| 摘要: |
| 【】 目的 构建并验证基于Lasso-Nomogram模型的儿童重症肺炎支原体(Severe Mycoplasma Pneumonia)的预测模型。方法 回顾性选取2023年5月至2024年5月期间新疆医科大学第一附属医院儿科收治的723例肺炎支原体肺炎(MPP)患儿的病历资料。根据随机数字表法(7:3)将病例分为建模集(506例)和验证集(217例),并依据病情严重程度进一步划分为轻症组与重症组。对建模集数据进行Lasso回归,筛选出具有差异的变量,采用向前逐步法行多因素 logistic 回归分析,构建 Nomogram 预测模型,并绘制受试者工作特征曲线(ROC)及其曲线下面积(AUC)计算、校准曲线分析以评估Nomogram模型的内部和外部验证效能,临床决策曲线(DCA)评估模型临床价值。结果 Lasso回归及多因素Logistic回归分析表明,年龄(OR=1.134,P=0.001)、听诊呼吸音减弱(OR=6.147,P=0.006)、合并肺内并发症(OR=5.466,P=0.000)、血乳酸升高(BLA,OR=2.164,P=0.041)、乳酸脱氢酶升高(LDH,OR=10.766,P=0.000)是导致SMPP的独立危险因素。基于此构建的Nomogram模型,在建模集中预测SMPP的ROC曲线下面积(AUC)为0.810(95%CI:0.772～0.847,P <0.01),验证集的AUC为0.773(95%CI:0.708～0.838,P <0.001)。建模集和验证集的校准曲线均显示出良好的拟合度和一致性。临床决策曲线分析(DCA)表明,Lasso-Nomogram模型在预测SMPP方面有较高的临床收益。结论 儿童SMPP的发生与年龄、听诊呼吸音减弱、合并肺内并发症、BLA、LDH升高等因素密切相关。基于上述因素构建的Lasso-Nomogram预测模型具有较高的准确度,有助于早期识别患儿的SMPP风险,并采取相应的预防和治疗措施。 |
| 关键词: Lasso-Nomogram模型 儿童肺炎支原体肺炎 重症 危险因素 |
| DOI： |
| 分类号: |
| 基金项目:“天山英才”医药卫生高层次人才培养计划(TSYC202301B003) |
|
| Research on Risk Prediction of Mycoplasma Pneumonia in Children with Severe Pneumonia Based on Lasso-Nomogram Model |
|
wangyingru
|
|
Department of Pediatrics, Friendship Hospital of Yili Kazak Autonomous Prefecture, Xinjiang Uygur Autonomous Region
|
| Abstract: |
| 【】 Objective:To build and validate a prediction model for severe mycoplasma pneumonia (SMPP) in children based on a Lasso-Nomogram model. Methods:Retrospectively selected medical records of 723 children with Mycoplasma pneumonia (MPP) admitted to the First Affiliated Hospital of Xinjiang Medical University from May 2023 to May 2024. The cases were divided into a modeling set (506 cases) and a validation set (217 cases) based on a random number table method (7:3), and further categorized into mild and severe groups according to disease severity. Lasso regression was performed on the modeling set data to screen for different variables, and a forward stepwise method was used for multivariate logistic regression analysis to construct a Nomogram forecasting model. The receiver operating characteristic curve (ROC) and its area under the curve (AUC) were plotted, and calibration curve analysis was conducted to evaluate the internal and external validation efficacy of the Nomogram model, while the clinical decision curve (DCA) assessed the clinical value of the model. Results: Lasso regression and multivariate logistic regression analysis indicate that age (OR=1.134, P=0.001), weakened respiratory sound (OR=6.147, P=0.006), pulmonary complications (OR=5.466, P=0.000), elevated blood lactic acid (BLA, OR=2.164, P=0.041), and increased lactate dehydrogenase (LDH, OR=10.766, P=0.000) are independent risk factors for SMPP. The Nomogram model constructed based on this predicts the area under the ROC curve (AUC) for SMPP in the modeling set to be 0.810 (95% CI: 0.772–0.847, P <0.01), and the AUC for the validation set to be 0.773 (95% CI: 0.708–0.838, P <0.001). The calibration curves for both the modeling set and the validation set show good fit and consistency. Clinical decision curve analysis (DCA) indicates that the Lasso-Nomogram model has high clinical benefit in predicting SMPP.Conclusion: The occurrence of SMPP in children is closely related to factors such as age, weakened respiratory sounds, pulmonary complications, BLA, and elevated LDH. The Lasso-Nomogram prediction model constructed based on the above factors has high accuracy, which helps in the early identification of SMPP risk in children and the implementation of corresponding preventive and therapeutic measures. |
| Key words: Lasso-Nomogram model pediatric mycoplasma pneumonia severe risk factors |
