HOMI, Fundación Hospital Pediátrico la Misericordia, Bogotá, Colombia.
Hospital General de Medellín & Hospital Pablo Tobón Uribe, Medellín, Colombia.
J Antimicrob Chemother. 2023 Oct 3;78(10):2462-2470. doi: 10.1093/jac/dkad255.
Risk factors for carbapenem resistance in Enterobacterales bloodstream infections among children with cancer or post-HSCT have not been thoroughly explored.
All children with cancer or post-HSCT who developed Enterobacterales bloodstream infections in two cancer referral centres in major Colombian cities between 2012 and 2021 were retrospectively examined. When the infection episode occurred, carbapenem resistance mechanisms were evaluated according to the available methods. Data were divided in a training set (80%) and a test set (20%). Three internally validated carbapenem-resistant Enterobacterales (CRE) prediction models were created: a multivariate logistic regression model, and two data mining techniques. Model performances were evaluated by calculating the average of the AUC, sensitivity, specificity and predictive values.
A total of 285 Enterobacterales bloodstream infection episodes (229 carbapenem susceptible and 56 carbapenem resistant) occurred [median (IQR) age, 9 (3.5-14) years; 57% male]. The risk of CRE was 2.1 times higher when the infection was caused by Klebsiella spp. and 5.8 times higher when a carbapenem had been used for ≥3 days in the previous month. A model including these two predictive variables had a discriminatory performance of 77% in predicting carbapenem resistance. The model had a specificity of 97% and a negative predictive value of 81%, with low sensitivity and positive predictive value.
Even in settings with high CRE prevalence, these two variables can help early identification of patients in whom CRE-active agents are unnecessary and highlight the importance of strengthening antibiotic stewardship strategies directed at preventing carbapenem overuse.
儿童癌症或造血干细胞移植后发生肠杆菌血流感染时，碳青霉烯类耐药的风险因素尚未得到充分探讨。
在哥伦比亚两个主要城市的两家癌症转诊中心，回顾性分析了 2012 年至 2021 年间发生肠杆菌血流感染的所有癌症或造血干细胞移植后患儿。在感染发作时，根据现有方法评估碳青霉烯类耐药机制。数据分为训练集（80%）和测试集（20%）。创建了三个经内部验证的碳青霉烯类耐药肠杆菌科（CRE）预测模型：多变量逻辑回归模型和两种数据挖掘技术。通过计算 AUC、敏感性、特异性和预测值的平均值来评估模型性能。
共发生 285 例肠杆菌血流感染（229 例碳青霉烯敏感和 56 例碳青霉烯耐药）[中位数（IQR）年龄，9（3.5-14）岁；57%为男性]。当感染由肺炎克雷伯菌引起时，CRE 的风险增加 2.1 倍，当在上个月中使用碳青霉烯类药物≥3 天时，CRE 的风险增加 5.8 倍。包含这两个预测变量的模型在预测碳青霉烯类耐药方面具有 77%的区分性能。该模型的特异性为 97%，阴性预测值为 81%，敏感性和阳性预测值均较低。
即使在 CRE 流行率较高的环境中，这两个变量也可以帮助早期识别出不需要使用 CRE 活性药物的患者，并强调加强针对预防碳青霉烯类药物过度使用的抗生素管理策略的重要性。
