摘要
目的基于糖尿病专病数据库开发并验证一种糖尿病足溃疡(DFU)预测模型。方法该研究为横断面研究。基于华中科技大学同济医学院附属武汉中心医院糖尿病专病数据库,采用标准化流程收集2432例糖尿病患者的临床资料,如体重指数(BMI)及BMI≥24.0 kg/m^(2)、足部皮肤颜色异常、足动脉搏动减弱或消失、胼胝形成和足溃疡史的患者比例。采用简单随机抽样方法,按照7∶3的比例将糖尿病患者分为模型开发队列(1702例)和外部验证队列(730例)。通过多因素logistic回归分析法分析DFU的影响因素,基于多因素logistic回归分析筛选出的DFU独立影响因素构建列线图和预测模型,模型的区分度通过受试者工作特征(ROC)曲线的曲线下面积(AUC)评估,计算Brier评分以评估模型的整体性能,Brier评分越低表示预测越准确。此外,使用C指数评估模型的区分能力,其值越接近1,表明模型的判别能力越强。结果在模型开发队列中,DFU患者ROC曲线的AUC为0.815(95%CI 0.807~0.823,P<0.001),Brier评分为0.127;在外部验证队列中,DFU患者的ROC曲线的AUC为0.807(95%CI 0.794~0.820,P<0.001),Brier评分为0.131。多因素logistic回归分析结果显示,BMI≥24.0 kg/m^(2)(OR=1.56,95%CI 1.32~1.85,P<0.001)、足部皮肤颜色异常(OR=2.14,95%CI 1.76~2.60,P<0.001)、足动脉搏动减弱或消失(OR=1.89,95%CI 1.58~2.26,P<0.001)、胼胝形成(OR=2.37,95%CI 1.95~2.88,P<0.001)和足溃疡史(OR=3.42,95%CI 2.81~4.16,P<0.001)均是DFU的独立影响因素。基于上述识别的独立影响因素,构建DFU预测模型列线图。对列线图评估模型预测性能的区分能力分析结果显示,列线图在模型开发队列和外部验证队列中均表现出较好的区分能力,C指数分别为0.816(95%CI 0.808~0.824,P<0.001)和0.809(95%CI 0.796~0.822,P<0.001)。根据列线图得分,将糖尿病患者划分为低危(<160分)、中危(160~240分)和高危(>240分)3组,3组糖尿病患者DFU的实际发生率分别为2.5%(10/400)、8.8%(22/250)和23.2%(58/250),组间差异有统计学意义(χ^(2)=73.80,P<0.001)。结论该研究基于糖尿病专病数据库构建了一种准确性高、性能优良预测DFU的模型。
Objective To develop and validate a prediction model for diabetic foot ulcer(DFU)based on a diabetes database.Methods This cross-sectional study was based on the diabetes database in Central Hospital of Wuhan,Tongji Medical College,Huazhong University of Science and Technology.Clinical data of 2432 diabetic patients were collected using standardized procedures,including body mass index(BMI),the proportion of patients of BMI≥24.0 kg/m^(2),abnormal foot skin color,weak or absent foot artery pulsation,callus formation,and history of foot ulcers.Using simple random sampling method,2432 diabetic patients were divided into a model development cohort(n=1702)and an external validation cohort(n=730)in a ratio of 7∶3.Multivariate logistic regression analysis was used to identify independent risk factors for DFU.The discriminative ability of the model was assessed using the area under curve(AUC)of the receiver operating characteristic(ROC).Brier score was calculated to evaluate the overall performance of the model,with lower scores indicating better prediction accuracy.In addition,concordance index(C-index)was used to assess the discriminative ability of the model,with values closer to 1 indicating stronger discriminative ability.Results In the model development cohort,the AUC of DFU patients was 0.815(95%CI 0.807-0.823,P<0.001),and the Brier score was 0.127.In the external validation cohort,the AUC was 0.807(95%CI 0.794-0.820,P<0.001),and the Brier score was 0.131.Multivariate logistic regression analysis identified five independent risk factors for DFU:BMI≥24.0 kg/m^(2)(OR=1.56,95%CI 1.32-1.85,P<0.001),abnormal foot skin color(OR=2.14,95%CI 1.76-2.60,P<0.001),weak or absent foot artery pulsation(OR=1.89,95%CI 1.58-2.26,P<0.001),callus formation(OR=2.37,95%CI 1.95-2.88,P<0.001),and history of foot ulcers(OR=3.42,95%CI 2.81-4.16,P<0.001).A prediction nomogram for DFU was constructed based on these identified independent risk factors.The nomogram showed good discriminative ability in both the model development cohort[C-index:0.816(95%CI 0.808-0.824,P<0.001)]and external validation cohort[C-index:0.809(95%CI 0.796-0.822,P<0.001)].According to the nomogram scores,patients were classified into low-risk(<160 points),moderate-risk(160-240 points),and high-risk(>240 points)groups.The actual incidence rates of DFU in these three groups were 2.5%(10/400),8.8%(22/250),and 23.2%(58/250),respectively,with statistically significant differences between groups(χ^(2)=73.80,P<0.001).ConclusionThis study has developed a highly accurate and well-performing model for predicting DFU based on a diabetes database.
作者
杨帆
徐子辉
赵玉
付秀立
谭琴
王中京
Yang Fan;Xu Zihui;Zhao Yu;Fu Xiuli;Tan Qin;Wang Zhongjing(Department of Endocrinology,the Central Hospital of Wuhan,Tongji Medical College,Huazhong University of Science and Technology,Diabetic Foot Center,Wuhan Diabetes Clinical Research Center,Wuhan 430014,China;School of Medicine,Jianghan University,Wuhan 430056,China)
出处
《中华糖尿病杂志》
北大核心
2025年第7期824-829,共6页
CHINESE JOURNAL OF DIABETES MELLITUS
基金
武汉市卫生健康委员会科研项目(WX23A55)。
关键词
糖尿病
糖尿病足溃疡
专病数据库
预测模型
Diabetes mellitus
Diabetic foot ulcer
Specialized disease database
Prediction model
