Hallucination in artificial intelligence and its implications on anesthesiology practice and patient outcomes

Abstract

Artificial intelligence (AI) is revolutionizing anesthesiology by enhancing patient monitoring, optimizing drug dosing, and predicting adverse intraoperative events. AI-driven models, particularly those utilizing deep learning, are increasingly used for anesthesia depth monitoring, hemodynamic control, and perioperative risk stratification. However, a significant challenge in AI-driven healthcare is AI hallucination (AIH)—a phenomenon where AI generates misleading, incorrect, or fabricated information. In anesthesiology, hallucinations can lead to severe consequences, such as incorrect dosing recommendations, misinterpretation of patient monitoring data, and flawed clinical decision support, all of which pose risks to patient safety. This article explores the concept of AIH, its causes, real-world examples of its impact in healthcare, and its potential consequences for anesthesiology practice. We also discuss mitigation strategies, including improving data quality, implementing clinician-in-the-loop models, and ensuring regulatory oversight. As AI becomes increasingly integrated into anesthetic practice, recognizing and addressing the risks of AIH is crucial for improving patient safety and maintaining the integrity of anesthetic care.

Abbreviations: AI: Artificial intelligence, BIS: Bispectral index, DDA: Data-driven analytics, ABM: Algorithm-based management, XAI: explainable Artificial intelligence

Keywords: Artificial intelligence; BIS; Data-driven analytics; Algorithm-based management; Machine learning; AI in anesthesiology

Citation: Nawaz S, Ahmad K, Khamash O, Khan E, Mendonca R. Hallucination in artificial intelligence and its implications on anesthesiology practice and patient outcomes. Anaesth. pain intensive care 2025;29(5):376-381. DOI: 10.35975/apic.v29i5.2873

Received: May 09, 2024; Revised: October 26, 2024; Accepted: January 01, 2025