statistically significant for transition from state 3 to state 2.Age,high density lipoprotein,triglycerides,creatinine and free fatty acids are statistically significant for transition from state 3 to state 4.Triglycerides,totalcholesterol,creatinine and fasting insulin are statistically significant for transition from state 4 to state 3.Fasting blood glucose,fasting insulin and triglycerides are statistically significant for transition from state 4 to state 5.Results of multi state Markov model shows that hypertension,glycosylated hemoglobin,free fatty acids,lipoprotein a,urinary albumin / urinary creatinine ratio and urinary albumin are statistically significant for transition from state 1 to state 2.Hypertension,urinary albumin / creatinine ratio,age,fasting insulin,low density lipoprotein,apolipoprotein A1 are statistically significant for transition from state 2 to state 1.Hypertension,urinary albumin / urinary creatinine ratio,glycosylated hemoglobin,free fatty acid,lipoprotein a,urinary microalbumin are statistically significant for transition from state 2 to state 3.Hypertension,free fatty acids,apolipoprotein A1,high density lipoprotein,creatinine,urea are statistically significant for transition from state 3 to state 2.Hypertension,free fatty acids,apolipoprotein A1,status at admission are statistically significant for transition from state 3 to state 4.Free fatty acids,creatinine,urinary albumin,fasting insulin,triglycerides are statistically significant for transition from state 4 to state 3.Free fatty acids,creatinine,triglycerides,low density lipoprotein,fasting blood glucose are statistically significant for transition from state 4 to state 5.ConclusionMultivariate Markov model and Logistic regression showed that thenumber of chronic complications of diabetes mellitus was affected by age,blood glucose,lipid metabolism,blood pressure and renal dysfunction,and the effect of different variables on different transfer is not the same.However,the multi-state Markov model is based on the global perspective,and considers the influence of time on the appearance of target outcome,thus its results are more accurate and scientific than the traditional Logistic regression.Thus,the multi-state Markov model can only be used as a supplement to the traditional Logistic regression model.At the same time,the transition intensity matrix and transition probability matrix suggests that patients in the state1,state2 and state 3 are prone to deteriorate and patients in state 4 are prone to improve.And results of residence time prompts that patient have longer stay in state 3 before enter state5,suggesting that the time window could be used clinically to reverse the progression of the disease...