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
作者:
作者单位:

作者简介:

通讯作者:

基金项目:

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


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

Fund Project:

  • 摘要
  • |
  • 图/表
  • |
  • 访问统计
  • |
  • 参考文献
  • |
  • 相似文献
  • |
  • 引证文献
  • |
  • 资源附件
  • |
  • 文章评论
    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 1 003 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1 003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 1 000 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。

    Abstract:

    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 1 003 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 1 003 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 1 000 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 the validation group. The calibration curve showed that the predicted probability of PSD by the nomogram was basically consistent with the actual probability of occurrence. The decision curve results showed that when the threshold probability was between 5% and 75%, using this predictive model to screen stroke patients would result in higher net benefits. Conclusions The nomogram of PSD constructed has a good predictive performance, which can be used for early PSD risk screening in stroke patients to help physicians make better treatment decisions.

    参考文献
    相似文献
    引证文献
引用本文

胡填,王陶陶,叶玉焊,陈楚霈,古剑雄.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[J].神经疾病与精神卫生,2023,23(3):
DOI :10.3969/j. issn.1009-6574.2023.03.001.

复制
文章指标
  • 点击次数:
  • 下载次数:
历史
  • 收稿日期:
  • 最后修改日期:
  • 录用日期:
  • 在线发布日期: 2023-05-06