目的 基于支持向量机模型分析老年甲型流感（influenza A，H1N1）合并病毒性肺炎的危险因素并构建风险预测模型，为临床提供早期预警工具。方法 2024年2月至2025年2月，以便利抽样法选取某医院收治的老年H1N1患者206例为研究对象，结合入院影像学、临床及病原学结果将其分为H1N1合并病毒性肺炎组（以下简称肺炎组，155例）和非H1N1合并病毒性肺炎组（以下简称非肺炎组，51例），比较基础资料、临床症状、入院首次实验室检查结果以及H1N1病毒载量差异，分析实验室指标与H1N1病毒载量联系。经Modeler软件支持向量机模型筛选影响因素，构建预测模型，并进行内部验证。结果 与非肺炎组患者相比，肺炎组患者咳嗽、气喘占比更高；C反应蛋白（C-reactive protein，CRP）、降钙素原（procalcitonin，PCT）、血清淀粉样蛋白A（serum amyloid A，SAA）水平更高，差异均有统计学意义（均 P ＜0.05）。CRP、PCT、SAA水平均与病毒CT值呈负相关（ P ＜0.05）。特征选择结果显示，SAA、PCT、CT值、CRP和咳嗽具有高预测价值。在不同核函数模型中，径向基函数核模型综合性能最优，其准确度为90.29%。结论 SAA和PCT是老年H1N1合并病毒性肺炎的影响因素，与病毒载量负相关；基于RBF核的支持向量机模型预测效能最优值得推广。

Objective To identify the risk factors of viral pneumonia in elderly patients with influenza A (H1N1) and construct a risk prediction model based on the Support Vector Machine (SVM) algorithm,thereby providing an early warning tool for clinical practice.Methods A total of 206 elderly patients with H1N1 admitted to a hospital from February 2024 to February 2025 were selected using convenience sampling.Based on admission imaging,clinical,and etiological results,patients were divided into an influenza A viral pneumonia group (hereafter referred as pneumonia group, n =155) and a non-influenza A viral pneumonia group (hereafter referred as non-pneumonia group, n =51).Differences in baseline data,clinical symptoms,initial laboratory test results,and H1N1 viral load were compared.The relationship between laboratory indicators and H1N1 viral load was analyzed.Influencing factors were screened using the SVM model in Modeler software to construct a prediction model,which was then internally validated.Results Compared with the non-pneumonia group,the pneumonia group had significantly higher proportions of cough and wheezing,as well as significantly higher levels of C-reactive protein (CRP),procalcitonin (PCT),and serum amyloid A (SAA) (all P ＜0.05).CRP,PCT,and SAA levels were negatively correlated with viral CT values ( P ＜0.05).Feature selection results indicated that SAA,PCT,CT value,CRP,and cough had high predictive value.Among different kernel function models,the radial basis function (RBF) kernel model exhibited the best comprehensive performance,with an accuracy of 90.29%. Conclusions SAA and PCT are influencing factors for H1N1 complicated with viral pneumonia in the elderly and are negatively correlated with viral load.The SVM model based on the RBF kernel demonstrates optimal predictive efficacy and is worthy of clinical promotion.

R473.51；R823