基于最小绝对值压缩选择模型的青春期多囊卵巢综合征抑郁风险预警模型构建及验证  

Development and validation of a least absolute shrinkage and selection operator-based prediction model for depression in adolescents with polycystic ovary syndrome

作　　者：丁锐 谭惠文 刘颖 严欣 郭云梅 王连红 DING Rui;TAN Hui-wen;LIU Ying;YAN Xin;GUO Yun-mei;WANG Lian-hong(Department of Nursing,Affiliated Hospital of Zunyi Medical University,Zunyi,Guizhou 563003,China)

机构地区：[1]遵义医科大学附属医院护理部,贵州遵义563003 [2]重庆医科大学附属第二医院感染病科 [3]遵义医科大学护理学院

出　　处：《现代预防医学》2024年第10期1787-1794,1832,共9页Modern Preventive Medicine

基　　金：2023年贵州省卫生健康委科学技术基金项目(gzwkj2023-243);遵义市科技计划项目(遵市科合HZ字(2021)21号)。

摘　　要：目的构建青春期多囊卵巢综合征(Polycystic Ovary Syndrome,PCOS)抑郁风险预警模型并验证。方法收集2021年10月—2022年9月到遵义医科大学附属医院妇科门诊就诊的青春期PCOS患者资料。使用R 4.2.1软件的最小绝对值压缩选择模型(Least Absolute Shrinkage and Selection Operator,LASSO)筛选危险因素,纳入logistic回归建立青春期PCOS抑郁风险预警模型,绘制列线图并进行内外部验证,使用区分度、特异度和灵敏度评价模型,通过临床决策曲线图分析临床效用。结果最终筛选出6个预测变量构建模型:抑郁风险=1/(1+exp-(-4.055+0.221×睡眠+0.729×服用激素类避孕药+0.920×多毛+0.079×疾病认知-0.058×社会支持+1.049×促黄体激素/促卵泡激素(≥2)))。该模型ROC曲线下面积0.881,最佳截断值0.278时,特异度和灵敏度分别为76.2%和88.0%。内部验证后ROC曲线下面积0.867。临床决策曲线结果示模型可为临床决策提供证据支持,外部验证ROC曲线下面积0.871。结论本研究构建的青春期PCOS抑郁风险预警模型,可早期识别青春期PCOS抑郁高风险人群,为实施综合有效的风险防范措施提供理论依据。Objective To establish a depression prediction model for adolescents with Polycystic Ovary Syndrome(PCOS)and validate the model.Methods Patients’data were collected from the gynecological clinic of Affiliated Hospital of Zunyi Medical University according to the item pool of risk factors for depression in adolescents with PCOS.Data collected between October 2021 and September 2022.In this study,R software(version 4.2.1)was used to perform regression analysis by the least absolute shrinkage and selection operator(LASSO),so as to screen out strong risk factors related to depression in adolescents with PCOS.These risk factors were then incorporated into logistic regression to develop a depression warning model in adolescents with PCOS.The model has been visualized by nomogram and has been verified both internally and externally.The predicted effect of the model was evaluated through discrimination,specificity and sensitivity.Decision curve analysis was used to analyze the clinical effect of the model.Results The model was as follows:depression risk=1/(1+exp-(-4.055+0.221×sleep+0.729×hormonal contraceptive use+0.920×hirsutism+0.079×illness perception-0.058×social support+1.049×(luteinizing hormone/follicle stimulating hormone)≥2)).The area under the ROC curve for this model was 0.881.The optimal cut-off value on the ROC curve was 0.278,corresponding to a high specificity and sensitivity of 76.2%and 88.0%,respectively.The corrected area under the ROC curve obtained was 0.867.In addition,the result of decision curve analysis showed that the model could provide effective evidence support for clinical decision-making.The area under the ROC curve obtained from external validation was 0.871.Conclusion In this study,an early warning model of depression risk in adolescents with PCOS was constructed.It can effectively identify people at high risk of depression in adolescents with PCOS at an early stage,thus providing a theoretical basis for the implementation of comprehensive and effective risk prevention measures.

关 键 词：多囊卵巢综合征 青春期 抑郁 最小绝对值压缩选择模型 预警模型 

分 类 号：R711.75[医药卫生—妇产科学]