中国全科医学 ›› 2026, Vol. 29 ›› Issue (15): 1998-2005.DOI: 10.12114/j.issn.1007-9572.2025.0236

• 论著 • 上一篇    下一篇

亚临床高风险人群夜间高血压筛查模型的构建与验证:一项单中心队列研究

刘帆1, 陈秋雨2, 李婧3,*()   

  1. 1.010050 内蒙古呼和浩特市,内蒙古医科大学附属医院
    2.010050 内蒙古呼和浩特市,内蒙古医科大学附属医院血液内科
    3.010050 内蒙古呼和浩特市,内蒙古医科大学附属医院心血管内科
  • 收稿日期:2025-07-01 修回日期:2025-12-01 出版日期:2026-05-20 发布日期:2026-04-14
  • 通讯作者: 李婧
  • 刘帆与陈秋雨为共同第一作者


    作者贡献:

    刘帆负责论文数据搜集、论文撰写;陈秋雨负责论文格式、图表制作;李婧负责论文设计、撰写指导,对文章整体负责。

  • 基金资助:
    国家自然科学基金资助项目(82260075); 内蒙古医科大学附属医院高层次人才培养项目"航行系列"(QH202402); 内蒙古医科大学创客培育项目(101322024101)

Development and Validation of a Nighttime Hypertension Screening Model for Subclinical High-risk Populations: a Single-center Cohort Study

LIU Fan1, CHEN Qiuyu2, LI Jing3,*()   

  1. 1. The Affiliated Hospital of Inner Mongolia Medical University, Hohhot 010050, China
    2. Hematology Department, the Affiliated Hospital of Inner Mongolia Medical University, Hohhot 010050, China
    3. Cardiovascular Medicine Department, the Affiliated Hospital of Inner Mongolia Medical University, Hohhot 010050, China
  • Received:2025-07-01 Revised:2025-12-01 Published:2026-05-20 Online:2026-04-14
  • Contact: LI Jing
  • About author:

    LIU Fan and CHEN Qiuyu are co-first authors

摘要: 背景 夜间高血压(NH)是心脑肾等多器官损伤的重要因素,为全因死亡的预测因子,其预测价值高于日间血压及诊室血压,因其在夜间睡眠中发生的隐蔽特点被归为隐匿性高血压的一种,早日筛查并给予个体化治疗能够降低心脑肾疾病的患病风险。 目的 基于常规体检的实验室检查指标、个人健康资料探究NH患者的相关危险因素,建立NH的临床预测Nomogram模型。 方法 纳入2021-01-01—2024-06-30于内蒙古医科大学附属医院就诊的完善24 h动态血压监测的406例患者为研究对象,收集患者临床基线资料、实验室检查结果、心脏超声结果,进行动态血压监测。将患者按7∶3比例随机划分为训练集(n=284)和验证集(n=122)。采用LASSO回归分析与多因素Logistic回归分析构建NH风险预测模型并绘制列线图,绘制ROC曲线,计算ROC曲线下面积(AUC)验证预测模型的准确性。绘制校准曲线衡量模型的预测能力,反映预测风险与实际风险的一致性。 结果 依据24 h动态血压监测结果将患者分为NH组(n=254)和非NH组(n=152)。以LASSO回归分析筛选出的4个预测变量[体质量、总胆固醇(TC)、高血压、脑卒中]为自变量进行多因素Logistic回归分析,结果显示,体质量增大(OR=1.029,95%CI=1.006~1.053)、TC升高(OR=1.496,95%CI=1.136~1.972)、高血压(OR=2.372,95%CI=1.214~4.632)、脑卒中(OR=7.850,95%CI=4.157~14.824)均为NH的危险因素(P<0.05)。绘制列线图,脑卒中(得分0或62)和高血压(得分0或26)对诊断率的影响显著高于体质量和TC(得分随变量值线性变化),模型满分240分,当得分超过176分时,NH的风险为95%。绘制ROC曲线对上述预测模型进行评价,训练集AUC为0.791(95%CI=0.739~0.843),灵敏度为0.698,特异度为0.786;验证集AUC为0.820(95%CI=0.742~0.899),灵敏度为0.817,特异度为0.725。Hosmer-Lemeshow校准曲线显示模型拟合度较好,决策曲线分析结果显示验证集在阈值概率范围为0.2~0.6表现出较高的净收益,具有最佳的临床实用性。 结论 本研究建立了包含体质量、TC、高血压、脑卒中4项临床指标的NH风险预测模型,可用于预测筛查疑似患者的NH发生风险,模型具有良好的拟合度、区分度和临床应用价值。

关键词: 高血压, 夜间高血压, 风险预测模型, 列线图

Abstract:

Background

Nocturnal hypertension (NH) is a significant contributor to multi-organ damage (cardiovascular, cerebral, and renal) and serves as a predictor of all-cause mortality. Its predictive value surpasses that of daytime blood pressure and office blood pressure. Classified as a form of masked hypertension due to its occult nature during nocturnal sleep, early screening and individualized treatment can mitigate the risk of cardiovascular, cerebral, and renal diseases.

Objective

To investigate risk factors associated with NH by leveraging routine health checkup parameters and personal health data, and to develop a clinical predictive nomogram model for NH.

Methods

A total of 406 patients who underwent 24-hour ambulatory blood pressure monitoring (ABPM) at the Affiliated Hospital of Inner Mongolia Medical University between January 1, 2021, and June 30, 2024, were included. Baseline clinical data, laboratory test results, and echocardiographic findings were collected. Patients were randomly divided into a training set (n=284) and a validation set (n=122) in a 7∶3 ratio. A risk prediction model for NH was constructed using LASSO regression analysis and multivariate Logistic regression analysis, followed by Nomogram development. The ROC curve was plotted, and the area under the ROC curve (AUC) was calculated to validate the model's accuracy. Calibration curves were generated to assess the model's predictive capability and consistency between predicted and observed risks.

Results

Based on 24-hour ABPM results, patients were categorized into an NH group (n=254) and a non-NH group (n=152). Four predictors identified via LASSO regression "body weight, total cholesterol (TC), hypertension, and stroke "were used as independent variables in multivariate Logistic regression analysis. The results indicated that increased body weight (OR=1.029, 95%CI=1.006-1.053), elevated TC (OR=1.496, 95%CI=1.136-1.972), hypertension (OR=2.372, 95%CI=1.214-4.632), and stroke (OR=7.850, 95%CI=4.157-14.824) were all risk factors for NH (P<0.05). The Nomogram revealed that stroke history (score: 0 or 62) and hypertension (score: 0 or 26) had a more pronounced impact on diagnostic rates compared to body weight and TC (scores varied linearly with variable values). The total model score was 240, with a 95%risk of NH when the score exceeded 176. ROC curve analysis demonstrated an AUC of 0.791 (95%CI=0.739-0.843) in the training set, with a sensitivity of 0.698 and specificity of 0.786. In the validation set, the AUC was 0.820 (95%CI=0.742-0.899), with a sensitivity of 0.817 and specificity of 0.725. The Hosmer-Lemeshow calibration curve indicated good model fit, and decision curve analysis showed that the validation set achieved high net benefit within a threshold probability range of 0.2-0.6, confirming optimal clinical utility.

Conclusion

This study established an NH risk prediction model incorporating four clinical indicators: body weight, TC, hypertension history, and stroke history. The model demonstrates robust calibration, discrimination, and clinical applicability for screening NH risk in suspected patients.

Key words: Hypertension, Nocturnal hypertension, Risk prediction model, Nomogram