Development and validation of a prognostic prediction model for carbapenem-resistant gram-negative bacteria bloodstream infection in patients with hematological malignancies
10.11816/cn.ni.2025-246925
- VernacularTitle:恶性血液病碳青霉烯类耐药革兰阴性菌血流感染预后预测模型构建及验证
- Author:
Xiaqin HE
1
;
Meng LIU
1
;
Yi ZHANG
1
;
Weiqi WANG
1
;
Zhe LIU
1
;
Xiaoqian WANG
1
;
Xiaoyan ZENG
1
Author Information
1. 西安交通大学第一附属医院检验科,陕西西安 710061
- Publication Type:Journal Article
- Keywords:
Hematological malignancie;
Bloodstream infection;
Gram-negative bacteria;
Carbapenem resistance;
Prognosis;
Prediction model
- From:
Chinese Journal of Nosocomiology
2025;35(12):1787-1792
- CountryChina
- Language:Chinese
-
Abstract:
OBJECTIVE To investigate the risk factors for carbapenem-resistant gram-negative bacteria(GNB)bloodstream infection(BSI)in patients with hematological malignancies(HMs)and their prognosis,and to devel-op a nomogram prediction model.METHODS A total of 316 patients with HMs and GNB-BSI admitted to the First Affiliated Hospital of Xi'an Jiaotong University from Jan.2017 to Dec.2022 were selected as the training set,and 106 patients admitted from Jan.2023 to Oct.2024 were selected as the validation set.Variables were selected by lasso regression and multifactor logistic regression,and a nomogram model was constructed.The prediction model was internally validated by the receiver operating characteristic(ROC)curve,calibration curve and decision curve analysis(DCA),respectively.RESULTS Granulocytopenia for ≥7 days(OR=14.525),use of cephalosporins/β-lactamase inhibitors within 30 days before BSI(OR=3.510),exposure history of carbapenem antibacterial drug(OR=4.840)and albumin＜30 g/L(OR=2.697)were risk factors for CR-GNB BSI in patients with HMs(P＜0.05).Septic shock(OR=6.934),central venous catheterization(OR=5.586),inappropriate empirical antibac-terial drug therapy(OR=4.744),CR-GNB infection(OR=2.916)and albumin＜30 g/L(OR=3.324)were risk factors for 30-day mortality in patients with HMs and GNB-BSI(P＜0.05).Based on these indicators,two nomogram models were constructed.The areas under the ROC curve(AUC)for the internal validation set were 0.775 and 0.849,respectively.The calibration curves demonstrated high predictive performance for the pre-diction models(P=0.998 and 0.660,respectively),and DCA showed high clinical application value for both models.CONCLUSION The nomogram prediction model constructed in this study based on multifactor analy-sis not only demonstrates good predictive value but also exhibits significant clinical efficacy,aiding in the early i-dentification of high-risk patients for targeted therapy.
