基于随机森林算法的糖尿病周围神经病变预测模型构建与验证

Blood hypercoagulation state lower limb deep vein thrombosis construction and validation of an early warning model based on random forest algorithm in diabetic peripheral neuropathy

目的探讨T2DM合并糖尿病周围神经病变(DPN)的影响因素,构建基于随机森林算法的DPN预测模型并进行验证。方法选取2019年1月至2021年12月于我院住院治疗的T2DM患者512例,根据是否合并DPN分为单纯T2DM组(n=292)和T2DM合并DPN组(DPN,n=220)。比较两组一般资料及生化指标,Logistic回归分析T2DM合并DPN的影响因素,构建随机森林模型。结果与T2DM组比较,DPN组白细胞计数、HbA1c、体重下降速度、DR患病率升高(P<0.05),DM病程≥10年、TG、HDL-C降低(P<0.05)。Logistic回归分析显示,年龄≥60岁、HbA1c、TG、HDL-C、体重下降速度、DR是T2DM合并DPN的影响因素。随机森林模型显示,树的数量为387时错误率最低,T2DM合并DPN的影响因素重要性排序为体重下降速度、TG、DR、HDL-C、HbA1c及年龄≥60岁。结论年龄≥60岁、HbA1c、TG、HDL-C、体重下降速度是T2DM合并DPN的影响因素,有助于临床早期诊治。

Objective To explore the influencing factors of type 2 diabetes mellitus(T2DM)merging with diabetic peripheral neuropathy(DPN),and to construct and verify a prediction model based on random forest algorithm.Methods 512 T2DM patients who were hospitalized in our hospital from January 2019 to December 2021 were divided into simple T2DM group(n=292)and T2DM combined with DPN group(DPN,n=220)based on whether or not DPN was present.The general data and biochemical indicators of the two groups were compared.Logistic regression analysis was conducted to identify the influencing factors of DPN in T2DM patients.A random forest model was constructed.Results Compared with the T2DM group,the DPN group showed an increase in weight loss rate,incidence of diabetic retinopathy(DR),WBC and HbA1c(P<0.05),with decrease in DM duration≥10 years,TG and HDL-C(P<0.05).Logistic regression analysis showed that age≥60 years,HbA1c,TG,HDL-C,rate of weight loss,DR were influencing factor for T2DM combined with DPN.The random forest model showed that when the number of trees was 387,the error rate was the lowest.The importance ranking of the influencing factors of T2DM combined with DPN were the rate of weight loss,TG,DR,HDL-C,HbA1c and age≥60 years.Conclusions Age≥60 years,HbA1c,TG,HDL-C,rate of weight loss and DR are influencing factors for T2DM combined with DPN,that can be used for early clinical diagnosis and treatment.