儿童重症细菌感染死亡风险预测模型的建立及评价

Development and evaluation of a mortality risk prediction model for severe bacterial infections in children

目的构建儿童重症细菌感染死亡风险模型并同儿科早期预警评分(pediatric early warning score,PEWS)、小儿危重病例评分(pediatric critical illness score,PCIS)和小儿死亡风险评分Ⅲ(pediatric risk of mortalityⅢ,PRISMⅢ)进行比较。方法选取南京医科大学附属儿童医院PICU2017年5月至2022年6月收治的178名危重患儿。在获得父母/监护人的知情同意后,收集患儿的性别、年龄、身高、体重等基本信息,以及心率、收缩压、呼吸频率等指标。在患儿进入PICU 24 h后进行评分,根据入院后28 d生存状态将患儿分为存活组和死亡组。构建死亡风险模型,并绘制列线图。使用受试者工作特征(receiver operating characteristic,ROC)曲线,并通过ROC曲线下面积(area under curve,AUC)评估和比较死亡风险模型、PEWS、PCIS和PRISMⅢ预测死亡风险的价值。结果178例危重患儿中11例因资料严重缺失、住院未超过24 h被排除。共167例患儿纳入分析,其中28 d存活组134例,死亡组33例。用瞳孔改变、意识状态、皮肤颜色、是否行机械通气、总胆固醇、凝血酶原时间构建了儿童重症细菌感染死亡风险预测模型。ROC曲线分析表明,死亡风险模型预测重症细菌感染儿童死亡AUC是0.888(P<0.05),PEWS、PCIS和PRISMⅢ预测重症细菌感染儿童死亡的AUC分别为0.769(P<0.05)、0.575(P<0.05)和0.759(P<0.05)。Hosmer-Lemeshow拟合优度检验显示死亡风险模型和PEWS预测病死率与实际病死率的一致性最好(χ^(2)=5.180,P=0.738;χ^(2)=4.939,P=0.764),PCIS和PRISMⅢ预测病死率与实际病死率的拟合效果尚可(χ^(2)=9.110,P=0333;χ^(2)=8.943,P=0.347)。结论死亡风险模型对重症细菌感染患儿的预后价值较PEWS、PCIS和PRISMⅢ更好。

Objective To establish a mortality risk prediction model of severe bacterial infection in children and compare it with the pediatric early warning score(PEWS),pediatric critical illness score(PCIS)and pediatric risk of mortality scoreⅢ(PRISMⅢ).Methods A total of 178 critically ill children were selected from the PICU of the Children's Hospital of Nanjing Medical University from May 2017 to June 2022.After obtaining the informed consent of the parents/guardians,basic information such as sex,age,height and weight,as well as indicators such as heart rate,systolic blood pressure and respiratory rate were collected from all children.A standard questionnaire was used to score the child 24 h after admission to the PICU.The children were divided into the survival and death groups according to their survival status at 28 d after admission.A mortality risk prediction model was constructed and nomogram was drawn.The value of the mortality risk prediction model,PEWS,PCIS and PRISM in predicting the risk of death was assessed and compared using the receiver operating characteristic(ROC)curve and the area under the ROC curve(AUC).Results Among the 178 critically ill children,11 cases were excluded due to severe data deficiencies and hospitalization not exceeding 24 h.A total of 167 children were included in the analysis,including 134 in the survival group and 33 in the death group.A mortality risk prediction model for children with severe bacterial infection was constructed using pupillary changes,state of consciousness,skin color,mechanical ventilation,total cholesterol and prothrombin time.ROC curve analysis showed that the AUCs of mortality risk prediction model was 0.888(P<0.05).The AUCs of PEWS,PCIS and PRISMⅢin predicting death in children with severe bacterial infection were 0.769(P<0.05),0.575(P<0.05)and 0.759(P<0.05),respectively.Hosmer-Lemeshow goodness-of-fi t test showed the best agreement between risk of death and PEWS predicted morbidity and mortality and actual morbidity and mortality(χ^(2)=5.180,P=0.738;χ^(2)=4.939,P=0.764),and the PCIS and PRISMⅢpredicted mortality rates fi tted reasonably well with actual mortality rates(χ^(2)=9.110,P=0333;χ^(2)=8.943,P=0.347).Conclusions The mortality risk prediction model for predicting the death risk has better prognostic value than PEWS,PCIS and PRISMⅢfor children with severe bacterial infection.