An agentic LLM system outperformed a single LLM system in 4 out of 6 models for binary frailty classification and in 5 out of 6 models for generating patient Risk Assessment Index scores.
Observational (n=204)
No
Does an agentic LLM system improve the accuracy of automated preoperative frailty assessment compared to a single LLM system in cancer patients undergoing major abdominal surgery?
An agentic LLM workflow enhances the automated assessment of preoperative frailty from EHR data compared to a single LLM approach, potentially improving clinical efficiency for surgical oncology patients.
Assessment of preoperative frailty is essential for risk stratification and optimization of patients undergoing major abdominal surgery for hepatobiliary (HPB) and gastrointestinal (GI) malignancies. Large language model (LLM)-based agents may facilitate automated frailty scoring, thereby supporting targeted prehabilitation efforts to improve surgical outcomes. We compared the performance of an agentic LLM system with a single LLM system in calculating frailty scores and classifying frailty status. Using real-world preoperative notes documented within 60 days prior to surgery, we evaluated both approaches. The agentic LLM system outperformed the single LLM system in 4 out of 6 models for a binary frailty classification task and in 5 out of 6 models for generating patient Risk Assessment Index (RAI) scores. The greatest improvements were observed among lower-parameter models (Llama 3.1 8b, Qwen 2.5 7b). These findings suggest than an agentic workflow may enhance frailty assessment performance. Automating frailty score calculation from Electronic Health Record (EHR) data may enhance clinical efficiency and enable targeted prehabilitation strategies for patients with cancer undergoing surgery.
Lee et al. (Wed,) conducted a observational in Abdominal malignancy (n=204). Agentic LLM system vs. Single LLM system was evaluated on Accuracy of binary frailty classification and generation of Risk Assessment Index (RAI) scores across 6 LLM models. An agentic LLM system outperformed a single LLM system in 4 out of 6 models for binary frailty classification and in 5 out of 6 models for generating patient Risk Assessment Index scores.