Abstract: Objective To discuss the construction of a multilevel model for binary outcome measure with hierarchical cluster data based on an example of type 2 diabetes. Methods Data was collected from a survey on chronic disease of Longzihu district of Bengbu city in 2015. One-level and multilevel Logistic regression models were conducted to investigate the influencing factors of type 2 diabetes using R software. Results The influencing factors of type 2 diabetes analyzed by one-level and multilevel Logistic regression were not exactly the same, including age, family history of diabetes, history of hypertension, frequency of fruit intake, glycosylated hemoglobin, body mass index (BMI). One-level Logistic regression showed that the prevalence of diabetes in non-overweight group was significantly different from that in overweight group (all P<0.05), which was not significant between non-overweight and overweight group in multilevel Logistic regression (all P>0.05). The Akaike information criterions of one-level and multilevel Logistic regression were 1 550.2 and 1 535.3, respectively. Conclusions The conclusions from one-level and multilevel Logistic regression model are not exactly the same, and the multilevel model has advantages in data analysis for hierarchical cluster data. Based on hierarchical cluster of type 2 diabetes, the prevention and control of type 2 diabetes should be carried out on the basis of community service center.