基于LASSO-logistic回归的TB/HIV患者抗结核治疗结局影响因素研究  

Influencing factors of anti-tuberculosis treatment outcome in TB/HIV co-infected patients based on LASSO-logistic regression model

作　　者：严菁敏 陈莹 于洁 胡琳 廖如珺[2] 沈卓之[3] 魏荣杰 姜恬 王若冰 王菊[4] 张韬[1] YAN Jing-min;CHEN Ying;YU Jie;HU Lin;LIAO Ru-jun;SHEN Zhuo-zhi;WEI Rong-jie;JIANG Tian;Wang Ruo-bing;WANG Ju;ZHANG Tao(Department of Epidemiology and Health Statistics,West China School of Public Health/West China Fourth Hospital,Sichuan University,Chengdu,Sichuan 610041,China)

机构地区：[1]四川大学华西公共卫生学院/四川大学华西第四医院,流行病与卫生统计学系,四川成都610041 [2]四川省疾病预防控制中心 [3]重庆市疾病预防控制中心 [4]凉山州疾病预防控制中心 [5]四川大学学报(医学版)编辑部

出　　处：《现代预防医学》2023年第22期4039-4045,4051,共8页Modern Preventive Medicine

基　　金：四川省科技厅项目(2022YFS0229,2022YFS0641);重庆市科技局项目(cstc2020jscx-cylhX0003);凉山州科技计划重点研发项目(22ZDYF0125)。

摘　　要：目的基于LASSO-logistic模型分析凉山彝族自治州TB/HIV患者抗结核治疗结局的影响因素。方法利用凉山州结核病专报系统中2019年1月1日至2022年12月31日上报的TB/HIV患者2130例,采用LASSO-logistic回归模型分析TB/HIV患者抗结核治疗结局的影响因素,并与常规的logistic回归结果进行比较,用AIC和BIC评价两种模型的效果。结果研究结果表明年龄≥60岁(OR=3.67,95%CI:1.46~8.81)、离异或丧偶(OR=1.77,95%CI:1.16~2.66)、血行播散(OR=2.85,95%CI:1.54~5.05)、确诊结核病时病原学结果为阳性(OR=1.84,95%CI:1.15~2.89)、延迟治疗≥1天(OR=1.96,95%CI:1.41~2.72)、复治(OR=2.26,95%CI:1.42~3.52)是凉山彝族自治州TB/HIV患者抗结核治疗结局的危险因素,而确诊结核时的CD4细胞计数>50cells/μl是一个保护因素(OR=0.54,95%CI:0.38~0.76)。LASSO-logistic回归模型的AIC和BIC均小于常规logistic回归模型。结论研究发现,年龄、婚姻状况、患者的结核类型、确诊结核病时的病原学结果、延迟治疗、患者分类、确诊结核时的CD4细胞计数水平与四川省凉山彝族自治州TB/HIV患者的抗结核治疗结局有关。本文采用LASSO-logistic回归分析影响因素,能纳入足够多的变量进行筛选且不会受到各变量间共线性的影响,最终的模型具有良好的解释性和稳定性。Objective To analyze the influencing factors of anti-tuberculosis treatment outcomes in TB/HIV patients in Liangshan Yi Autonomous Prefecture based on LASSO-logistic model.Methods A total of 2130 TB/HIV patients reported from the TB Special Report System of Liangshan Prefecture from January 1,2019 to December 31,2022 were selected.LASSO-logistic regression model was used to analyze the influencing factors of anti-tuberculosis treatment outcome of TB/HIV patients.The results were compared with those of the conventional logistic regression model,then AIC and BIC were used to evaluate the performance of the two models.Results The results showed that age≥60(OR=3.67,95%CI:1.46-8.81),divorced or widowed(OR=1.77,95%CI:1.16-2.66),hematogenous spread(OR=2.85,95%CI:1.54-5.05),the positive etiology when tuberculosis was diagnosed(OR=1.84,95%CI:1.15-2.89),delayed treatment≥1 day(OR=1.96,95%CI:1.41-2.72),patients who were retreatment(OR=2.26,95%CI:1.42-3.52)were risk factors for the outcome of anti-tuberculosis treatment in TB/HIV patients in Liangshan Yi Autonomous Prefecture.The CD4 cell count>50cells/μl at the time of diagnosis of tuberculosis was a protective factor(OR=0.54,95%CI:0.38-0.76).The AIC and BIC of the LASSO-logistic regression model were lower than the AIC and BIC of the conventional logistic regression model.Conclusion Our results show that age,marital status,type of tuberculosis,results of etiological diagnosis,delayed treatment,patient classification and CD4 cell count at the time of diagnosis of tuberculosis were associated with the outcome of anti-tuberculosis treatment in TB/HIV co-infected patients in Liangshan Yi Autonomous Prefecture,Sichuan Province.In our study,LASSO-logistic regression is used to analyze the influencing factors,which could include enough variables for screening without being affected by collinearity between the variables,and the final model has better explanatory and stability.

关 键 词：TB/HIV患者 抗结核治疗结局 LASSO-logistic 影响因素 

分 类 号：R512.91[医药卫生—内科学]