川崎病患儿冠状动脉病变的影响因素分析及预测模型构建

doi:10.3760/cma.j.cn441417-20240910-01013

摘要/Abstract

摘要：

目的　分析川崎病患儿冠状动脉病变（CAL）的影响因素，并构建预测模型。方法　回顾性分析2020年1月至2024年1月神木市医院收治的306例川崎病患儿临床资料，根据患儿是否合并CAL将其分为CAL组（108例）和未合并CAL组（198例）。CAL组中男73例，女35例，年龄≤2岁69例，2岁39例。未合并CAL组中男117例，女81例，年龄≤2岁105例，2岁93例。采用多因素logistic回归分析法分析川崎病患儿CAL的影响因素。按照7∶3比例将306例川崎病患儿随机分为训练集（214例）和测试集（92例），基于训练集数据构建决策树模型，基于测试集数据验证决策树模型的预测效能。绘制受试者操作特征曲线（ROC）评价多因素logistic回归模型和决策树模型的预测效果。采用χ²检验进行统计分析。结果　CAL组持续发热时间10 d、血小板计数（PLT）300×10⁹/L、C反应蛋白50 mg/L、氨基末端脑钠肽前体（NT-proBNP）≥300 ng/L、肌酸激酶同工酶（CK-MB）18 U/L、红细胞沉降率（ESR）40 mm/h、静脉注射免疫球蛋白（IVIg）治疗2次占比分别为41.67%（45/108）、60.19%（65/108）、49.07%（53/108）、67.59%（73/108）、47.22%（51/108）、34.26%（37/108）、47.22%（51/108），未合并CAL组上述指标分别为14.65%（29/198）、45.96%（91/198）、35.86%（71/198）、36.87%（73/198）、32.83%（65/198）、16.67%（33/198）、26.77%（53/198），CAL组上述指标均高于未合并CAL组（均P0.05）。多因素logistic回归分析结果显示，持续发热时间10 d（比值比=3.449）、PLT300×10⁹/L（比值比=2.088）、C反应蛋白50 mg/L（比值比=1.883）、NT-proBNP≥300 ng/L（比值比=3.267）、CK-MB18 U/L（比值比=2.284）、ESR40 mm/h（比值比=2.599）、IVIg治疗2次（比值比=2.930）均为川崎病患儿发生CAL的危险因素（均P0.05）。基于训练集数据建立包括持续发热时间、NT-proBNP、IVIg治疗次数、ESR、CK-MB、PLT 6个变量的川崎病患儿CAL决策树模型，共提取7条分类规则，其中持续发热时间是该模型的首要影响因素。基于测试集数据对决策树模型和logistic回归模型进行验证，结果显示川崎病患儿CAL的决策树模型预测灵敏度、特异度、曲线下面积（AUC）分别为81.25%、86.67%、0.818，logistic回归模型预测灵敏度、特异度、AUC分别为62.50%、90.00%、0.662。结论　川崎病患儿CAL的决策树模型包括6个变量，分别为持续发热时间、NT-proBNP、IVIg治疗次数、ESR、CK-MB、PLT，其中持续发热时间是该模型的首要影响因素；该模型具有良好的预测效能。

关键词:

川崎病, 冠状动脉病变, 影响因素, 决策树, 儿童

Abstract:

Objective　To analyze the influencing factors of coronary artery lesion (CAL) in children with Kawasaki disease (KD) and construct predictive models. Methods The clinical data of 306 children with KD admitted to Shenmu City Hospital from January 2020 to January 2024 were retrospectively analyzed. According to whether the children had concurrent CAL, they were divided into a CAL group (108 cases) and a non-CAL group (198 cases). In the CAL group, there were 73 boys and 35 girls, 69 cases ≤2 years old and 39 cases 2 years old. In the non-CAL group, there were 117 boys and 81 girls, 105 cases ≤2 years old and 93 cases 2 years old. Multivariate logistic regression analysis was used to analyze the influencing factors of CAL in children with KD. According to the ratio of 7:3, 306 children with KD were randomly divided into a training set (214 cases) and a test set (92 cases). The decision tree model was constructed based on the training set data, and the prediction efficiency of the decision tree model was verified based on the test set data. The receiver operating characteristic curve (ROC) was drawn to evaluate the predictive effect of multivariate logistic regression model and decision tree model. χ² test was used for statistical analysis. Results The proportions of duration of fever 10 days, platelet count (PLT) 300×10⁹/L, C-reactive protein (CRP)50 mg/L, N-terminal brain natriuretic peptide precursor (NT-proBNP)≥ 300 ng/L, creatine kinase isoenzyme (CK-MB) 18 U/L, erythrocyte sedimentation rate (ESR)40 mm/h, and twice intravenous immunoglobulin (IVIg) treatment in the CAL group [41.67% (45/108), 60.19% (65/108), 49.07% (53/108), 67.59% (73/108), 47.22% (51/108), 34.26% (37/108), and 73.15% (79/108)] were higher than those in the non-CAL group [14.65% (29/198), 45.96% (91/198), 35.86% (71/198), 36.87% (73/198), 32.83% (65/198), 16.67% (33/198), and 52.02% (103/198)] (all P0.05). Multivariate logistic regression analysis showed that the duration of fever 10 days (odds ratio=3.449), PLT 300×10⁹/L (odds ratio=2.088), CRP 50 mg/L (odds ratio=1.883), NT-proBNP ≥300 ng/L (odds ratio=3.267), CK-MB 18 U/L (odds ratio=2.284), ESR 40 mm/h (odds ratio=2.599), and twice IVIg treatment (odds ratio=2.930) were all risk factors for CAL in children with KD (all P0.05). Based on the training set data, a decision tree model of CAL in children with KD was established, including 6 variables of duration of fever, NT-proBNP, IVIg treatment times, ESR, CK-MB, and PLT, and 7 classification rules were extracted, among which duration of fever was the primary influencing factor of the model. The logistic regression model and decision tree model were verified based on the test set data. The results showed that the sensitivity, specificity, and area under curve (AUC) of the decision tree model for predicting CAL in KD children were 81.25%, 86.67%, and 0.818, respectively; the sensitivity, specificity, and AUC of the logistic regression model were 62.50%, 90.00%, and 0.662, respectively. Conclusions The decision tree model for CAL in children with KD includes 6 variables, namely duration of fever, NT-proBNP, IVIg treatment times, ESR, CK-MB, and PLT. The duration of fever is the primary influencing factor of the model. The model has good predictive performance, which can provide reference for clinical screening of high-risk KD children with CAL.

Key words:

Kawasaki disease, Coronary artery lesion, Influencing factors, Decision tree, Children

刘娟王凤兰.

川崎病患儿冠状动脉病变的影响因素分析及预测模型构建 [J]. 国际医药卫生导报, 2025, 31(1): 60-65.

Liu Juan, Wang Fenglan.

Analysis of influencing factors and construction of predictive models for coronary artery lesion in children with Kawasaki disease [J]. International Medicine and Health Guidance News, 2025, 31(1): 60-65.

参考文献

[1] 谷惠茹. 0～6月龄婴儿川崎病血清BNP、PCT、IL-6的变化及临床意义[J]. 国际医药卫生导报,2022,28(18):2549-2552. DOI:10.3760/cma.j.issn.1007-1245.2022. 18.007.
[2] Yang C, Yang Y, Cao S, et al. Kawasaki disease coronary artery lesions prediction with monocyte-to-high-density lipoprotein ratio[J]. Pediatr Res, 2023, 94(1):246-251. DOI: 10.1038/s41390-022-02401-4.
[3] 刘翠,刘小红,吴岳,等. 川崎病儿童冠脉病变的影响因素分析[J]. 西安交通大学学报(医学版),2022,43(6):845-849. DOI:10.7652/jdyxb202206008.
[4] 许艳鹏,赵京,刘彦平. 基于C反应蛋白表达预测川崎病并发冠状动脉损伤患儿风险的列线图模型构建[J]. 岭南心血管病杂志,2023,29(2):156-161,179. DOI:10.3969/j.issn.1007-9688.2023.02.08.
[5] Zhang X, Ya-Wang Shao, Ya-Lan Zhang, et al. Serum Lp-PLA2 level predicting coronary artery lesions in children with Kawasaki disease[J]. Heart Surg Forum, 2021, 24(4): E611-E618. DOI: 10.1532/hsf.3833.
[6] Ma XB, Xu QL, Li N, et al. A decision tree model to distinguish between benign and malignant pulmonary nodules on CT scans[J]. Eur Rev Med Pharmacol Sci, 2023, 27(12): 5692-5699. DOI: 10.26355/eurrev_202306_32809.
[7] Banihashem SY, Shishehchi S. Ontology-based decision tree model for prediction of fatty liver diseases[J]. Comput Methods Biomech Biomed Engin, 2023, 26(6): 639-649. DOI: 10.1080/10255842.2022.2081502.
[8] McCrindle BW, Rowley AH, Newburger JW, et al. Diagnosis, treatment, and long-term management of Kawasaki disease: a scientific statement for health professionals from the American Heart Association[J]. Circulation, 2017, 135(17):e927-e999. DOI: 10.1161/CIR.0000000000000484.
[9] 中华医学会儿科学分会心血管学组,中华儿科杂志编辑委员会. 川崎病冠状动脉病变的临床处理建议(2020年修订版)[J]. 中华儿科杂志,2020,58(9):718-724. DOI:10.3760/cma.j.cn112140-20200422-00421.
[10] 陕西省川崎病诊疗中心/陕西省人民医院儿童病院,国家儿童医学中心/首都医科大学附属北京儿童医院,上海交通大学医学院附属儿童医院,等. 中国儿童川崎病诊疗循证指南(2023年)[J]. 中国当代儿科杂志,2023,25(12):1198-1210. DOI:10.7499/j.issn.1008-8830.2309038.
[11] 唐利瑛,徐震,陈晓霞. 川崎病患儿不同程度冠脉病变的相关因素分析及预测模型构建[J]. 中国妇幼保健,2022,37(16):2976-2980. DOI:10.19829/j.zgfybj.issn.1001- 4411.2022.16.020.
[12] Philip S, Jindal A, Krishna Kumar R. An update on understanding the pathophysiology in Kawasaki disease: possible role of immune complexes in coronary artery lesion revisited[J]. Int J Rheum Dis, 2023, 26(8):1453- 1463. DOI: 10.1111/1756-185X.14816.
[13] 杨艳章,王莉,杨艳芳,等. 川崎病患儿并发冠状动脉病变相关危险因素与治疗分析[J]. 中国妇幼保健,2019,34(14):3244-3246. DOI:10.7620/zgfybj.j.issn.1001-4411.2019. 14.31.
[14] Chen J, Li J, Yue YH, et al. Nomogram for predicting coronary artery lesions in patients with Kawasaki disease[J]. Clin Cardiol, 2023, 46(11): 1434-1441. DOI: 10.1002/clc.24113.
[15] 邴振,王先东,单光颂,等. C反应蛋白、降钙素原升高及血钾降低与重症川崎病发生发展高危因素及临床预测价值评价[J]. 中国实用儿科杂志,2023,38(2):129-135. DOI:10.19538/j.ek2023020611.
[16] 龙超,张睿,黄飞,等. 川崎病并发冠状动脉病变风险的列线图模型的构建与分析[J]. 实用临床医药杂志,2022,26(10):31-35. DOI:10.7619/jcmp.20213367.
[17] Ae R, Abrams JY, Maddox RA, et al. Platelet count variation and risk for coronary artery abnormalities in Kawasaki disease[J]. Pediatr Infect Dis J, 2020, 39(3): 197-203. DOI: 10.1097/INF.0000000000002563.
[18] 何永,莫国梁,康迁,等. 川东北地区川崎病患者并发冠状动脉病变的预测因素[J]. 西部医学,2024,36(2):286-290. DOI:10.3969/j.issn.1672-3511.2024.02.024.

[19] Shuai S, Zhang H, Zhang R, et al. Prediction of coronary artery lesions based on C-reactive protein levels in children with Kawasaki disease: a retrospective cohort study[J]. J Pediatr (Rio J), 2023, 99(4):406-412. DOI: 10.1016/j.jped.2023.02.005.
[20] Song HB, Zhang YD, Dong QW, et al. Significance of serum NT-proBNP and endogenous H₂S for predicting coronary artery lesions in pediatric Kawasaki disease[J]. J Coll Physicians Surg Pak, 2020, 30(1):37-40. DOI: 10.29271/jcpsp.2020.01.37.
[21] 雷玉琳,刘翠兰,熊伶俐,等. 肌酸激酶同工酶、C反应蛋白、尿酸与新生儿窒息心肌损伤严重程度的关系[J]. 临床和实验医学杂志,2023,22(11):1194-1197. DOI:10.3969/j.issn.1671-4695.2023.11.020.
[22] 董楠,祁冬,孙景巍,等. N末端脑钠肽前体、肌酸激酶同工酶MB与冠状动脉Z值联合检测在早期评估川崎病合并冠状动脉损伤的临床价值[J]. 联勤军事医学,2023,37(4):297-300. DOI:10.13730/j.issn.2097-2148.2023.04.005.
[23] 张亮,徐玮玮,郑钦亮,等. BNP、CRP及CK-MB检测在川崎病患儿合并冠状动脉病变中的预测价值分析[J]. 解放军医药杂志,2022,34(3):69-72. DOI:10.3969/j.issn.2095- 140X.2022.03.016.
[24] 王春笛,赵丽,秦雯,等. 川崎病患儿并发冠状动脉损伤的危险因素及预测模型的构建与评价[J]. 现代生物医学进展,2023,23(19):3636-3641. DOI:10.13241/j.cnki.pmb.2023.19.007.
[25] Liu J, Su D, Yuan P, et al. Risk Factors for coronary artery aneurysm in a Chinese pediatric population with Kawasaki disease at low risk of intravenous immunoglobulin resistance: a retrospective cohort study[J]. Cardiology, 2023, 148(5):457-468. DOI: 10.1159/000530708.
[26] Wang K, Liu Q, Mo S, et al. A decision tree model to help treatment decision-making for severe spontaneous intracerebral hemorrhage[J]. Int J Surg, 2024, 110(2): 788-798. DOI: 10.1097/JS9.0000000000000852.