| 摘要: |
| 目的 建立并验证一种基于多项血常规参数的风险预测模型，用于初步鉴别诊断甲型流行性感冒（以下简称甲流）和流感样疾病（ILI）患儿。方法 选择2022年7月1日至2023年6月30日在海军军医大学（第二军医大学）第一附属医院儿科治疗的甲流患儿（n＝2 686）及ILI患儿（n＝1 369）为研究对象，收集其临床、实验室特征进行回顾性分析。根据年龄将患儿分为1岁≤年龄≤6岁、6岁＜年龄≤16岁2个亚组，各亚组中患儿被随机分为训练集（占70%）和内部验证集（占30%）。选择2022年7月1日至2023年6月30日在海军军医大学（第二军医大学）第二附属医院和中国人民解放军东部战区海军医院儿科治疗的甲流患儿（n＝204）及ILI患儿（n＝404）作为外部验证集。对训练集进行多因素logistic回归分析以获得甲流的独立影响因素，将基于这些因素构建的预测模型以列线图展示，采用ROC曲线、Hosmer-Lemeshow检验、决策曲线分析分别从区分度、校准度及临床实用性3个方面评估模型的性能，并在内部验证集和外部验证集中对模型的诊断性能进行验证。结果 在1岁≤年龄≤6岁亚组，年龄、白细胞计数、淋巴细胞计数、CRP是甲流的独立影响因素（均P＜0.01），建立的列线图预测模型识别甲流的AUC值在训练集中为0.746，在内部验证集中为0.771，在外部验证集中为0.753；模型的预测概率和实际概率高度一致（P＝0.216）；在16%～60%的阈值概率范围内采取干预措施可以获得净收益。在6岁＜年龄≤16岁亚组，性别、白细胞计数、淋巴细胞计数是甲流的独立影响因素（均P＜0.01），建立的列线图预测模型识别甲流的AUC值为0.733，在内部验证集中为0.747，在外部验证集中为0.753；模型的预测概率和实际概率高度一致（P＝0.06）；在12%～58%的阈值概率范围内采取干预措施可以获得净收益。结论 这种基于易获取的血常规参数建立的风险预测模型对甲流显示出良好的诊断性能，具有较高的准确性和临床实用性。 |
| 关键词: 儿童 甲型流行性感冒 预测模型 列线图 血常规 |
| DOI：10.16781/j.CN31-2187/R.20240372 |
| 投稿时间：2024-05-29修订日期：2024-08-27 |
| 基金项目:海军计划生育专项（21JSZ05），海军军医大学（第二军医大学）校级课题（2023MS031）. |
|
| Multi-parameter prediction model based on blood routine in children with influenza A |
| SHI Yuxia1,ZHOU Lin1,LEI Lei1*,LIU Weina2,3,XU Tong2 |
(1. Department of Pediatrics, The First Affiliated Hospital of Naval Medical University (Second Military Medical University), Shanghai 200433, China;
2. Department of Pediatrics, The Second Affiliated Hospital of Naval Medical University (Second Military Medical University), Shanghai 200003, China;
3. Department of Pediatrics, Naval Hospital of PLA Eastern Theater Command, Zhoushan 316000, Zhejiang, China
*Corresponding author) |
| Abstract: |
| Objective To establish and validate a risk prediction model based on multiple blood routine parameters for preliminary differential diagnosis of influenza A and influenza like illness (ILI) in children. Methods Children with influenza A (n＝2 686) and ILI (n＝1 369) who were treated in Department of Pediatrics, The First Affiliated Hospital of Naval Medical University (Second Military Medical University) from Jul. 1, 2022 to Jun. 30, 2023 were enrolled, and their clinical and laboratory features were collected for retrospective analysis. According to age, patients were divided into 2 subgroups: 1 year≤age≤6 years and 6 years＜age≤16 years. Patients in each subgroup were randomly divided into training set (70%) and internal validation set (30%). Children with influenza A (n＝204) and ILI (n＝404) who were treated in Department of Pediatrics of The Second Affiliated Hospital of Naval Medical University (Second Military Medical University) and Naval Hospital of PLA Eastern Theater Command from Jul. 1, 2022 to Jun. 30, 2023 were selected as the external validation set. Multivariate logistic regression analysis was performed on the training set to obtain the independent influencing factors of influenza A. The prediction model based on these factors were displayed as a nomogram. Receiver operating characteristic (ROC) curve, Hosmer-Lemeshow test, and decision curve analysis (DCA) were used to evaluate the performance of the model from 3 aspects: discrimination, calibration, and clinical practicality, respectively. The diagnostic performance of the model was verified in both internal validation set and external validation set. Results In the subgroup of 1 year≤age≤6 years, age, white blood cell count, lymphocyte count and C reactive protein were the independent influencing factors of influenza A (all P＜0.01); the area under the curve (AUC) value of the established nomogram prediction model for identifying influenza A was 0.746 in the training set, 0.771 in the internal validation set, and 0.753 in the external validation set; the predicted probability of the model was highly consistent with the actual probability (P＝0.216); and taking intervention measures within a threshold probability range of 16%-60% could yield net benefits. In the subgroup of 6 years＜age≤16 years, gender, white blood cell count and lymphocyte count were the independent influencing factors of influenza A (all P＜0.01); the AUC value of the established nomogram prediction model for identifying influenza A was 0.733, 0.747 in the internal validation set, and 0.753 in the external validation set; the predicted probability of the model was highly consistent with the actual probability (P＝0.06); and taking intervention measures within a threshold probability range of 12%-58% could yield net benefits. Conclusion This risk prediction model based on easily obtainable blood routine parameters shows good diagnostic performance for influenza A, with high accuracy and clinical practicality. |
| Key words: child influenza A prediction model nomograms blood routine |
