实体肿瘤患者住院化疗期间侵袭性真菌感染危险因素与评分模型的建立

Analysis of risk factors and establishment of scoring model for invasive fungal infection in patients with solid tumors during chemotherapy

目的 探讨实体肿瘤患者住院化疗期间侵袭性真菌感染(IFI)危险因素,并建立预测模型及分析其预测效能。方法 选择绍兴市柯桥区中医医院肿瘤内科2014年1月—2018年1月收治的实体肿瘤患者301例为模型组,选择2018年1月—2021年1月收治的实体肿瘤患者285例为验证组。通过电子病历收集其临床资料,并以此建立实体肿瘤患者IFI预测模型,采用受试者工作特征曲线(ROC)分析模型对模型组及对照组IFI的诊断价值。结果 单因素分析结果显示,住院时间、合并糖尿病、真菌感染史、预防性抗真菌用药、长期糖皮质激素应用与模型组实体肿瘤患者IFI发生有关(P<0.05);多因素Logistic回归分析结果显示,住院时间>14 d、合并糖尿病、有真菌感染史、无预防性抗真菌用药、长期糖皮质激素应用为实体肿瘤患者发生IFI的独立危险因素(P<0.05);ROC分析结果显示,本研究建立预测模型对模型组及验证组IFI诊断曲线下面积分别为0.931、0.907, SE分别为0.026、0.031,95%CI分别为0.881~0.982,0.846~0.967,均具有统计学意义(P<0.001);Hosmer-Lemeshow拟合优度检验结果显示本研究建立模型在观测值与实际值之间差异无统计学意义(Hosmer-lemeshowχ^(2)=2.153,P=0.565),提示该模型对医院实体肿瘤患者IFI发生有良好的预测效能。结论 基于临床资料建立的实体肿瘤预测模型对于患者IFI具有良好的预测效能,可用于临床中IFI的预防、管理及治疗参考。

OBJECTIVE To explore the risk factors for invasive fungal infection(IFI) in patients with solid tumors during inpatient chemotherapy, to establish a predictive model and analyze its predictive efficacy. METHODS Total of 301 solid tumor patients admitted in the Department of Oncology of Shaoxing City Keqiao District Hospital of Traditional Chinese Medicine General Hospital from Jan 2014 to Jan 2018 were recruited as the model group, and 285 solid tumor patients admitted from January 2018 to January 2021 were in the verification group. The clinical data was collected by using electronic medical records, and an IFI prediction model for solid tumor patients was established. The receiver operating characteristic curve(ROC) analysis model was used to analyze the diagnostic value of IFI between the model group and the control group. RESULTS Univariate analysis showed that the length of hospital stay, diabetes mellitus, history of fungal infection, preventive antifungal medication, and long-term glucocorticoid use were related to the incidence of IFI in solid tumor patients of the model group(P<0.05). but were related to gender, age, and tumor type, tumor staging, combined with hypertension, combined with chronic obstructive pulmonary disease(COPD), combined with chronic kidney disease), history of stroke is irrelevant;Multivariate logistic regression analysis showed that the length of hospital stay>14 d, combined with diabetes, history of fungal infections, no preventive use of antifungals and long-term use of glucocorticoid were independent risk factors for IFI in patients with solid tumors(P<0.05). ROC analysis showed that the areas under the diagnostic curve of the predictive model for IFI prediction were 0.931 and 0.907, with the SE of 0.026, 0.031. and the 95% Cis of 0.881-0.982, 0.846-0.967, all of which were significant(P<0.001). Hosmer-Lemeshow goodness-of-fit test showed that the difference between the model established in this study and the actual value was not significant(Hosmer-lemeshow χ^(2)=2.153, P=0.565), indicating the effective value of the model in the prediction of IFI in patients with solid tumors in our hospital. CONCLUSION The solid tumor prediction model established based on clinical data is effective in the prediction of IFI, which can be used as a reference for the clinical prevention, management and treatment of IFI.