文献速递免疫规划慢病防控与筛查卫生政策与健康促进

50岁及以上成人带状疱疹疫苗犹豫的驱动因素:机器学习方法

Drivers of herpes zoster vaccine hesitancy in adults aged 50 and above: A machine learning approach.
PubMed 文献 · Vaccine · 2025-10-12

2024年4月至8月在上海四家社区卫生服务中心对50岁及以上成人开展横断面研究,共纳入1152名参与者(平均年龄66.1±8.8岁,男性49.7%),73.95%报告对带状疱疹疫苗犹豫。采用LASSO回归筛选变量后比较四种机器学习模型,XGBoost模型表现最佳(AUC 0.960,95%CI:0.937-0.983),SHAP分析识别出信心、疫苗素养、自满、疾病知识和计算为主要预测因素。

要点

  • 1152名50岁及以上成人中73.95%对带状疱疹疫苗犹豫。
  • XGBoost模型预测性能最佳,AUC为0.960(95%CI:0.937-0.983)。
  • 信心、疫苗素养、自满、疾病知识和计算是主要预测因素。

核心结论

  • SHAP-XGBoost模型对带状疱疹疫苗犹豫表现出强预测性能,疫苗素养和5C心理前因被识别为关键预测因素。

研究设计与样本

  • 横断面研究,于2024年4月至8月在上海四家社区卫生服务中心收集50岁及以上个体数据。
  • 数据收集包括人口学信息、带状疱疹疾病和疫苗知识、5C量表和疫苗健康素养。
  • 采用LASSO回归进行变量选择,随后使用逻辑回归、随机森林、支持向量机和极端梯度提升(XGBoost)四种机器学习算法分析关键变量。
  • 模型性能使用AUC和校准图评估,采用Shapley加法解释(SHAP)解释和识别关键预测因素。

主要发现

  • 共纳入1152名参与者(平均年龄66.1±8.8岁;男性49.7%),73.95%报告对带状疱疹疫苗犹豫。
  • 21个特征中有10个被选入建模。
  • XGBoost模型表现最佳,AUC为0.960(95%CI:0.937-0.983)。
  • SHAP分析识别出信心、疫苗素养、自满、疾病知识和计算为主要预测因素。

临床怎么用

  • 作者认为该工具可为针对性健康干预提供信息,未来工作可整合社区数据库进行大规模识别和定制化群体策略。

局限性

  • 原文未报告局限性。

以上解读依据论文摘要(全文未开放获取或许可不允许翻译转载)。

原文

BACKGROUND: Herpes zoster (HZ) poses a growing public health challenge among adults aged 50 and above, with vaccine hesitancy being a major barrier to improving immunization rates. Understanding the factors driving HZ vaccine hesitancy is essential for developing strategies to enhance vaccination uptake in this population. This study aims to identify and analyze the key determinants of HZ vaccine hesitancy using machine learning models, providing insights to guide targeted educational strategies and interventions. METHODS: A cross-sectional study was conducted from April to August 2024, collecting data from individuals aged 50 and older at four community health service centers in Shanghai. Data collection included demographic information, HZ disease and vaccine knowledge, the 5C scale, and the vaccine health literacy. We employed LASSO regression for variable selection, followed by analysis of key variables using four machine learning algorithms: logistic regression, random forest, support vector machine, and eXtreme Gradient Boosting (XGBoost). Model performance was evaluated using AUC and calibration plots. Shapley Additive Explanations (SHAP) were used to interpret and identify key predictors. RESULTS: A total of 1152 participants (mean age 66.1 ± 8.8 years; 49.7 % male) were included, with 73.95 % reporting hesitancy toward the HZ vaccine. Ten out of 21 features were selected for modeling. XGBoost model showed the best performance with an AUC of 0.960 (95 % CI: 0.937-0.983). SHAP analysis identified confidence, vaccine literacy, complacency, disease knowledge, and calculation as the primary predictors. CONCLUSIONS: The SHAP-XGBoost model showed strong predictive performance for herpes zoster vaccine hesitancy, with vaccine literacy and the 5C psychological antecedents identified as key predictors. This tool can inform targeted health interventions, and future work may integrate community databases for large-scale identification and tailored group strategies.

来源:Vaccine,查看原文。版权归原作者所有。

引用

  1. Wang X, Shang S, Zou T, et al. Drivers of herpes zoster vaccine hesitancy in adults aged 50 and above: A machine learning approach. Vaccine. 2025. PMID: 41082814. doi:10.1016/j.vaccine.2025.127838

供执业医师工作参考,不替代临床判断;数值与分级以指南原文为准。