Stop Asking Why the Person Failed. Ask Why the System Made Failure Easy.
A Human Factors Perspective on Medical Errors
When a nurse gives the wrong dose, when a physician misses a critical lab value, when a pharmacist dispenses the wrong medication, the instinct is to look for the person who made the mistake. Who was careless? Who was inattentive? Who needs retraining?
This instinct is understandable. It is also, according to decades of research in human factors science, fundamentally misguided.
The field of human factors engineering asks a different question: not who failed, but why the system allowed, or even encouraged, failure. When we conduct a rigorous human factors analysis, we consistently find that clinicians and other healthcare workers are not the root cause of errors. They are the final actor in a chain of events set in motion by poorly designed workflows, inadequate tools, excessive cognitive burdens, and environments hostile to safe performance (Carayon et al., 2006). In other words, people were doing their best in systems that made mistakes easy.
This article explores five key domains of human factors analysis: cognitive load, interruptions, fatigue, interface design, and mental models, and explains what the evidence tells us about building safer healthcare environments.
What Is Human Factors Analysis?
Human factors, also known as ergonomics, is the scientific discipline that examines how people interact with elements of a system. In healthcare, human factors involves understanding how clinicians interact with their environment, equipment, technology, and other team members, with the goal of optimizing performance and minimizing errors (Carayon et al., 2006). The field encompasses cognitive load, workflow design, communication strategies, and workplace culture.
Research has shown that numerous adverse events affecting patient safety are tied to insufficient consideration of human factors and ergonomics in the development of technologies, processes, workflows, and the broader sociotechnical systems in which healthcare workers operate (Carayon et al., 2006). Approximately 10% of patients experience at least one adverse event during their care, and a substantial portion of these events are preventable (Panagioti et al., 2019).
A useful framework for understanding how errors reach patients is the Swiss Cheese Model, developed by British psychologist James Reason in 1990. The model conceptualizes a healthcare system as multiple layers of defense, each represented as a slice of Swiss cheese. Every layer has holes: vulnerabilities or gaps, that shift and vary over time. An adverse event occurs when the holes in multiple layers align simultaneously, allowing a hazard to pass through every barrier unchecked (Reason, 1990). The model powerfully illustrates that no single individual is responsible for most adverse events; errors are systemic, not personal.
Cognitive Load: When the Brain Runs Out of Room
Cognitive load refers to the total amount of mental effort being used in working memory at any given moment. Working memory is the cognitive workspace where we hold and manipulate information in real time — and it is severely limited in capacity. When working memory becomes overwhelmed, errors are not a matter of negligence; they are a predictable consequence of exceeding a biological limit (Sweller, 1988).
Healthcare environments are uniquely suited to push clinicians beyond that limit. Diagnostic reasoning in a busy emergency department, for instance, demands that a clinician simultaneously manage incoming patient data, recall relevant medical knowledge, monitor ongoing tasks, and respond to the needs of a team, all while the environment generates a continuous stream of alarms, questions, and updates. Poorly integrated and visually complex clinical decision support systems, rather than assisting in diagnostic accuracy, can instead lead to cognitive overload and inaccurate decision making (Agency for Healthcare Research and Quality [AHRQ], 2024a).
Cognitive load theory identifies two types of load particularly relevant to clinical error. Intrinsic load stems from the complexity of the task itself, managing a deteriorating patient is simply more cognitively demanding than performing a routine check. Extraneous load, by contrast, arises from the design of the environment and tools, a confusing interface, an unnecessary alert, an awkward workflow step (Young et al., 2014). Critically, extraneous load can be reduced through better system design. The intrinsic difficulty of clinical care cannot be eliminated, but the unnecessary burdens piled on top of it can be.
One prominent example of cognitively offloading extraneous burden is the surgical safety checklist. By externalizing procedural memory requirements, checklists reduce extraneous cognitive load and ensure that critical safety steps, such as antibiotic prophylaxis and site confirmation, are consistently performed even under high-pressure conditions. Implementation of the World Health Organization (WHO) Surgical Safety Checklist has been associated with significant reductions in mortality and complication rates, demonstrating how cognitive principles applied to system design can directly enhance patient safety (Haynes et al., 2009).
When systems are designed without regard for cognitive load, clinicians are left to adapt however they can; and those adaptations, while creative, introduce their own risks.
Interruptions: The Hidden Architecture of Error
If there is one finding in human factors research that should reshape how healthcare units are organized, it is this: interruptions are not merely annoying. They are measurably dangerous.
A landmark observational study by Westbrook and colleagues (2010) directly observed nurses during medication administration and found that each interruption was associated with a 12.1% increase in procedural failures and a 12.7% increase in clinical errors. Interruptions occurred in more than half of all medication administrations, and error severity increased with interruption frequency. The risk of a major error without interruption was estimated at 2.3%; with four interruptions, that risk doubled to 4.7% (Westbrook et al., 2010).
The problem is widespread. Research in Chinese hospitals found a mean of nearly 13 nursing interruptions per hour during medication administration, arising from sources including colleagues seeking assistance, phone calls, equipment failures, and patient requests (Liu et al., 2019). A scoping review published in 2026 examining quantitative studies across global hospital settings confirmed that a majority of studies report a positive association between interruptions and medication administration errors (Schroers et al., 2026).
The mechanism behind this association is well understood. When a nurse is interrupted mid-task, they must hold the interrupted task in working memory while attending to the new demand. This is not a simple “pause and resume” operation. Research shows that following an interruption, nurses more frequently suspended the medication task to attend to the interrupting demand (51% of the time) or attempted to multitask (40% of the time), rather than completing the primary task before responding (Westbrook et al., 2010). Each of these responses places additional load on a working memory that is already taxed.
Interruption-reduction strategies such as: designated no-interruption zones, visual “do not disturb” signals for nurses during medication administration, and redesigned physical workflows, are system-level solutions that directly address this problem. They do not require individual clinicians to become more resilient. They remove the hazard from the environment.
Fatigue: A Physiological Reality, Not a Character Flaw
Healthcare workers, particularly nurses and physicians, routinely operate under conditions of significant sleep deprivation and fatigue. This is not a coincidence; it is a structural feature of healthcare labor norms. Long shifts, night duties, back-to-back rotas, and inadequate rest between shifts are common, and their consequences are measurable.
Fatigue can be defined as a physiological state of reduced mental or physical performance capability resulting from sleep loss, extended wakefulness, circadian disruption, and workload: a state that impairs alertness and the ability to perform safety-related duties (Caldwell et al., 2019). In a clinical audit across high-acuity NHS hospitals, 38% of clinical staff reported sleeping fewer than six hours before a shift, and 44% reported frequently missing protected breaks. Every single participant in the study acknowledged experiencing fatigue-related performance lapses during patient care (Mukherjee et al., 2025).
Among nurses, sleep deprivation and chronic fatigue associated with shift work have been linked to increased risk of patient care error, occupational injury, burnout, and long-term health consequences (James, 2024). For physicians in training, the data are similarly stark: a policy change limiting resident work hours was associated with a 32% reduction in resident-reported medical errors, a 34% reduction in preventable adverse events, and a 63% reduction in errors resulting in patient death (AHRQ, 2024b).
The most prevalent category of errors reported to patient safety systems in which worker fatigue is cited as a contributing factor is medication errors and errors related to procedures, treatments, or tests (Pennsylvania Patient Safety Authority, 2014). Yet healthcare organizations have been slow to treat fatigue as the structural patient safety risk it is.
The critical insight here is that fatigue is not primarily a problem of individual discipline or professional commitment. It is a predictable physiological response to the working conditions healthcare systems create. When a clinician makes a mistake after a 14-hour shift with no protected break, the appropriate question is not why they did not try harder. It is why the system required them to be there at all.
Interface Design: When Tools Work Against Their Users
Healthcare is increasingly a technology-mediated profession. Electronic health records (EHRs), infusion pumps, clinical decision support systems, medication dispensing cabinets — clinicians interact with dozens of digital and physical interfaces every shift. When these tools are well designed, they support safe performance. When they are poorly designed, they actively generate errors.
The evidence on EHR usability is particularly troubling. Research has consistently found that the inadequate presentation of large amounts of clinical data, combined with complex, non-intuitive user interfaces, adds unnecessary cognitive burden to every clinical task (Zahabi et al., 2024). A narrative review published in JMIR Medical Informatics found that the research findings suggest that inadequate efforts to present clinical data in a manner that allows users to control their cognitive burden can lead to cognitive overload and clinician burnout (Zahabi et al., 2024).
A scoping review examining EHR usability challenges found that clinicians frequently experienced significant workflow disruptions caused by poorly designed interfaces, which led to task-switching, excessive and prolonged screen navigation, and critical information fragmented across multiple screens (Moy et al., 2025). When an EHR’s task sequence misaligns with clinical workflow, clinicians create workarounds: using paper notes, spreadsheets, or text messaging to keep pace (Moy et al., 2025). These workarounds are not failures of professionalism. They are rational adaptations to tools that do not fit the work.
Alarm fatigue illustrates the same principle from a different angle. When monitoring equipment generates a high volume of low-priority alerts — as it does in most modern intensive care units — clinicians habituate to the sound. Frequent alarms generate substantial extraneous cognitive load and contribute to clinicians ignoring or delaying responses to alerts, including clinically important ones (AHRQ, 2024a). The failure is not inattention on the part of the clinician. It is a signal-to-noise ratio problem created by interface design.
Improving interface design requires involving frontline clinicians in the development and testing of tools, a process that remains far less common than it should be. As one systematic review concluded, involving clinicians in EHR development to meet their needs may help enhance usability and consequently improve their well-being and quality of care (Alobayli et al., 2023).
Mental Models: When Reality Diverges from the Map
A mental model is an internal representation of how a system works: the assumptions, expectations, and frameworks that guide behavior in the absence of explicit information. Clinicians carry mental models of how their equipment operates, how workflows proceed, how information is organized, and what is likely to happen next. These models are enormously valuable, but they are also a source of error when they do not match reality.
Mismatches between a clinician’s mental model and actual system behavior are particularly dangerous during high-stakes, time-pressured situations. When a nurse assumes that a newly installed infusion pump operates the same way as the previous model, or when a physician assumes that a particular EHR function works as it did at their prior institution, they are operating on a mental model that may be wrong. The error, when it occurs, is not stupidity. It is a predictable consequence of insufficient alignment between the design of the tool and the assumptions it inevitably creates in the user.
Human factors research emphasizes that the burden of alignment should fall on system designers, not users. A well-designed interface makes correct use intuitive and incorrect use difficult; what human factors engineers call “forcing functions.” Poorly designed systems, by contrast, allow, and sometimes actively invite, misuse by relying on users to remember exceptions, understand unintuitive operations, or maintain vigilance against easily avoided hazards.
Simulation-based training, which exposes clinicians to high-risk scenarios in controlled environments, can help correct faulty mental models and reduce the cognitive load of rare but critical events. By making the unfamiliar familiar before it occurs in real life, simulation reduces the intrinsic load associated with novel situations and builds the automaticity that allows experts to respond effectively even under pressure (Young et al., 2014).
The Systemic View: People Are Adapting, Not Failing
The clearest lesson from human factors analysis is one that safety scientists have articulated for decades but healthcare culture has been slow to absorb: when errors are frequent, the problem is almost never the people. It is the system.
James Reason, whose Swiss Cheese Model remains foundational to patient safety thinking, drew a critical distinction between what he called the “person approach” and the “systems approach” to error. The person approach attributes accidents to individual negligence, carelessness, or moral failure and responds with blame, retraining, and disciplinary action. The systems approach recognizes that human beings are fallible under predictable conditions and focuses instead on creating defenses, barriers, and safeguards that intercept errors before they reach patients (Reason, 2000).
The person approach feels satisfying because it identifies a culprit and signals that action has been taken. But it does not make systems safer, because it does not change the conditions that generated the error. The next person in the same role, in the same environment, facing the same cognitive burdens, will make the same mistake.
A systems approach asks: what was it about the task design, the environment, the tools, the workload, or the workflow that made this error likely? And then it changes those things.
When nurses make medication errors near the end of a 12-hour shift, fatigue is a system-level problem, not a personal one. When physicians miss alerts buried in an overcrowded EHR inbox, interface design is the culprit. When interruptions double the rate of clinical errors, the physical and social organization of care units must change. When cognitive load overwhelms working memory, the solution is not to demand more from clinicians — it is to design tools and workflows that demand less.
Human factors analysis often reveals that people were not failing. They were adapting, heroically and continuously, to systems that were never designed with their cognitive and physiological limits in mind. The question is whether healthcare organizations are ready to stop expecting that adaptation — and start building systems that do not require it.
References
Agency for Healthcare Research and Quality. (2024a). Cognitive load theory and its impact on diagnostic accuracy: Interplay between cognitive load and diagnostic accuracy. U.S. Department of Health and Human Services. https://www.ahrq.gov/diagnostic-safety/resources/issue-briefs/dxsafety-cognitive-load3.html
Agency for Healthcare Research and Quality. (2024b). Fatigue and sleepiness of clinicians due to hours of service. In Making healthcare safer IV. National Center for Biotechnology Information. https://www.ncbi.nlm.nih.gov/books/NBK603621/
Alobayli, F., O’Connor, S., Holloway, A., & Cresswell, K. (2023). Electronic health record stress and burnout among clinicians in hospital settings: A systematic review. SAGE Open Medicine, 11, 20552076231220241. https://doi.org/10.1177/20552076231220241
Caldwell, J. A., Caldwell, J. L., Thompson, L. A., & Lieberman, H. R. (2019). Fatigue and its management in the workplace. Neuroscience & Biobehavioral Reviews, 96, 272–289. https://doi.org/10.1016/j.neubiorev.2018.10.024
Carayon, P., Hundt, A. S., Karsh, B.-T., Gurses, A. P., Alvarado, C. J., Smith, M., & Brennan, P. F. (2006). Work system design for patient safety: The SEIPS model. Quality and Safety in Health Care, 15(Suppl. 1), i50–i58. https://doi.org/10.1136/qshc.2005.015842
Haynes, A. B., Weiser, T. G., Berry, W. R., Lipsitz, S. R., Breizat, A. H. S., Dellinger, E. P., Herbosa, T., Joseph, S., Kibatala, P. L., Lapitan, M. C. M., Merry, A. F., Moorthy, K., Reznick, R. K., Taylor, B., & Gawande, A. A. (2009). A surgical safety checklist to reduce morbidity and mortality in a global population. New England Journal of Medicine, 360(5), 491–499. https://doi.org/10.1056/NEJMsa0810119
James, S. (2024). The impact of shift-accumulated fatigue on patient care. Agency for Healthcare Research and Quality. https://www.ahrq.gov/sites/default/files/2024-10/james-report.pdf
Liu, W., Shan, X., Chen, Y., & Li, Y. (2019). Current status and influencing factors of nursing interruption events. American Journal of Managed Care, 25(4). https://www.ajmc.com/view/current-status-and-influencing-factors-of-nursing-interruption-events
Moy, E., Farooq, H., & Chen, I. (2025). Usability challenges in electronic health records: Impact on documentation burden and clinical workflow. Journal of Evaluation in Clinical Practice. https://doi.org/10.1111/jep.70189
Mukherjee, S., Clarke, T., & Agrawal, S. (2025). Sleep deprivation and fatigue in healthcare staff: A clinical audit on the risk to patient safety. PMC. https://www.ncbi.nlm.nih.gov/pmc/articles/PMC12619980/
Panagioti, M., Khan, K., Keers, R. N., Abuzour, A., Phipps, D., Kontopantelis, E., Bower, P., Campbell, S., Haneef, R., Avery, A. J., & Ashcroft, D. M. (2019). Prevalence, severity, and nature of preventable patient harm across medical care settings: Systematic review and meta-analysis. BMJ, 366, l4185. https://doi.org/10.1136/bmj.l4185
Pennsylvania Patient Safety Authority. (2014). Healthcare worker fatigue: Current strategies for the caregiver and the organization. https://patientsafety.pa.gov/ADVISORIES/Pages/201406_53.aspx
Reason, J. (1990). Human error. Cambridge University Press.
Reason, J. (2000). Human error: Models and management. BMJ, 320(7237), 768–770. https://doi.org/10.1136/bmj.320.7237.768
Schroers, G., Ross, J. G., & Moriarty, H. (2026). Associations between interruptions and medication administration errors among nurses in hospital settings: A scoping review of quantitative studies. Journal of Advanced Nursing. https://doi.org/10.1111/jan.70032
Sweller, J. (1988). Cognitive load during problem solving: Effects on learning. Cognitive Science, 12(2), 257–285. https://doi.org/10.1207/s15516709cog1202_4
Westbrook, J. I., Woods, A., Rob, M. I., Dunsmuir, W. T. M., & Day, R. O. (2010). Association of interruptions with an increased risk and severity of medication administration errors. Archives of Internal Medicine, 170(8), 683–690. https://doi.org/10.1001/archinternmed.2010.65
Young, J. Q., Van Merrienboer, J., Durning, S., & Ten Cate, O. (2014). Cognitive load theory: Implications for medical education. Academic Medicine, 89(12), 1564–1569. https://doi.org/10.1097/ACM.0000000000000426
Zahabi, M., Valdez, R. S., & others. (2024). Impact of electronic health record use on cognitive load and burnout among clinicians: Narrative review. JMIR Medical Informatics, 12, e55499. https://doi.org/10.2196/55499
