小儿急性穿孔性阑尾炎的列线图预测模型构建及验证

Nomogram prediction model construction and verification forpediatric acute perforation appendicitis

目的探讨小儿急性穿孔性阑尾炎的危险因素,构建列线图预测模型并验证。方法选取2020年6月30日至2022年6月30日该院行阑尾切除术的426例急性阑尾炎患儿为研究对象,2020年6月30日至2022年2月28日住院的340例急性阑尾炎患儿为训练集,2022年3月1日至2022年6月30日住院的86例急性阑尾炎患儿为验证集。采用单因素和多因素logistic回归模型分析小儿急性穿孔性阑尾炎的独立危险因素,构建列线图预测模型。受试者工作特征(ROC)曲线及校准曲线评估模型的预测效能,决策曲线分析(DCA)评价模型的应用价值。结果426例患儿中穿孔198例,未穿孔228例。单因素及多因素logistic回归分析显示,C反应蛋白(CRP)升高、阑尾腔有粪石、起病至就诊时间≥2 d和体温≥37.3℃是小儿急性穿孔性阑尾炎的独立危险因素(P<0.05)。Hosmer-Lemeshow检验显示,列线图预测模型拟合良好(P=0.869)。训练集和验证集的曲线下面积(AUC)分别为0.808和0.860,模型具有良好的预测能力。校准曲线接近理想的对角线,模型有较好的区分度、一致性及准确性。决策曲线分析(DCA)显示,曲线远离斜线及水平线,模型具有较好的临床实用性。结论构建的小儿急性穿孔性阑尾炎列线图预测模型具有良好的预测能力,可帮助临床尽早识别。

Objective To investigate the risk factors for pediatric acute perforation appendicitis,and to construct a nomogram predictive model and conduct the verification.Methods A total of 426 children patients with appendectomy in this hospital from June 30,2020 to June 30,2022 were selected as the study subjects 340 children with acute appendicitis admitted to the hospital from 30 June 2020 to 28 February 2022 were the training set and 86 children patients with appendicitis hospitalized in this hospital from March 1,2022 to June 30,2022 conducted the external validation(verification set).The univariate and multivariate logistic regression models were employed to analyze the independent risk factors of pediatric acute perforation appendicitis.The nomograms predictive model was constructed.The receiver operating characteristic(ROC)curve and calibration curve were used to evaluate the predictive efficiency of the model.The decision curve analysis(DCA)was used to evaluate the application value of the model.Results Of the 426 children,198 were perforated and 228 were not perforated.The univariate and multivariate logistic regression analyses revealed that elevated C-reactive protein(CRP),presence of stercorolith in appendiceal cavity,time of onset to visiting hospital≥2 d and body temperature≥37.3℃were the independent risk factors for pediatric acute perforation appendicitis(P<0.05).The Hosmer-Lemeshow test demonstrated that the nomogram predictive model had good fitting(P=0.869),and the area under the curve(AUC)for the training and validation sets were 0.808 and 0.860 respectively,showing the good predictive ability of the model.The calibration curve closely approach the ideal diagonal.The model showed good discrimination,consistency and accuracy.The DCA revealed that the curve was far away from oblique and horizontal lines,and the model had good clinical practicability.Conclusion The constructed nomogram model of pediatric acute perforation appendicitis has good predictive ability and may help clinic to identify as early as possible.