发热门诊血培养阳性率的影响因素分析及预测模型

Analysis of Influencing Factors and Prediction Model of Positive Rate of Blood Culture in Fever Clinic

摘要: 目的：分析发热门诊血培养阳性率的相关影响因素并构建血培养阳性预测模型，为规范开展血培养、迅速识别血流感染高危人群提供更多支持数据。方法：收集2020年1月至12月陕西省人民医院发热门诊的371例发热患者的相关临床资料及血培养结果，采用回顾性研究方法，分析血培养结果及与血培养阳性率有关的相关影响因素，并建立预测模型。结果：371例患者均进行血培养，其中73例培养阳性，阳性率为19.7%；血培养阳性确诊为血流感染者45例；单因素分析显示性别、年龄、心率、平均动脉压、免疫抑制剂、血培养套数、抗生素使用 ≥ 2种、侵入性操作与血培养阳性无关(P > 0.05)；而基础疾病、单一广谱抗生素的使用、体温数值与血培养阳性有关(P < 0.05)，多因素分析显示上述三个因素亦是血培养阳性的相关影响因素(P < 0.05)；其曲线下面积AUC为0.703，标准误为0.033，95%置信区间为0.639~0.767，根据约登指数计算公式，以敏感度 + 特异度 − 1最大为标准，确定P值的截断值为0.28，其诊断的敏感度为54.8%、特异度76.8%。结论：基础疾病、单一广谱抗生素的使用、体温数值是血培养阳性率的相关影响因素，利用上述因素构建模型可为临床预测血培养结果、提高积极实施血培养意愿、快速识别发生血流感染高危人群提供重要依据。

Abstract: Objective: To analyze the factors influencing the positive rate of blood culture in fever clinic and construct a positive prediction model of blood culture, so as to provide more supporting data for standardizing blood culture and quickly identifying the high-risk population of bloodstream infec-tion. Methods: The clinical data and blood culture results of 371 patients with fever in the fever clinic of Shaanxi Provincial People’s Hospital from January to December 2020 were collected. The results of blood culture and the influencing factors related to the positive rate of blood culture were analyzed retrospectively, and the prediction model was established. Results: All 371 patients un-derwent blood culture, of which 73 were positive (19.7%). Forty-five patients with positive blood culture were diagnosed as bloodstream infection. Univariate analysis showed that gender, age, heart rate, mean arterial pressure, immunosuppressive agents, number of blood culture sets, two kinds of antibiotic use, and invasive operation were not associated with positive blood culture (P > 0.05). The underlying diseases, the use of single broad spectrum antibiotics and body temperature were related to the positive blood culture (P < 0.05). Multivariate analysis showed that the above three factors were also related to the positive blood culture (P < 0.05). The AUC of the area under the curve was 0.703, the standard error was 0.033, and the 95% confidence interval was 0.639~0.767. According to the calculation formula of Jordan index, the cut-off value of P value was determined to be 0.28 based on the maximum standard of sensitivity + specificity − 1, and the sen-sitivity and specificity of the diagnosis were 54.8% and 76.8%. Conclusion: The underlying diseases, the use of single broad spectrum antibiotics and body temperature are the relevant influencing factors for the positive rate of blood culture. Using the above factors to construct a model can pro-vide an important basis for clinical prediction of blood culture results, improving the willingness to actively implement blood culture, and quickly identifying the high-risk population of bloodstream infection.