Artificial Intelligence-Based Clinical Decision Support Systems in Nursing Practice: Applications, Benefits, and Challenges — A Systematic Review

Abstract

Artificial intelligence (AI)-based clinical decision support systems (CDSS) are increasingly embedded in nursing workflows, promising earlier detection of patient deterioration, more efficient documentation, and more consistent risk stratification, yet the evidence base is fragmented across clinical domains and ethical, workforce, and implementation concerns remain insufficiently synthesised from a nursing-specific perspective. This review therefore aimed to systematically synthesise current evidence on the applications, benefits, and challenges of AI-based CDSS in nursing practice, and to derive implications for nursing education, regulation, and future research. A systematic literature search conceptually following PRISMA 2020 reporting guidance was conducted across five databases (PubMed/MEDLINE, CINAHL, Scopus, Web of Science, and IEEE Xplore), supplemented by backward citation tracking of nine recent systematic or scoping reviews, covering publications largely from 2018 to 2026; findings were synthesised narratively across four application domains. Thirty-eight unique sources met the inclusion criteria. AI-CDSS applications clustered into four domains: deterioration and sepsis early-warning systems; pressure injury and fall-risk prediction; natural language processing and generative AI for nursing documentation and care planning; and workflow, triage, and staffing optimisation. Reported benefits included earlier deterioration detection, reduced false-alarm rates, reduced documentation burden, and more standardised risk assessment, while reported challenges included algorithmic bias, alert fatigue, automation bias, deskilling risk, accountability ambiguity, integration barriers, and persistent AI-literacy gaps. AI-based CDSS therefore hold meaningful, evidence-supported potential to augment nursing practice, but current evidence is dominated by single-site, retrospective validation studies; realising this potential will require nurse involvement in design and governance, explainable and auditable algorithms, bias-mitigation standards, formal accountability frameworks, and integration of AI literacy into nursing curricula.