Question Answering for Electronic Health Records: Scoping Review of Datasets and Models
Author Information
Author(s): Mavragani Amaryllis, Hu Danqing, Kuo Kuang-Ming, Bae Seongsu, Bardhan Jayetri BE, MS, Roberts Kirk BS, MS, PhD, Wang Daisy Zhe BS, MS, PhD
Primary Institution: University of Florida
Hypothesis
This study aims to provide a methodological review of existing works on QA for EHRs.
Conclusion
EHR QA research faces multiple challenges, such as the limited availability of clinical annotations and the need for realistic EHR QA datasets.
Supporting Evidence
- A total of 4111 papers were identified for the study, and 47 papers were selected for review.
- 53% of the selected papers focused on EHR QA datasets, while 79% focused on EHR QA models.
- The emrQA dataset was identified as the most popular EHR QA dataset in terms of citations and usage.
Takeaway
This study looks at how to ask questions about patient records stored in electronic health records, and it finds that there are many challenges to making this work well.
Methodology
The study involved a systematic review of articles from January 1, 2005, to September 30, 2023, using PRISMA guidelines to screen and classify relevant publications.
Limitations
The search process was limited to a handful of EHR and QA-related keywords, which may have resulted in missing relevant studies.
Digital Object Identifier (DOI)
Want to read the original?
Access the complete publication on the publisher's website