OBJECTIVES : To study the risk probability of influencing blood glucose control in patients with type 2 diabetes mellitus (T2DM) in the real world and establish a Nomogram that can predict the patients' HbA1c control compliance, so as to improve the compliance rate of HbA1c control.

METHODS : From March 2018 to October 2019, a registration study was conducted in a third-grade grade a hospital, and all information of 720 T2DM patients was collected, including demographic characteristics, blood glucose status, hypoglycemic treatment information, exercise and diet status, medication compliance, etc. In March 2018 to June 2019, collecting 503 cases of training set is used to establish model, in June 2018 to October 2019 collected 217 cases of validation set is used to validate model, using logistic regression analysis for variable selection, and AIC method was used to construct the prediction model. Setting up the logistics of regression model with the AUC value evaluation model of degree of differentiation, the evaluation model to predict consistency, calibration diagram and draw Nomogram prediction model. All statistical analyses were performed using Empower Stats.

RESULTS : After screening, 9 independent variables were used to establish the prediction model. Two logistics regression prediction models suitable for different clinical conditions were established, in which the AUC value of model 1 in the validation group was 0.828.The AUC value of model 2 was 0.801 in the validation group. The calibration diagram shows that the predictive power of model 1 and model 2 is good.

CONCLUSIONS : Combined with clinically available indicators, the two predictive models and Nomogram can make accurate predictions of glycated hemoglobin control in patients, helping to improve the glycemic compliance rate in patients with type 2 diabetes.