关键词: 糖尿病, 2型；老年人；糖化血红蛋白；预测模型；并发症

Abstract: Background　Achieving hemoglobin control can reduce the risk of complications of type 2 diabetes among elderly patients and the hemoglobin management is crucial in the daily management of type 2 diabetes.However，there has been increasingly evidence shows that the achieving rate of hemoglobin-control among elderly patients with type 2 diabetes are generally poor.However，the risk prediction models of hemoglobin can be generally used to predict high level hemoglobin risk for individual patients.Objective　To establish target predictive model and scale of HbA1c by exploring influencing factors of hemoglobin-control in elderly patients with type 2 diabetes，in order to provide a kind of versatile and reliable tool for diabetes management.Methods　The basic information and related laboratory indexes of 483 elderly patients with type 2 diabetes referred to the endocrinology department，Sichuan Provincial People's Hospital from March 2018 to December 2019 were collected by face-to-face questionnaire survey，including gender，age，occupation，nationality，marital status，educational level，body size，family history of diabetes，central obesity comparison，previous HbA1c values，the HbA1c test result，the duration of diabetes，fasting blood sugar monitoring situation，fasting blood glucose values，the present stage treatment，medication compliance，type of oral medication，whether to use insulin，glucose-lowering drugs daily cost，high blood pressure，high cholesterol，whether merger diabetes complications，daily exercise time，quality of sleep，diet control，and depression.According to the results of HbA1c examination，the patients were divided into：HbA1c up to standard group （224 cases） and HbA1c not up to standard group（259 cases）.Univariate and multivariate Lasso-Logistic regression were used for variable selection，and hemoglobin values were used as independent variables to build the predictive models and rating scales.Bootstrap method were used for internal validity analysis of the prediction models.The sensitivity，specificity as well as the area under the receiver operating characteristic（ROC）curve（AUC）was evaluated for predict accuracy.The score-occurrence probability graph was used to verify the performance of the rating scales and the best out of cut-points was determined.Results　The statistically significant differences was found in age，central obesity comparison，previous HbA1c value，duration of diabetes，fasting blood glucose，medication compliance，type of oral medication，whether to use insulin，diabetes complications，daily exercise time，diet control，and depression between normal hemoglobin value and high hemoglobin value between the two group（P<0.05）.Univariate Logistic regression analysis revealed that age，central obesity comparison，previous HbA1c value，duration of diabetes，fasting blood glucose，medication compliance，insulin use，daily exercise time and diet control was significantly correlated with hemoglobin-control（P<0.05）.Lasso-Logistic regression analysis revealed that age，previous HbA1c value，duration of diabetes，fasting blood glucose，medication compliance，insulin use，daily exercise time and diet control was significantly correlated with hemoglobin-control（P<0.05）.According to the result of Lasso-Logistic analysis，the logistic regression model is logit（P）=3.89+1.72×〔previous HbA1c（> 7%）〕+0.97×〔（2 years to 10 years）duration of diabetes〕 + 1.41×〔diabetes duration（>10 years）〕+1.51×〔fasting glucose values（≥7mmol）〕 +1.02×〔medication compliance（or poor）］ + 0.85×〔whether to use insulin（yes）〕 + 0.58×〔daily movement time（>0~0.5 h）〕+1.21×〔Daily exercise time（no exercise）］+1.09×〔diet controlled（no）〕 （if the condition in 〔〕 is satisfied，it is 1；if it is not satisfied，it is 0）.The discriminatory ability of this model and the results of internal validation showed that the model performed well.The area under the ROC curve was 0.86〔95%CI（0.83，0.89）〕.To derive a simple-to-compute risk score，regression coefficients were converted to weighted scores by dividing each regression coefficient by rounding to the nearest integer，then，a HbA1c rating score with a maximum score of 9 was obtained，ratings increased from 0 to 9 with increasing probability of high level hemoglobin risk.When the cut-off value was set at 0.56 according to the max value of the Youden index，the cut point of the rating table is 5 points，the sensitivity was 79.54%，specificity was 76.79%，and accuracy was equaled to 78.26%.Conclusion　By applying hemoglobin scoring model，clinicians can predict which elderly patients with type 2 diabetes may have a high risk of hemoglobin-control and thus help patients make the right decisions，so this scoring model has clinical application and promotion value.

Key words: Diabetes mellitus, type 2；Aged；Hemoglobin control；Predicting model；Complications