{
  "id": "patient-safety",
  "title": "Patient Safety",
  "lang": "en",
  "source_lang": "en",
  "summary": "Patient safety is the prevention of avoidable harm in health care through safer clinical practices, organizational systems, technologies, and patient participation.",
  "category": "health",
  "subcategory": "public-health",
  "type": "concept",
  "aliases": [
    "Safety of patients",
    "Health care safety"
  ],
  "generated_at": "2026-11-03",
  "model": "pro/gpt-6.1-sol",
  "url": "https://aiwiki.page/en/patient-safety/",
  "translations": {
    "zh-Hans": {
      "url": "https://aiwiki.page/zh/patient-safety/",
      "json": "https://aiwiki.page/zh/patient-safety.json"
    }
  },
  "sections": [
    {
      "heading": "",
      "text": "Patient safety is the organized effort to prevent avoidable harm during health care and to reduce the risk of unnecessary harm to an acceptable minimum. It concerns both what clinicians do and the conditions under which care is delivered: staffing, equipment, medicines, information systems, communication, physical environments, and institutional governance. Within a health system, safety is a fundamental dimension of quality, alongside effectiveness, timeliness, efficiency, equity, and person-centred care.\n\nThe field addresses injuries associated with treatment, missed or delayed diagnoses, infections acquired during care, medication errors, unsafe procedures, and failures in follow-up or transitions between services. Such events occur in hospitals, outpatient clinics, pharmacies, long-term care facilities, and patients’ homes. Some cause immediate injury; others produce delayed disability, psychological distress, financial costs, or loss of trust. Not every undesirable outcome is preventable, and not every error causes harm.\n\nModern patient safety combines clinical knowledge with human factors and ergonomics, organizational science, engineering, and methods of measurement and improvement. Its central premise is that reliable care cannot depend solely on individual vigilance: systems must make appropriate actions easier, detect emerging hazards, and limit the consequences when failures occur."
    },
    {
      "heading": "Meaning, boundaries, and historical development",
      "text": "A patient safety incident is an event or circumstance that could have resulted, or did result, in unnecessary harm to a patient. An adverse event is an incident that causes harm associated with care rather than solely with the underlying illness. A near miss does not reach the patient, while a no-harm incident reaches the patient without producing discernible injury. For example, a pharmacist intercepting a tenfold prescribing error constitutes a near miss; administration of an incorrect dose without an observed adverse effect constitutes a no-harm incident. These distinctions matter because injury alone does not reveal how frequently unsafe processes occur.\n\nError and harm are related but separate concepts. An error may involve failure to carry out an intended action correctly or selection of an unsuitable plan. An adverse drug reaction can occur despite appropriate prescribing and administration, whereas a prescribing error may be intercepted before any medicine is given. Preventability is therefore a separate judgment, usually based on the knowledge, available safeguards, and circumstances at the time. It is not equivalent to negligence, which is a legal determination involving jurisdiction-specific standards and evidence.\n\nConcern about harm from treatment long predates the modern movement. In 1847, Ignaz Semmelweis introduced chlorinated hand cleansing in a Vienna maternity clinic and documented a marked decline in deaths from puerperal fever. Florence Nightingale used mortality data and observations of hospital conditions during and after the Crimean War to argue for environmental and organizational reform. The later development of germ theory, antisepsis, safer anesthesia, and blood banking transformed the risks of clinical care, although each new intervention also introduced hazards requiring control.\n\nPatient safety emerged as a distinct international policy field in the late twentieth century. The Harvard Medical Practice Study, published in 1991 using records from New York hospitals in 1984, found adverse events in approximately 3.7% of hospitalizations. In 1999, the United States Institute of Medicine released To Err Is Human, which estimated that preventable adverse events in hospitals contributed to between 44,000 and 98,000 deaths annually. These estimates were extrapolations from earlier studies, not a direct national census, but the report brought system failures to public attention.\n\nThe World Health Organization launched the World Alliance for Patient Safety in 2004. Its subsequent challenges included Clean Care Is Safer Care in 2005, Safe Surgery Saves Lives in 2008, and Medication Without Harm in 2017. In 2019, the World Health Assembly established World Patient Safety Day on 17 September. The Global Patient Safety Action Plan 2021–2030 subsequently provided an international framework for reducing avoidable harm across settings, linking clinical safety with leadership, education, patient engagement, and learning systems."
    },
    {
      "heading": "Burden of harm and the difficulties of measurement",
      "text": "There is no single, universally comparable measure of unsafe care. Studies differ in the populations examined, settings included, definitions of injury, observation periods, and judgments about preventability. The WHO’s 2023 patient safety fact sheet described harm affecting around one in ten patients receiving health care and more than three million deaths annually associated with unsafe care. These figures summarize heterogeneous evidence; they should not be interpreted as the result of one worldwide surveillance system or as a uniform risk for every health service.\n\nHospital record reviews commonly identify medication-related injury, infection, surgical complications, pressure injuries, and diagnostic problems. Yet hospital-based research captures only part of the burden. In primary health care, harm may arise through delayed referral, inadequate monitoring, lost test results, or medication use across several prescribers. Home care introduces risks involving storage, equipment maintenance, caregiver training, and the suitability of the physical environment. These events may become visible only when a patient attends an emergency department or is admitted elsewhere.\n\nDifferent measurement methods reveal different parts of the problem. Voluntary incident reports can provide detailed descriptions of unusual hazards, but reporting depends on staff awareness, time, organizational culture, and confidence that reports will lead to learning. Reports therefore cannot usually establish the true incidence of harm. An increase in reporting may indicate improved openness rather than deteriorating safety. Administrative data cover large populations but are affected by coding practices and may not distinguish a pre-existing condition from a complication acquired during care.\n\nStructured record review and trigger tools search for signals such as antidote administration, unexpected return to an operating room, abrupt laboratory changes, or unplanned transfer to intensive care. Reviewers then determine whether harm occurred and whether care contributed. These methods detect events that are absent from incident-reporting systems, but they require resources and remain vulnerable to incomplete documentation and disagreement between reviewers. Patient interviews and complaints add information about communication failures, symptoms after discharge, and problems occurring outside institutional records.\n\nReliable interpretation requires appropriate denominators and attention to case mix. A device-associated infection rate may be expressed as:\n\n\\[\\text{Infections per 1,000 device-days} = \\frac{\\text{Number of qualifying infections}}{\\text{Total device-days}} \\times 1{,}000.\\]\n\nComparisons can still be misleading if hospitals differ in patient severity, diagnostic testing, or surveillance intensity. Outcome measures are therefore commonly combined with process measures, such as compliance with an insertion checklist, and balancing measures, such as delays caused by a new verification procedure. Estimates of deaths attributable to unsafe care are particularly uncertain because severe illness, treatment complications, and multiple contributing causes often overlap. Classifying medical error as a leading cause of death is not directly comparable with conventional cause-of-death statistics."
    },
    {
      "heading": "Human factors, system design, and organizational reliability",
      "text": "Clinical work occurs within interacting social and technical systems. A nurse administering an infusion depends on the prescription, pharmacy preparation, pump interface, tubing configuration, patient identification process, staffing level, and availability of information about renal function or allergies. An incident may appear to be a single incorrect action at the bedside while actually reflecting several earlier failures. Human factors analysis examines these interactions rather than treating human performance as an isolated property of an individual.\n\nJames Reason’s “Swiss cheese” model, developed from work on organizational accidents, distinguishes active failures at the point of action from latent conditions embedded in systems. Multiple protective layers may each contain weaknesses. Harm occurs when those weaknesses align sufficiently for a hazard to pass through. The model helps explain why adding another instruction is often inadequate: an unreadable label, an unsafe default, and repeated interruptions may defeat even knowledgeable staff. It is a conceptual model, not a quantitative law, and real incidents often involve feedback loops and changing conditions rather than a simple linear chain.\n\nHuman errors also differ in mechanism. A slip occurs during execution, such as selecting an adjacent item from a menu; a lapse involves memory, such as forgetting a pending task. A mistake involves an incorrect interpretation or plan. Fatigue, workload, distraction, ambiguous displays, and poorly coordinated responsibilities can increase the likelihood of all three. Violations—departures from a rule or procedure—require separate analysis because staff may be responding to impractical procedures, conflicting demands, or normalized shortcuts rather than intending harm.\n\nThe strength of a safeguard depends partly on how much it relies on attention and memory. Training and warnings may be necessary but are often weaker than redesigning a process. A forcing function prevents a hazardous action under specified conditions; a connector that cannot fit an incompatible route is one example. Standardized concentrations, clear labeling, separation of look-alike products, and well-designed order sets can reduce ambiguity. Nevertheless, excessive standardization can create new risks when legitimate clinical variation is not accommodated, and poorly designed barriers may encourage workarounds.\n\nInvestigative methods include root cause analysis, which examines an incident retrospectively, and failure mode and effects analysis, which prospectively identifies how a process might fail and prioritizes safeguards. Neither automatically produces improvement. An investigation can generate a detailed report without changing the conditions that caused the event. Effective corrective action therefore requires an accountable owner, implementation resources, testing under realistic conditions, and follow-up measurement. Contemporary approaches also examine how staff successfully adapt to difficult conditions, recognizing that the same adaptations that preserve care on one occasion may become hazardous on another."
    },
    {
      "heading": "Medication safety across the treatment pathway",
      "text": "Medication safety covers prescribing, dispensing, preparation, administration, monitoring, and discontinuation. Errors can arise at any stage and propagate across organizational boundaries. An inaccurate medication history may become an incorrect admission prescription, then be copied into a discharge list and continued in community care. Conversely, an intended treatment change may fail to reach the patient, pharmacist, or next prescriber. Safe medication use is therefore a property of the entire pathway, not simply of the final administration check.\n\nCertain medicines can cause substantial injury when used incorrectly. Commonly designated high-alert groups include insulin, anticoagulants, opioids, and concentrated electrolytes. Insulin dosing errors can produce severe hypoglycemia; anticoagulant errors can cause major bleeding; opioid-related respiratory depression can become life-threatening. Risk depends on the drug, dose, route, patient characteristics, and monitoring arrangements. Impaired kidney function, for example, may reduce elimination of some medicines or their metabolites, making an otherwise familiar dose inappropriate for a particular patient.\n\nPrescribing safeguards include access to allergy information, clinically relevant interaction checking, standardized dose expressions, and decision support linked to current patient data. Electronic prescribing removes handwriting ambiguity but can introduce selection errors, duplicate orders, and inappropriate default doses. Frequent low-value warnings contribute to alert fatigue, in which clinicians override notifications because most appear irrelevant. The safety of an alert depends on its specificity, timing, clarity, and capacity to support an appropriate response, not merely on whether the software generates it.\n\nMedication reconciliation compares an accurate account of a patient’s medicines with orders at a transition of care and resolves discrepancies. The account includes prescribed medicines and relevant nonprescription products, as well as the actual pattern of use. Its purpose is not to preserve every previous prescription unchanged: intentional stopping or dose adjustment may be appropriate. The safety task is to distinguish intended changes from omissions, duplications, or misunderstandings and to communicate the resulting plan clearly. Polypharmacy and multimorbidity make this process more complex because several clinicians may be treating overlapping conditions.\n\nAt dispensing and administration, barcode verification, appropriately configured infusion pumps, standardized preparation, and independent checks for selected high-risk tasks can provide additional protection. These controls remain vulnerable to bypasses, incorrect database configuration, and checks performed without genuine independence. Monitoring after administration is equally important, particularly when effects change with illness, interactions, or organ function. Pharmacovigilance complements local safety work by detecting and evaluating adverse effects across larger populations. It addresses suspected harmful effects of medicines, whereas medication-error analysis focuses on failures in their use; the two fields overlap but are not interchangeable."
    },
    {
      "heading": "Infection prevention and the safety of the care environment",
      "text": "Health care-associated infections develop in connection with the delivery of care and were not present or incubating at the relevant starting point, according to the surveillance definition used. They include surgical-site infections, bloodstream infections, catheter-associated urinary infections, and some pneumonias. Infection may follow transmission from another person or the environment, but it can also arise when a patient’s own microorganisms enter normally sterile tissues through an incision or device. Consequently, infection prevention involves both interrupting transmission and reducing opportunities for invasion.\n\nHand hygiene is a central control because hands connect patients, equipment, surfaces, and invasive procedures. The WHO’s “Five Moments for Hand Hygiene” identifies opportunities before touching a patient, before a clean or aseptic procedure, after body-fluid exposure risk, after touching a patient, and after touching patient surroundings. Performance depends on more than staff knowledge: accessible supplies, workflow, skin tolerance, staffing, and feedback all affect adherence. Alcohol-based hand rub is widely used, while soap-and-water washing has particular roles when hands are visibly soiled and in settings requiring removal of certain contaminants.\n\nInvasive devices require coordinated safeguards. Central venous catheter programs commonly combine hand hygiene, aseptic insertion, appropriate skin antisepsis, careful site care, and review of continuing necessity. Urinary catheter safety similarly addresses whether a device is indicated, how it is inserted, maintenance of a closed drainage system, and timely removal when no longer needed. These measures are often implemented as bundles: small groups of evidence-based practices delivered together. Their effectiveness depends on reliable execution and surveillance, not on the checklist’s existence alone.\n\nEnvironmental controls include cleaning, disinfection, sterilization, safe water, ventilation, waste management, and appropriate handling of reusable equipment. Each acts through different mechanisms. Sterilization aims to eliminate all viable microorganisms, including bacterial spores, from appropriate instruments; disinfection has a more limited scope determined by the agent and procedure. Airborne transmission requires controls distinct from those used for contaminated surfaces. Problems in design or maintenance, such as inadequate ventilation or difficult-to-clean equipment, cannot be fully compensated for by exhorting staff to be more careful.\n\nAntimicrobial resistance connects infection prevention with medicine use. Resistant organisms can make infections harder to treat, while unnecessary antibiotic exposure selects for resistance and can contribute to Clostridioides difficile infection. Antimicrobial stewardship seeks appropriate antimicrobial selection, dosing, route, and duration while preserving effective treatment. Infection prevention and stewardship are complementary: preventing infections reduces treatment requirements, and appropriate antimicrobial use limits avoidable drug harm and ecological pressure. Facilities with unreliable water, sanitation, electricity, or supply chains face additional structural constraints, making basic infrastructure part of patient safety rather than a separate concern."
    },
    {
      "heading": "Surgical, procedural, and device-related safety",
      "text": "Procedural safety encompasses the decision to intervene, preparation, performance, recovery, and follow-up. An operation can be technically successful yet unsafe if performed on the wrong patient, at the wrong site, or without adequate preparation for a foreseeable complication. Relevant hazards include bleeding, infection, anesthesia complications, retained items, positioning injuries, and failure to recognize postoperative deterioration. The appropriateness of the procedure itself also matters: technical safeguards cannot remove the harms of an intervention that offers no reasonable expected benefit.\n\nThe WHO Surgical Safety Checklist, introduced in 2008, organizes team checks around three points: before induction of anesthesia, before skin incision, and before the patient leaves the operating room. Its elements include identity and procedure confirmation, assessment of airway and blood-loss risks, timely antibiotic prophylaxis when indicated, instrument and sponge counts, specimen labeling, and discussion of recovery concerns. It is intended as a shared verbal process rather than a clerical exercise. Team members must be able to identify discrepancies and pause the workflow when necessary.\n\nA 2009 multicentre before-and-after study of checklist implementation in eight hospitals reported a decline in inpatient complications from 11.0% to 7.0% and in deaths from 1.5% to 0.8%. The study helped establish international interest, but its design could not fully separate the checklist from accompanying changes in communication, attention, or practice. Later implementation studies have produced varying results. Differences in baseline care, participation, leadership, and fidelity help explain why distributing a checklist does not reliably reproduce the effects of an intensive implementation program.\n\nAnesthesia safety illustrates the interaction of technology, specialist training, and organizational standards. Monitoring with pulse oximetry and capnography can reveal physiological changes before they are obvious through observation alone, but monitoring requires meaningful interpretation and timely response. Safe equipment, airway planning, drug identification, and recovery-area observation address different failure pathways. Similar principles apply to sedation outside operating rooms, where procedural teams may work with different staffing, equipment, and rescue arrangements. Blood transfusion safety adds requirements for specimen identification, compatibility testing, component handling, and bedside verification.\n\nMedical devices introduce risks throughout procurement, installation, configuration, use, maintenance, and disposal. A device may function as designed yet be unsafe in a particular workflow because controls are confusing, connectors permit incompatible connections, or alarms are difficult to distinguish. Alarm fatigue develops when numerous signals, many of limited clinical importance, compete for attention. Safety measures therefore include appropriate alarm configuration, maintenance, staff competence, and clear responsibility for response. Postoperative care adds further system demands: recognition of bleeding, respiratory compromise, sepsis, or thromboembolism must connect promptly to assessment and treatment rather than depend on an isolated measurement or score."
    },
    {
      "heading": "Diagnostic safety, deterioration, and transitions of care",
      "text": "Diagnostic safety concerns whether a timely and appropriate explanation of a patient’s health problem is established and communicated. In its 2015 report Improving Diagnosis in Health Care, the United States National Academies described diagnostic error as failure to establish an accurate and timely explanation of a patient’s health problem, or failure to communicate that explanation to the patient. This definition places communication within the diagnostic process rather than treating it as an optional step after clinical reasoning is complete.\n\nDiagnosis unfolds over time through history-taking, examination, testing, interpretation, consultation, and reassessment. Problems may arise through premature closure on an initial explanation, failure to consider a plausible alternative, inaccurate test interpretation, or fragmented information. However, diagnostic uncertainty is not itself an error. Many illnesses present with nonspecific findings, and some diagnoses become apparent only as a condition evolves. Assessment of a possible failure must therefore consider what information was available at the time rather than relying solely on knowledge of the eventual outcome.\n\nTesting systems create distinct organizational vulnerabilities. A specimen can be mislabeled, a report routed to the wrong clinician, or a significant result left unacknowledged after discharge. Closed-loop communication assigns responsibility for reviewing a result, determining its significance, informing the patient or relevant professional, and documenting subsequent action. The existence of an electronic inbox does not establish that this loop has closed. More testing is not automatically safer: false positives, incidental findings, unnecessary procedures, and overdiagnosis can create additional harm without improving outcomes.\n\nRecognition of deterioration depends on both observation and escalation. Changes in respiratory rate, mental status, blood pressure, or other findings can precede a major event. Early warning scores combine selected observations to prompt review, while rapid response systems organize assistance for patients whose condition is worsening. Their performance depends on measurement quality, appropriate thresholds, staff confidence in escalation, and the availability of a capable response. Scores supplement rather than replace clinical judgment, and their accuracy can vary between populations and care settings.\n\nTransitions create risk because information, responsibility, and treatment plans move between people or organizations. A shift handover may omit a pending investigation; discharge instructions may conflict with the prescription; a referral may not reach the receiving service. Structured handoff methods can specify illness severity, current problems, anticipated developments, and actions required. Read-back or confirmation helps establish a shared understanding. For patients moving between hospital, community services, and home, safe transfer also involves equipment, medication access, transport, and the practical ability to follow the plan. A completed discharge document is therefore only one component of continuity."
    },
    {
      "heading": "Safety culture, patient participation, and responses to harm",
      "text": "Safety culture refers to the shared values, expectations, and behaviors that influence how an organization manages risk. Its visible features include whether staff report hazards, ask for help, challenge unsafe decisions, and receive meaningful responses. Written policies may endorse openness while day-to-day practices discourage it through humiliation, retaliation, or indifference. Surveys can assess perceptions of teamwork, management support, workload, and speaking up, but scores require interpretation alongside observed behavior and working conditions.\n\nA just culture seeks fair accountability while preserving the ability to learn from error. It distinguishes inadvertent human error from risk-taking behavior and reckless disregard, considering both individual actions and the conditions shaping them. It does not mean that all conduct is acceptable or that disciplinary action is never appropriate. Conversely, automatically blaming the last person involved can conceal design failures and deter reporting. Consistency and procedural fairness matter because staff judge organizational commitments through how actual incidents are handled.\n\nPatients and families contribute knowledge that may be absent from records, including usual medication use, previous reactions, changes from baseline, and practical difficulties after discharge. Their observations can reveal misidentification, omitted information, or deterioration. Shared decision-making also supports safety by clarifying the expected benefits, harms, uncertainties, and alternatives of an intervention. Participation must not shift responsibility for safe care onto patients. Severe illness, disability, language barriers, fear, and unequal power can limit a person’s ability to question a professional or detect an error.\n\nCommunication practices therefore have an equity dimension. Professional interpretation, accessible formats, and methods that check understanding can reduce failures associated with language or health literacy. Reliance on a signature or the delivery of a leaflet does not establish comprehension. Patients with communication disabilities, cognitive impairment, or complex care arrangements may require adapted processes and involvement of authorized supporters. Measures of safety that omit these experiences can conceal differences between groups, making health equity relevant to both surveillance and improvement.\n\nWhen harm occurs, organizational responses include clinical treatment, timely communication, investigation, and practical support. Open disclosure generally involves explaining known facts, acknowledging the event, expressing regret or apology as appropriate, and describing the next steps without premature speculation. Patients and families may seek recognition and information as well as financial redress. Staff involved in serious incidents can also experience distress and impaired confidence, sometimes described by the term “second victim,” although the terminology is debated because patients and families remain those directly harmed. Support for staff and transparent accountability serve different functions and need not conflict."
    },
    {
      "heading": "Governance, improvement methods, and digital systems",
      "text": "Patient safety governance assigns authority and responsibility for identifying risks, providing resources, monitoring performance, and acting on evidence of harm. Governing bodies and senior leaders influence safety through staffing, procurement, maintenance, information infrastructure, and organizational priorities. A safety office cannot independently correct shortages or incentives controlled elsewhere. Regulatory systems may require incident notification, minimum standards, professional licensing, and oversight of medicines and devices, while accreditation assesses selected organizational processes. None of these mechanisms alone establishes that care is consistently safe.\n\nImprovement work translates an identified problem into a change that can be tested and sustained. Small-cycle methods, including plan–do–study–act, examine whether a proposed alteration works in the local setting. Statistical process-control charts can help distinguish routine variation from signals consistent with a meaningful change. A decline after an intervention is not by itself proof of causation: seasonal trends, changes in case mix, concurrent programs, and regression to the mean may explain part of the result. Randomized controlled trials, including cluster designs, provide stronger causal evidence for some interventions, though they may be difficult to conduct for organization-wide changes.\n\nEvaluation also examines unintended effects. A mandatory double check may reduce one type of medication error but delay urgent treatment or divert staff from other patients. A fall-prevention policy may reduce recorded falls while unnecessarily restricting mobility. A digital warning may improve one prescribing decision while adding enough workload to obscure another important signal. Measures therefore need to capture patient outcomes, implementation quality, resource use, and relevant balancing effects. Sustainability requires attention to staff turnover, changing equipment, and whether the intervention remains workable under routine pressures.\n\nElectronic records and decision support can improve access to information, identify contraindications, and standardize selected processes. They can also propagate copied errors, obscure important details, encourage wrong-patient selection, or make responsibility unclear across interfaces. Downtime and cybersecurity incidents can interrupt access to records, prescribing, laboratory services, and scheduling. Digital safety consequently includes usable design, access controls, interface testing, contingency procedures, and recovery planning, alongside protection of data privacy. A technically available system may still be unsafe if users cannot reliably locate the information needed for a decision.\n\nArtificial intelligence introduces additional questions of validation, oversight, and performance over time. A model may perform well in retrospective data yet fail in another institution because disease patterns, documentation, equipment, or workflows differ. Automation bias can lead users to accept an incorrect output, while excessive distrust can cause useful warnings to be ignored. Evaluation must distinguish predictive accuracy from actual effects on care and patient outcomes. Relevant safeguards include clear intended uses, external validation, monitoring for performance drift, assessment across patient groups, and explicit responsibility for reviewing and acting on outputs. Systems that generate fluent but unsupported clinical text require particular attention to factual verification and error propagation."
    }
  ],
  "links": [
    "health-system",
    "germ-theory-of-disease",
    "anesthesia",
    "world-health-organization",
    "world-health-assembly",
    "primary-health-care",
    "incidence",
    "insulin",
    "opioid",
    "hypoglycemia",
    "kidney",
    "pharmacovigilance",
    "antimicrobial-resistance",
    "blood-transfusion",
    "sepsis",
    "diagnosis",
    "overdiagnosis",
    "shared-decision-making",
    "health-literacy",
    "health-equity",
    "randomized-controlled-trial",
    "cybersecurity",
    "data-privacy",
    "artificial-intelligence"
  ],
  "sources": [],
  "license": "https://creativecommons.org/licenses/by/4.0/"
}