|
| 开发并验证慢性阻塞性肺疾病急性加重病毒感染风险预测模型 |
| Development and validation of a risk prediction model for viral infection in acute exacerbation of chronic obstructive pulmonary disease |
| 投稿时间：2025-10-15 |
| DOI：10.3969/j.issn.1000-0399.2026.07.007 |
| 中文关键词: 慢性阻塞性肺疾病急性加重 病原体靶向测序 病毒感染 预测模型 在线APP |
| 英文关键词: Acute exacerbation of chronic obstructive pulmonary disease Pathogen-targeted sequencing Viral infection Predictive model Web application |
| 基金项目:湖南省卫生健康委科研计划项目(编号:D202314019409) |
|
| 摘要点击次数: 249 |
| 全文下载次数: 0 |
| 中文摘要: |
| 目的 基于支气管肺泡灌洗液病原体宏基因组二代测序（NGS）检测结果，建立慢性阻塞性肺疾病急性加重期（AECOPD）患者病毒感染风险预测模型。方法 回顾性收集2024年9月1日至2025年2月28日怀化市中心医院呼吸与危重症医学科1区接受支气管镜检查并行支气管灌洗液病原体二代测序（BALF-NGS）检测的86例AECOPD患者临床资料，分为未检测到病毒组（n=36）和检测到病毒组（n=32）。通过随机抽样将数据按8∶2比例分为训练集与验证集。采用R 4.4.0软件进行多因素logistic回归分析确定病毒感染独立危险因素，构建列线图预测模型。通过混淆矩阵及相关指标评估了预测模型的性能，并开发基于Shiny框架的在线预测工具。结果 在68例训练集数据中，单因素分析显示，与未检测到病毒组比较，检测到病毒组（n=32）患者的淋巴细胞百分比明显升高[40.00%(35.00%,51.55%)比35.65%(31.07%,40.10%),P=0.003]，长期使用激素比例明显增加（71.88%比25.00%,P<0.001），且影像学磨玻璃影检出率更高（56.25%比13.89%,P<0.001），差异均有统计学意义。多因素logistic回归分析进一步证实，肺部影像提示磨玻璃影（OR=6.028,95%CI:1.528～23.775,P=0.010）、长期糖皮质激素使用（OR=5.916,95%CI:1.612～21.717,P=0.007）及淋巴细胞比例升高（OR=1.105,95%CI:1.022～1.196,P=0.012）是AECOPD患者合并病毒感染的独立危险因素。模型在训练集与验证集的曲线下面积分别为0.872(0.783～0.960)和0.827(0.618～1.000)。基于研究成果开发的在线预测工具已部署于：https：//polymyxinaki.shinyapps.io/dynnomapp/。结论 本研究建立的AECOPD病毒感染预测模型展现出良好的区分效能（AUC>0.8）与临床适用性，为早期识别病毒感染相关AECOPD提供了可靠的量化工具。 |
| 英文摘要: |
| Objective To establish and validate a risk prediction model for viral infection in patients with acute exacerbation of chronic obstructive pulmonary disease(AECOPD) based on pathogen detection results from bronchoalveolar lavage fluid next-generation sequencing(BALF-NGS). Methods The clinical data of 86 AECOPD patients who underwent bronchoscopy and BALF-NGS testing in the Department of Respiratory and Critical Care Medicine, Huaihua Central Hospital, between September 1, 2024, and February 28, 2025, were retrospectively collected. The data were randomly split into a training set and a validation set at an 8∶2 ratio. Multivariate logistic regression analysis using R 4.4.0 software was performed to identify independent risk factors for viral infection and to construct a nomogram prediction model. The performance of the prediction model was evaluated using a confusion matrix and related metrics, and an online prediction tool based on the Shiny framework was developed. Results Among the 68 cases in the training set, univariate analysis showed that compared to the non-viral detection group(n=36), the viral detection group(n=32) had a significantly higher lymphocyte percentage [40.00%(35.00%, 51.55%) vs. 35.65%(31.07%, 40.10%), P=0.003], a significantly increased proportion of long-term glucocorticoid use(71.88% vs. 25.00%, P<0.001), and a higher detection rate of ground-glass opacity on imaging(56.25% vs. 13.89%, P<0.001), with all differences being statistically significant. Multivariate logistic regression analysis further confirmed that ground-glass opacity on pulmonary imaging(OR=6.028, 95%CI:1.528～23.775, P=0.010), long-term glucocorticoid use(OR=5.916, 95%CI:1.612～21.717, P=0.007), and elevated lymphocyte proportion(OR=1.105, 95%CI: 1.022～1.196, P=0.012) were independent risk factors for viral co-infection in AECOPD patients. The area under the curve(AUC) of the model was 0.872(95%CI: 0.783～0.960) in the training set and 0.827(95%CI: 0.618～1.000) in the validation set. An online prediction tool developed based on the research findings has been deployed at: https://polymyxinaki.shinyapps.io/dynnomapp/. Conclusion The prediction model for viral infection in AECOPD established in this study demonstrates good discriminative performance(AUC > 0.8) and clinical applicability, providing a reliable quantitative tool for the early identification of AECOPD associated with viral infection. |
|
查看全文
查看/发表评论 下载PDF阅读器 |
| 关闭 |
|
|
|
