Key result
GPT-4o preanesthetic assessment shows high agreement with clinician-assigned ASA classifications using NHI cloud data.
Why the study?
Conventional preanesthetic assessment tools such as ASA classification and PONV risk scores can be limited by subjective judgment, incomplete documentation, and fragmented clinical data.
Does a GPT-4o-powered AI system accurately predict clinician-assigned ASA physical status classifications and PONV risk in adult surgical patients?
Observational (n=600)
No
Does a GPT-4o-powered AI system accurately predict clinician-assigned ASA physical status classifications and PONV risk in adult surgical patients?
Effect estimate: Cramér V 0.531
p-value: p=<.001
A GPT-4o-powered AI system demonstrated high agreement with clinician-assigned ASA classifications, supporting its potential as a decision-support tool for preanesthetic assessment.
Captured external expert commentary on this paper, strongest first. Original sources are linked where available.
“At the 2025 IARS Annual Meeting, researchers unveiled a pilot study showing that GPT-4o matched board-certified anesthesiologists in assigning ASA Physical Status scores—a key tool for perioperative risk assessment. With a Cohen's Kappa of 0.89, the AI demonstrated good-to-excellent agreement with human experts, suggesting that large language models could one day serve as valuable adjuncts in preoperative evaluation, training, and even billing support.”
LLM-assisted preanesthetic assessment may improve risk stratification; leaves open whether it outperforms conventional tools in prospective validation.
Background: Accurate preanesthetic assessment is essential for perioperative risk stratification, but conventional tools such as the American Society of Anesthesiologists (ASA) physical status classification and postoperative nausea and vomiting (PONV) risk scores may be affected by subjective judgment, incomplete documentation, and fragmented clinical data. Large language models may support preanesthetic assessment by integrating structured and unstructured clinical information. Objective: This study aimed to develop and retrospectively validate a GPT-4o (OpenAI)-powered AI system for preanesthetic assessment. The primary validation focus was the agreement between AI-generated and clinician-assigned ASA physical status classifications. PONV risk stratification was evaluated as an additional clinically relevant performance outcome. A secondary objective was to assess the incremental contribution of National Health Insurance (NHI) cloud data to model performance. Methods: This single-center retrospective validation study included 600 adult surgical patients randomly sampled from 4404 eligible inpatient surgical patients between January and May 2025. The system processed structured and unstructured electronic health record data, with and without NHI cloud data, to generate preanesthetic assessment outputs. ASA agreement was assessed using Cohen κ with 95% CIs. PONV prediction was evaluated using sensitivity, specificity, overall accuracy, positive predictive value, negative predictive value, confusion matrix analysis, the 2-sided Fisher-Freeman-Halton exact test, and Cramér V. Results: <.001, Cramér V=0.531). Conclusions: The GPT-4o-powered preanesthetic assessment system demonstrated feasibility and high agreement with clinician-assigned ASA classifications when NHI cloud data were incorporated. For PONV risk stratification, the system showed high specificity and a high negative predictive value but modest sensitivity, indicating that it identified a small high-risk subgroup with a high observed PONV incidence but did not capture all patients who developed PONV. These findings support the use of large language model-based decision-support tools as adjuncts to, rather than replacements for, anesthesiologist judgment.
No takes yet. Share an insight, caveat, or question.
Lin et al. (2026) conducted an observational in Preanesthetic assessment (n=600). GPT-4o-powered AI system vs. Clinician-assigned assessment was evaluated on Agreement between AI-generated and clinician-assigned ASA physical status classifications (Cramér V 0.531, p=<.001). The GPT-4o-powered preanesthetic assessment system demonstrated high agreement with clinician-assigned ASA classifications when NHI cloud data were incorporated (P<.001, Cramér V=0.531).
Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context: