云南省边境地区早期指征预测重症登革热及登革热伴预警发生的风险模型:基于LASSO-logistic回归  

Risk model for predicting severe dengue and dengue with warning signs by early indications in border areas in Yunnan province:based on LASSO-logistic regression

作　　者：傅瀚文 沈加员[2] 吴超[2] 张晓灿 俞皓尹 李静 FU Hanwen;SHEN Jiayuan;WU Chao;ZHANG Xiaocan;YU Haoyin;LI Jing(Key Laboratory of Cross-border Infectious Diseases and New Drug Development of Yunnan Province,Yunnan Key Laboratory of Public Health and Biosafety,School of Public Health,Kunming Medical University,Kunming,Yunnan 650500,China;Yunnan Institute of Parasitic Diseases,Pu'er,Yunnan 665099,China)

机构地区：[1]云南省跨境传染病与新药创制重点实验室,云南省公共卫生与生物安全重点实验室,昆明医科大学公共卫生学院,云南昆明650500 [2]云南省寄生虫病防治所,云南普洱665099

出　　处：《中国热带医学》2025年第3期309-315,共7页China Tropical Medicine

基　　金：国家重点研发计划项目(No.2023YFC2307400)。

摘　　要：目的 在登革热疾病进展的早期阶段通过建立预测模型来评估重症登革热及登革热伴预警出现的风险,以避免重症登革热在疾病发展早期得不到有效防治,并降低登革热死亡率。方法 回顾性收集云南省瑞丽市人民医院2019—2023年临床及实验室检查等831例患者资料,按7∶3比例分为训练集和验证集。训练集进行统计描述、单因素分析,LASSO回归筛选变量,logistic回归开发登革热重症风险预警模型;训练集和验证集进行ROC曲线模型性能验证。结果 本研究共纳入831名登革热患者,年龄为(44.20±15.02)岁,52.59%为男性,5.42%为缅甸籍;发生重症登革热或登革热伴预警122例,占比14.68%,女性为主(58.20%)。训练集采用LASSO回归筛选发生重症登革热或登革热伴预警的相关变量11个:年龄、头晕、呕吐、凝血酶原时间、部分活化凝血活酶时间、红细胞压积、血小板、单核细胞百分比、单核细胞绝对值、血红蛋白、C反应蛋白(λmin=0.011 59);logistic回归建立重症登革热及登革热伴预警模型,具有统计学意义的变量为年龄[OR=1.034(95%CI:1.016~1.053)]、红细胞压积[OR=1.258(95%CI:1.143~1.519)]、血小板[OR=0.991(95%CI:0.985~0.997)]、血红蛋白[OR=0.919(95%CI:0.873~0.950)]、C反应蛋白[OR=1.019(95%CI:1.004~1.034)]。训练集ROC曲线下面积(area under the curve, AUC)值为0.894(95%CI:0.796~0.867),验证集中AUC值为0.862(95%CI:0.709~0.827)。最佳阈值点(Cut-off)取0.197,灵敏度为0.850,特异度为0.743。结论 本研究建立了LASSO-logistic回归风险预测模型,可在登革热患者患病早期预测发生重症登革热及登革热伴预警风险,提高医院重症登革热的防治能力,有助于指导临床治疗决策。Objective A predictive model should be established during the early stages of dengue progression to evaluate the likelihood of severe dengue and dengue with warning signs,thereby preventing delayed clinical management and reducing dengue-related mortality.Methods Clinical and laboratory examination data of 831 patients admitted to Ruili People's Hospital of Yunnan Province during 2019-2023 were retrospectively collected.The dataset was divided into a training set and a validation set in a 7∶3 ratio.Statistical description and univariate analysis were performed on the training set,with LASSO regression employed to screen variables,followed by logistic regression to develop a risk prediction model for severe dengue.Model performance was validated using ROC curves on both the training set and validation set.Results A total of 831 dengue patients were included in the study,with a mean age of(44.20±15.02)years.Among them,52.59%were male and 5.42%were Myanmar nationality.In total,122 cases(14.68%)exhibited severe dengue or dengue with warning signs,predominantly female(58.20%).LASSO regression was used in the training set to screen 11 variables related to the risk of severe dengue and dengue with warning signs:Age,dizziness,vomiting,prothrombin time,partial activated thromboplastin time,hematocrit,platelet,monocyte percentage,absolute value of monocytes,hemoglobin,C-reactive protein(λmin=0.01159);Logistic regression identified statistically significant variables for the risk model of severe dengue and dengue with warningsigns as follows:age[OR=1.034(95%CI:1.016-1.053)],red blood cells deposited[OR=1.258(95%CI:1.143-1.519)],platelet[OR=0.991(95%CI:0.985-0.997)],hemoglobin[OR=0.919(95%CI:0.873-0.950)],C-reactive protein[OR=1.019(95%CI:1.004-1.034)].The model achieved an AUC of 0.894(95%CI:0.796-0.867)in the training set and 0.862(95%CI:0.709-0.827)in the validation set.At a cut-off threshold of 0.197,sensitivity and specificity were 0.850 and 0.743,respectively.Conclusion This study established a LASSO-logistic regre

关 键 词：重症登革热 临床特征 预测模型 LASSO回归 LOGISTIC回归 边境地区 云南省 

分 类 号：R373.33[医药卫生—病原生物学]