The Role of Artificial Intelligence in Improving Healthcare Quality, Patient Safety, and Service Delivery: A Systematic Review of Real-World Applications
Abstract
Background: Artificial intelligence (AI) is increasingly being integrated into healthcare systems to support clinical decision-making, improve patient safety, enhance quality of care, and optimize healthcare service delivery. Despite rapid technological development, evidence regarding the real-world impact of AI on healthcare outcomes remains fragmented across different clinical specialties and organizational settings. Objective: This systematic review aimed to evaluate the impact of real-world AI applications on healthcare quality, patient safety, and healthcare service delivery, while identifying implementation challenges and areas requiring further research. Methods: A systematic search was conducted across PubMed/MEDLINE, Embase, Scopus, Web of Science, CINAHL, IEEE Xplore, and the Cochrane Library for studies published between January 2015 and September 2026. The review followed the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA 2020) guidelines. The database search identified 4,286 records, with an additional 37 records identified through reference-list searching and other sources. After removing 1,247 duplicates, 3,076 records underwent title and abstract screening. Of these, 2,824 records were excluded, and 252 full-text articles were assessed for eligibility. Following full-text assessment, 187 articles were excluded, resulting in 65 studies being included in the final systematic review. Eligible studies evaluated AI applications implemented within real healthcare settings and reported outcomes related to healthcare quality, patient safety, or service delivery. Study quality was assessed using appropriate Joanna Briggs Institute critical appraisal tools, and findings were synthesized narratively because of substantial methodological and clinical heterogeneity. Results: The 65 included studies represented hospital, outpatient, emergency, diagnostic, and primary-care settings across multiple countries. AI applications included machine learning, deep learning, natural language processing, computer vision, predictive analytics, and intelligent clinical decision-support systems. Overall, 28 studies evaluated healthcare quality outcomes, 21 studies primarily examined patient safety, and 16 studies focused mainly on healthcare service delivery and operational performance, although several studies assessed outcomes across more than one domain. AI-supported interventions were frequently associated with improvements in diagnostic and clinical decision support, early identification of patient deterioration, medication-related safety, workflow efficiency, patient prioritization, and resource utilization. Several studies also reported reductions in waiting time, length of stay, documentation burden, or avoidable adverse events. However, findings were not consistently positive across all settings. Important limitations included heterogeneous outcome measures, retrospective study designs, limited external validation, algorithmic bias, workflow-integration difficulties, lack of transparency, data-quality concerns, and limited evidence regarding long-term clinical and organizational outcomes. Conclusions: Real-world AI applications show considerable potential to improve healthcare quality, patient safety, and service delivery, particularly when integrated into clinical workflows as decision-support rather than standalone technologies. However, evidence of benefit remains uneven across healthcare settings, and strong technical performance does not necessarily translate into improved patient or organizational outcomes. Future research should prioritize prospective multicenter evaluations, standardized outcome measures, economic assessment, equity analysis, long-term monitoring, and robust governance frameworks to ensure that AI implementation produces safe, effective, and sustainable improvements in healthcare