摘要目的:利用多中心数据对本课题组前期单中心大样本的回顾性研究建立的基于多因素逻辑回归的Nomogram模型进行验证及优化。方法:回顾性分析国内7家儿童医学中心接受诊治的1 314例新生儿梗阻性黄疸患儿的临床资料。收集患儿的一般资料及术前临床指标，包括性别、体重、直接胆红素(direct bilirubin，DB)、碱性磷酸酶(alkaline phosphatase，ALP)、γ-谷氨酰转肽酶(gamma-glutamyl transferase，GGT)。根据患儿术中胆道造影结果及术后病理学检查结果，将其分为胆道闭锁(biliary atresia，BA)组(BA组)和非BA组，其中BA组1 131例，非BA组183例。另外，在1 314例患儿资料中以其中500例患儿的资料作为模型组，建立优化的、基于多因素逻辑回归的Nomogram模型，余814例患儿资料作为验证组，同时在此次建模的基础上对前期模型及本次优化后的模型进行外部验证。结果:采用单因素逻辑回归分析发现，BA组和非BA组在性别、体重、DB和GGT等4个变量上存在显著差异，差异具有统计学意义( P<0.05)，故以此作为预测因子，建立了优化的Nomogram诊断模型，其具有较好的鉴别效能，截断值为0.745，受试者工作特征(receiver operating characteristic，ROC)曲线下面积(area undercurve，AUC)值为0.8 358，灵敏度为80.8%，特异度为71.72%，阳性预测值(positivepredictive value，PPV)为92.55%，优于GGT单项指标的诊断效能。外部验证的结果提示，与前期模型相比，其更稳定，诊断效能更高，灵敏度为81.51%，特异度为90.48%，PPV为98.67%。 结论:基于多中心的优化Nomogram诊断模型在BA的术前诊断中有更稳定的预测效能，具有很好的临床应用前景，可基于此开发计算工具投入使用，为基层非专科医生提供早期鉴别诊断的新方法。

abstractsObjective:The overlapping features of biliary atresia (BA) and neonatal cholestasis (NC) with different causes (non-BA) have posed practical challenges for making a preoperative diagnosis of BA. A novel Nomogram model was developed on the basis of a retrospective analysis at a single center. The goal was validating and developing a better Nomogram model for BA using multi-center data.Methods:Retrospective analysis was performed for clinical data of 1, 314 children with neonatal obstructive jaundice (NOJ) at 7 domestic clinical centers. Multivariate logistic regression-based nomogram for BA was created and externally validated.Results:As previously, four predictors, including gender, weight, direct bilirubin (DB), alkaline phosphatase (ALP) and gamma-glutamyl transpeptidase (GGT), were significantly different between BA and non-BA groups ( P<0.05). The area under receiver operating characteristic (ROC) curve (AUC) value for Nomogram was 0.8358 at a cut-off value of 0.745. And it was greater than that of a single biomarker in predicting BA. The values of sensitivity, specificity and PPV were 80.8%, 71.72% and 92.55% respectively. External validation showed that optimized Nomogram was better and stable in predicting BA. The values of sensitivity, specificity and PPV were 81.51%, 90.48% and 98.67% respectively. Conclusions:Optimized nomogram offers better performance for predicting BA and holds a promise for future clinical application.