Large Language Model Symptom Identification From Clinical Text: Multicenter Study

View Full Paper

Ask AI

Bookmark

Authors

Discussion

Multicenter analysis shows LLMs improve symptom identification accuracy, indicating better generalizability than ICD-10 methods.

Accuracy of symptom identification improved significantly with large language models over traditional methods.
GPT-4 exhibited the highest performance (F1-score=94.0%) in identifying infectious respiratory symptoms compared to ICD-10.
LLMs demonstrated strong generalizability across multisite evaluations without the need for extensive customization.
The findings suggest that large language models could replace traditional chart review methods in clinical settings.

McMurry et al. (2025) studied this question.

View Full Paper

Ask AI

Bookmark

View Full Paper

Ask AI

Bookmark

Authors

Multicenter analysis shows LLMs improve symptom identification accuracy, indicating better generalizability than ICD-10 methods.

Accuracy of symptom identification improved significantly with large language models over traditional methods.
GPT-4 exhibited the highest performance (F1-score=94.0%) in identifying infectious respiratory symptoms compared to ICD-10.
LLMs demonstrated strong generalizability across multisite evaluations without the need for extensive customization.
The findings suggest that large language models could replace traditional chart review methods in clinical settings.

McMurry et al. (2025) studied this question.

View Full Paper

Ask AI

Bookmark