Abstract:

Objective　To explore the risk factors for chemotherapy-associated infections in childhood acute leukemia and to construct a predictive model. Methods　By using the convenience sampling method, 213 children with acute leukemia who were hospitalized at Shenzhen Children's Hospital from March 2021 to October 2024 were selected as the research subjects. According to whether the children had chemotherapy-associated infections, they were divided into an infection group (69 cases) and a non-infection group (144 cases). In the infection group, there were 35 boys and 34 girls, 30 cases aged <6 years and 39 cases aged 6-14 years. In the non-infection group, there were 78 boys and 66 girls, 65 cases aged <6 years and 79 cases aged 6-14 years. Univariate analysis (χ² test and independent sample t test) and multivariate logistic regression analysis were used to identify the influencing factors of chemotherapy-associated infections in childhood acute leukemia. Based on the results of regression analysis, a nomogram was drawn using R language, a risk prediction model was constructed, and its predictive value was analyzed using the receiver operating characteristic curve (ROC). Results　The incidence of chemotherapy-associated infections in childhood acute leukemia was 32.39% (69/213). The results of univariate analysis showed that there were statistically significant differences between the infection group and the non-infection group in terms of glucocorticoid treatment, prophylactic antibiotics, induction remission period, length of hospital stay, minimum neutrophil count, duration of neutrophil deficiency, and number of chemotherapy (all P<0.05), while there was no statistically significant difference in terms of gender, age, minimum hemoglobin level, or hospitalization season (all P>0.05). The results of logistic regression analysis showed that prophylactic antibiotics, length of hospital stay, duration of neutrophil deficiency, and number of chemotherapy were the influencing factors for chemotherapy-associated infections in children with acute leukemia (all P<0.05). Based on the results of logistic regression analysis, a nomogram model was constructed. The area under the curve of this model was 0.875 (95%CI: 0.713 - 0.946), with a sensitivity of 89.2% and a specificity of 90.7%, which indicated that the model had a high discriminatory ability. The Hosmer-Lemeshow test was conducted, with χ²=0.769 and P=0.993, indicating that the model fit well. Conclusion　The constructed risk prediction model has a high degree of discrimination and helps clinical medical staff quickly identify children with high risk of chemotherapy-associated infections in acute leukemia.

Key words:

Children, Acute leukemia, Chemotherapy-associated infections, Nomogram, Prediction model construction

摘要：

目的　探讨儿童急性白血病化疗相关性感染的风险因素，并构建预测模型。方法　采用便利抽样法，选取2021年3月至2024年10月在深圳市儿童医院住院的急性白血病患儿213例作为研究对象。根据患儿是否发生化疗相关性感染，将患儿分为感染组69例和非感染组144例。感染组中，男35例，女34例，年龄<6岁30例，6～14岁39例；非感染组中，男78例，女66例，年龄<6岁65例，6～14岁79例。采用单因素分析（χ²检验、独立样本t检验）和多因素logistic回归分析儿童急性白血病化疗相关性感染的危险因素，根据回归分析结果用R语言绘制列线图，构建风险预测模型，并采用受试者操作特征曲线（ROC）分析其预测价值。结果　儿童急性白血病化疗相关性感染的发生率为32.39%（69/213）。单因素分析结果显示，感染组与非感染组在糖皮质激素治疗、预防性抗菌药物、诱导缓解期、住院时间、最低中性粒细胞计数、中性粒细胞缺乏持续时间、化疗次数方面差异均有统计学意义（均P<0.05），而在性别、年龄、最低血红蛋白、住院季节方面差异均无统计学意义（均P>0.05）。logistic回归分析结果显示，预防性抗菌药物、住院时间、中性粒细胞缺乏持续时间、化疗次数是儿童急性白血病化疗相关性感染的影响因素（均P<0.05）。根据回归分析结果构建列线图模型，该模型的曲线下面积为0.875（95%CI为0.713～0.946），灵敏度为89.2%，特异度为90.7%，说明模型具有较高的区分能力。经Hosmer-Lemeshow检验，χ²=0.769，P=0.993，表示模型拟合度良好。结论　构建的风险预测模型具有较高的区分度，可以帮助临床医护人员快速识别儿童急性白血病化疗相关性感染高风险患儿。

关键词:

儿童, 急性白血病, 化疗相关性感染, 列线图, 预测模型构建