The article highlights demographic differences in patient cohorts at MSHS and WCM, noting that MSHS had a younger and more diverse population. In terms of system performance, the Rule-Based System (RBS) significantly outperformed the LLM-based system in classifying social support (SS) and social isolation (SI) categories, achieving higher macro-averaged f-scores. Notably, there were no documented ICD codes associated with social isolation in the clinical notes, emphasizing a gap in structured data capturing.
The patient composition at MSHS was noted to be younger and more diverse compared to patients at WCM, highlighting demographic differences within the cohort.
At MSHS, the Rule-Based System (RBS) outperformed the LLM-based system, achieving macro-averaged f-scores of 0.90 for fine-grained categories and 0.89 for coarse-grained.
The results demonstrated that the RBS classification had consistently higher macro-averaged f-scores compared to the LLM, particularly in fine-grained category classification.
Notably, the analysis revealed there were no visits associated with the clinical notes where social isolation was captured by the structured ICD codes.
Collection
[
|
...
]