ARTIFICIAL INTELLIGENCE (AI) FOR HOSPITAL QUALITY IMPROVEMENT AND PATIENT SAFETY : A BIBLIOMETRIC-SYSTEMATIC LITERATURE REVIEW

Abstract

Purpose/Background: Artificial intelligence (AI) is increasingly applied in hospital settings to support risk prediction, clinical decision support, medication safety, adverse event detection, and operational improvement. However, evidence regarding whether AI provides added value over manual processes, rule-based alerts, non-AI systems, or standard clinical judgement remains fragmented and insufficiently consolidated.

Design/Methodology/Approach: This bibliometric-systematic literature review analyzed 65 Scopus-indexed articles published between 2021 and 2026. VOSviewer was used to map keyword co-occurrence, while Biblioshiny supported descriptive and thematic analyses. The synthesis examined AI technologies, hospital service domains, quality and safety outcomes, implementation patterns, study limitations, and evidence gaps.

Findings: Publication output increased after 2023 and peaked in 2024, with health informatics journals and Chinese institutions contributing prominently. Seven major clusters were identified: predictive analytics, algorithmic risk modelling, clinical decision support and early warning, EHR/EMR- and NLP-based adverse event detection, medication safety, deep learning in critical and specialty care, and operational efficiency. The strongest evidence involved prospective risk stratification for sepsis, falls, postoperative complications, mortality, readmission, length of stay, adverse drug events, and medication errors. Nevertheless, the evidence remains stronger for model accuracy and discrimination than for direct reductions in preventable harm.

Research limitations/implications: This review was limited to English-language Scopus-indexed articles and may not capture evidence from other databases or non-journal sources. Future studies should prioritize multicentre validation, pragmatic implementation trials, workflow evaluation, fairness, explainability, privacy, and post-deployment monitoring.

Practical implications: Hospitals should evaluate AI based on workflow integration, clinical actionability, governance, and measurable quality and safety outcomes rather than accuracy metrics alone.

Originality/value: This review integrates bibliometric mapping and systematic synthesis to position AI as an infrastructure for prospective risk governance in hospitals, rather than a replacement for clinical judgement.