Published in BMC Nursing on 12 September 2026 as early-access, peer-reviewed, open-access research, the review examined evidence concerning identifiable AI applications in nursing practice or nursing workflows. The authors said AI is increasingly used in activities including care planning, assessment, documentation, handover and patient education. Eight studies met the inclusion criteria. Seven focused on large language models or other generative AI systems, and most evaluated these tools in experimental or prototype settings rather than routine clinical workflows.

The authors synthesised three categories of mechanisms associated with the reported risks: unreliable knowledge bases and model performance; misalignment with nursing work systems; and compromised information integrity and nursing representation. Safeguards described in the literature fell into three broad groups: model optimisation and context-specific validation; human review and workflow control; and governance and competency building. These proposed safeguards reflect recommendations and approaches in the included literature; the review did not establish their effectiveness in routine nursing practice.

For the review, the authors searched PubMed, Embase, CINAHL, Scopus, Web of Science, PsycINFO, IEEE Xplore and the Cochrane Library. Database searches returned 6,295 records through 28 April 2026, while supplementary searches identified 12 further records through 10 May 2026. Eligible sources described AI applications in nursing practice or workflows and reported safety risks, mechanisms or safeguards. Sources limited to non-clinical applications, nursing education or research, and technical model development were excluded, as were sources without extractable safety-relevant data.

The findings are based on a small and largely experimental evidence base, so they should not be generalised to all forms of AI or to routine nursing practice. The review authors called for evaluation of these tools in real-world or simulated nursing workflows, as well as prospective testing of safeguards that incorporate context-specific validation, nurse-led review, workflow controls and organisational governance. The supplied publisher record states that the early-access version remains subject to further editing and will be replaced by the final Version of Record.

Subject of Research: Scoping review of artificial intelligence-related patient safety risks in nursing practice

Article Title: Artificial intelligence-related patient safety risks in nursing practice: a scoping review

Article References: Zhang, P., Li, Z., She, S., Wu, H., Zheng, A., Wei, S., Ding, X., & Xiao, M. (2026). Artificial intelligence-related patient safety risks in nursing practice: a scoping review. BMC Nursing. https://doi.org/10.1186/s12912-026-05336-x

Image Credits: AI Generated

DOI: 10.1186/s12912-026-05336-x

Keywords: artificial intelligence, patient safety, nursing practice, scoping review, generative AI, large language models, clinical workflows, human review, workflow governance, Artificial, intelligence-related, patient

Vascular access is the circulatory lifeline of neonatal intensive care. Premature and critically ill infants depend on intravenous lines for nutrition, medications, and fluids, often for weeks at a time. Yet the devices that deliver these therapies fail frequently. Peripheral intravenous catheters in neonates are prone to infiltration, dislodgement, phlebitis, and occlusion, and each failure means another painful insertion attempt in an infant whose veins may be barely visible. Previous work by the same research group, including retrospective cohort studies such as the ABBA project, documented the scale of device-related complications and the factors associated with them. What remained poorly understood was the human and organizational side of the problem: how nurses actually decide when to insert, monitor, escalate, or abandon a line.

To answer that question, Matheus van Rens of Radboud University Medical Center and colleagues conducted semi-structured interviews with twelve neonatal nurses involved in vascular access and infusion therapy across three tertiary neonatal intensive care units in the Netherlands. The team analyzed the transcripts using reflexive thematic analysis, a method that treats the researcher’s perspective as an active ingredient in interpretation rather than a bias to be eliminated. Sample adequacy was guided by the concept of information power, an approach that determines whether enough interviews have been conducted based on the richness of the data and the specificity of the study aims, rather than by a fixed numerical target.

Four themes emerged from the analysis, and together they sketch a system that consistently rewards reaction over anticipation. The dominant finding was that vascular access care was characterized by reactive decision-making, with escalation to more secure or more appropriate devices typically occurring only after a complication had already appeared. Nurses described delayed escalation, repeated peripheral intravenous attempts, and inconsistent monitoring practices across all three centers. In other words, a catheter often had to fail before the team reconsidered whether a different strategy was needed from the start.

Behind this reactive pattern, the interviews revealed a web of interacting pressures. Organizational factors, including workload, staffing constraints, and the absence of structured decision-support embedded in daily workflows, shaped how much attention vascular access received at any given moment. Knowledge varied between individual nurses, and so did the implementation of existing guidelines, producing inconsistency both within and between units. The researchers also identified relational factors: parents were largely left out of decisions about vascular access, despite growing evidence from family integrated care research that involving parents in neonatal care improves outcomes for infants and families alike.

Perhaps the most striking theme was a persistent gap between risk awareness and clinical action. The nurses interviewed clearly understood which infants were at high risk of catheter failure and could articulate the consequences of repeated attempts. Knowing, however, did not translate into doing. This disconnection echoes classic patient safety literature, including James Reason’s models of human error, which emphasize that accidents rarely stem from individual carelessness alone but from systems that allow known risks to go unaddressed. The study’s authors frame their findings in exactly this spirit, pointing to organizational and behavioral factors that reinforce reactive practice regardless of individual competence or good intentions.

The technical literature has long documented the consequences of this gap. Studies of peripheral intravenous catheter failure in adult and pediatric populations have described the problem as accepted but unacceptable, a phrase coined by researchers reviewing the burden of device failure in hospitals. In neonates, multicenter observational studies have reported substantial complication rates for peripheral cannulation, and consensus tools such as the neonatal DAV-expert algorithm developed by European vascular access groups now exist to guide the selection of the most appropriate venous access for each newborn. The Dutch study suggests that the bottleneck is not a lack of technical knowledge or decision tools, but the absence of structures that trigger their use at the right moment.

The authors argue that the way forward lies in structured, anticipatory approaches integrated directly into clinical workflows. Rather than relying on individual nurses to remember to reassess a line or to escalate concerns through informal channels, units could build anticipatory planning into routine care, ensuring that the choice of device, the monitoring plan, and the escalation pathway are considered together at the moment of insertion and revisited systematically thereafter. The team has previously proposed a patient-centered framework based on the 7-Rights of vascular access, and the new findings support the idea that such frameworks must be embedded in practice rather than left as optional guidance.

The implications extend beyond the neonatal unit. Catheter-related bloodstream infections in intensive care were dramatically reduced in landmark studies when checklists and standardized bundles transformed culture and behavior, demonstrating that structured interventions can convert awareness into consistent action. The Dutch researchers point to similar potential in neonatal vascular access: if the reactive cycle of failure and rescue can be interrupted by anticipatory planning, the payoff could include fewer painful procedures, fewer infections, shorter stays, and more consistent care across centers. Practice variation in neonatal care is well documented, and standardizing the behavioral and organizational dimensions of vascular access may be as important as standardizing the technical ones.

The study also carries a methodological lesson for patient safety research. By listening to nurses rather than measuring devices, the investigators surfaced factors that retrospective audits of complication rates could never capture: the hesitation to escalate, the uneven distribution of knowledge, the silence of parents at the bedside. The authors acknowledge the limitations inherent in qualitative work, including the interpretive nature of thematic analysis and the potential identifiability of interview data, which is why the underlying datasets are available only on reasonable request. Yet the consistency of the four themes across three separate tertiary centers lends weight to the central conclusion. Vascular access care in neonatal intensive care, the nurses suggest, is shaped by interacting organizational, behavioral, knowledge-related, and relational forces that quietly reinforce reaction over prevention. Recognizing those forces, the researchers argue, is the first step toward designing workflows in which the smallest patients receive lines that are planned, monitored, and escalated with the same rigor that the technology itself deserves.

Subject of Research: Nurses' decision-making and reactive versus proactive practice in neonatal vascular access care

Article Title: Reactive rather than proactive: nurses’ perspectives on decision-making in neonatal vascular access

Article References: van Rens, M. F., Huis, A., Hugill, K., van der Lee, R., & de Boode, W. P. (2026). Reactive rather than proactive: nurses’ perspectives on decision-making in neonatal vascular access. Pediatric Research. https://doi.org/10.1038/s41390-026-05570-8

Image Credits: AI Generated

DOI: 10.1038/s41390-026-05570-8

Keywords: neonatal intensive care, vascular access, nursing decision-making, peripheral intravenous catheters, patient safety, qualitative research, thematic analysis, catheter failure, escalation delays, family integrated care, clinical workflows, Pediatric Research

The research team, drawn from the University of Leipzig, the German Research Center for Artificial Intelligence, Charité in Berlin, SurgiTAIX AG, Vodafone’s Tech Innovation Center in Dresden and RPTU Kaiserslautern-Landau, conducted semi-structured interviews between August 2023 and May 2025. Participants came from clinical fields including anesthesiology, cardiology, internal medicine, pulmonology and sleep medicine, as well as from the medical technology and health IT sectors. Each interview lasted between 32 and 47 minutes, followed the COREQ reporting checklist for qualitative research, and was transcribed with OpenAI’s Whisper speech recognition system. For the analysis, the researchers used a combined deductive-inductive thematic approach, employing the Llama 3.1 8B Instruct language model to generate preliminary transcript summaries that were then manually reviewed and corrected against the original recordings. Five initial categories came from the interview guide itself, and three more emerged from the data.

The starting point of the study is a sobering diagnosis of the present. Even before 6G arrives, hospitals struggle with insufficient interoperability, fragmented IT systems and unreliable network coverage. Devices are frequently tethered by cables, data must often be moved by hand between incompatible systems, and documentation consumes time that clinical staff, working under chronic shortages, do not have. These problems matter because the demographic shift between patients and practitioners is intensifying, making efficiency gains from digital transformation not a luxury but a necessity. Remote health solutions, automation and artificial intelligence all depend on communication networks capable of moving massive data volumes in real time with guaranteed reliability.

Against that backdrop, the stakeholders’ expectations for 6G are strikingly concrete. They anticipate higher data rates, ultra-low latency, broader coverage, the capacity to connect vastly more devices simultaneously, enhanced encryption and lower energy consumption. In practical terms, they envision real-time transmission of vital signs, imaging data and other medical information supporting live consultations, remote diagnostics and tele-supervision in operating rooms and emergency departments. Continuous home-based monitoring could accelerate clinical decision-making and improve patient safety. They also see 6G as a driver of workflow optimization: standardized device communication, mobile ward rounds, integrated patient and equipment tracking, and automated documentation assisted by large language models. Improved connectivity, they argue, is essential for efficient hospital operations and optimal resource allocation.

The study drilled into three specific use cases. The first involves artificial heart systems such as mobile ventricular assist devices, where stakeholders demand maximum fail-safety, reliable alarms and stable data transmission supported by proactive network monitoring and seamless coverage beyond hospital walls. They stressed transparency with patients about network stability and ensuring emergency connectivity with ambulance services, but flagged unresolved liability questions when automated systems fail. The second covers augmented and virtual reality applications, viewed positively for training and surgical support, though participants worried about distraction and information overload in daily practice. The third is smart hospital tracking: asset and material tracking drew broad enthusiasm for efficiency gains, patient tracking was deemed useful in specific scenarios such as surgical logistics, fall prevention and dementia care, but staff tracking was predominantly regarded critically, touching a raw nerve about surveillance in the workplace.

The concerns expressed were equally pointed. Data protection and privacy emerged as the single most critical challenge, encompassing fears of data misuse, cyberattacks and ambiguous patient consent for data use, along with skepticism about centralized data storage. Some participants found the European GDPR overly restrictive for medical innovation, while others defended strict local data handling and access controls. Responsibility and liability in the context of increasing automation and AI-supported decision-making remain unsolved, and stakeholders called for a clear framework defining who is accountable when technical malfunctions occur. There was also anxiety about technological dependency itself: if practitioners lose the skills to treat patients without technical assistance systems, any system shutdown could degrade care quality. Manual backup procedures, paradoxically, were seen as necessary to sustain trust in automated ones.

Acceptance, the study found, is deeply individual and hinges less on the technology itself than on perceived usefulness and honest communication of benefits. Reactions ranged from outright rejection to enthusiastic adoption, largely determined by reliability. Key barriers included limited transparency, poor communication and restrictive regulation, with misinformation amplifying uncertainty. Notably, many interviewees worried more about user errors caused by complex, unintuitive interfaces than about external cyberattacks, particularly under staff shortage and time pressure. This led to a strong emphasis on intuitive system design, redundancy for monitoring and diagnostic systems, and graceful degradation that preserves essential functionality during failures. Stakeholders also prioritized low latency over raw bandwidth for timing-critical scenarios such as remote surgery and remote control, and demanded deterministic, real-time communication with immediate alerts and automated emergency responses during network outages.

Synthesizing these findings, the authors derived five recommendations for technology providers. Interoperability comes first: open interfaces, shared data models and transparent standards are needed to end redundant documentation and proprietary lock-in. Second, network coverage and reliability are prerequisites for everything else, including robust connectivity in rural areas where telemedicine is growing; as one participant put it, the infrastructure must guarantee a stable and fast connection outside major cities. Third, awareness and trust must be actively built through transparent, comprehensible information about functionality, limitations and risks, coupled with clear explanations of system behavior during failures. Fourth, usability must be a central design criterion, adaptable to varying levels of technical proficiency. Fifth, technological performance must be balanced against sustainable energy use, because, as one technical stakeholder observed, it would be impractical if every device required an additional power bank. Continuous real-time monitoring can drain battery-powered sensors, so power-aware scheduling must be weighed against clinical needs for uninterrupted data transmission.

Ultimately, the study frames the hospital of the future not as a hyper-connected machine but as a resilient, people-centered ecosystem in which technology serves as an enabler of care. The authors emphasize that purely technical performance metrics, such as higher data rates or lower latency, only matter when they solve real clinical problems. The introduction of 6G should be understood as a co-evolutionary process between technology and healthcare practice, requiring ongoing dialog and co-design; as one physician put it, clinical users should develop more IT skills while IT developers should spend more time in clinics. The researchers acknowledge limitations: the LLM-assisted analysis saved less time than expected due to validation overhead, demographic characteristics were not systematically collected, and all interviews took place within the German healthcare system, which may limit transferability. Future work should adopt a multicentric, international scope and extend beyond the hospital to general practitioners and emergency medical services. But the core message stands: if engineers want 6G to transform medicine, they must start listening to medicine now, while the standard is still being written.

Subject of Research: Stakeholder perspectives on 6G technology requirements for healthcare and future hospitals

Article Title: Anticipating the hospital of the future: stakeholder perspectives on 6G technology requirements in health care

Article References: Schatz, A., Rüb, M., Hegermann, H., Hiddemann, M., Wieschebrock, D., Stegemann, D., Herbst, J., Lipps, C., Möllenhoff, C., Petzold, T., Irmer, R., Köhler, F., Rockstroh, M., Pabst, T., Neumuth, T., Schotten, H. D., & Treskatsch, S. (2026). Anticipating the hospital of the future: stakeholder perspectives on 6G technology requirements in health care. International Journal of Computer Assisted Radiology and Surgery. https://doi.org/10.1007/s11548-026-03776-z

Image Credits: AI Generated

DOI: 10.1007/s11548-026-03776-z

Keywords: 6G, smart hospitals, healthcare, interoperability, telemedicine, clinical workflows, data protection, network reliability, medical technology, artificial intelligence, remote monitoring, standardization