中国全科医学 ›› 2023, Vol. 26 ›› Issue (12): 1437-1443.DOI: 10.12114/j.issn.1007-9572.2022.0695

所属专题: 神经系统疾病最新文章合集 脑健康最新研究合集

• 论著 • 上一篇    下一篇

基于贝叶斯多变量联合模型的体检人群脑卒中发病风险因素的纵向研究

杨毅1, 丛慧文1, 王廉源1, 杨丽萍2, 包绮晗1, 王浩桦1, 李承圣1, 周立雯1, 丁子琛1, 石福艳1,*(), 王素珍1,*()   

  1. 1261053 山东省潍坊市,潍坊医学院公共卫生学院卫生统计学系
    2710032 陕西省西安市,空军军医大学西京医院健康医学中心
  • 收稿日期:2022-09-26 修回日期:2023-01-01 出版日期:2023-04-20 发布日期:2023-01-18
  • 通讯作者: 石福艳, 王素珍

  • 作者贡献:杨毅负责选题、清洗保留研究数据、模型构建、计算机代码和支持算法的实现、原稿写作;丛慧文和王廉源负责模型构建、计算机代码和支持算法的实现;杨丽萍负责调查开展、提供研究数据;包绮晗负责数据可视化展示;王浩桦和李承圣负责验证研究结果;周立雯和丁子琛负责清洗和整合研究数据;通信作者石福艳和王素珍对选题进行指导,对文章涉及观点及立论依据进行审阅和修订;所有作者确认了论文的最终稿。
  • 基金资助:
    国家自然科学基金资助项目--基于双稳健共享参数Joint模型的脑卒中早期关键风险因素推断研究(81803337); 国家自然科学基金资助项目--广义提升模型和机器学习下基于逆概率加权的纵向数据因果推断研究(81872719); 国家统计局课题--基于机器学习和逆概率加权的高维数据降维方法及应用研究(2018LY79); 山东省自然科学基金资助项目--基于广义提升模型下逆概率加权法的纵向数据因果推断研究(ZR2019MH034); 山东省高等学校青创人才引育计划--卫生统计理论方法及应用研究创新团队(No.2019-6-156,Lu-Jiao); 潍坊医学院博士启动基金项目--心脑血管病纵向数据关键风险因素推断及共享参数模型研究(2017BSQD51)

Longitudinal Study on the Risk Factors of Stroke in Check-up Population Based on Bayesian Multivariate Joint Model

YANG Yi1, CONG Huiwen1, WANG Lianyuan1, YANG Liping2, BAO Qihan1, WANG Haohua1, LI Chengsheng1, ZHOU Liwen1, DING Zichen1, SHI Fuyan1,*(), WANG Suzhen1,*()   

  1. 1Department of Health Statistics, School of Public Health, Weifang Medical University, Weifang 261053, China
    2Center for Health and Medicine, Xijing Hospital, Air Force Military Medical University, Xi'an 710032, China
  • Received:2022-09-26 Revised:2023-01-01 Published:2023-04-20 Online:2023-01-18
  • Contact: SHI Fuyan, WANG Suzhen

摘要: 背景 脑卒中是目前影响人类健康的主要公共卫生问题之一;健康体检纵向数据累积了大量的健康信息,由于缺失数据多、样本量小等诸多问题,导致其利用率低、重要信息未能得到充分挖掘,进而对常见慢性病的有效防控等工作带来一定困难。目的 基于贝叶斯多变量联合模型,探讨体检人群脑卒中发病风险因素,为慢性病风险因素分析提供新的方法。方法 本研究使用空军军医大学西京医院健康医学中心2008—2015年的体检资料。随访情况:以首次发生脑卒中为结局事件,发生结局事件立即停止随访;若未发生,到2015年体检信息收集完成后结束随访;体检间隔时间为1年。依据随访过程中是否发生脑卒中分为脑卒中组和非脑卒中组。纵向观察变量包括总胆固醇(TC)、三酰甘油(TG)、低密度脂蛋白胆固醇(LDL-C)、高密度脂蛋白胆固醇(HDL-C)、体质指数(BMI)和收缩压(SBP)。采用多因素Cox回归模型分析基线情况对脑卒中结局事件的影响;采用贝叶斯多变量联合模型,分析随访过程中TC、TG、LDL-C、HDL-C、BMI和SBP的纵向变化轨迹对脑卒中发病的影响。结果 本研究共纳入234例研究对象,1 581条纵向随访记录,平均随访时间为(7.4±1.2)年,其中70例(29.9%)在随访过程中发生脑卒中。多因素Cox回归模型结果显示:基线TC、TG、LDL-C、HDL-C、BMI、SBP对脑卒中发病均无影响(P>0.05)。贝叶斯多变量联合模型结果显示:TG每纵向升高1 mmol/L,脑卒中发病风险升高1.863倍[95%CI(1.018,3.294),P=0.042];LDL-C每纵向升高1 mmol/L,脑卒中发病风险升高1.347倍[95%CI(1.045,1.863),P=0.046]。结论TG、LDL-C水平随时间变化的纵向升高是体检人群脑卒中发病的危险因素;贝叶斯多变量联合模型可用于体检人群的慢性病风险因素探讨研究中。

关键词: 卒中, 体格检查, 贝叶斯联合模型, Cox回归模型, 血脂异常, 危险因素

Abstract:

Background

Stroke is one of the major public health problems affecting human health in current. Longitudinal check up data has accumulated a large amount of health information. However, the utilization rate of the longitudinal check up data is low and important information has not been fully extracted due to many problems such as missing data and small sample size, which brings difficulties to the effective prevention and control of common chronic diseases.

Objective

To explore the risk factors of stroke in check-up population based on Bayesian multivariate joint model, so as to provide a new approach for the analysis of risk factors for chronic diseases.

Methods

In this study, the data were collected from the Center for Health and Medicine, Xijing Hospital, Air Force Military Medical University from 2008 to 2015. Follow-up status: the follow up was conducted with the first occurrence of stroke as the outcome event and stopped at the occurrence of outcome event or ended when the collection of medical examination information was completed by 2015 if the outcome event did not occur. The interval between physical examinations was 1 year. The participants were divided into the stroke group and the non-stroke group according to whether stroke occurred during follow-up. Longitudinal variables observed in this study included total cholesterol (TC) , triglyceride (TG) , low density lipoprotein cholesterol (LDL-C) , high density lipoprotein cholesterol (HDL-C) , body mass index (BMI) and systolic blood pressure (SBP) . Multivariate Cox regression model was used to analyze the influence of baseline conditions on stroke outcome events. Bayesian multivariate joint model was used for analyzing the effect of longitudinal trajectory of TC, TG, LDL-C, HDL-C, BMI and SBP on the incidence of stroke during follow-up.

Results

A total of 234 subjects with 1 581 longitudinal follow-up records were included in this study, with the mean follow-up time of (7.4±1.2) years, of which 70 cases (29.9%) developed stroke during the follow-up. The results of multivariate Cox proportional hazards model showed that there was no effect of baseline values including TC, TG, LDL-C, HDL-C, BMI and SBP on the incidence of stroke (P>0.05) . The results of Bayesian multivariate joint model showed that the risk of stroke was 1.863 times higher for per longitudinal increase of 1 mmol/L TG level 〔95%CI (1.018, 3.294) , P=0.042〕 and 1.347 times higher for per longitudinal increase of 1 mmol/L LDL-C level〔95%CI (1.045, 1.863) , P=0.046〕.

Conclusion

The longitudinal increase of TG and LDL-C levels over time is a risk factor for stroke in check-up population. Bayesian multivariate joint model can be used to explore the risk factors of chronic diseases in check-up population.

Key words: Stroke, Physical examination, Bayesian joint model, Cox regression model, Dyslipidemias, Risk factors