文献速递免疫规划卫生政策与健康促进

美国50岁及以上成人带状疱疹疫苗未接种机器学习模型的时间验证

Temporal validation of machine-learning models for shingles vaccine nonvaccination among U.S. adults aged 50 years and older.
PubMed 文献 · Vaccine · 2026-09-03

基于2019-2024年美国国家健康访谈调查(NHIS)中96,663名50岁及以上成人数据,用2019-2023年数据开发调查加权逻辑回归与加权XGBoost模型,在2024年留出样本中验证预测未来一年未接种状态的表现。未接种比例从2019年73.7%降至2024年56.1%,XGBoost区分度仅略优于逻辑回归(AUC 0.771 vs 0.764),两模型均系统性高估未接种概率,提示应用前需时间验证与模型更新。

要点

  • 带状疱疹疫苗未接种比例从2019年73.7%降至2024年56.1%,但仍常见。
  • 2024年留出样本中XGBoost的AUC为0.771,逻辑回归为0.764,差异很小。
  • 两模型均系统性高估2024年未接种概率,需时间验证与模型更新。

核心结论

  • 美国50岁及以上成人带状疱疹疫苗未接种比例大幅下降但仍常见。
  • XGBoost较调查加权逻辑回归仅小幅提升区分度。
  • 两模型存在时间校准漂移,应用前需时间验证、模型更新及外部验证。

研究设计与样本

  • 分析2019-2024年NHIS样本成人数据,纳入50岁及以上且疫苗、预测变量和调查设计数据完整者(N=96,663)。
  • 用2019-2023年数据开发调查加权逻辑回归与加权XGBoost模型,使用相同的15个协调预测变量。
  • 在2024年留出调查年评估模型,采用AUC、Brier评分、校准、训练衍生分类阈值及模型重要性指标。
  • 进行年龄分层分析和敏感性分析。

主要发现

  • 调查加权未接种比例从2019年73.7%降至2024年56.1%。
  • 2024年留出样本中逻辑回归AUC为0.764(95%CI 0.757-0.771),XGBoost AUC为0.771(95%CI 0.764-0.778),配对AUC差为0.0065(p=1.10×10-9)。
  • Brier评分分别为0.210和0.207。
  • 两模型均系统性高估2024年未接种概率,校准斜率接近1。
  • 两模型中重要性排名前四的预测变量均为流感疫苗接种、年龄、教育程度和种族/族裔。

临床怎么用

  • 机器学习模型可用于识别带状疱疹疫苗未接种高风险人群,但当前校准漂移限制直接公共卫生应用。
  • 应用前应使用更近期数据更新模型并进行外部验证。
  • 流感疫苗接种、年龄、教育程度和种族/族裔是重点考虑的风险因素。

局限性

  • 模型基于NHIS数据,预测变量限于调查所收集的15个协调变量。
  • 两模型在2024年均系统性高估未接种概率,存在时间校准漂移。
  • 未报告外部验证结果,应用前需进一步验证。

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 vaccination is recommended for U.S. adults aged 50 years and older, yet many remain unvaccinated. Prior National Health Interview Survey (NHIS) studies have described vaccination coverage and associated factors, but temporal validation of models predicting future-year unvaccinated status using interpretable machine-learning methods is limited. METHODS: The 2019-2024 NHIS Sample Adult data were analyzed for adults aged 50 years and older with complete vaccination, predictor, and survey-design data (N = 96,663). Models were developed using 2019-2023 data and evaluated in the held-out 2024 survey year. Survey-weighted logistic regression and weighted XGBoost used the same 15 harmonized predictors. Performance was assessed using area under the receiver operating characteristic curve (AUC), Brier score, calibration, training-derived classification thresholds, and model-importance measures. Age-stratified and sensitivity analyses were also conducted. RESULTS: The survey-weighted proportion unvaccinated declined from 73.7% in 2019 to 56.1% in 2024. In the held-out 2024 sample, logistic regression achieved an AUC of 0.764 (95% CI, 0.757-0.771) and XGBoost an AUC of 0.771 (95% CI, 0.764-0.778); the paired AUC difference was 0.0065 (p = 1.10 × 10-9). Brier scores were 0.210 and 0.207, respectively. Both models systematically overpredicted the probability of being unvaccinated in 2024, while calibration slopes remained near 1. Influenza vaccination, age, education, and race/ethnicity were the four highest-ranked predictors in both models. Survey-weighted nonvaccination was 71.4% among adults aged 50-59 years, 59.6% among those aged 60-64 years, and 44.6% among those aged 65 years and older. Restricting model development to 2022-2023 produced similar discrimination but reduced systematic overprediction. CONCLUSIONS: Shingles nonvaccination declined substantially but remained common among U.S. adults aged 50 years and older. XGBoost provided only a small improvement in discrimination over survey-weighted logistic regression. Temporal calibration drift in both models and its reduction with more recent training data emphasize the importance of temporal validation, model updating, and external validation before applied public-health use.

来源:Vaccine,查看原文。版权归原作者所有。

引用

  1. Costa NS. Temporal validation of machine-learning models for shingles vaccine nonvaccination among U.S. adults aged 50 years and older. Vaccine. 2026. PMID: 42691823. doi:10.1016/j.vaccine.2026.129099

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。