利用自然语言处理工具从电子健康记录中获取病例以加强大流行期间基于自发报告的信号检测
本研究开发电子健康记录不良事件(EHR-AE)搜索方法,利用自然语言处理(NLP)文本挖掘工具在两家荷兰医院电子健康记录中检索COVID-19疫苗相关潜在不良事件,并将识别病例报告至Lareb。2023年1月1日至12月1日共进行13次检索,6个关联至少识别1例,共新增41例报告,促成2个安全信号检测,若大流行期间实施可分别提前约18个月和2个月检出。
要点
- EHR-AE搜索方法利用NLP文本挖掘在电子健康记录中识别COVID-19疫苗相关潜在不良事件病例。
- 13次检索中6个关联识别至少1例,共41例额外病例报告至Lareb。
- 这些病例促成2个COVID-19疫苗安全信号检测,若大流行期间实施可提前约18个月和2个月检出。
核心结论
- EHR-AE搜索方法可加强并加速COVID-19疫苗接种相关潜在不良事件的信号检测过程。
- 未来研究应扩展至更多医院,以进一步评估该方法的附加价值,包括该方法对哪些药物-不良事件关联最有益。
研究设计与样本
- 开发EHR-AE搜索方法,使用具有文本挖掘功能的NLP工具在电子健康记录中进行针对性搜索以识别额外潜在病例。
- 在两家荷兰医院的电子健康记录中,针对COVID-19疫苗相关的若干已确立和未确立(潜在)不良事件进行检索。
- 识别病例报告至Lareb,并与自发报告一起用于安全信号检测分析。
主要发现
- 2023年1月1日至2023年12月1日期间共进行13次检索,涉及COVID-19疫苗相关的不同(潜在)不良事件。
- 6个关联至少识别1例病例,共向Lareb报告41例额外病例。
- 这些病例促成2个COVID-19疫苗安全信号的检测。
- 若EHR-AE方法在COVID-19大流行期间实施,2个安全信号可分别提前约18个月和2个月检出。
临床怎么用
- 原文未据此给出推荐。
局限性
- 自发报告系统的一个局限性是报告不良事件为自愿行为。
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
INTRODUCTION: During the COVID-19 pandemic, vaccines were rapidly developed, but some potential adverse drug reactions (ADRs) were still undetected at the time of market authorization. The analysis of ADR reports received through the spontaneous reporting system (SRS) remains the cornerstone for safety signal detection. A limitation is that reporting ADRs is voluntary. In this study the added value of electronic health records (EHRs) was explored as an additional source to spontaneous reports, to strengthen safety signal detection concerning COVID-19 vaccines. METHODS: The electronic health record adverse event (EHR-AE) search method was developed to enable targeted searches in EHRs using a natural language processing (NLP) tool with text-mining functionalities to identify additional potential cases. Searches were performed in EHRs of two Dutch hospitals concerning several established and non-established (potential) ADRs associated with COVID-19 vaccines. Identified cases were reported to Lareb and analyzed in addition to spontaneous reports for safety signal detection. RESULTS: Thirteen searches were conducted between January 1, 2023, and December 1, 2023, concerning different (potential) ADRs associated with COVID-19 vaccines. For 6 associations at least 1 case was identified, resulting in a total of 41 additional cases reported to Lareb. These cases contributed to the detection of two safety signals concerning COVID-19 vaccines. Two safety signals could have been detected approximately 18 and 2 months earlier, if the EHR-AE method had been implemented during the COVID-19 pandemic. CONCLUSION: The EHR-AE search method can strengthen and accelerate the signal detection process for potential ADRs associated with COVID-19 vaccination. Future studies should expand to more hospitals, aiming to further evaluate the added value of the method, including for which drug-ADR associations the EHR-AE search method is most beneficial.
引用
- van der Weg W, von Kreijfelt G, Davidson L, et al. Strengthening spontaneous reporting-based signal detection during a pandemic with cases from electronic health records using a natural language processing tool. Vaccine. 2025. PMID: 40714526. doi:10.1016/j.vaccine.2025.127549
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。