基于大样本的2型糖尿病Nomogram预测模型构建

Nomogram Prediction Model of Type 2 Diabetes Mellitus Based on Large Samples

目的:构建本地区2型糖尿病风险预测模型,对糖尿病进行预防和控制。方法:选择2011年1月至2016年12月在某综合性医院体检中心190 435例体检人群作为研究对象,采用Logistic回归分析法筛选2型糖尿病的影响因素,并应用Nomogram法对各个因素进行评分,构建预测模型,同时运用决策曲线分析(decision curve analysis,DCA)对简单评价模型和复杂评价模型进行比较。结果:多因素Logistic回归分析显示性别(OR=1.336,95%CI:1.315~1.366)、年龄(OR=1.065,95%CI:1.063~1.067)、体重指数(OR=1.117,95%CI:1.108~1.126)、高血压(OR=1.331,95%CI:1.263~1.403)、三酰甘油(OR=1.397,95%CI:1.366~1.430)、总胆固醇(OR=1.101,95%CI:1.068~1.135)、高密度脂蛋白胆固醇(OR=0.796,95%CI:0.729~0.869)、红细胞分布宽度(OR=0.786,95%CI:0.761~0.811)、尿素氮(OR=1.140,95%CI:1.120~1.160)为2型糖尿病独立的影响因素(P<0.01)。通过多因素Logistic回归结果建立Nomogram模型,经验证预测模型一致性指数良好(C-index=0.802,P<0.01)。DCA显示在阈值概率为0~0.22范围内,简单评价模型的净利益率高于复杂评价模型;在阈值概率为0.23~0.74范围内,复杂评价模型的净利益率则高于简单评价模型。结论:成功建立预测2型糖尿病的Nomogram模型。

Objective：To construct a risk prediction model of type 2 diabetes mellitus to prevent and control diabetes.Methods： A total of 190 435 population in the physical examination center of a comprehensive hospital were selected as the subjects from Jan 2011 to Dec 2016. The influencing factors of type 2 diabetes mellitus were screened by Logistic regression analysis,and each factor was scored by Nomogram method to construct the prediction model. The decision curve analysis （DCA） was applied to compare the simple and complex evaluation models. Results：Multivariate logistic regression analysis showed that sex （OR=1.336,95%CI：1.315-1.366）,age （OR=1.065,95%CI：1.063-1.067）,BMI （OR=1.117,95%CI：1.108-1.126）,HBP （OR=1.331,95%CI：1.263-1.403）,TC （OR=1.397,95%CI：1.366-1.430）,TG （OR=1.101,95%CI：1.068-1.135）,HDL-C OR=0.796,95%CI：0.729-0.869）,RDW-SD （OR=0.786,95%CI：0.761-0.811） and BUN （OR=1.140,95%CI：1.120-1.160） were independent risk factors for type 2 diabetes mellitus （P 〈0.01） . The Nomogram model was established based on the results of multivariate Logistic regression analysis. The consistency index of the model was good （C-index=0.802, P〈0.01） . DCA showed that the net benefit of the simple evaluation model was higher than that of complex evaluation model within the risk threshold of 0-0.22. The net benefit of the complex evaluation model was higher than that of the simple evaluation model within the risk threshold of 0.23-0.74. Conclusion：The Nomogram model for predicting type 2 diabetes mellitus is successfully established.