Large Language Model–Assisted Clinical Decision Support for Regional Anesthesia Selection in Knee Arthroscopy Patients

Abstract

Regional anesthesia has become a fundamental component of perioperative pain management for knee arthroscopy procedures due to its ability to reduce opioid consumption, improve postoperative recovery, and enhance patient satisfaction. Nevertheless, selecting an appropriate regional anesthesia strategy remains a complex clinical decision-making process that requires consideration of patient-specific characteristics, surgical factors, institutional resources, practitioner expertise, and evolving evidence-based guidelines. Recent advances in large language models (LLMs) have created opportunities to augment clinical decision support systems through sophisticated reasoning capabilities, natural language interaction, and integration of heterogeneous medical information sources. This study explores a system-level framework for LLM-assisted clinical decision support in regional anesthesia selection for knee arthroscopy patients. Rather than focusing solely on predictive performance, the paper examines the architectural, organizational, and governance dimensions associated with integrating LLM technologies into perioperative care environments. The proposed framework combines structured electronic health record data, clinical practice guidelines, historical anesthesia outcomes, and real-time clinical documentation within a multi-layer decision support architecture. Particular attention is given to explainability, reliability, fairness, workflow integration, and human oversight mechanisms necessary for safe deployment. The study analyzes how LLM-assisted systems can support individualized recommendations among femoral nerve blocks, sciatic nerve blocks, adductor canal blocks, combined approaches, and alternative analgesic pathways while maintaining clinician accountability. Furthermore, the paper evaluates implementation challenges including data quality, institutional interoperability, regulatory compliance, model drift, and infrastructure sustainability. Findings suggest that LLM-assisted decision support can substantially enhance consistency and knowledge accessibility in perioperative decision-making when embedded within robust governance structures. The research contributes a comprehensive socio-technical perspective on the future role of generative artificial intelligence in regional anesthesia management and clinical decision support ecosystems.