Lai Qixun, Liao Kaifu, Kuang Guangzhi, Liao Weijie, Zhang Shengrui
Department of Thoracic Surgery, Ganzhou Fifth People's Hospital, Ganzhou City, 341000, People's Republic of China.
Department of Critical Care Medicine, Ganzhou People's Hospital, Ganzhou City, 341000, People's Republic of China.
Infect Drug Resist. 2025 Jun 26;18:3137-3147. doi: 10.2147/IDR.S526221. eCollection 2025.
To construct a nomogram model for individualized prediction of pulmonary fungal infection risk in lung cancer patients.
A total of 483 lung cancer patients hospitalized between August 2021 and August 2024 were retrospectively analyzed and randomly divided into a modeling group (n=338) and validation group (n=145). Patients in the modeling group were categorized based on the presence or absence of pulmonary fungal infection. Clinical data were analyzed using logistic regression, and a nomogram was developed using R software. Model performance was assessed using ROC curves, the Hosmer-Lemeshow (H-L) test, and Decision Curve Analysis (DCA).
Pulmonary fungal infections occurred in 99 out of 483 patients (20.50%). In the modeling group, the infection rate was 21.30%. Multivariate logistic regression identified age, smoking history, diabetes, glucocorticoid use, type of antimicrobial agents, invasive procedures, and length of hospitalization as independent risk factors (P<0.05). The Area Under the Curve (AUC) was 0.933 in the modeling group and 0.954 in the validation group. H-L tests indicated good model calibration (P>0.05). DCA demonstrated high clinical utility when the predicted probability ranged from 0.08 to 0.93.
The nomogram based on key clinical factors effectively predicts the risk of pulmonary fungal infection in lung cancer patients and is a promising tool for assisting in early identification and intervention.
构建用于个体化预测肺癌患者肺部真菌感染风险的列线图模型。
回顾性分析2021年8月至2024年8月期间住院的483例肺癌患者，并将其随机分为建模组（n = 338）和验证组（n = 145）。建模组患者根据是否存在肺部真菌感染进行分类。采用逻辑回归分析临床数据，并使用R软件绘制列线图。使用受试者工作特征曲线（ROC曲线）、霍斯默-莱梅肖（H-L）检验和决策曲线分析（DCA）评估模型性能。
483例患者中有99例发生肺部真菌感染（20.50%）。建模组的感染率为21.30%。多因素逻辑回归分析确定年龄、吸烟史、糖尿病、糖皮质激素使用、抗菌药物类型、侵入性操作和住院时间为独立危险因素（P < 0.05）。建模组的曲线下面积（AUC）为0.933，验证组为0.954。H-L检验表明模型校准良好（P > 0.05）。当预测概率在0.08至0.93之间时，DCA显示出较高的临床实用性。
基于关键临床因素的列线图能有效预测肺癌患者肺部真菌感染的风险，是协助早期识别和干预的有前景的工具。
