Artificial Intelligence in Acute Care Environments: A Multidisciplinary Review of Its Impact on Clinical Decision-Making, Workflow Integration, and the Challenges to Widespread Clinical Adoption
TL;DR
A narrative review synthesizes evidence from 2015–2025 on AI applications across the acute care continuum, examining multidisciplinary implications involving emergency medicine, radiology, laboratory medicine, general practice, nursing, pharmacy, and physical therapy.
Abstract
Background: Acute care environments—emergency departments (EDs) and intensive care units (ICUs)—demand rapid, precise clinical decisions under extreme time pressure and uncertainty. Artificial intelligence (AI) offers transformative potential across multiple healthcare disciplines by providing real-time data synthesis, advanced pattern recognition, and personalized decision support. Aim: This narrative review synthesizes evidence from 2015–2025 on AI applications across the acute care continuum, examining multidisciplinary implications involving emergency medicine, radiology, laboratory medicine, general practice, nursing, pharmacy, and physical therapy. Methods: A systematic literature search identified 40 peer-reviewed articles, systematic reviews, and consensus statements for narrative synthesis. Results: AI enhances triage accuracy through machine learning algorithms, improves diagnostic precision in imaging via deep learning, enables early detection of sepsis and deterioration through predictive analytics, optimizes referral decisions in primary care, supports nursing surveillance and patient monitoring, enhances medication safety through clinical decision support systems, and enables personalized rehabilitation planning. Predictive models demonstrate AUC-ROC values of 0.75–0.95 across outcomes, with sepsis mortality reductions of up to 20% and medication accuracy improvements of 30%. However, fewer than 2% of AI models achieve clinical deployment, with barriers including interoperability gaps, poor model generalizability, algorithmic bias, and limited clinician trust. Conclusion: AI holds substantial promise for acute patient management across all healthcare disciplines, but realizing its potential requires standardized validation, explainable models, workflow integration, and comprehensive interprofessional clinician education.