基于常规指标的糖尿病足合并坏死性筋膜炎的风险评估

Risk assessment of diabetic foot with necrotizing fasciitis based on conventional indicators

目的探讨常规指标在糖尿病足(DF)合并坏死性筋膜炎(NF)风险评估模型中的应用价值。方法本研究为横断面研究。回顾性分析2018年10月至2022年4月就诊于重庆医科大学附属第一医院的815例DF患者,根据是否合并NF分为DF组(703例)和DF合并NF组(DNF组,112例),分析两组患者的临床资料(年龄、体重指数、病程、糖尿病并发症、足部皮肤坏死、体温、血常规、肌酐、纤维蛋白原、白蛋白、糖化血红蛋白、超敏C反应蛋白等),将具有统计学意义及临床意义的指标转换为分类变量,以DNF为因变量进行多因素logistic回归建立新的DNF风险评估(NRADNF)模型并绘制列线图,采用受试者工作特征(ROC)曲线下面积(AUC)评估模型区分度,Hosmer-Lemeshow拟合优度检验评估模型准确度,决策曲线分析评估模型有效性,Bootstrap重复抽样评估模型稳定性。结果构建的NRANDF模型包含6个指标:年龄≥60岁(OR=0.378,P<0.001)、体温≥38℃(OR=3.119,P=0.001)、足部皮肤坏死(OR=1.974,P=0.012)、中性粒细胞/淋巴细胞比值≥8.5(OR=2.500,P=0.002)、超敏C反应蛋白>20 mg/L(OR=2.455,P=0.006)、血红蛋白≤100 g/L(OR=1.833,P=0.026)。该模型AUC(95%CI)为0.815(0.773,0.857),Hosmer-Lemeshow拟合优度良好(χ^(2)=8.736,P=0.272);临床决策曲线分析显示,该模型具有临床净获益,且采用Bootstrap进行内部验证显示重复抽样前后ROC曲线基本重合。结论构建的NRADNF模型具有良好的预测效能和临床获益,且所纳入的指标易于获得,可在临床中运用进行DNF的初步筛查。

Objective To explore the application value of common indicators in the risk assessment model for diabetic foot(DF)complicated with necrotizing fasciitis(NF).Methods This was a cross-sectional study.A retrospective analysis was performed on 815 patients with DF,who were admitted to the First Affiliated Hospital of Chongqing Medical University from October 2018 to April 2022.Patients were divided into DF group(703 cases)and DF with NF group(DNF,112 cases)according to whether NF was combined.Clinical data(age,body mass index,course of disease,diabetic complications,necrosis of the foot skin,body temperature,blood routine,creatinine,fibrinogen,albumin,glycated hemoglobin A 1c,hypersensitive C-reactive protein,etc.)of the two groups were analyzed,and indicators of statistical and clinical significance were converted into categorical variables.Multivariate logistic regression was used to develop a new risk assessment model for DNF(NRADNF)and to construct a nomogram.Model discriminability was assessed by area under receiver operating characteristic(ROC)curve(AUC),accuracy was evaluated by Hosmer-Lemeshow goodness-fit test,effectiveness was evaluated by decision curve analysis,and stability was evaluated by bootstrap repeated sampling.Results The constructed NRANDF model includes 6 indicators:age≥60 years old(OR=0.378,P<0.001),body temperature≥38℃(OR=3.119,P=0.001),foot skin necrosis(OR=1.974,P=0.012),neutrophil-lymphocyte ratio≥8.5(OR=2.500,P=0.002),hs-CRP>20 mg/L(OR=2.455,P=0.006),hemoglobin≤100 g/L(OR=1.833,P=0.026).The AUC(95%CI)was 0.815(0.773,0.857),and Hosmer-Lemeshow goodness-fit test showed that the model performed well overall(χ^(2)=8.736,P=0.272).The result of decision curve analysis suggested that the model had a net clinical benefit,and internal verification using bootstrap showed that ROC curves were essentially identical before and after repeated sampling.Conclusion The NRADNF model has good predictive efficiency,clinical utility and the included indicators are easy to obtain,it can be used in the clinic for preliminary screening of DNF.