小儿川崎病诊断模型的建立与诊断效能研究

Establishment and research of Kawasaki disease diagnosis model

目的建立小儿川崎病鉴别诊断模型,为临床诊断决策提供参考。方法选取上海市儿童医院2013年1月至2017年7月入院接受治疗的发热患儿,根据最终临床诊断将其分为川崎病组与发热对照组,回顾性分析比较两组间一般临床信息及实验室指标,通过Logistic回归分析建立数学模型并评价其效能。结果本研究共纳入1916例患儿,平均年龄为(3.47±2.83)岁。其中,男性1085(56.6%)例,女性831(43.4%)例,川崎病组患儿479(25.0%)例,发热对照组患儿共1099(75.0%)例。采用向后逐步回归的方法,将因变量和自变量纳入Logistic回归分析中。结果显示,此模型的Hosmer and Lemeshow检验P=0.944,说明拟合方程与真实方程无偏差;年龄、发热天数、红细胞沉降率、C反应蛋白、白细胞计数、白蛋白、D-D二聚体水平是川崎病的独立危险因素。Logistic回归的预测方程为:ln(P/1-p)=-7.337+2.163×CRP+1.56×DD+1.612×ESR+1.392+年龄+1.724×发热天数+2.295×WBC+0.808×ALB。计算患者模型得分,并做ROC曲线,曲线下面积为0.927(95%CI:0.905-0.950),当得分为9分时,约登指数最高,为72.9%,将得分为9分设置为截断点,预测是否川崎病的灵敏度和特异度分别为89.7%和83.2%。结论本研究建立的小儿川崎病鉴别诊断模型具有较好的诊断效能,有待于进一步大规模、多中心研究加以证实。

Objective To establish a Kawasaki disease mathematical diagnosis model in order to support clinical decision-making.Methods Children with fever admitted to Shanghai Children's Hospital from January 2013to July 2017were recruited and were divided into Kawasaki disease group and other disease control groups according to the final clinical diagnosis.The general clinical information and laboratory indicators were compared,a mathematical model was established and evaluated through the logistic regression analysis.Results A total of 1916children were enrolled in this study,with an average age of 3.47±2.83years.Of these,1085 (56.6%)were male,831(43.4%)were female,479(25.0%)were diagnosed with Kawasaki disease and 1099(75.0%)were with other diagnosis.Logistic regression analysis included dependent variables and independent variables,and the results showed that the Hosmer and Lemeshow test of this model was P =0.944,the difference was not significant,indicating that the fitting equation and the true equation without deviation;age , fever days,ESR,CRP,WBC,ALB and DD dimers were independent risk factors for Kawasaki disease.The predictive equation of Logistic regression is:ln(P/1-p)=-7.337+2.163×CRP+1.56×DD+1.612×ESR+1.392+age+1.724×days of fever+2.295×WBC+0.808×ALB .The patient model score and the ROC curve was calculated.The area under the curve was 0.927(95%CI:0.905-0.950).When the score was 9,the Youden index was the highest(72.9%),the sensitivity and specificity were 89.7%and 83.2%.Conclusion The Kawasaki disease diagnosis mathematical model established in this study has good diagnostic efficacy,which need to be confirmed by further large-scale,multicenter studies.