目的 结合肺炎患儿早期肺部超声征象与临床症状，构建儿童难治性肺炎风险预测模型。 方法 回顾性选取2024年6月—2025年5月在江苏省启东市妇幼保健院住院的152例肺炎患儿作为研究对象。依据是否符合难治性肺炎诊断标准将患儿分为两组：难治性肺炎组（n=50）和普通肺炎组（n=102），比较两组的临床特征及肺部超声检查结果。使用多因素logistic回归分析筛选独立预测因素，构建难治性肺炎的列线图预测模型。采用受试者操作特征曲线、校准曲线及决策曲线分析评估该模型的性能。 结果 多因素logistic回归分析显示，发热（OR=4.193，95%CI：1.422~12.362）以及肺部超声表现，即肺实变面积（OR=1.071，95%CI：1.012~1.133）、胸腔积液（OR=3.794，95%CI：1.571~9.165）、气胸征象（OR=1.818，95%CI：1.014~3.261）为难治性肺炎的独立预测因素（均P<0.05）。基于这4个因素构建的难治性肺炎风险预测模型的C指数为0.772，受试者操作特征曲线下面积为0.772（95%CI：0.690~0.854），提示该模型具有良好的风险判别能力。决策曲线分析显示，当风险阈值设定在15%~80%之间时，模型的临床适用性好。 结论 主要基于早期肺部超声特征构建的儿童难治性肺炎风险预测模型的预测性能良好，有助于早期评估难治性肺炎的发生风险。

Objective To develop a risk prediction model for refractory pneumonia in children by combining early lung ultrasound features with clinical symptoms. Methods Data from 152 children with pneumonia hospitalized at Jiangsu Qidong Maternal and Child Health CareHospital between June 2024 and May 2025 were retrospectively collected. The children were divided according to whether the diagnostic criteria for refractory pneumonia were met into a refractory pneumonia group (n=50) and a non-refractory pneumonia group (n=102). Clinical characteristics and lung ultrasound findings were compared between groups. Independent predictors were identified using multivariable logistic regression analysis, and a nomogram-based prediction model for refractory pneumonia was developed. Model performance was evaluated using receiver operating characteristic curve, calibration curve, and decision curve analyses. Results Multivariable logistic regression analysis showed that fever (OR=4.193, 95%CI: 1.422-12.362) and lung ultrasound findings—area of pulmonary consolidation (OR=1.071, 95%CI: 1.012-1.133), pleural effusion (OR=3.794, 95%CI: 1.571-9.165), and signs of pneumothorax (OR=1.818, 95%CI: 1.014-3.261)—were independent predictors of refractory pneumonia (all P<0.05). The prediction model based on these four factors had a C-index of 0.772, and the area under the receiver operating characteristic curve was 0.772 (95%CI: 0.690-0.854), indicating good discrimination. Decision curve analysis showed favorable clinical utility when the threshold probability was between 15% and 80%. Conclusions A risk prediction model for refractory pneumonia in children mainly based on early lung ultrasound features shows good predictive performance and helps in the early assessment of refractory pneumonia risk.