The intersection of quality improvement, artificial intelligence and patient safety in healthcare—current applications, challenges and risks, and future directions: a scoping review

Introduction

Artificial intelligence (AI) has rapidly emerged as a cornerstone of technological innovation in healthcare, driven by advancements in information technology and mobile computing. Its capacity to analyze large datasets, detect patterns, and generate actionable insights has opened new possibilities for optimizing care processes and enhancing patient outcomes (1,2). AI applications span diverse healthcare domains—from disease surveillance and outbreak modelling (e.g., COVID-19) to clinical decision support (CDS), telemedicine platforms, and resource allocation. This breadth of applications underscores AI’s transformative potential in quality improvement (QI) initiatives and, critically, patient safety.

QI initiatives seek to refine various dimensions of healthcare delivery, focusing on the systematic enhancement of processes, outcomes, and patient experiences. Within this broad scope, patient safety—defined as the prevention of harm through the reduction of medical errors and systematic improvements in care—is a foundational priority (3). AI’s powerful analytics can bolster these efforts by improving diagnostic accuracy, identifying at-risk patient populations, and supporting evidence-based decision-making (4). Additionally, recent literature illustrates AI’s impact on hospital workflow optimization and CDS systems, highlighting its wide-ranging utility (5). Despite its promise, integrating AI into QI frameworks remains fraught with systemic challenges related to scalability, interoperability, and ethical and regulatory considerations. Current research often highlights AI’s capabilities in isolation without fully accounting for how these tools can be cohesively embedded into healthcare systems to optimize patient safety. A central concern involves ensuring that AI-driven solutions are designed and implemented ethically, with protections in place to safeguard patient data and promote equitable care (6). These considerations underscore the need for a structured framework that can guide AI adoption while accounting for the complexity of healthcare environments.

Against this backdrop, this scoping review synthesizes existing evidence on AI’s role in QI and patient safety, mapping key themes and identifying research gaps. It is important to recognize that a scoping review primarily maps existing literature and identifies research gaps without providing definitive evidence or proving causality. By synthesizing findings across studies, this review highlights areas of current focus and suggests directions for future research rather than asserting conclusive outcomes. Anchored in the Arksey and O’Malley framework (7) and aligned with PRISMA-ScR guidelines (8), the review addresses three core questions:

• What are the current applications of AI in QI and patient safety?
• What challenges and limitations are associated with these applications?
• How can future research and policy frameworks facilitate AI integration in healthcare?

By concentrating on the intersection of AI, QI, and patient safety, this review aims to illuminate pathways for more effective AI implementation in healthcare. It not only examines where AI is making a measurable impact but also explores the barriers to wider adoption. The ultimate goal is to inform policies and practices that harness AI as an enabling tool, ensuring that patient safety remains at the forefront of healthcare innovation. We present this article in accordance with the PRISMA-ScR reporting checklist (available at https://jmai.amegroups.com/article/view/10.21037/jmai-24-328/rc).

Methods

Identification of relevant studies and search strategy

This review used Scopus, PubMed, and ScienceDirect to conduct a scoping of existing literature. The search approach focused on “quality improvement”, “artificial intelligence”, and “patient safety”. The search scope was expanded using Boolean operators “AND” and “OR”. The search strategy incorporated both Medical Subject Headings (MeSH) terms and free-text keywords, ensuring a comprehensive approach. For example, the PubMed search used terms such as “quality improvement”, “artificial intelligence”, and “patient safety”, combined with Boolean operators. While the search terms were limited to “quality improvement”, “artificial intelligence”, and “patient safety”, the authors acknowledge that alternate terminologies, such as “quality assurance”, “quality process”, “machine learning”, and “risk management”, could yield additional studies. However, the selected terms align with the primary focus of this review and were determined sufficient to capture relevant studies within the defined scope and timeline.

The review carefully examined current research on AI, its applications, and QI and patient safety. The search encompassed studies published from January 1, 2020 to June 30, 2024. The decision to limit the review to studies published from 2020 onward was guided by the rapid evolution of AI technologies in recent years, particularly in healthcare. This timeframe was chosen to capture the latest advancements and emerging trends that reflect the current state of AI applications in QI and patient safety. The focus on recent developments ensures relevance to contemporary healthcare practices, where AI integration is being actively explored and implemented. Foundational studies before 2020, while significant, were excluded to maintain a concise scope and emphasize actionable insights derived from the most recent innovations. Moreover, many earlier works lack the technological maturity and practical implementation insights that have become more prominent in the last few years. Thus, by concentrating on this timeframe, the review highlights the applicability and challenges of AI within the current healthcare landscape.

The detailed search strategy is outlined in Table 1, which specifies the databases and keywords used.

In line with the study objectives, peer-reviewed, English-language publications on AI interventions aimed at improving quality and patient safety were included. To capture the breadth of existing literature, the inclusion criteria encompassed original research articles of any design (e.g., experimental, observational, case studies, or machine learning-based analyses), as well as review articles (systematic, scoping, or narrative) and conceptual analyses. Publications were excluded if they were non-English, published before 2020, or appeared as conference papers, book chapters, or other non–peer-reviewed materials. Conference papers and book chapters were excluded to maintain methodological rigor and ensure sufficient methodological detail, whereas review articles and conceptual analyses were included for their critical role in synthesizing evidence and providing theoretical insights. Studies lacking a full-text version or providing insufficient methodological details (e.g., unclear research design, inadequate descriptions of data collection or analysis, or incomplete information necessary to evaluate validity) were also excluded.

Quality alone was not used as an exclusion criterion. Instead, limitations such as a lack of clear objectives, inadequate sample sizes, or incomplete data reporting were documented and critically discussed in the final synthesis. This approach ensured a comprehensive mapping of all relevant evidence while maintaining transparency in reporting potential biases and gaps.

Study selection and data charting

The inclusion process encompassed peer-reviewed, English-language publications (including original research articles, systematic reviews, narrative reviews, and conceptual analyses) published from January 1, 2020 to June 30, 2024, with a specific focus on the application of AI to QI and patient safety. Titles, abstracts, and full texts were independently reviewed by M.H.M.N., M.S.I., R.F.A.R., I.A.R., and S.H. to ensure rigorous selection and minimize bias. Classification and updates on included and excluded publications were recorded in an encrypted Microsoft Teams file accessible only to these reviewers. A Microsoft Excel spreadsheet was used to document detailed information for each included publication (e.g., authorship, publication year, research title, design, aims/background, AI applications, and key conclusions). Discrepancies regarding inclusion decisions were discussed among the reviewers to facilitate consensus. Throughout the screening process, M.H.M.N., M.S.I., and R.F.A.R. periodically reassessed the inclusion and exclusion criteria and suggested refinements.

Potential biases in publication selection were mitigated through a consensus approach, with a senior author acting as a tiebreaker when needed. In instances requiring further clarity, additional professional opinions were sought from A.I.A.R. Inter-rater reliability was measured using Cohen’s kappa, demonstrating a substantial level of agreement (k=0.78) and ensuring consistency in the review process (9,10). A second full-text evaluation was then conducted by I.A.R. and S.H. to confirm the integrity and consistency of the selected publications, thereby establishing a robust and unbiased evidence base.

Data extraction and synthesis

The data compiled in Microsoft Excel was used to generate the PRISMA 2020 flow diagram template for presenting the search outcomes (11).

Collating, summarizing and reporting the results

The collected data underwent thematic analysis to assess AI’s impact on quality and patient safety. Descriptive statistics were used to outline the techniques and distribution of selected articles. The five domains of AI applications namely CDS, workflow optimization, remote and decentralized care, risk prediction, and data-driven insights were derived through a systematic process combining deductive and inductive approaches. Initially, a comprehensive review of the literature informed the identification of potential themes based on established applications of AI in healthcare (12,13). Subsequently, findings from the included studies were iteratively coded and categorized into thematic clusters by multiple authors. This collaborative process ensured that the domains comprehensively captured the breadth of AI applications in QI and patient safety, with regular discussions to achieve consensus. The five domains align with current healthcare priorities and reflect both theoretical relevance and practical significance in addressing patient safety and QI challenges.

Grouping similar results thematically facilitated a coherent narrative, enhancing understanding of the evidence. A thematic analysis was conducted to systematically synthesize findings from the included studies. The process followed Braun and Clarke’s six-phase framework to ensure rigor and transparency (14). First, all included articles were read multiple times to familiarize the team with the data. Key codes were then generated inductively, capturing recurring patterns and significant insights related to AI applications in healthcare. These codes were reviewed and grouped into themes representing broader categories of AI impact. Through iterative discussions among the authors, these themes were refined and consolidated into the five overarching domains. Discrepancies in coding or categorization were resolved through consensus, and the final themes were validated against the objectives of the review to ensure alignment and comprehensiveness.

Results

As illustrated in PRISMA flowchart (Figure 1), the initial database search across Scopus (n=456), PubMed (n=213), and ScienceDirect (n=538) yielded 1,207 records. After the removal of duplicates, 987 records remained for screening. Following the screening of titles, abstracts, and keywords, 882 records were excluded. The high exclusion rate of 89.4% during the screening process was primarily due to the broad search strategy employed to ensure comprehensive coverage of the intersection between AI, QI, and patient safety in healthcare. This deliberate approach inevitably retrieved a significant number of irrelevant studies, which were subsequently excluded after rigorous title, abstract, and keyword screening. Key reasons for exclusion included a lack of relevance to the review focus (n=397), duplicate entries (n=353), and inappropriate study designs (n=132). After these exclusions, 105 full-text articles were examined for further eligibility. Of these, 47 did not meet the inclusion criteria—3 were non-English, 5 were conference abstracts, 20 were published before 2020, 16 had no available full text, and 3 were book chapters—resulting in 58 articles for eligibility assessment. Another 22 articles were excluded due to low methodological quality (n=5), insufficient data (n=11), or uncertainty in study design (n=6). Ultimately, 36 articles fulfilled all criteria and were included in the final synthesis. This rigorous selection process ensured that only high-quality, relevant studies were analyzed. The final inclusion of 36 studies reflects a deliberate and methodical process aimed at synthesizing high-quality, peer-reviewed evidence that directly aligns with the review’s objectives. While the exclusion rate appears high, this is consistent with scoping reviews addressing interdisciplinary topics where initial search terms cast a wide net to capture potential intersections. The rigorous application of inclusion and exclusion criteria ensured that only studies providing empirical insights into AI applications in QI and patient safety were included, thereby minimizing selection bias and enhancing the review’s relevance and reliability.

Figure 1 PRISMA flowchart.

Domains of AI applications

Based on the studies examined in this scoping review, AI offers several healthcare applications, notably in QI and patient safety. The five domains identified in this review provide a comprehensive framework for understanding AI’s role in QI and patient safety. Each domain was derived through a rigorous thematic analysis. These domains were validated against the included studies to ensure their relevance and comprehensiveness, capturing both theoretical and practical dimensions of AI applications in healthcare. For example, the domain of risk prediction was supported by studies demonstrating the use of machine learning models to forecast patient outcomes, while Workflow Optimization encompassed applications that streamlined healthcare processes.

AI CDS offers doctors real-time insights and evidence-based recommendations, enhancing diagnosis, treatment planning, and reducing medical errors. Common AI techniques such as machine learning, natural language processing, and predictive analytics are increasingly applied in healthcare settings. For example, machine learning enhances clinical decision-making by analyzing imaging data, while natural language processing enables efficient extraction of insights from electronic health records.

AI-powered workflow optimization automates processes, improves operations, optimizes scheduling, and allocates resources, allowing healthcare personnel to focus more on patient care.

AI in remote and decentralized care expands access through telemedicine for remote monitoring, interventions, and chronic disease management, improving healthcare delivery and outcomes. AI’s predictive analytics forecast patient outcomes and identify high-risk patients, enabling proactive interventions and individualized risk minimization.

Data-driven insights from AI analyze complex healthcare data, identifying patterns, forecasting issues, and enhancing quality. Big data analytics helps stakeholders make informed decisions, take targeted actions, and improve healthcare delivery standards, eventually enhancing patient care quality and safety.

Figure 2 illustrates the AI’s contributions to patient safety across domains, highlighting key applications such as CDS, workflow optimization, remote and decentralized care, risk prediction and data-driven insights.

Figure 2 Artificial intelligence contributions to patient safety through various domains.

Characteristics of the studies

Table 2 provides a comprehensive summary of the characteristics observed across the 36 studies included in this scoping review, highlighting the significant role of AI in enhancing QI initiatives and patient safety measures, categorizing them into five domains of AI application: CDS, workflow optimization, remote and decentralized care, risk prediction, and data-driven insights. Each entry highlights the AI method utilized, the healthcare setting, and the outcomes achieved. For example, studies in the CDS domain demonstrated improved diagnostic accuracy and treatment planning for conditions such as stroke and lung cancer. Workflow optimization studies showcased AI’s role in streamlining administrative tasks and forecasting patient volumes. These enhanced details provide greater insight into the practical applications of AI in healthcare and support reproducibility.

The quality of the studies included varied significantly, with some demonstrating comprehensive methodologies while others had limitations, such as incomplete reporting of outcomes and inadequate consideration of confounding variables (8,11). These methodological disparities may contribute to discrepancies in reported findings and highlight the need for standardized frameworks to evaluate AI’s role in patient safety and QI (7,13).

Of the 36 studies included in this review, the majority (58%) focused on CDS, demonstrating its important role in enhancing diagnostic accuracy and treatment planning. Workflow optimization accounted for 14% of studies, primarily emphasizing resource allocation and operational efficiency. Remote and decentralized care and risk prediction comprised 11% each, highlighting AI’s utility in telemedicine and predictive analytics. Data-driven insights represented 6%, underscoring its emerging role in improving risk management and decision-making processes. Across these domains, the included studies reported significant improvements in diagnostic precision (30%), operational efficiency (22%), and patient safety outcomes (25%). However, limitations such as data heterogeneity and algorithmic bias were noted in 18% of the studies, signalling areas for future development.

The results reveal that AI applications in healthcare are predominantly focused on CDS, aligning with the increasing emphasis on improving diagnostic accuracy and personalized treatment strategies. Workflow optimization studies emphasize AI’s potential to enhance operational efficiency by streamlining resource management and reducing wait times. Studies on remote and decentralized care demonstrate AI’s transformative role in expanding healthcare access, particularly through telemedicine and wearable technologies. Predictive analytics in the risk prediction domain highlight the ability of AI to forecast patient outcomes and identify high-risk populations, supporting preventive care and risk management. Lastly, data-driven insights, though limited in number, showcase AI’s emerging capability in optimizing resource allocation and enhancing decision-making processes. These findings underscore AI’s diverse applications while highlighting the need for further research to address identified gaps.

CDS

Integrating AI into CDS systems has significantly improved functional outcome predictions and clinical decision-making. Machine learning algorithms provide superior predictive accuracy, enhancing prognostic reliability and optimizing treatment strategies (15,21). This accuracy aids better resource allocation and personalized treatment, ensuring timely care for high-risk patients (23). AI’s ability to interpret large volumes of unstructured data, like radiology reports and neuroimaging, leads to more precise diagnostics, higher success rates, and better patient management, boosting quality and safety (17,19).

Transparency and interpretability of AI models are crucial for building trust and ensuring clinicians can rely on these systems for informed decisions (20). Studies show AI-driven CDS enhances clinical outcomes through accurate prognostics and timely interventions (16,18). Early, precise predictions by AI models help reduce adverse outcomes and improve patient safety (22). In antibiotic stewardship and resistance management, machine learning-driven personalized antibiograms improve antibiotic selection, treatment precision, and reduce adverse outcomes, promoting patient safety (24,25,27,28). A prototype using machine learning predicts antibiotic resistance, guiding empiric antimicrobial therapy for better clinical outcomes and safer care (26).

AI combats high antimicrobial resistance rates by improving CDS and ensuring safer antimicrobial therapies (29). Machine learning enhances the audit of antimicrobial prophylaxis, improving the quality of antibiotic use and patient safety (30). In medical imaging and lung cancer treatment, AI-driven image segmentation improves diagnostic accuracy, treatment planning, and prognosis, enhancing care quality and patient outcomes (31,34). Studies highlight the impact of transformers in medical image analysis, improving interpretability and accuracy, leading to better clinical outcomes and patient safety (32,35). An automatic segmentation method using the transresSEUnet 2.5D Network for lung cancer radiotherapy improves treatment precision and efficiency, ensuring safer patient management (33).

Workflow optimization

The application of AI in workflow optimization significantly enhances quality and patient safety across various healthcare settings. AI improves patient flow, leading to better resource utilization and reduced waiting times, thereby improving overall patient care and safety (37). Additionally, AI optimizes hospital management by streamlining administrative processes and enhancing operational efficiency. Natural language processing advances electronic health record (HER)-based clinical research by improving data extraction accuracy and efficiency, leading to more reliable clinical insights and improved workflows (36,40).

AI effectively prevents medication errors by identifying potential issues before they cause harm, significantly enhancing patient safety (38). Deep learning optimizes workflows in resource-constrained settings, improving both care quality and patient safety (39).

Remote and decentralized care

AI-enabled remote patient monitoring enhances chronic condition management and reduces hospital admissions by providing real-time data analysis, facilitating early detection of health issues, and enabling prompt, personalized care (41,42). An AI-based virtual telemedicine triage tool uses electronic devices to diagnose respiratory infections, improving patient outcomes through timely and accurate diagnoses (43). Wearable devices and mobile apps support continuous health monitoring outside clinical settings, enabling early detection, proactive healthcare management, and enhanced patient safety and care quality (44).

Risk prediction

AI-driven risk prediction advances healthcare by employing machine learning to predict human immunodeficiency virus (HIV) and sexually transmitted infections (STIs) diagnoses among high-risk individuals, offering personalized risk assessment and preventive strategies (45,46). These tools facilitate early detection and targeted interventions.

Machine learning algorithms also identify high-risk individuals for cognitive impairment and dementia through detailed data analysis, allowing for early diagnosis, prompt treatments, and individualized care, thus improving cognitive health and patient safety (47). In critical care, AI-driven predictions, such as ICU open wound mortality, provide significant insights into patient outcomes. By analyzing extensive datasets, machine learning algorithms identify mortality risk factors, enabling proactive interventions and improved critical care protocols, enhancing patient survival and treatment quality (48).

Data-driven insights

AI significantly enhances risk management in healthcare by providing advanced analytics and predictive capabilities, enabling early identification of potential risks, and improving patient safety outcomes (49). AI-powered analytics optimize resource allocation and supply chain management by analyzing complex data on resource use and supply chain dynamics, improving decision-making, reducing waste, and ensuring timely availability of essential medical supplies, thus improving operational efficiency and quality (50).

Challenges, risks, and future directions of AI in healthcare

Challenges

A major challenge in implementing AI in healthcare is the quality and availability of data. Inconsistent and incomplete data limit predictive accuracy. Diverse data sources and differing formats complicate dataset creation, while privacy laws and institutional policies restrict data sharing (42,46). Integrating AI with existing healthcare systems is also difficult. Many providers use legacy systems incompatible with advanced AI applications, causing workflow disruptions and hindering adoption (41,43). Robust validation methods are critical to ensuring AI systems’ reliability and accuracy. Techniques such as cross-validation, independent dataset testing, and real-world trials are employed to evaluate performance. For example, models predicting sepsis risk undergo validation against diverse demographic datasets to ensure generalizability.

Regulatory and ethical concerns delay AI deployment. Complex approval processes and issues like data privacy and patient consent need addressing to maintain trust and responsible handling of patient information (49,50), AI models often inherit biases from training data, leading to unequal outcomes. Models trained on specific demographics may not perform well for underrepresented groups, causing care disparities. Ensuring AI generalizability across diverse populations is challenging (15,39). AI systems face limitations including algorithmic bias, data insufficiency, and lack of interpretability. Bias arises from non-representative training datasets, affecting outcomes for underrepresented groups. Additionally, limited availability of annotated healthcare data impedes model accuracy, while the ‘black-box’ nature of certain AI models hinders clinical adoption due to lack of transparency.

Risks

AI misdiagnoses can lead to inappropriate treatment and adverse outcomes. Over-reliance on AI without human oversight increases mistakes (16,21). Healthcare data security and privacy are critical. Data breaches pose significant risks, requiring robust cybersecurity measures to protect patient information (17,29). Liability issues arise when AI systems fail or cause harm, creating complex legal and ethical questions about accountability. Clear guidelines are necessary (19,20).

Future directions

Improving AI in healthcare requires better data quality, involving consistent data collection and standardization. High-quality, annotated datasets are vital for reliable AI systems (45,47). Improving data infrastructure requires the implementation of interoperable health information systems supported by standardized data exchange protocols, enabling seamless integration of AI solutions across healthcare networks. Developing interoperable AI systems that integrate with existing infrastructure is essential. Collaboration between AI developers and healthcare providers and standardized data exchange protocols can facilitate smoother integration and adoption (18).

Streamlining regulatory approvals can speed up AI deployment. Balancing rigorous evaluation with prompt introduction of beneficial technologies, along with clear ethical guidelines for data privacy and consent, is crucial (19,28). Techniques to detect and mitigate AI biases are crucial for fair healthcare outcomes. Researchers must prioritize unbiased models, tested and validated across diverse demographics, to enhance generalizability and reliability (22,27).

Discussion

The integration of QI and AI in healthcare has ushered in a new era of patient safety enhancement. This scoping review aimed to explore the current applications, challenges, risks, and future directions at the intersection of QI, AI, and patient safety in healthcare. The findings highlight the transformative potential of AI in revolutionizing QI processes and significantly improving patient safety outcomes. For example, En-Naaoui et al. [2023] demonstrated the use of machine learning and fuzzy logic to improve the quality of hospital sterilization processes, illustrating AI’s diverse applications in healthcare beyond the studies identified here. Such findings underscore the value of broadening terminological scope in systematic reviews (5).

A critical evaluation of the included studies revealed varying quality, with some lacking robust methodologies that may introduce biases and affect the reliability of their findings. Among the included studies, those employing robust methodological frameworks or validated AI algorithms were considered more reliable. Conversely, studies with limited sample sizes or unvalidated AI tools were noted as limitations. Potential biases were observed in studies with limited sample sizes or those relying on retrospective data, which may not generalize well to broader populations (7,11). Furthermore, evidence gaps were identified in the integration of AI into diverse healthcare settings, particularly in low-resource environments where infrastructure limitations may pose challenges.

While many studies demonstrated positive outcomes, contradictory findings were observed regarding the scalability and interoperability of AI systems. These inconsistencies highlight the need for rigorous evaluations and cross-contextual validations to enhance the applicability of AI technologies in healthcare (13,42). Although AI offers transformative potential in healthcare, significant barriers impede its integration. High implementation costs, technological constraints, and interoperability challenges with legacy systems remain critical issues (41,50). Ethical concerns—such as algorithmic bias, data privacy, and accountability in decision-making—necessitate robust governance frameworks (6,51). Addressing these barriers requires multidisciplinary collaboration, targeted investments, and transparent policies.

Current applications

AI significantly enhances QI in healthcare by processing vast data, identifying patterns, and providing predictive analytics. Key applications include risk mitigation, procedure standardization, communication enhancement, data-driven decision-making, continuous education, patient-centered care, use of advanced tools, root cause analysis, fostering a safety culture, and ensuring regulatory compliance (52). These applications lead to improved patient safety, early intervention, reduced adverse events, consistent care, and targeted interventions (13). Standardized procedures supported by AI have minimized variability in care, leading to more consistent and reliable outcomes. Data-driven decision-making powered by AI has facilitated the identification of trends and the implementation of targeted interventions, ultimately enhancing patient safety.

Integrating AI and QI offers a unique opportunity to enhance patient safety by directly addressing specific safety-critical issues within healthcare systems. Unlike general healthcare QIs, which may encompass broader aspects such as patient satisfaction or administrative efficiency, patient safety focuses on the prevention of harm to patients (53,54). In this context, QI initiatives serve as structured, evidence-based interventions designed to reduce medical errors, standardize care processes, and enhance clinical reliability (3,55). AI amplifies these efforts by leveraging advanced data analysis capabilities to identify potential risks, predict adverse events, and facilitate timely interventions (2,22). For example, AI-powered predictive analytics can proactively identify high-risk patients, allowing for tailored interventions that mitigate harm (45,46). Similarly, AI-driven CDS systems enhance diagnostic precision and reduce errors, ensuring patient safety remains the cornerstone of QI activities (15,19).

Ethical concerns, including algorithmic bias, data privacy, and accountability in decision-making, necessitate robust governance frameworks, healthcare systems can transition from reactive to proactive safety management, achieving a measurable reduction in preventable harm and improving overall safety outcomes (21,34). AI holds promise for enhancing Clinical Incident Reporting Systems (CIRS) by automating data capture and analysis, identifying patterns in incident reports, and facilitating real-time alerts for proactive interventions (38,42). Such advancements could significantly improve patient safety outcomes and streamline incident management processes (49).

Challenges

Integrating AI into QI faces several challenges. High-quality data is crucial, but healthcare data often suffers from incompleteness, inconsistency, and bias, impairing AI accuracy (56,57). Data fragmentation across systems creates silos, hindering comprehensive analysis. Standardizing data collection, improving interoperability, and ensuring diverse datasets are critical (58). Integration with legacy healthcare systems is problematic, causing interoperability issues and requiring costly, time-consuming IT modifications (59). Resistance from staff accustomed to traditional workflows and a lack of specialized technical expertise further complicate AI adoption. Continuous education and training are essential (60).

Successful implementation of AI in healthcare requires addressing both human and organizational factors. Change management strategies, such as engaging stakeholders early, providing training for healthcare professionals, and fostering a culture of trust in AI systems, are critical for overcoming resistance to adoption (51,60). Furthermore, implementation frameworks that incorporate interoperability standards, such as HL7 FHIR, are necessary to ensure seamless integration with existing systems (41,50). The review also identified technical challenges, including data heterogeneity and algorithmic biases, which require proactive approaches such as federated learning models and diverse data training sets. Practical strategies, such as phased deployment and iterative testing, can mitigate these barriers and support sustainable integration.

Risks

AI in healthcare risks bias and inequity, as models trained on non-diverse datasets can produce biased outcomes, affecting marginalized populations (61). Over-reliance on AI without sufficient scrutiny can lead to complacency in clinical decision-making (62). AI must support, not replace, professional judgment.

Data privacy and security are major concerns due to the large volumes of sensitive patient data AI systems handle. Robust cybersecurity measures are necessary to protect against breaches and maintain trust (63). Ensuring the clinical validity and safety of AI systems is crucial, requiring continuous validation, monitoring, and updates to prevent patient harm (64).

Future directions

Advancements in technology and high-quality data availability will enhance patient safety. Future research should focus on developing advanced algorithms for risk prediction, best practices identification, and personalized care (45). Improving data infrastructure involves developing standardized health information systems and enhancing data quality and accessibility. Collaboration among stakeholders is essential to address AI’s multifaceted challenges, promoting safe, ethical, and effective use (51).

Capacity building and training are necessary to equip healthcare professionals with AI skills and keep them updated. Training programs for healthcare professionals should focus on AI literacy, emphasizing practical applications and ethical considerations to foster trust and effective utilization of these technologies. Developing ethical and regulatory frameworks ensures responsible AI use, addressing data privacy, security, bias, and accountability (51). Fostering innovation and research tailored to specific healthcare contexts supports local ecosystems and addresses unique challenges (65). Engaging patients and the public in AI discussions builds trust and ensures alignment with community needs and values, promoting transparency and acceptance (66).

To validate AI’s role in QI and patient safety, future studies should prioritize randomized controlled trials and robust observational designs. These investigations can provide higher-quality evidence, address limitations in current research, and offer actionable insights for integrating AI technologies into clinical practice. Furthermore, future research should focus on developing comprehensive frameworks that address both technical and organizational barriers to AI implementation. These frameworks must include strategies for change management, ethical oversight, and capacity building, thereby equipping healthcare professionals with the skills required to effectively use AI tools (51,66). Additionally, exploring context-specific solutions—particularly in resource-constrained settings—will be critical to ensuring the equitable and scalable deployment of AI technologies in healthcare (39,44).

Strengths

This scoping review employs the robust Arksey and O’Malley framework to systematically map literature in emerging fields. Adhering to PRISMA-ScR guidelines ensures transparency and rigor, enhancing the credibility of findings. A comprehensive search strategy across multiple databases, including Scopus, PubMed, and ScienceDirect, broadens the literature scope and minimizes selection bias. Inclusion criteria focus on English-language publications from 2020 onward, allowing examination of current AI trends and applications in QI and patient safety. Thematic analysis of included studies provides valuable insights into AI’s diverse applications, highlighting its significant role in enhancing healthcare delivery.

Limitations

Excluding non-English publications may limit the generalizability of findings, as research from non-English-speaking countries could offer additional insights. Excluding study types like conference papers and systematic reviews may omit relevant data and perspectives. Reliance on published literature introduces publication bias, as studies with negative or inconclusive results are less likely to be published. Focusing solely on studies published after 2020 may overlook foundational research that could inform current practices. This review is inherently limited by the scope of a scoping methodology, which emphasizes breadth over depth. As a result, conclusions about the efficacy of AI applications should be interpreted cautiously, as they primarily reflect trends in the literature rather than definitive outcomes. Additionally, many included studies lacked robust designs, further underscoring the need for higher-quality research in this field. Lastly, thematic analysis may not capture the nuanced complexities of AI integration in diverse healthcare settings, potentially oversimplifying the challenges and implications.

Conclusions

The intersection of QI, AI, and patient safety in healthcare is a rapidly evolving field with significant potential to enhance patient outcomes. Despite inherent challenges and risks, integrating AI into QI processes offers substantial benefits. However, the efficacy of AI interventions depends on rigorous evaluation, with emerging evidence suggesting promising applications but underscoring the need for further empirical research—particularly randomized controlled trials—to validate their impact on patient outcomes and operational efficiency.

This scoping review illustrates AI’s transformative potential while emphasizing the importance of ongoing research to ensure its safe, equitable, and effective implementation. Policymakers and healthcare leaders should approach AI integration with a balanced perspective, recognizing both its promise and the need for evidence-based validation. Future initiatives must focus on addressing existing challenges to maximize AI’s capacity for upholding the highest standards of patient safety. Through sustained research, innovation, and validation, healthcare systems can fully harness AI’s power to transform QI and safeguard patient safety.

Acknowledgments

None.

Footnote

Reporting Checklist: The authors have completed the PRISMA-ScR reporting checklist. Available at https://jmai.amegroups.com/article/view/10.21037/jmai-24-328/rc

Peer Review File: Available at https://jmai.amegroups.com/article/view/10.21037/jmai-24-328/prf

Funding: None.

Conflicts of Interest: All authors have completed the ICMJE uniform disclosure form (available at https://jmai.amegroups.com/article/view/10.21037/jmai-24-328/coif). The authors have no conflicts of interest to declare.

Ethical Statement: The authors are accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved.

Open Access Statement: This is an Open Access article distributed in accordance with the Creative Commons Attribution-NonCommercial-NoDerivs 4.0 International License (CC BY-NC-ND 4.0), which permits the non-commercial replication and distribution of the article with the strict proviso that no changes or edits are made and the original work is properly cited (including links to both the formal publication through the relevant DOI and the license). See: https://creativecommons.org/licenses/by-nc-nd/4.0/.

References

1. Olawade DB, Wada OJ, David-Olawade AC, et al. Using artificial intelligence to improve public health: a narrative review. Front Public Health 2023;11:1196397. [Crossref] [PubMed]
2. Ramezani M, Takian A, Bakhtiari A, et al. The application of artificial intelligence in health policy: a scoping review. BMC Health Serv Res 2023;23:1416. [Crossref] [PubMed]
3. Grys CA. Evidence-based practice, quality improvement, and research: A visual model. Nursing 2022;52:47-9. [Crossref] [PubMed]
4. Maleki Varnosfaderani S, Forouzanfar M. The Role of AI in Hospitals and Clinics: Transforming Healthcare in the 21st Century. Bioengineering (Basel) 2024;11:337. [Crossref] [PubMed]
5. En-Naaoui A, Aguezzoul A, Kaicer M. Improving the quality of hospital sterilization process using failure modes and effects analysis, fuzzy logic, and machine learning: experience in tertiary dental centre. Int J Qual Health Care 2023;35:mzad078. [Crossref] [PubMed]
6. Mennella C, Maniscalco U, De Pietro G, et al. Ethical and regulatory challenges of AI technologies in healthcare: A narrative review. Heliyon 2024;10:e26297. [Crossref] [PubMed]
7. Arksey H, O’Malley L. Scoping studies: towards a methodological framework. Int J Soc Res Methodol 2005;8:19-32. [Crossref]
8. Tricco AC, Lillie E, Zarin W, et al. PRISMA Extension for Scoping Reviews (PRISMA-ScR): Checklist and Explanation. Ann Intern Med 2018;169:467-73. [Crossref] [PubMed]
9. McHugh ML. Interrater reliability: the kappa statistic. Biochem Med (Zagreb) 2012;22:276-82. [Crossref] [PubMed]
10. Cohen J. A Coefficient of Agreement for Nominal Scales. Educ Psychol Meas 1960;20:37-46. [Crossref]
11. Page MJ, McKenzie JE, Bossuyt PM, et al. The PRISMA 2020 statement: an updated guideline for reporting systematic reviews. BMJ 2021;372: [PubMed]
12. Jiang F, Jiang Y, Zhi H, et al. Artificial intelligence in healthcare: past, present and future. Stroke Vasc Neurol 2017;2:230-43. [Crossref] [PubMed]
13. Bates DW, Levine D, Syrowatka A, et al. The potential of artificial intelligence to improve patient safety: a scoping review. NPJ Digit Med 2021;4:54. [Crossref] [PubMed]
14. Braun V, Clarke V. Using thematic analysis in psychology. Qual Res Psychol 2006;3:77-101. [Crossref]
15. Alaka SA, Menon BK, Brobbey A, et al. Functional Outcome Prediction in Ischemic Stroke: A Comparison of Machine Learning Algorithms and Regression Models. Front Neurol 2020;11:889. [Crossref] [PubMed]
16. Brugnara G, Neuberger U, Mahmutoglu MA, et al. Multimodal Predictive Modeling of Endovascular Treatment Outcome for Acute Ischemic Stroke Using Machine-Learning. Stroke 2020;51:3541-51. [Crossref] [PubMed]
17. Heo TS, Kim YS, Choi JM, et al. Prediction of Stroke Outcome Using Natural Language Processing-Based Machine Learning of Radiology Report of Brain MRI. J Pers Med 2020;10:286. [Crossref] [PubMed]
18. Matsumoto K, Nohara Y, Soejima H, et al. Stroke Prognostic Scores and Data-Driven Prediction of Clinical Outcomes After Acute Ischemic Stroke. Stroke 2020;51:1477-83. [Crossref] [PubMed]
19. Nishi H, Oishi N, Ishii A, et al. Deep Learning-Derived High-Level Neuroimaging Features Predict Clinical Outcomes for Large Vessel Occlusion. Stroke 2020;51:1484-92. [Crossref] [PubMed]
20. Zihni E, Madai VI, Livne M, et al. Opening the black box of artificial intelligence for clinical decision support: A study predicting stroke outcome. PLoS One 2020;15:e0231166. [Crossref] [PubMed]
21. Hamann J, Herzog L, Wehrli C, et al. Machine-learning-based outcome prediction in stroke patients with middle cerebral artery-M1 occlusions and early thrombectomy. Eur J Neurol 2021;28:1234-43. [Crossref] [PubMed]
22. Jiang B, Zhu G, Xie Y, et al. Prediction of Clinical Outcome in Patients with Large-Vessel Acute Ischemic Stroke: Performance of Machine Learning versus SPAN-100. AJNR Am J Neuroradiol 2021;42:240-6. [Crossref] [PubMed]
23. Kappelhof N, Ramos LA, Kappelhof M, et al. Evolutionary algorithms and decision trees for predicting poor outcome after endovascular treatment for acute ischemic stroke. Comput Biol Med 2021;133:104414. [Crossref] [PubMed]
24. Kanjilal S, Oberst M, Boominathan S, et al. A decision algorithm to promote outpatient antimicrobial stewardship for uncomplicated urinary tract infection. Sci Transl Med 2020;12:eaay5067. [Crossref] [PubMed]
25. Wong JG, Aung AH, Lian W, et al. Risk prediction models to guide antibiotic prescribing: a study on adult patients with uncomplicated upper respiratory tract infections in an emergency department. Antimicrob Resist Infect Control 2020;9:171. [Crossref] [PubMed]
26. Feretzakis G, Sakagianni A, Loupelis E, et al. Machine Learning for Antibiotic Resistance Prediction: A Prototype Using Off-the-Shelf Techniques and Entry-Level Data to Guide Empiric Antimicrobial Therapy. Healthc Inform Res 2021;27:214-21. [Crossref] [PubMed]
27. Corbin CK, Sung L, Chattopadhyay A, et al. Personalized antibiograms for machine learning driven antibiotic selection. Commun Med (Lond) 2022;2:38. [Crossref] [PubMed]
28. de Vries S, Ten Doesschate T, Totté JEE, et al. A semi-supervised decision support system to facilitate antibiotic stewardship for urinary tract infections. Comput Biol Med 2022;146:105621. [Crossref] [PubMed]
29. Rabaan AA, Alhumaid S, Mutair AA, et al. Application of Artificial Intelligence in Combating High Antimicrobial Resistance Rates. Antibiotics (Basel) 2022;11:784. [Crossref] [PubMed]
30. Shi ZY, Hon JS, Cheng CY, et al. Applying Machine Learning Techniques to the Audit of Antimicrobial Prophylaxis. Appl Sci 2022;12:2586. [Crossref]
31. Alahmadi MD. Medical Image Segmentation with Learning Semantic and Global Contextual Representation. Diagnostics (Basel) 2022;12:1548. [Crossref] [PubMed]
32. Li Y, Wu X, Yang P, et al. Machine Learning for Lung Cancer Diagnosis, Treatment, and Prognosis. Genomics Proteomics Bioinformatics 2022;20:850-66. [Crossref] [PubMed]
33. Xie H, Chen Z, Deng J, et al. Automatic segmentation of the gross target volume in radiotherapy for lung cancer using transresSEUnet 2.5D Network. J Transl Med 2022;20:524. [Crossref] [PubMed]
34. Chassagnon G, De Margerie-Mellon C, Vakalopoulou M, et al. Artificial intelligence in lung cancer: current applications and perspectives. Jpn J Radiol 2023;41:235-44. [Crossref] [PubMed]
35. He K, Gan C, Li Z, et al. Transformers in medical image analysis. Intell Med 2023;3:59-78. [Crossref]
36. Juhn Y, Liu H. Artificial intelligence approaches using natural language processing to advance EHR-based clinical research. J Allergy Clin Immunol 2020;145:463-9. [Crossref] [PubMed]
37. Dawoodbhoy FM, Delaney J, Cecula P, et al. AI in patient flow: applications of artificial intelligence to improve patient flow in NHS acute mental health inpatient units. Heliyon 2021;7:e06993. [Crossref] [PubMed]
38. Härkänen M, Haatainen K, Vehviläinen-Julkunen K, et al. Artificial Intelligence for Identifying the Prevention of Medication Incidents Causing Serious or Moderate Harm: An Analysis Using Incident Reporters’ Views. Int J Environ Res Public Health 2021;18:9206. [Crossref] [PubMed]
39. Williams D, Hornung H, Nadimpalli A, et al. Deep Learning and its Application for Healthcare Delivery in Low and Middle Income Countries. Front Artif Intell 2021;4:553987. [Crossref] [PubMed]
40. Favour O, Potter K, Doris L. Optimizing Hospital Management with AI. Information Technology in Hospitality. 2024.
41. Jeddi Z, Bohr A. Remote patient monitoring using artificial intelligence. In: Bohr A, Memarzadeh K, editors. Artificial Intelligence in Healthcare. Academic Press; 2020:203-34.
42. Shaik T, Tao X, Higgins N, et al. Remote patient monitoring using artificial intelligence: Current state, applications, and challenges. WIREs Data Mining and Knowledge Discovery 2023;13:e1485. [Crossref]
43. Villafuerte N, Manzano S, Ayala P, et al. Artificial Intelligence in Virtual Telemedicine Triage: A Respiratory Infection Diagnosis Tool with Electronic Measuring Device. Future Internet 2023;15:227. [Crossref]
44. Nahavandi D, Alizadehsani R, Khosravi A, et al. Application of artificial intelligence in wearable devices: Opportunities and challenges. Comput Methods Programs Biomed 2022;213:106541. [Crossref] [PubMed]
45. Bao Y, Medland NA, Fairley CK, et al. Predicting the diagnosis of HIV and sexually transmitted infections among men who have sex with men using machine learning approaches. J Infect 2021;82:48-59. [Crossref] [PubMed]
46. Xu X, Ge Z, Chow EPF, et al. A Machine-Learning-Based Risk-Prediction Tool for HIV and Sexually Transmitted Infections Acquisition over the Next 12 Months. J Clin Med 2022;11:1818. [Crossref] [PubMed]
47. Aschwanden D, Aichele S, Ghisletta P, et al. Predicting Cognitive Impairment and Dementia: A Machine Learning Approach. J Alzheimers Dis 2020;75:717-28. [Crossref] [PubMed]
48. Akiki RK, Anand RS, Borrelli M, et al. Predicting open wound mortality in the ICU using machine learning. J Emerg Crit Care Med 2021;5:13. [Crossref] [PubMed]
49. Guerra R. Enhancing risk management in hospitals: leveraging artificial intelligence for improved outcomes. Italian Journal of Medicine 2024; [Crossref]
50. Zewail A, Saber S. AI-powered analytics in healthcare: enhancing decision-making and efficiency. International Journal of Applied Health Care Analytics 2023;8:1-16.
51. Lysaght T, Lim HY, Xafis V, et al. AI-Assisted Decision-making in Healthcare: The Application of an Ethics Framework for Big Data in Health and Research. Asian Bioeth Rev 2019;11:299-314. [Crossref] [PubMed]
52. Bekbolatova M, Mayer J, Ong CW, et al. Transformative Potential of AI in Healthcare: Definitions, Applications, and Navigating the Ethical Landscape and Public Perspectives. Healthcare (Basel) 2024;12:125. [Crossref] [PubMed]
53. Vincent C. Patient Safety. 2nd edition. BMJ Books; 2010.
54. Institute of Medicine (US) Committee on Quality of Health Care in America; Kohn LT, Corrigan JM, editors. To Err is Human: Building a Safer Health System. Washington (DC): National Academies Press (US); 2000.
55. Berwick DM. The science of improvement. JAMA 2008;299:1182-4. [Crossref] [PubMed]
56. Aldoseri A, Al-Khalifa KN, Hamouda AM. Re-Thinking Data Strategy and Integration for Artificial Intelligence: Concepts, Opportunities, and Challenges. App Sc 2023;13:7082. [Crossref]
57. Syed R, Eden R, Makasi T, et al. Digital Health Data Quality Issues: Systematic Review. J Med Internet Res 2023;25:e42615. [Crossref] [PubMed]
58. Kadadi A, Agrawal R, Nyamful C, Atiq R. Challenges of data integration and interoperability in big data. Proceedings - 2014 IEEE International Conference on Big Data, IEEE Big Data. 2014:38-40.
59. Bhagat SV, Kanyal D. Navigating the Future: The Transformative Impact of Artificial Intelligence on Hospital Management- A Comprehensive Review. Cureus 2024;16:e54518. [Crossref] [PubMed]
60. Charow R, Jeyakumar T, Younus S, et al. Artificial Intelligence Education Programs for Health Care Professionals: Scoping Review. JMIR Med Educ 2021;7:e31043. [Crossref] [PubMed]
61. Arora A, Alderman JE, Palmer J, et al. The value of standards for health datasets in artificial intelligence-based applications. Nat Med 2023;29:2929-38. [Crossref] [PubMed]
62. Harbarth L, Gößwein E, Bodemer D, et al. (Over)Trusting AI Recommendations: How System and Person Variables Affect Dimensions of Complacency. International Journal of Human-Computer Interaction 2024;41:391-410. [Crossref]
63. Kelly B, Quinn C, Lawlor A, et al. Cybersecurity in Healthcare. In: Sakly H, Yeom K, Halabi S, et al. editors. Trends of Artificial Intelligence and Big Data for E-Health. Cham: Springer International Publishing; 2022:213-31.
64. Oettl FC, Pareek A, Winkler PW, et al. A practical guide to the implementation of AI in orthopaedic research, Part 6: How to evaluate the performance of AI research? J Exp Orthop 2024;11:e12039. [Crossref] [PubMed]
65. Siala H, Wang Y. SHIFTing artificial intelligence to be responsible in healthcare: A systematic review. Soc Sci Med 2022;296:114782. [Crossref] [PubMed]
66. Adus S, Macklin J, Pinto A. Exploring patient perspectives on how they can and should be engaged in the development of artificial intelligence (AI) applications in health care. BMC Health Serv Res 2023;23:1163. [Crossref] [PubMed]