1、现代生物医学进展Progress in Modern Biomedicine Vol.23NO.2JAN.2023doi:10.13241/ki.pmb.2023.02.023支原体生殖道感染妇女阴道微生态改变对临床转归的影响及 XGboost 模型构建*赵思琪王焱张俊俊咸倩倩赵珍（河南科技大学临床医学院/河南科技大学第一附属医院妇科 河南 洛阳 471003）摘要 目的：探讨支原体生殖道感染妇女阴道微生态改变对临床转归的影响，并构建 XGboost 模型。方法：选取 2019 年 1 月 2020年 12 月于我院妇科门诊确诊的支原体生殖道感染妇女 186 例。根据治疗后 3 个月的临床转归
2、分为有效组 145 例和无效组 41例。比较两组患者的临床资料、阴道微生态形态学指标和功能学指标、微生态类型。使用 Cox 比例风险回归森林图筛选影响支原体生殖道感染妇女临床转归的因素，利用筛选出的影响因素构建 XGboost 模型并对影响因素按重要度排序，ROC 曲线分析 XG-boost 模型对支原体生殖道感染妇女临床转归的预测效能，校准曲线评价 XGboost 模型的准确度，临床决策曲线评价 XGboost 模型的有效性。结果：两组患者的流产次数、学历、年龄、避孕方式、生产次数对比有差异（P0.05）。与有效组相比，无效组患者阴道微生态形态学和功能学各指标异常发生率、阴道微生态失调率明显
3、较高（P0.05），无效组的需氧型阴道炎（AV）、外阴阴道假丝酵母菌病（VVC）、细菌性阴道病（BV）、BV+VVC、滴虫性阴道炎（TV）、BV 中间型+VVC 的检出率均明显较高（P0.05），无效组正常微生态及菌群正常、功能下降所占人数比例均明显更低，差异均有统计学意义（P0.05）。年龄 1次、生产产次3 次以及阴道微生态失调是影响支原体生殖道感染转归的重要因素（P0.05）。构建的 XGboost 模型具有较高的预测效能，准确度和有效性均较高。结论：年龄、学历、流产次数、生产次数以及阴道微生态失衡是影响支原体生殖道感染临床转归的重要因素，本研究构建的 XGboost 模型具有较好的预测
4、效能。关键词：支原体生殖道感染；阴道微生态；临床转归；影响因素；XGboost 模型中图分类号：R711.3文献标识码：A文章编号：1673-6273（2023）02-324-06Effect of Vaginal Microecological Changes on the Clinical Outcomein Women with Mycoplasma Genital Tract Infection and Construction ofXGboost Model*ZHAO Si-qi,WANG Yan,ZHANG Jun-jun,XIAN Qian-qian,ZHAO Zhen(Depa
5、rtment of Gynaecology,School of Clinical Medicine,Henan University of Science and Technology/The First Affiliated Hospital ofHenan University of Science and Technology,Luoyang,Henan,471003,China)ABSTRACT Objective:To investigate the effect of vaginal microecological changes on the clinical outcome i
6、n women withMycoplasma genital tract infection,and to construct XGboost model.Methods:186 women with Mycoplasma genital tract infection whowere diagnosed in the gynecological clinic of our hospital from January 2019 to December 2020 were selected.According to the clinicaloutcome 3 months after treat
7、ment,they were divided into effective group with 145 cases and ineffective group with 41 cases.The clinicaldata,vaginal microecological morphological and functional indexes and microecological types of the two groups were compared.TheCox proportional hazards regression forest map was used to screen
8、the factors affecting the clinical outcome of women with Mycoplasmagenital tract infection.The selected influencing factors were used to construct the XGboost model and rank the influencing factorsaccording to the importance.The ROC curve was used to analyze the prediction efficiency of XGboost mode
9、l on the clinical outcome ofwomen with Mycoplasma genital tract infection,the calibration curve was used to evaluate the accuracy of XGboost model,and theclinical decision curve was used to evaluate the effectiveness of XGboost model.Results:There were significant differences in thenumber of abortio
10、ns,educational background,age,contraceptive methods and the number of births between the two groups(P0.05).Compared with the effective group,the incidence of abnormal vaginal microecological morphology and function indexes and the rate ofvaginal microecological imbalance in the ineffective group wer
11、e significantly higher(P0.05).The detection rates of aerobic vaginitis(AV),vulvovaginal candidiasis(VVC),bacterial vaginosis(BV),BV combined with VVC,trichomonal vaginitis(TV),BV intermediatecombined with VVC in the ineffective group were significantly higher(P0.05).The proportion of normal microeco
12、logy,normal flora andfunctional decline in the ineffective group was significantly lower,and the differences were statistically significant(P0.05).Age 1,number of births 3 and vaginal microecologicalimbalance were important factors affecting the prognosis of Mycoplasma genital tract infection(P0.05)
13、.The XGboost model has highprediction efficiency,high accuracy and effectiveness.Conclusions:Age,educational background,number of abortions,number of birthsand vaginal microecological imbalance are important factors affecting the clinical outcome of Mycoplasma genital tract infection.TheXGboost mode
14、l constructed in this study has good prediction efficiency.Key words:Mycoplasma genital tract infection;Vaginal microecological;Clinical outcome;Influencing factors;XGboost modelChinese Library Classification(CLC):R711.3Document code:AArticle ID:1673-6273(2023)02-324-06前言生殖道支原体感染是临床上女性常见的一种感染，极易导致患者
15、出现不孕不育、流产、宫颈炎、宫颈内膜炎等疾病，严重危害患者的生殖健康1-3。因此，分析支原体生殖道感染临床转归的影响因素并据此进行防治，对治疗妇科生殖系统疾病有重要意义。在健康状态下，女性阴道微生物多样性要比人体黏膜和表面等其他部位的微生态体系低，其中乳酸杆菌占主导地位4,5。当乳酸杆菌数量降低时，致病菌群大量繁殖，导致阴道微生态失衡，从而增加病菌感染风险。准确了解女性阴道微生态环境，有助于防止阴道微生态失衡。已有研究报道指出6，阴道微生态失衡是女性生殖道高危型人乳头瘤病毒（HR-HPV）感染转归的主要危险因素，然而关于女性阴道微生态改变对支原体生殖道感染转归的影响鲜有研究报道。本研究通过探讨
16、支原体生殖道感染患者临床转归的影响因素，分析阴道微生态改变对患者临床转归的影响，并构建 XGboost 模型，旨在为临床上支原体生殖道感染患者治疗后临床转归评估提供理论依据，从而改善支原体生殖道感染患者的预后。1 对象与方法1.1 研究对象选取 2019 年 1 月 2020 年 12 月于我院妇科门诊确诊的支原体生殖道感染妇女 186 例。纳入标准：（1）完成为期 3 个月随访；（2）已婚或有性交史；（3）年龄 1850 周岁；（4）符合 细菌性阴道病诊治指南（草案）相关标准7。排除标准：（1）无法配合本研究；（2）临床资料缺失；（3）妊娠期、哺乳期或已绝经妇女；（4）合并严重肝肾功能障碍；（5）合并内分泌疾病、免疫性疾病、恶性肿瘤及其他全身性感染等疾病。本研究经我院伦理委员会批准（批号：Z1301160026），开展研究前，患者签署知情同意书。1.2 研究方法1.2.1 临床转归评估根据支原体培养及药敏结果正确使用抗生素。评估治疗后 3 个月的病情恢复情况，每月进行 1 次支原体培养。疗效评价标准8：痊愈：患者经治疗后，外阴灼热、小便不适、瘙痒、异味和白带增多等临床症状和体征完全消