基于机器学习构建传染性单核细胞增多症患儿持续发热预测模型-现代医学

基于机器学习构建传染性单核细胞增多症患儿持续发热预测模型

单位：1. 南京医科大学附属儿童医院感染科, 江苏南京 210008;
2. 南京医科大学附属儿童医院护理部, 江苏南京 210008

分类号：R725.1

出版年·卷·期（页码）：2026·54·第二期（191-200）

摘要：

目的： 基于多维临床与实验室数据，利用多种机器学习算法构建传染性单核细胞增多症(IM)患儿持续性发热的风险预测模型，并对模型性能及关键危险因素进行解释与可视化分析，为临床早期识别与个体化管理提供参考。方法： 回顾性收集南京医科大学附属儿童医院2024年4月至2025年3月期间收治的499例确诊IM患儿入院时的临床及实验室资料。根据发热持续时间将患儿分为持续性发热组(>7 d)和短暂性发热组(≤7 d)。按8∶2比例随机划分为训练集(399例)和测试集(100例)。在训练集中采用最小绝对收缩与选择算子(LASSO)回归并结合交叉验证筛选关键特征，构建Logistic回归、支持向量机(SVM)、决策树、随机森林、梯度提升决策树(GBDT)、极端梯度提升(XGBoost)及轻量级梯度提升机(LightGBM)共7种预测模型。采用受试者工作特征曲线下面积(AUC)及多项分类性能指标评估模型判别能力，并通过决策曲线分析(DCA)评价模型的临床应用价值；利用沙普利可加性解释(SHAP)方法对最优模型进行可解释性分析。结果： 499例IM患儿中，持续性发热212例(42.48%)，短暂性发热287例(57.52%)。LASSO回归最终筛选出17个与持续性发热相关的特征用于模型构建。训练集中，XGBoost和LightGBM模型的AUC较高；在测试集中，XGBoost模型表现最佳，AUC为0.753(95% CI 0.69～0.82)。SHAP分析显示，发热峰值、白蛋白、中性粒细胞绝对计数、年龄、白蛋白/球蛋白比值、血小板计数、CD4⁺ T细胞百分比及肝肿大是预测IM患儿持续性发热的重要特征。结论： 基于机器学习构建的IM患儿持续性发热风险预测模型中，XGBoost模型具有较优的判别性能和良好的可解释性。多项临床表现、炎症及免疫功能相关指标与持续性发热密切相关，该模型有望辅助临床医生进行早期风险评估和分层管理，为IM患儿的精准诊疗提供量化支持。

Objective: To develop machine learning-based models for predicting the risk of persistent fever in children with infectious mononucleosis(IM) using multidimensional clinical and laboratory data, and to interpret key risk factors through visualization, providing a reference for early clinical identification and individualized management. Methods: Clinical and laboratory data obtained at hospital admission from 499 children diagnosed with IM at Nanjing Children's Hospital between April 2024 and March 2025 were retrospectively collected. Persistent fever was defined as fever lasting more than 7 days, while fever lasting ≤7 days was classified as transient. Patients were randomly divided into a training set(n=399) and a test set(n=100) at an 8∶2 ratio. In the training set, key features were selected using least absolute shrinkage and selection operator(LASSO) regression combined with cross-validation. Seven machine learning models were developed, including logistic regression, support vector machine(SVM), decision tree, random forest, gradient boosting decision tree(GBDT), extreme gradient boosting(XGBoost), and light gradient boosting machine(LightGBM). Model performance was evaluated using the area under the receiver operating characteristic curve(AUC) and other classification metrics. Clinical utility was assessed by decision curve analysis(DCA). Shapley additive explanations(SHAP) were applied to interpret the optimal model. Results: Among the 499 children with IM, 212(42.48%) experienced persistent fever, and 287(57.52%) had transient fever. Seventeen features identified by LASSO regression were included for model construction. In the training set, XGBoost and LightGBM models showed relatively high discriminative performance. In the test set, the XGBoost model achieved the best performance, with an AUC of 0.753(95% CI 0.69-0.82). SHAP analysis revealed that peak body temperature, serum albumin, absolute neutrophil count, age, albumin-to-globulin ratio, platelet count, CD4⁺ T-cell percentage, and hepatomegaly were the most important contributors to persistent fever. Conclusion: Among the evaluated machine learning, the XGBoost model demonstrated superior discriminative performance and good interpretability for predicting persistent fever in children with IM. Multiple clinical features and inflammation-and immune-related indicators were closely associated with fever persistence. This model may serve as a quantitative tool to assist early risk assessment and stratified management in clinical practice.

参考文献：

[1] DUNMIRE S K,VERGHESE P S,BALFOUR H H Jr.Primary Epstein-Barr virus infection[J].J Clin Virol,2018,102:84-92.
[2] SYLVESTER J E,BUCHANAN B K,SILVA T W.Infectious mononucleosis:rapid evidence review[J].Am Fam Physician,2023,107(1):71-78.
[3] LIU M,WANG X,ZHANG L,et al.Epidemiological characteristics and disease burden of infectious mononucleosis in hospitalized children in China:a nationwide retrospective study[J].Virol Sin,2022,37(5):637-645.
[4] HU H,DENG H,BI J,et al.Clinical characteristics and effectiveness of antiviral agents in hospitalized children with infectious mononucleosis in China:a multicenter retrospective study[J].Pediatr Investig,2021,5(3):188-194.
[5] XIONG G,ZHANG B,HUANG M Y,et al.Epstein-Barr virus(EBV) infection in Chinese children:a retrospective study of age-specific prevalence[J].PLoS One,2014,9(6):e99857.
[6] TIAN S,ZHENG J,ZHOU Z,et al.Retrospective review of children hospitalized for Epstein-Barr virus-related infectious mononucleosis[J].Pathogens,2025,14(7):702.
[7] ZHOU J,CAO J,DENG B.Risk factors for persistent fever in children with infectious mononucleosis[J].Pediatr Int,2023,65(1):e15476.
[8] 邱敏,季然.传染性单核细胞增多症患儿60例临床观察[J].蚌埠医学院学报,2017,42(8):1083-1086.
[9] 倪树芳,周凯.儿童传染性单核细胞增多症的临床特征及预后[J].中华实验和临床感染病杂志(电子版),2015,9(6):51-54.
[10] HUANG C J,DRAZEN J M.Artificial intelligence and machine learning in clinical medicine[J].N Engl J Med.2023;388(13):1201-1208.
[11] OBERMEYER Z,EMANUEL E J.Predicting the future-big data,machine learning,and clinical medicine[J].N Engl J Med,2016,375(13):1216-1219.
[12] ISHIHARA S,OKADA S,WAKIGUCHI H,et al.Chronic active Epstein-Barr virus infection in children in Japan[J].Acta Paediatr,1995,84(11):1271-1275.
[13] XIAO R,GENG L,CHANG K,et al.Clinical characteristics and analysis of risk factors for concomitant liver damage of infectious mononucleosis in children:a single-center retrospective study[J].Altern Ther Health Med,2024,30(9):124-133.
[14] ZHANG L,ZHOU P,MENG Z,et al.Infectious mononucleosis and hepatic function[J].Exp Ther Med,2018,15(3):2901-2909.
[15] 石惠,周夏袷,何伟,等.抗疱疹病毒核苷类药物对成人传染性单核细胞增多症有效性的研究[J].现代医学,2023,51(5):660-665.
[16] ZHANG C,CUI S,MAO G,et al.Clinical characteristics and the risk factors of hepatic injury in 221 children with infectious mononucleosis[J].Front Pediatr,2022,9:809005.
[17] ZHANG Y,ZHAO Y,JIANG Y,et al.Effects of Epstein-Barr virus infection on liver function in children[J].J Infect Public Health,2020,13(2):260-265.
[18] WANG C,SABOOR S,ZHANG Y,et al.Comprehensive insights into pediatric infectious mononucleosis:a retrospective study[J].J Infect Dev Ctries,2025,19(9):1359-1369.
[19] HUANG S C,CHEN J S,CHENG C N,et al.Hypoalbuminaemia is an independent predictor for hemophagocytic lymphohistiocytosis in childhood Epstein-Barr virus-associated infectious mononucleosis[J].Eur J Haematol,2012,89(5):417-422.
[20] 吕洁,金莲花,孙景辉,等.儿童传染性单核细胞增多症临床特点与发病年龄的关系:附312例分析[J].临床儿科杂志,2011,29(6):518-521.
[21] LI Y,WANG K.Clinical analysis of 163 pediatric patients with infectious mononucleosis:a single-center retrospective analysis[J].Immun Inflamm Dis,2024,12(9):e70020.
[22] NAUMENKO V,TURK M,JENNE C N,et al.Neutrophils in viral infection[J].Cell Tissue Res,2018,371(3):505-516.
[23] LAROCHELLE B,FLAMAND L,GOURDE P,et al.Epstein-Barr virus infects and induces apoptosis in human neutrophils[J].Blood,1998,92(1):291-299.
[24] DE MELO SILVA J,DE CASTRO ALVES C E,PONTES G S.Epstein-Barr virus:the mastermind of immune chaos[J].Front Immunol,2024,15:1297994.
[25] SON K H,SHIN M Y.Clinical features of Epstein-Barr virus-associated infectious mononucleosis in hospitalized Korean children[J].Korean J Pediatr,2011,54(10):409-413.
[26] 齐双月.儿童传染性单核细胞增多症中性粒细胞的变化及影响因素[D].大连:大连医科大学,2023.

服务与反馈：

【文章下载】【发表评论】【查看评论】【加入收藏】

提示：您还未登录，请登录！点此登录