某三甲医院门诊不明原因发热患者中感染性因素的列线图预测模型构建与评估

Construction of nomogram prediction model for infectious factors in outpatients with fever of unknown origin in a tertiary hospital and its evaluation

目的 调查某三甲医院综合门诊不明原因发热(FUO)的病因类型,探讨与感染性疾病病因相关的因素,建立预测感染性疾病的列线图模型并评估。方法 回顾性分析2019年1月-2021年6月某三甲医院综合门诊收治的FUO患者445例作为开发队列,收集患者基本特征、实验室检查等,应用LASSO回归优化筛选变量,采用logistic回归模型筛选感染性疾病的独立危险因素并建立预测模型,使用R软件绘制出列线图。内部验证采用Bootstrap法,使用受试者工作特征(ROC)曲线、校准曲线和Hosmer-Lemeshow拟合优度检验验证模型的区分度与校准度,决策曲线分析(DCA)评估模型的临床有效性。另选取2021年7-12月该院综合门诊收治的110例FUO患者作为验证队列对模型进行外部验证。结果 本研究经LASSO回归分析,最终纳入病程<35 d、畏寒/寒战、血清铁蛋白<520μg/L、降钙素原≥0.32μg/L、中性粒细胞<76%及碱性磷酸酶<82 U/L作为FUO患者感染性疾病病因的独立预测因素(P<0.05);依据上述独立预测变量绘制列线图,模型预测开发队列ROC曲线下面积为0.824,最大Youden指数所对应的风险预测值34.62%,预测临界值评分为145分,因此FUO患者得分≥145分的病因很可能是感染性疾病;Hosmer-Lemeshow拟合优度检验χ2=6.223,P=0.514;校准曲线的实际曲线与理想曲线几乎重合;验证队列的ROC曲线下面积为0.768,校准曲线的实际曲线位于理想曲线附近。结论 本研究基于某三甲医院门诊FUO患者的感染性疾病病因的风险因素构建的列线图预测模型具有良好的一致性与区分度,风险预测值评分为145分。

Objective To investigate the etiological types of fever of unknown origin(FUO)in the comprehensive outpatient department of a tertiary hospital,to explore the factors related to the etiology of infectious diseases,and to establish and evaluate a nomogram model for predicting infectious diseases.Methods From January 2019 to June 2021,445 FUO patients admitted to the comprehensive outpatient department of a tertiary hospital were retrospectively analyzed as the development cohort.The basic characteristics of patients,laboratory examinations,and other data were collected.Then LASSO regression was used to optimize the screening variables,and the logistic regression model was used to screen independent risk factors of infectious diseases and establish a predictive model,and a nomogram was drawed by R software.Bootstrap was used for internal validation,receiver operating characteristic(ROC)curve,calibration curve,and Hosmer-Lemeshow goodness-of-fit test were used to verify the discrimination and calibration of the model,and decision curve analysis(DCA)was used to evaluate the clinical effectiveness of the model.In addition,110 FUO patients admitted to the comprehensive outpatient department of a tertiary hospital from July to December 2021 were selected as the validation cohort to conduct external validation of the model.Results After LASSO regression analysis,the disease duration<35 d,chills or chills,serum ferritin<520μg/L,procalcitonin≥0.32μg/L,neutrophil<76%,and alkaline phosphatase<82 U/L in this study were ultimately included as independent predictors of the etiology of infectious diseases in patients with FUO(P<0.05).According to the above independent predictive variables,a nomogram was drawed,and the model predicted that the area under the ROC curve of the development cohort was 0.824,the risk prediction value corresponding to the maximum Youden index was 34.62%,and the predicted critical value score was 145 points.Therefore,the etiology of FUO patients'score≥145 points was likely to be infectious disease.The result of Hosmer-Lemeshow goodness of fit test wasχ2=6.223,P=0.514;The actual curve of the calibration curve nearly coincided with the ideal curve;The area under ROC curve of the validation cohort was 0.768,and the actual curve of the calibration curve was located near the ideal curve.Conclusion The nomogram prediction model constructed based on the risk factors of infectious disease etiology in FUO outpatients in a tertiary hospital has good consistency and discrimination.The predicted critical value score was 145 points.