基于NHANES数据库开发和验证卒中后抑郁风险的临床预测模型  

Development and validation a clinical prediction model for post-stroke depression based on NHANES database

作　　者：胡填 王陶陶 叶玉焊 陈楚霈 古剑雄 Hu Tian;Wang Taotao;Ye Yuhan;Chen Chupei;Gu Jianxiong(School of Medical Technology,Guangdong Medical University,Zhanjiang 524000,China;Department of Rehabilitation Medicine,Zhongshan City People's Hospital,Zhongshan 528400,China;Child Developmental and Behavior Center,the Third Affiliated Hospital of Sun Yat-sen University,Guangzhou 510000,China;The First Clinical Medical College,Guangdong Medical University,Zhanjiang 524000,China;Department of Rehabilitation Medicine,Affiliated Hospital of Guangdong Medical University,Zhanjiang 524000,China)

机构地区：[1]广东医科大学医学技术学院,湛江524000 [2]中山市人民医院康复医学科,528400 [3]中山大学附属第三医院儿童发育行为中心,广州510000 [4]广东医科大学第一临床医学院,湛江524000 [5]广东医科大学附属医院康复医学科,湛江524000

出　　处：《神经疾病与精神卫生》2023年第3期153-160,共8页Journal of Neuroscience and Mental Health

基　　金：国家自然科学基金(61431007);广东医科大学青年培育基金(GDMUQ2021034)。

摘　　要：目的基于美国国家健康和营养检查调查(NHANES)数据库开发、验证以非侵入方法评估卒中幸存者发生卒中后抑郁(PSD)的列线图,为临床早期筛选高危人群提供有价值的参考。方法从NHANES数据库中选取2007—2018年调查的1003例卒中幸存者为研究对象,将2007—2014年调查的659例卒中幸存者纳入建模组,将2015—2018年调查的344例卒中幸存者纳入验证组。采用患者健康问卷(PHQ-9)评估卒中幸存者的抑郁症状。采用多因素Logistic回归分析卒中幸存者发生PSD的影响因素。将多因素Logistic回归分析中P<0.10的预测因子纳入列线图,采用受试者工作特征(ROC)曲线下面积(AUC)和校准曲线评价列线图的预测性能,采用决策曲线分析探究列线图的临床应用价值。结果在1003例卒中幸存者中,共190例(18.94%)被评估为有抑郁症状(PHQ-9≥10分),其中建模组124例(18.82%),验证组66例(19.19%)。多因素Logistic回归分析结果表明,女性(OR=1.671,95%CI=1.040~2.684)、有睡眠障碍(OR=2.797,95%CI=1.740~4.494)、有工作限制(OR=2.293,95%CI=1.362~3.861)和有行走障碍(OR=2.163,95%CI=1.304~3.588)是卒中幸存者发生PSD的危险因素(P<0.05);60~79岁(OR=0.321,95%CI=0.121~0.852)、≥80岁(OR=0.117,95%CI=0.032~0.426)是卒中幸存者发生PSD的保护因素(P<0.05)。构建基于性别、年龄、心血管疾病史、睡眠障碍、工作限制和行走障碍的列线图,ROC曲线显示,建模组的AUC为0.797(95%CI:0.756~0.838),经1000次Bootstrap重抽样法内部验证得C指数为0.782;验证组AUC为0.752(95%CI:0.684~0.820)。校准曲线表明,列线图预测PSD发生概率与实际发生概率基本吻合。决策曲线结果显示,当阈值概率为5%~75%时,使用该预测模型筛查卒中患者将获得更高的净收益。结论本研究构建的PSD风险列线图预测性能良好,可用于卒中患者进行早期PSD风险的筛查,以帮助医生做出更好的治疗决策。Objective To develop and validate a nomogram by non-invasive method to assess the risk of post-stroke depression(PSD)based on National Health and Nutrition Examination Survey(NHANES)database,to provide a valuable reference for early clinical screening of high-risk individuals.Methods A total of 1003 stroke survivors surveyed from 2007 to 2018 were selected from the NHANES database for the study.A total of 659 cases from 2007 to 2014 were assigned to development group,and 344 cases from 2015 to 2018 were assigned to validation group.Depressive symptoms in stroke survivors were assessed by the Patient Health Questionnaire-9(PHQ-9).Multivariate Logistic regression was applied to analyze the influencing factors of PSD in stroke survivors.Predictive factors with P<0.10 in multivariate Logistic regression analysis was incorporated into the nomogram.The prediction performance of nomogram was evaluated by area under curve(AUC)of receiver operator characteristic(ROC)curve.The clinical application value of nomograms was analyzed by decision curve.Results Of all the 1003 patients,there were 190 cases(18.94%)were assessed with depression symptoms(the score of PHQ-9 greater or equal to 10),with 124 cases in development group(18.82%)and 66 cases in validation group(19.19%).Multivariate Logistic analysis showed that female(OR=1.671,95%CI=1.040-2.684),sleep disorder(OR=2.797,95%CI=1.740-4.494),work limitation(OR=2.293,95%CI=1.362-3.861)and difficulty walking(OR=2.163,95%CI=1.304-3.588)were independent risk factors for PSD in stroke survivors(P<0.05),and 60 to 79 years old(OR=0.321,95%CI=0.121-0.852),≥80 years old(OR=0.117,95%CI=0.032-0.426)were the protective factor for PSD in stroke survivors(P<0.05).The nomogram was constructed based on gender,age,history of cardiovascular disease,sleep disorders,work restrictions and walking disorders.The AUC was 0.797(95%CI=0.756-0.838)in the development group.After 1000 times internal validation with bootstrap resampling methods,the C-index was 0.782.The AUC was 0.752(95%CI=0.684-0.820)in

关 键 词：卒中后抑郁 临床预测模型 列线图 风险因素 

分 类 号：R743.3[医药卫生—神经病学与精神病学] R749.4[医药卫生—临床医学]