小儿细菌性肺炎的高效识别模型及临床价值研究

A Model for Effective Identification of Pediatric Bacterial Pneumonia and Its Clinical Value

背景目前国内尚缺乏对发热就诊的肺炎患儿是否为细菌感染做出快速判断的简易方法,容易引起漏诊及抗生素的滥用。目的建立预判发热就诊的肺炎患儿是否为细菌感染的简单模型。方法回顾性选取2012—2013年温州医科大学附属第二医院育英儿童医院符合纳入标准的以发热就诊的肺炎患儿538例为研究对象。根据疾病原因将患儿分为细菌感染组(133例)和非细菌感染组(405例)。从538例患儿中随机选取54例作为验证集(细菌性肺炎13例,非细菌性肺炎41例)。收集患儿一般资料、实验室检测结果,建立5个诊断细菌性肺炎的模型〔F1=C反应蛋白(CRP)×降钙素原(PCT)、F2=CRP2×PCT、F3=CRP×PCT2、F4=性别权重×就诊季节权重×喘息症状系数×(CRP×PCT)、F5=性别权重×就诊季节权重×喘息症状系数×(CRP×PCT2)〕,绘制其诊断细菌性肺炎的ROC曲线,确定最优模型。结果两组患儿性别、就诊季节、寒战发生率、呼吸加快发生率、喘息发生率、呕吐发生率、腹泻发生率、哭闹发生率、干啰音发生率、湿啰音发生率、发热持续天数、最高体温、白细胞计数(WBC)、CRP水平、PCT水平比较,差异有统计学意义(P<0.05)。单独CRP诊断细菌性肺炎的ROC曲线下面积(AUC)为0.969,95%CI(0.955,0.979),临界值为48.5 mg/L,灵敏度为88.0%,特异度为93.6%;单独PCT诊断细菌性肺炎的AUC为0.974,95%CI(0.959,0.989),临界值为0.5 g/L,灵敏度为92.5%,特异度为84.0%;F1诊断细菌性肺炎的AUC为0.983,95%CI(0.973,0.993),临界值为17.4,灵敏度为92.5%,特异度为96.3%;F2诊断细菌性肺炎的AUC为0.981,95%CI(0.971,0.992),临界值为241.1,灵敏度为97.7%,特异度为90.6%;F3诊断细菌性肺炎的AUC为0.983,95%CI(0.973,0.993),临界值为6.3,灵敏度为94.0%,特异度为96.3%;F4诊断细菌性肺炎的AUC为0.987,95%CI(0.980,0.996),临界值为1.1,灵敏度为94.7%,特异度为95.6%;F5诊断细菌性肺炎的AUC为0.988,95%CI(0.981,0.997),临界值为0.2,灵敏度为97.7%,特异度为94.3%。根据单独CRP、单独PCT、F5的临界值,对验证集患儿进行诊断,结果显示,单独CRP诊断验证集患儿细菌性肺炎的灵敏度为76.9%,特异度为97.6%,正确率为92.6%;单独PCT诊断验证集患儿细菌性肺炎的灵敏度为84.6%,特异度为97.6%,正确率为94.4%;F5诊断验证集患儿细菌性肺炎的灵敏度为92.3%,特异度为97.6%,正确率为96.3%。结论对于因发热就诊的肺炎患儿,可以通过F5模型〔F5=性别权重×就诊季节权重×喘息症状系数×(CRP×PCT2)〕计算得到相应的结果,若结果大于0.2,可诊断细菌性肺炎,建议早期使用抗生素治疗。

Background Interiorly, there is no simple and convenient way to make a quick judgment on whetherthere is bacterial infection in pneumonia children receiving treatment due to fever. This will easily cause missedabuses of antibacterial drug. Objective To establish a simple model to diagnose whether the pneumonia children with fever arecaused by bacterial infection. Methods According to the inclusion criteria, 538 children with pneumonia, who were underobservation because of fever in the 2nd Affiliated Hospital and Yuying Children^ Hospital of Wenzhou Medical University between January 2012 and December 2013 , were retrospectively selected. Based on their causes of disease, the children were divided into bacterial infection group （ n = 133） and non - bacterial infection group （ n = 405） . Fifty - four children were randomly selected from 538 children as a validation set （13 with bacterial pneumonia and 41 with non - bacterial pneumonia） .The generalinformation of children and laboratory test results were collected. Based on the study, five models of bacterial pneumonia diagnosis were established： F1 = C - reactive protein （ CRP） x procalcitonin （ PCT）, F2 = CRP2 x PCT, F3 = CRPx PCT2, F4 =gender （index） x season （index） x breathing symptoms （index） x （ CRP x PCT） , F5 = gender （index） x season （index） x breathing symptoms （index） x （ CRP x PCT2 ） . The receiver operating characteristic （ ROC ） curve of thesemodels for diagnosing bacterial pneumonia was drawn to determine the optimal model. Results There was significant difference in gender , seasonal changes , shiver occurrence , rapid respiration occurrence , possibility of asthma , vomiting , diarrhea ,tantrum , dry rale , moist rale , number of days in fever , the highest temperature , WBC , CRP level and PCT level of children intwo groups （P 〈0. 05） . The AUC of separate CPR for diagnosis of bacterial pneumonia was 0. 969 , 95% C l （0. 955 , 0. 979） , the critical value was 48. 5 mg/L , the sensitivity was 88. 0% and the specificity was 93. 6% ; the AUC of separate PCT for diagnosis of bacterial pneumonia was 0. 974 , 95% C l （0. 959,0. 989 ） , the critical value was 0. 5 gL , the sensitivity was 92.5% and the specificity was 84.0% ; the AUC, critical value, the sensitivity and specificity of model F1 for diagnosis of bacterial pneumonia were 0. 983,95% C/ （0. 973, 0. 993 ）, 17. 4,92. 5% and 96. 3% respectively; the above four indicators of model F were 0. 981, 95% C l （0. 971, 0. 992）, 241. 1, 97. 7% and 90. 6% respectively; the above four indicators of model F3 were 0. 983, 95% C/ （0. 973,0. 993）, 6. 3, 94. 0% and 96. 3% respectively; the above four indicators of modelF4 were 0. 987, 95% Cl （0. 980, 0. 996）, 1. 1, 94. 7%, and 95. 6% respectively; the above four indicators of model F5were 0. 988, 95% C/ （0. 981,0. 9 9 7 ）,0. 2, 97. 7%, and 94. 3% respectively. The children in validation set were diagnosed according to the critical value of separate CPR, separate PCT and model F5. The results shCRP for diagnosis of bacterial pneumonia in validation set was 76. 9%, the specificity was 97. 6% and the accuracy was 92. 6% ; while these of separate PCT were 84. 6%, 97. 6% and 94. 4% respectively; and these of model F5 were 92. 3%, 97. 6% and 96. 3% respectively. Conclusion For pneumonia children whose visiting reason is fever, we can get the result by the F5 model[F5 = gender （index） x season （index） x breathing symptoms （index） x （ CRP x PCT2 ）〕 . If the result is greater than0. 2,bacterial pneumonia can be diagnosed, and the early use of antibiotics is suggested.