EUVIMEDEUVIMED entdecken
Crossref

Nearest-Neighbor Case Retrieval for Emergency Department Urgency Prediction with a Large Language Model

Ismail Bouajaja, Omar Bouattane, Oussama Barakat, Abdelaziz Daaif

International Journal of Advanced Computer Science and Applications · 2026 · 17 (10)

DOI: 10.14569/ijacsa.2026.0171064 ↗

Kurz gefasst

Fachartikel in der Zeitschrift International Journal of Advanced Computer Science and Applications, 2026. Verfasst von Ismail Bouajaja und 3 weiteren Autor:innen.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

This study evaluates whether prior emergency department visits improve a large language model’s (LLM) prediction of recorded urgency. Four methods were compared: an always-urgent baseline, conventional k-nearest-neighbor (kNN) voting, an LLM without retrieved cases, and the same LLM with 30 kNN-retrieved labeled cases. The 2022 National Hospital Ambulatory Medical Care Survey yielded 10,207 eligible visits: 7,167 for training, 1,522 for validation, and 1,518 for testing. The final case library combined the 8,689 training and validation visits. On the commonly held-out test set, the kNN-search LLM achieved 74.70% accuracy and 66.86% balanced accuracy, compared with 68.97% and 65.68% for the no-search LLM and 62.52% and 64.30% for conventional kNN. Relative to no-search prompting, retrieved cases increased urgent sensitivity from 73.87% to 86.37% but reduced nonurgent specificity from 57.49% to 47.36%. A paired 95% interval excluded zero for the accuracy difference but included zero for balanced accuracy. API-reported model-call charges were USD 0.0786 without cases and USD 0.8577 with cases. The results demonstrate a change in the balance of errors, not established clinical utility or uniformly better discrimination.

Zitieren

APA

Bouajaja, I., Bouattane, O., Barakat, O., & Daaif, A. (2026). Nearest-Neighbor Case Retrieval for Emergency Department Urgency Prediction with a Large Language Model. International Journal of Advanced Computer Science and Applications, 17(10). https://doi.org/10.14569/ijacsa.2026.0171064

Vancouver

Bouajaja I, Bouattane O, Barakat O, Daaif A. Nearest-Neighbor Case Retrieval for Emergency Department Urgency Prediction with a Large Language Model. International Journal of Advanced Computer Science and Applications. 2026;17(10). doi:10.14569/ijacsa.2026.0171064

BibTeX
@article{Bouajaja2026,
  author = {Bouajaja, Ismail and Bouattane, Omar and Barakat, Oussama and Daaif, Abdelaziz},
  title = {Nearest-Neighbor Case Retrieval for Emergency Department Urgency Prediction with a Large Language Model},
  journal = {International Journal of Advanced Computer Science and Applications},
  year = {2026},
  volume = {17},
  number = {10},
  doi = {10.14569/ijacsa.2026.0171064}
}
Kostenlos starten →