2型糖尿病足病预测模型的建立

Establishment of prediction model for type 2 diabetes mellitus complicated diabetic foot

目的使用常见的实验室及病例数据,建立一种中国人群中2型糖尿病患者的中晚期并发症糖尿病足病的预测模型,为糖尿病足病患者早期诊断提供一种改进的方法。方法选择2016年1月—2021年12月上海市第十人民医院收治的893例2型糖尿病患者为研究对象,符合纳入条件共405例,按是否患有糖尿病足病分为糖尿病足组(DF组,106例)和非糖尿病足组(非DF组,299例)两组。基于405例2型糖尿病患者的训练数据集建立预测模型,纳入39项常规实验室检查数据,采用单因素分析、LASSO回归、二元logistic回归分析,结合最小绝对收缩和选择算子回归模型中选择的变量建立预测模型。使用C指数、校正图和决策曲线分析评估预测模型的鉴别、校正和临床有用性。使用引导验证来评估内部验证。结果预测图中包含的预测因子有年龄、糖尿病病程、空腹血糖、糖化血红蛋白、甘油三酯、淋巴细胞计数、中性粒细胞计数。模型C指数为0.897,AUC为0.897(95%置信区间:0.863-0.932),具有较好的判别性和校正能力。结论结合年龄、糖尿病病程、空腹血糖、糖化血红蛋白、甘油三酯、淋巴细胞计数、中性粒细胞计数等因素的预测模型可方便2型糖尿病患者并发糖尿病足病的风险预测。

Objective To develop a predictive model for diabetic foot of intermediate and late complications of diabetic diseasein a Chinese population with type 2 diabetes by using common laboratory and case data,so as to provide an improved method for early diagnosis of patients with diabetic foot disease.Methods A total of 893 patients suffered with type 2 diabetes those were admitted to Shanghai tenth people’s hospital during January 2016 and December 2021 were enrolled in this study,and 405 cases were eligible for inclusion.They were divided into diabetic foot group(DF group)and non-diabetic foot group(non-DF group)according to whether they had diabetic foot disease.A predictive model was established based on the training data set of 405 patients with type 2 diabetes.Data from 39 routine laboratory examinations were included.Univariate analysis,LASSO regression,binary logistic regression analysis were used.The prediction model is established by combining the minimum absolute contraction and the selected variables in the selection operator regression model.The identification,calibration,and clinical usefulness of predictive models were evaluated using C-index,calibration diagram,and decision curve analysis.Use bootstrap validation to evaluate internal validation.Results Predictors included in the prognosticators were age,duration of diabetes,fasting GLU,HbA1c,TG,LY,NE.The C index of model was 0.897,andthe AUC was 0.897(95%confidence interval:0.863-0.932),which has good discriminative and corrective power.Conclusions Combined with age,duration of diabetes,fasting GLU,HbA1c,TG,LY,NE and other factors,the prediction model could facilitate the risk prediction of diabetic foot in patients with type 2 diabetes.