Author: Minehart RD et al.
Anesthesia & Analgesia 141(3):536–539, September 2025. doi:10.1213/ANE.0000000000007473
This commentary argues that anesthesiologists face escalating cognitive burden from workforce shortages, sicker patients, and a flood of new literature—conditions that increase the risk of cognitive overload, stress, and error-prone decisions. Traditional supports (journals, databases, protocols) help but don’t provide timely, verified “second checks,” so clinicians often rely on informal curbside consults that can be delayed, incomplete, or legally fraught.
The authors propose a “human-augmented and verified AI” model for decision support: clinicians submit questions; AI drafts an evidence-based response; vetted peer consultants review, edit, and verify it before returning guidance—combining speed and breadth of AI with human trust and accountability. They note LLM limitations (hallucinations, bias, opacity) and FDA concerns about automation bias, highlighting mitigations like retrieval-augmented generation, fine-tuning, and ensemble checks. The envisioned system preserves anonymity for question askers, scales expertise to smaller/rural sites, and may reduce consultant workload—while still requiring guardrails to avoid practicing outside a formal consult and to ensure timely, accurate answers.
What You Should Know
• Cognitive overload is rising in anesthesia; informal “curbsides” are common but inconsistent.
• Pure AI raises trust, bias, and hallucination risks—especially under time pressure.
• Proposed solution: AI drafts + peer expert verification (“human-augmented AI”) for faster, higher-confidence answers.
• FDA expects rigor for clinical decision support; safeguards (RAG, fine-tuning, ensembles) are key.
• Early use cases may be prep/follow-up; speed and reliability should improve as Q&A libraries grow.
References
Morriss WW, Enright AC. Anesth Analg. 2023;136:227–229.
Burden M et al. J Hosp Med. 2013;8:31–35.
U.S. FDA. Clinical decision support software guidance, 2024.
Kaplan AD et al. Hum Factors. 2023;65:337–359.
Estrada Alamo C et al. Anesth Analg. 2024;138:938–950.
Thank you to Anesthesia & Analgesia for making this work available.