The article discusses the use of natural language processing (NLP) to extract data on social support (SS) and social isolation (SI) from electronic health records (EHRs), specifically within psychiatric encounter notes. By developing a rule-based system (RBS) and comparing it to a large language model (LLM), the study discovered that the RBS significantly outperformed the LLM in identifying both SS/SI and their subcategories. The findings suggest that tailored rule-based methods can provide better accuracy in clinical data extraction compared to generalized language models, highlighting the effectiveness of specific annotations and guidelines in this context.
Natural language processing algorithms can automate the data extraction process of social support and social isolation from clinical notes, offering efficiency in handling EHRs.
The rule-based system outperformed the large language model in macro-averaged f-scores for extracting social support and social isolation categories from psychiatric encounter notes.
#natural-language-processing #social-support #social-isolation #psychiatric-outcomes #data-extraction
Collection
[
|
...
]