Psychology by BogdanovaThe Encyclopedia of Psychology for the Artificial Era

RESEARCH DOSSIER

Clinical Ethics, Iatrogenic Harm, and Patient Safety

Abstract. Clinical ethics, iatrogenic harm, and patient safety concern the ways mental health care can produce physical, psychological, relational, informational, and rights-related harm alongside intended therapeutic benefit. This critical narrative monograph traces the historical development of safety science, defines error, adverse events and preventability, and integrates ethical principles with empirical research on diagnostic uncertainty, informed consent, suicide-risk prediction, restrictive care, psychological treatment deterioration, and service transitions. It evaluates organizational error models, incident reporting, causal intervention designs, patient-defined outcomes, equity, and emerging digital clinical systems. Foundational observational studies, two large psychiatric inpatient safety reviews, contemporary systematic and scoping syntheses, and international guidance show that reliable safety evaluation requires population-specific measures, careful causal attribution, explicit patient agency, and learning across clinical and institutional boundaries. The synthesis distinguishes general medical estimates from psychiatric-specific evidence, observational associations from causal intervention effects, and rights-based commitments from empirical outcome claims. It concludes with methodological requirements for prospective, multimethod, longitudinal research and accountable safety governance.

Keywords: clinical ethics; iatrogenic harm; patient safety; mental health care; diagnostic error; informed consent; decision-making capacity; coercion; involuntary psychiatric care; restraint; patient rights; suicide risk prediction; psychotherapy adverse effects; patient-reported harm; systems safety; incident reporting; health equity; digital mental health

The Scientific Object of Clinical Ethics and Patient Safety

Clinical ethics, iatrogenic harm, and patient safety form a single research problem whenever an intervention intended to relieve suffering creates a foreseeable risk of physical injury, psychological deterioration, loss of liberty, informational injustice, or exclusion from effective care. The unit of analysis ranges from the individual encounter and the diagnostic decision to longitudinal care pathways, institutional rules, and population-level distributions of harm. A full account studies the mechanisms through which intentions are translated into clinical actions, how patients experience those actions, which outcomes are attributable to them, and which safeguards can alter the trajectory. The patient-safety literature supplies a vocabulary of adverse events, error, preventability, organizational learning, and accountable improvement; clinical ethics establishes the value commitments by which desirable outcomes and acceptable procedures are evaluated World Health Organization (2021a) Vincent et al. (1998) Donabedian (1988).

The central questions are consequential and empirically answerable in different degrees. How often does mental health care contribute to injury or deterioration? Which clinical decisions become unsafe because of uncertainty, communication breakdown, coercion, structural discrimination, or inadequate follow-up? How can a service establish causality when symptoms fluctuate independently of treatment? Which interventions reduce risk without imposing new psychological and social burdens? The answers require measurement of adverse outcomes, documentation of patient preferences, attention to counterfactual care pathways, and direct evaluation of safeguards. This dossier treats psychiatric inpatient care, ambulatory assessment, psychotherapy, medication-related care coordination, digital services, and transitions as connected environments while reserving modality-specific treatment mechanisms for their specialized research dossiers Singh et al. (2014) Crawford et al. (2016) WHO & OHCHR (2023).

The method is a critical narrative synthesis of foundational work, observational investigations, quantitative and scoping reviews, patient-reported outcome research, clinical guidelines, health-system policy documents, and methodological critiques. It interprets the scope of each evidence source according to its design and population; it supplies no pooled effect estimate, independently conducted trial, or prospective surveillance dataset. Its organizing thesis is that safety is jointly determined by the technical appropriateness of care, the conditions under which decisions are made, the patient's rights and agency, and the organization's ability to detect and correct errors. A service that improves symptoms while routinely concealing harm, denying supported decision-making, or discharging people into a fractured care system has left central dimensions of safety unmeasured Donabedian (1988) WHO & OHCHR (2023) NHS England (2022/2025).

Errors, Adverse Events, Iatrogenic Harm, and Preventability

An error concerns the planning or execution of an intended action, while an adverse event concerns harm associated with health care. A near miss is a circumstance that could have led to harm but did not; an unsafe condition is a hazard that increases risk before an event becomes observable. Iatrogenic harm is injury or other adverse consequence caused or contributed to by clinical management. This category includes complications arising under appropriate care, and its causal claim must therefore be separated from moral fault, negligence, and the preventability judgment. Historical hospital-record studies distinguish injuries attributable to medical management from the smaller subset deemed negligent, although judgments of causation and substandard care depend on reviewers, records, and prevailing standards Brennan et al. (1991) Leape et al. (1991) NHS England (2022/2025).

In mental health contexts, a wider outcome taxonomy is essential. Physical adverse events include medication complications, injuries during restraint, falls, or failure to detect a concurrent medical condition. Psychological harms include intervention-related worsening, traumatic care experiences, humiliation, loss of trust, and distress following coercion. Relational harms arise when clinical communication silences the patient or fractures important support relationships. Epistemic harms arise when a person's account is repeatedly dismissed or inaccurately encoded in the medical record, affecting subsequent decisions. Some domains are objectively observable, while others require systematic patient report; the validity of a measurement strategy depends on its match to the outcome under examination Crawford et al. (2016) Rozental et al. (2016) World Health Organization (2021b).

Preventability is a counterfactual assessment: under realistically available safer practices, could this injury have been avoided? It can be studied through case review, comparator treatments, failure-mode analysis, and natural experiments, but judgments may differ by professional background and clinical context. It is also possible for an initially justified decision to become harmful because safeguards, monitoring, or reassessment are absent. Safety science accordingly distinguishes a harmful outcome, an event causally associated with treatment, a breach of an established standard, and a preventable contribution by the organization. Conflating these categories exaggerates certainty and undermines learning. The systematic review by Panagioti and colleagues addresses preventable harm across medical care settings generally; its prevalence estimates are not a psychiatric-specific incidence measure Panagioti et al. (2019) Reason (2000).

Moral Foundations: Autonomy, Benefit, Harm, Justice, and Dignity

Clinical ethics integrates an obligation to promote well-being, a duty to reduce avoidable harm, respect for the person's authority over consequential decisions, and fair allocation of opportunities for safe care. These commitments have practical force because mental health services encounter uncertainty over capacity, predictions of future risk, and disagreements between clinicians, patients, families, and institutions. Beneficence requires meaningful therapeutic benefit; nonmaleficence directs attention to adverse consequences and burdens; autonomy concerns intelligible information and self-determination; justice addresses access, discrimination, and unequal exposure to coercion. The concepts may point toward different actions in a particular case, so ethical reasoning must identify the specific stakes and available options rather than reciting principles as if they generated mechanical answers Appelbaum (2007) WHO & OHCHR (2023).

Respect for autonomy has procedural and relational aspects. A signature on a form is a weak indicator when treatment alternatives, uncertainty, likely burdens, privacy implications, or the possibility of declining an intervention have not been meaningfully discussed. Equally, a patient who requires language interpretation, cognitive accommodations, accessible information, or supported decision-making still has preferences and values that structure responsible care. Shared decision-making models describe a process of presenting options, exchanging information, and helping the person deliberate according to what matters to them. These models provide testable process standards, although their transfer to emergency and involuntary settings requires special attention to pressure, available alternatives, and legal constraints Elwyn et al. (2012) World Health Organization (2021b).

The ethical frame extends from individual clinical choice to the architecture of institutions. Fair treatment entails access to adequate physical assessment, culturally intelligible explanations, continuity after crises, transparent mechanisms for complaint, and meaningful participation in improvement. Clinical responsibility also requires adequate staffing, supervision, and functioning handoffs; exhortations to individual compassion do not compensate for dangerous operating conditions. A rights-based service evaluates treatment outcomes alongside restriction of liberty, social participation, and patient-defined recovery. The World Health Organization and Office of the High Commissioner for Human Rights describe a policy direction toward voluntary, person-centered, community-based supports and the reduction and elimination of coercive practices in line with human rights standards WHO & OHCHR (2023) World Health Organization (2021b) World Health Organization (2021a).

Historical Development of Iatrogenesis and Safety Science

The possibility that treatment can injure is ancient, but contemporary patient-safety research became quantitatively institutionalized through record-review programs, quality assessment, and systems analysis. Avedis Donabedian's structure–process–outcome approach showed why a favorable clinical result cannot substitute for evaluating the way care is organized and delivered. The Harvard Medical Practice Study examined randomly selected historical hospital records and provided a landmark estimate of iatrogenic injury in nonpsychiatric acute care. Its significance lies in demonstrating how management-related injury can be systematically investigated, not in supplying contemporary estimates for every specialty, country, or era Donabedian (1988) Brennan et al. (1991) Leape et al. (1991).

Vincent, Taylor-Adams, and Stanhope placed adverse events within a causal network connecting patients, tasks, teams, physical conditions, organizational decisions, and institutional environments. Reason's analysis of human error likewise explained how failures at the point of care interact with latent conditions that make mistakes more likely and their consequences more severe. These perspectives changed the analytic question from which individual made the last mistake to what sequence of vulnerabilities allowed harm to reach the patient. They retained professional accountability while broadening the range of potentially effective remedies, including redesigned systems, effective communication, and detection of emerging hazards Vincent et al. (1998) Reason (2000).

Policy development subsequently linked patient safety to patient engagement and health-system governance. The Institute of Medicine's 2000 and 2001 reports focused public attention on preventable injury and broader dimensions of quality. The WHO Global Patient Safety Action Plan for 2021–2030 established an international framework for safer, respectful health systems. In psychiatry, the history is inseparable from disputes over institutional confinement, consent, rights, and the meaning of recovery; contemporary rights-based guidance changes both the evaluation criteria and the proposed direction of service reform. Historical milestones describe shifts in ideas and governance, while their empirical adequacy must still be assessed against modern studies of interventions and patient outcomes Institute of Medicine (2000) Institute of Medicine (2001) World Health Organization (2021a) WHO & OHCHR (2023).

Levels of Explanation and Causal Pathways

Patient harm develops along interacting clinical, psychological, interpersonal, organizational, and policy pathways. At the clinical level, incomplete diagnostic information or a contraindicated intervention may generate direct injury. At the cognitive level, anchoring, overconfidence, premature closure, and failure to reconsider a diagnosis can delay corrective treatment. At the interpersonal level, unequal authority, poor communication, or perceived stigma can suppress disclosure of relevant symptoms. At the organizational level, understaffing, uncoordinated records, and inaccessible aftercare permit manageable problems to become crises. A causal analysis needs explicit hypotheses for links between these levels, not a blanket assignment of every adverse outcome to clinician behavior Vincent et al. (1998) Reason (2000) Singh et al. (2014).

The distinction between proximal and systemic causes is important when interpreting prevention studies. A checklist may improve completion of an assessment yet leave the patient's ability to explain a concern unchanged. New technology may increase access to information while generating alert fatigue or false reassurance. Restricting a patient's movement may reduce one immediate exposure to risk while increasing fear, injury, and reluctance to seek future care. Such tradeoffs make benefit–harm evaluation multidimensional. A complete causal pathway specifies whose outcome is measured, over what interval, against which alternative action, and whether effects persist or are displaced to another setting Donabedian (1988) Corderoy et al. (2025) WHO & OHCHR (2023).

An overarching systems model can be represented as a sequence of exposures, clinical judgments, actions, safeguards, experiences, and outcomes with feedback. New information after a decision should update both the assessment and the safeguards; institutions should likewise revise operating rules when events reveal predictable failure modes. Because psychiatric symptoms and risks vary over time, claims about treatment-produced deterioration demand temporal information and plausible comparators. The evidence is particularly vulnerable to confounding by indication, whereby people in more severe crisis receive more intensive or compulsory care and subsequently have different outcomes for reasons preceding the intervention Corderoy et al. (2025) Panagioti et al. (2019).

Source Hierarchy and the Problem of Measurement

An original trial establishes a different kind of inference from a retrospective chart audit, qualitative interview, administrative dataset, cross-sectional patient survey, or expert recommendation. Randomized allocation can strengthen causal inference for an evaluated intervention but often excludes highly unstable patients or cannot be ethically implemented for coercive practices. Administrative records cover large service populations but usually reflect events that staff recognized and coded. Direct patient interviews reveal harms absent from institutional reporting, although recall, social desirability, and attribution uncertainty must be considered. A rigorous synthesis triangulates these evidential windows while preserving the distinctive blind spots of each Crawford et al. (2016) Honkalampi et al. (2025) Corderoy et al. (2025).

Operational definitions deserve particular attention. Symptom worsening measured on a validated longitudinal scale differs from a transient emotional reaction, an unwanted experience, failure to improve, a new diagnosis, or an adverse event causally attributable to a treatment. Similarly, high-risk labeling is a prediction rather than an observed outcome; safety planning is a process that requires an independent evaluation of results. Surveillance rates depend on denominators and observation periods, so a percentage per inpatient admission cannot be compared directly with an estimate per outpatient visit or person-year. Sampling design, case-mix adjustment, missing data, and between-service variability influence all claims of comparative safety Rozental et al. (2016) Rozental & Řiháček (2025) Carter et al. (2017).

Research quality therefore depends on a hierarchy of linked questions: whether the outcome is validly defined; whether its detection is systematic; whether the intervention plausibly preceded it; whether competing explanations have been addressed; and whether the estimates generalize to the population and setting under discussion. This dossier uses source-linked findings and explicit limits rather than combining heterogeneous harm estimates into a single apparent prevalence of psychiatric iatrogenesis. Patient safety is also a normative construct: a purely epidemiological account will miss the consequences of coercion, disrespect, and unequal decision-making that patients consider central to safe and acceptable care World Health Organization (2021b) Donabedian (1988) NHS England (2022/2025).

Diagnostic Error and the Dynamics of Clinical Uncertainty

Diagnostic safety concerns both the accuracy of clinical classification and the consequences of acting on a provisional explanation. Psychiatric presentations are heterogeneous, often evolve over time, and overlap with neurological, endocrine, substance-related, medication-induced, and social conditions. A diagnosis may support coordinated treatment, yet premature certainty can transform a working hypothesis into an enduring label that reshapes access, stigma, medication exposure, and how subsequent clinicians interpret new evidence. The National Academies' diagnostic-safety framework emphasizes the diagnostic process as a collaborative, longitudinal information-gathering activity rather than the isolated assignment of a code. Failures may arise during history taking, testing, follow-up, communication, or reinterpretation after a patient's condition changes National Academies (2015) Singh et al. (2014).

Historical epidemiology illustrates the scale of the general problem without directly estimating psychiatric error. Singh, Meyer, and Thomas combined three observational approaches to estimate diagnostic errors in approximately 5.08% of U.S. adults receiving outpatient care, with explicitly modeled assumptions and uncertainty. Those data were not a national census of misdiagnoses in mental health clinics. Their transferable insight is methodological: diagnostic error remains difficult to detect when the expected correct answer is only visible after further evaluation and when records do not capture uncompleted recommendations or lost follow-up. Clinical psychology requires additional caution because diagnostic classification is not equivalent to a biological gold standard, and interpretive disagreements may arise even under reasonable practice Singh et al. (2014) National Academies (2015).

Safer diagnostic reasoning uses longitudinal formulation, explicit uncertainty, review of physical and medication histories, and accessible routes for reassessment when the evidence changes. Differential diagnosis should include adverse effects of prior treatment, physical illness, trauma, environmental stressors, and a patient's own account of meaningful symptoms. Case conferences and second opinions can help, although their efficacy depends on independence, adequate information, and the absence of shared anchoring. A safety-oriented record distinguishes observed facts, the patient's reported experiences, hypotheses, and decisions. This practice reduces the chance that inherited assumptions are silently converted into apparently established findings across repeated encounters National Academies (2015) Vincent et al. (1998).

Cognitive Bias, Clinical Judgment, and Epistemic Authority

Clinical reasoning is vulnerable to selective attention, confirmation bias, diagnostic momentum, overreliance on familiar categories, and failures to update hypotheses. These tendencies do not by themselves establish misconduct or identify a specific causal error, because every clinical decision must occur under incomplete information. Their importance arises when processes reward speed or certainty at the expense of alternative hypotheses. A patient whose atypical symptoms are repeatedly interpreted only through a previous psychiatric diagnosis may receive less investigation of new physical problems. The analytic task is to reconstruct a decision sequence and show how information was collected, weighted, communicated, and revised National Academies (2015) Reason (2000).

Epistemic authority is distributed unevenly in clinical encounters. Clinicians contribute technical expertise, while patients contribute first-person experience, symptom history, goals, values, and knowledge of changes outside the clinic. Neither perspective is sufficient alone; the relationship requires procedures through which conflicting accounts can be examined without routine dismissal. Supported decision-making and effective informed consent provide one route to recognizing the patient's interpretive contribution. In psychological treatment, the experience of not knowing what intervention is being provided has been associated with reports of negative effects, but the cross-sectional design of that finding does not establish that poor information caused those effects Crawford et al. (2016) Elwyn et al. (2012) WHO & OHCHR (2023).

Improvement initiatives should test behavioral processes rather than merely educate clinicians about biases. Prompts for alternative hypotheses, review of unresolved test results, callbacks after acute visits, and structured opportunities for patient correction may reduce some pathways to error; any claimed benefit requires outcome evaluation in the relevant setting. Recording diagnostic uncertainty can also have costs if poorly explained to patients or used as an excuse for inaction. The productive balance is transparent provisionality coupled with accountable follow-up. The principle applies equally to human judgments and decisions supported by computational systems, whose apparent precision can intensify unwarranted confidence National Academies (2015) NHS England (2022/2025).

Suicide and Self-Harm Risk: Prediction, Formulation, and Safety

Suicide and self-harm safety decisions are high-stakes because false reassurance may contribute to harm while indiscriminate restrictions may cause trauma, exclusion, and loss of trust. Risk factors associated with future events in groups do not establish accurate prediction for an individual. Prediction depends on base rates, population, observation interval, model calibration, and the consequences of false positives and false negatives. The systematic review by Carter and colleagues found that pooled positive predictive value for suicide across studied risk scales was 5.5%, with substantial limits imposed by heterogeneity and rarity of the outcome. That estimate characterizes evaluated instruments and populations, not a universal performance guarantee for every clinical setting Carter et al. (2017).

England's NICE guideline NG225 explicitly recommends against using risk assessment scales to predict future suicide or repetition of self-harm, and against using global low-, medium-, or high-risk strata to determine treatment access or discharge. Instead it calls for attention to immediate and longer-term needs, collaborative assessment, and a clinical risk formulation as part of psychosocial evaluation. This guidance changes the meaning of safety work: the purpose of assessment is to understand the person and organize useful care, not to create a deceptively precise category that substitutes for action. The guideline is a recommendation for the population and health system it addresses, so other jurisdictions require their own legal and practice review NICE (2022) Carter et al. (2017).

A meaningful risk formulation integrates current distress, precipitants, relevant history, protective resources, access to support, and expected changes in circumstances. Its output should guide a coherent response with the person, including physical and psychological care, agreed follow-up, communication across services, and review when new information appears. Research needs to measure actual outcomes, patient understanding, safety planning fidelity, and adverse effects of restrictive management. Adverse events following an apparently low-risk categorization do not automatically establish that the prediction procedure was negligent; likewise a short-term absence of self-harm after restrictive care cannot prove that the restriction caused safety. Both interpretations demand counterfactual and longitudinal evidence NICE (2022) Corderoy et al. (2025) National Academies (2015).

Crisis Decisions, Admission, Discharge, and Continuity of Care

The transitions surrounding crisis services are vulnerable points in mental health care. Information may be incomplete on presentation, an immediate medical need may overshadow psychosocial evaluation, and discharge can separate a patient from professionals who know their circumstances. Admission is a major intervention with potential benefits and burdens that depend on available outpatient alternatives and the conditions of inpatient care. Discharge is likewise a clinical process rather than a purely administrative event: the safety question concerns the plan for follow-up, treatment access, medication reconciliation when relevant, and communication of what should happen if the situation changes NICE (2022) World Health Organization (2021b).

NICE NG225 specifies psychosocial assessment following self-harm and directs clinicians to consider the person's immediate circumstances and support network; it does not endorse mechanically allocating care through a numerical risk threshold. These recommendations illustrate the need to evaluate the sequence of care, including physical treatment and postdischarge arrangements, rather than judging one clinician's risk phrase in isolation. A hospital may have strong internal procedures yet expose patients to foreseeable harm when referrals are delayed, records unavailable, or responsibility for continuing care ambiguous. The organizational approach of Vincent and colleagues offers a framework for tracing these system connections NICE (2022) Vincent et al. (1998).

Evidence on safer transitions should identify the time period under surveillance, competing pathways, and implementation conditions. For example, an intervention that increases appointments offered after discharge may not increase appointments attended, and improved attendance may not translate into a demonstrable reduction in serious harm. Patients' experience of crisis admission and discharge can also affect willingness to return for help, a pathway often missing from routine administrative outcomes. A rigorous service evaluation therefore considers both clinical endpoints and patients' lived experience, along with equity of access and the possibility that harms shift to families, emergency departments, or community agencies World Health Organization (2021b) Donabedian (1988) NHS England (2022/2025).

Confidentiality, Documentation, and the Safety of Information

Accurate information sharing can reduce medication errors, missed diagnoses, and unsafe transitions, while inappropriate disclosure can damage trust, expose vulnerable persons, and discourage future care. Psychiatric records frequently contain sensitive experiences, third-party information, and interpretive notes that may persist across years. Clinical ethics therefore requires deliberate decisions about what is documented, how certainty is expressed, who may access it, and how inaccuracies can be corrected. The quality of health information is simultaneously a technical safety property and an issue of dignity, autonomy, and justice National Academies (2015) WHO & OHCHR (2023).

Confidentiality has justified exceptions under particular laws and circumstances, but the threshold and procedural protections vary substantially. Ethical assessment identifies the purpose of disclosure, whether it is necessary and proportionate, available alternatives, and how the person's views will be heard. Clinical teams should distinguish an explicit patient statement, a third-party allegation, professional inference, and an established fact. Failure to maintain those distinctions can propagate diagnostic momentum and injurious characterizations through handoffs and future assessments. Information design is consequently a part of diagnostic safety, not an ancillary administrative task National Academies (2015) Appelbaum (2007).

Digital platforms introduce additional pathways for leakage, unauthorized secondary use, misidentification, and the automated amplification of inaccurate notes. New data sharing tools should be studied for access control, auditability, correction procedures, incident response, and the possibility that clinician time saved in one process increases harm through mistaken information elsewhere. Decisions about machine-supported summaries require verification against original clinical evidence and should preserve the patient's agency to contest inaccuracies. These are plausible risk pathways and design requirements, not evidence that every automated clinical document is unsafe. The appropriate test combines security evaluation, workflow observation, measured errors, and patient-reported effects NHS England (2022/2025) World Health Organization (2021a).

Involuntary Treatment: Competing Aims, Evidence, and Rights

Involuntary psychiatric admission illustrates the intersection of care, public safety claims, institutional authority, and fundamental rights. Decisions are made under conditions of severe distress and uncertainty, often involving disagreement about treatment need, decision-making capacity, or immediate danger. The relevant outcomes include symptom change, physical injuries, duration of restriction, perceived coercion, confidence in services, social disruption, readmission, and long-term help-seeking. These outcomes belong to different moral and methodological domains; an apparent gain on a clinical symptom scale cannot by itself settle the ethical appraisal of confinement, and an adverse patient experience does not automatically establish that treatment failed to reduce an immediate risk Corderoy et al. (2025) WHO & OHCHR (2023).

Corderoy and colleagues' scoping review compared evidence on involuntary and voluntary inpatient populations and identified patterns of greater reported coercion, increased resource costs, and lower patient satisfaction associated with involuntary admission. Some comparisons indicated larger improvements in symptoms and functioning among involuntary patients, but those patients may have entered hospital with more severe symptoms, weakening simple causal interpretation. The review's findings must be read as heterogeneous, largely observational comparisons shaped by legal definitions, selection into care, baseline crisis severity, and outcome measurement. It supports simultaneous attention to benefits and burdens rather than a simple causal ranking of all admission pathways Corderoy et al. (2025).

The WHO–OHCHR rights guidance advocates reform toward voluntary, person-centered services and the removal of coercion, while clinical studies continue to examine how current compulsory systems affect outcomes. These materials operate at different levels: human rights guidance expresses normative and legal design aims; a quantitative comparison estimates observed associations under particular arrangements. Research should document actual available alternatives, rights information, independent review, the duration of restrictions, patient-defined outcomes, and postdischarge consequences. The resulting evidence can assess whether improved access to community support reduces crisis admissions and restriction without transferring risk elsewhere WHO & OHCHR (2023) World Health Organization (2021b) Corderoy et al. (2025).

Restraint, Seclusion, and Institutional Violence Prevention

Physical restraint and seclusion involve concentrated institutional control over an individual and raise immediate concerns about bodily injury, psychological trauma, dignity, and disproportionate use. Evaluating these practices requires more than counting episodes. Important denominators include occupied bed-days, patient admissions, and actual exposure time, together with severity of injuries and the circumstances surrounding each use. Changes in documentation rules, staffing, diagnostic case mix, and thresholds for incident reporting can alter recorded rates without a corresponding improvement in the experience of care. The rights-based safety perspective prioritizes alternatives that preserve liberty and meaningful engagement while maintaining access to urgent clinical attention WHO & OHCHR (2023) World Health Organization (2021b).

A systematic review by Pedersen and colleagues investigated interventions intended to reduce mechanical restraint in adult mental health inpatient settings. Studies in this field often combine staff training, policy change, environmental redesign, and de-escalation protocols, making the active ingredients difficult to isolate. Observed reductions may reflect improvements, changes in case mix, shifts toward other restrictive practices, or altered reporting. A causally informative evaluation therefore monitors injuries, repeated use per individual, substitute practices, perceived coercion, and patient-reported safety, using appropriate comparison conditions where available Pedersen et al. (2024) NHS England (2022/2025).

A service's response to aggression or extreme distress must consider early identification, staff communication, trauma-sensitive support, environmental pressures, and transparent incident review. Decisions taken under pressure should remain open to review by people affected and should be documented with clear observations, alternatives considered, and the factual circumstances. Organizational safety programs that prioritize a single target, such as fewer recorded seclusion episodes, may unintentionally incentivize informal coercion or concealment. Monitoring a balanced set of physical, psychological, relational, and rights-related outcomes offers a stronger account of harm reduction Reason (2000) WHO & OHCHR (2023) NHS England (2022/2025).

Psychotherapy-Associated Deterioration and Adverse Experiences

Psychotherapy is a major evidence-based element of mental health treatment, and its ethical evaluation includes the possibility of deterioration, unwanted experiences, broken confidentiality, dependency, and inappropriate handling of trauma or relationships. Adverse change can occur through an intervention, through the condition's natural course, through contextual factors, or through some combination of these. Research must distinguish an unpleasant yet potentially useful therapeutic experience from an enduring deterioration or an independently avoidable adverse event. Lilienfeld's discussion of potentially harmful psychological treatments helped establish that treatment-related injury deserves systematic investigation even when a modality is ordinarily considered noninvasive Lilienfeld (2007) Rozental et al. (2016).

Crawford and colleagues' cross-sectional survey of 14,587 people using psychological treatment in England and Wales found that 763 respondents, or 5.2%, reported lasting bad effects. Those reports are vital evidence about patient experience, but neither randomization nor an untreated counterfactual was built into the survey. The proportion is therefore not a causal prevalence estimate for psychotherapy-induced injury across services or countries. Participants who felt they had been given adequate information were less likely to report lasting adverse effects, an association that could reflect treatment communication, patient characteristics, retrospective appraisal, or other common causes Crawford et al. (2016).

Honkalampi, Urhonen, and Virtanen systematically reviewed negative-effect reporting in randomized studies of psychotherapies and psychological interventions. Rozental and colleagues developed the Negative Effects Questionnaire to provide an instrument for describing adverse and unwanted events, while later methodological reviews stress that instruments, study protocols, and reporting practices remain variable. A rigorous clinical safety program should measure deterioration on validated outcomes, unsolicited adverse experiences, early dropout, burden, and patient-defined harms using explicit time points and agreed procedures. This module summarizes cross-modal safety issues; mechanism-specific comparative analysis belongs to the psychotherapy adverse-outcome research dossier Honkalampi et al. (2025) Rozental et al. (2016) Rozental & Řiháček (2025) Strauss et al. (2026).

Medication Safety as a Cross-Modal Clinical Duty

Psychopharmacological safety includes evidence about drug indications, adverse reactions, interactions, monitoring, withdrawal, and regulatory status, which require dedicated pharmaceutical expertise. The present dossier investigates the interfaces where such risks become patient-safety failures: incomplete reconciliation of treatment history, poor explanation of alternatives, inadequate access to medical evaluation, unsafe transfer between prescribers, and failure to respond to reported adverse experiences. Medication safety and clinical ethics are joined when a patient cannot meaningfully understand the anticipated benefit, relevant uncertainty, or foreseeable burden of a proposed plan World Health Organization (2021a) Appelbaum (2007).

Complex psychiatric care often involves multiple simultaneous medications, physical conditions, intermittent adherence, and changes in prescribers or service settings. Causal interpretation of an adverse event depends on temporal relationships, alternative explanations, dose and exposure information, and the limitations of routine incident reporting. A patient-safety review should ask whether potential interactions were evaluated, whether an updated medication history was available, and whether monitoring and communication operated as planned. It cannot infer a particular pharmacological mechanism merely because an adverse symptom emerged after a prescription; the underlying medicine-specific evidence belongs to the separate Psychopharmacology Science and its specialized dossiers Panagioti et al. (2019) World Health Organization (2021a).

Ethical performance is therefore assessed at several points in the care pathway: selection of an intervention after considering the person's values and current evidence; explanation of alternatives; ability to report symptoms without dismissal; planned monitoring; and timely response to a suspected problem. Even if every individual prescriber meets narrow technical standards, discontinuous records and uncertain organizational responsibility can produce avoidable harm. This systems account avoids substituting generic claims about medication risk for the source-specific analyses required by actual drugs and populations Reason (2000) Vincent et al. (1998) NHS England (2022/2025).

Therapeutic Relationships, Professional Power, and Trust

The therapeutic relationship is a mechanism through which clinical information is elicited, difficult choices are negotiated, and safety signals become visible. Trust can support disclosure, engagement, and adherence to mutually understood plans, while breaches of confidence, stigmatizing language, dismissal, and coercive pressure can inhibit future contact. These effects require analysis at both interpersonal and institutional levels. Patients may report that an interaction was harmful even when symptom records show an apparent short-term improvement, and this discrepancy should trigger inquiry into the validity of the outcome measures rather than automatic dismissal of either source Crawford et al. (2016) Donabedian (1988).

Professional boundaries protect patients against exploitation and conflicts of interest but cannot be reduced to rule compliance detached from context. The present research object focuses on how relationships shape the detection, experience, and prevention of system-wide harm. Detailed boundary violations, therapist misconduct, and countertransference-related risks belong in a neighboring specialist dossier. Here the relevant questions are whether a service makes complaints safe, whether a patient can change a clinician or seek a second opinion, and how accountability persists when institutional loyalty discourages reporting NHS England (2022/2025) WHO & OHCHR (2023).

Patient-centered language and supported dialogue constitute operational practices that can be evaluated. For example, a service may invite correction of written summaries, offer interpreters, document agreed goals, and include patients in reviews of adverse events. A procedure that invites feedback but provides no response can deepen mistrust; the decisive metric is how reported concerns alter decisions and safeguards. Transparent explanation, listening, and independent channels for challenge must therefore be studied alongside clinical outcomes, not treated as optional attributes of a pleasant encounter Gallagher et al. (2003) Elwyn et al. (2012) World Health Organization (2021b).

Patient-Defined Harm, Recovery, and Longitudinal Consequences

The severity and significance of iatrogenic harm cannot be represented entirely by acute symptom scores or hospital incident counts. A restrictive intervention may affect housing, employment, social identity, family relationships, and the willingness to contact mental health services in a later crisis. An unexpected disclosure of confidential information may create a prolonged social risk even after immediate clinical symptoms subside. Conversely, a patient may report substantial clinical benefit from a difficult encounter and still reasonably request correction of the way it was conducted. Evaluating safety therefore requires outcome domains defined in partnership with patients and observed across meaningful time periods World Health Organization (2021b) WHO & OHCHR (2023).

Patient-reported adverse experiences, satisfaction, quality of life, functioning, and recovery goals give access to dimensions that clinical records often miss. These measures have limitations: the timing of evaluation affects recall, the meaning of an item may differ across cultures and care settings, and satisfaction can change with expectations or severity of illness. Triangulation with independently rated outcomes, administrative events, and prospective accounts helps determine which findings are robust. Psychological safety must be measured prospectively when possible, especially where adverse experiences alter patients' willingness to return and hence who is later represented in follow-up data Crawford et al. (2016) Rozental & Řiháček (2025).

Longitudinal models should represent treatment initiation, deterioration, improvement, discharge, recurrence, and exposure to subsequent services as a sequence of states. Comparing only people who completed treatment can conceal those who withdrew because of distress or loss of trust. Similarly, a service may report lower readmission rates because patients face barriers to reentry rather than because they are healthier. Methodologically responsible analysis specifies missingness, attrition, censoring, competing outcomes, and the social consequences of care. This allows safety science to measure a person's continuing opportunity to obtain effective help, an outcome central to recovery-oriented policy World Health Organization (2021b) Donabedian (1988) National Academies (2015).

Safety Culture, Organizational Design, and Accountability

Organizational culture influences whether clinicians disclose uncertainty, report near misses, challenge unsafe decisions, and learn from complaints. Safety culture is observable through patterns of communication, supervisory behavior, staffing practices, incident response, and the consequences that follow speaking up. A service in which errors are treated only as evidence of individual failure can discourage early reporting, while a service that treats every mistake as system-produced may weaken accountability for reckless behavior. The organizational aim is proportionate responsibility coupled with an analysis of the conditions that enabled injury. Reason's systems approach and Vincent's clinical risk framework provide durable explanations for this relationship Reason (2000) Vincent et al. (1998).

In mental health services, safety culture also includes the response to contested coercive practices and the extent to which patients and peer representatives participate in reviewing incidents. The institutional definition of an incident may exclude distress or threats to dignity that patients regard as substantial harms. Adding patient-defined outcomes changes what becomes visible and what staff are encouraged to prevent. The WHO Global Patient Safety Action Plan endorses learning and patient engagement across health systems, while rights-based mental health guidance requires services to examine structural conditions that produce restriction and exclusion World Health Organization (2021a) WHO & OHCHR (2023).

Accountability must remain traceable. An incident analysis should identify which decisions were taken, which information was available, what standards applied, and which remedial actions were assigned and later checked. Multiple individuals may have acted reasonably within an unsafe environment, while some events may also involve serious breaches requiring appropriate professional response. A credible institution makes those distinctions explicit and evaluates whether remedies address the actual mechanisms that caused harm. Training counts, policy circulation, and completed incident forms are implementation measures, not proof of safer patient outcomes NHS England (2022/2025) Reason (2000) Donabedian (1988).

Incident Reporting, Detection, and Underascertainment

Incident-reporting systems capture observations selected by institutional awareness, definitions, incentives, and time available for documentation. They are useful for learning about failure modes but cannot be treated as complete population registries of harm. Underreporting is particularly likely for delayed psychological deterioration, patient experiences of coercion, and harms occurring after discharge. Conversely, a rise in recorded incidents following the introduction of an accessible reporting channel can indicate better detection rather than deteriorating care. Changes in reported rates need interpretation alongside independent chart review, patient interviews, and stable definitions NHS England (2022/2025) Panagioti et al. (2019).

Trigger tools search records for features indicating possible adverse events, while prospective surveillance systematically asks what occurred in defined groups over specified periods. Chart audits require explicit adjudication criteria and independent reviewers when feasible; agreement between reviewers is itself a measurable property. Patient reports can reveal harms invisible to coded records but require distinctions among experienced distress, suspected treatment contribution, and established adverse events. Cross-source reconciliation should quantify overlap and disagreement rather than treating one source as the unquestioned truth Brennan et al. (1991) Crawford et al. (2016) Rozental et al. (2016).

A mature reporting program supports learning across teams while protecting confidentiality and meaningful participation by people affected. Reports should be classified by severity, type, setting, preventability assessment, and relevant service transitions. Monitoring should detect recurring processes and environmental risk factors rather than only isolated dramatic cases. The NHS Patient Safety Incident Response Framework explicitly connects proportionate incident responses with learning and improvement, and its guidance illustrates why an organizational response is evaluated by its ability to change risk, not by its capacity to produce longer investigations NHS England (2022/2025).

Causal Evaluation of Safety Interventions

Evaluating an intervention intended to reduce harm requires clarity about its causal mechanism, comparator, implementation fidelity, and target outcomes. A change may reduce reported events because of altered reporting, staff behavior, patient selection, or real improvement. Randomized trials can reduce confounding when feasible, while interrupted time-series analyses, controlled before–after comparisons, difference-in-differences, and natural experiments may provide evidence in complex services. Each design has assumptions regarding temporal trends, exchangeability, missingness, and measurement. A credible evaluation documents those assumptions and tests alternative explanations rather than attributing every favorable trend to the intervention Donabedian (1988) Panagioti et al. (2019) NHS England (2022/2025).

Safety initiatives often have multiple ingredients: staff education, new forms, a more accessible service, multidisciplinary review, and changes to escalation rules. When implemented together, the contribution of any component is difficult to isolate. Such programs may also alter treatment intensity or redistribute patients across institutions, affecting who appears in the monitored denominator. Designs should anticipate spillovers, learning effects, and unequal uptake. For coercion-reduction programs, the measured endpoints need to include substitution toward other forms of restriction and patient experience, not only the targeted procedure's frequency Pedersen et al. (2024) Corderoy et al. (2025).

One defensible evaluation strategy predefines desired outcomes, possible adverse consequences, measures of implementation, and a time horizon. It reports uncertainty intervals and case-mix information, separates primary from exploratory analyses, and checks whether results hold across demographic and clinical groups. It also evaluates whether benefits are maintained after resources or intensive supervision change. The ethical case for an intervention may be compelling even when its evidence base is developing, but stronger causal claims require correspondingly stronger designs. Transparency about this gradient of certainty supports better future research and responsible policy World Health Organization (2021a) National Academies (2015).

Quality Indicators and Balanced Safety Measurement

Donabedian's structure, process, and outcome distinctions provide a coherent architecture for evaluating mental health service safety. Structural indicators include staffing, supervision, access to medical consultation, physical environments, and continuity infrastructure. Process indicators include assessment completion, informed-choice documentation, medication reconciliation, timely communication, and involvement in discharge planning. Outcome indicators include verified adverse events, injuries, treatment deterioration, patient-reported harms, readmission, functioning, and restoration of meaningful opportunities for care. These classes are linked but are not interchangeable: a reliably completed process may fail to improve its intended outcome Donabedian (1988).

Balanced measurement avoids the temptation to reward a single easily available number. Lower inpatient admission rates may reflect better community support or diminished access; shorter stays may indicate effective care or premature discharge; fewer recorded restraint incidents may reflect safer practice or changing definitions. Complementary measures should examine both expected benefits and potential transfers of harm. Patient safety dashboards are more informative when they specify denominators, time periods, missing data, and uncertainty and when they allow review of individual trajectories behind aggregate trends World Health Organization (2021a) NHS England (2022/2025).

Indicators must be assessed for reliability, validity, feasibility, and susceptibility to gaming. Interrater agreement matters for reviews of preventability, while the responsiveness of patient-report scales affects evaluation of changes over time. Comparison between institutions requires attention to case mix, socioeconomic conditions, health-system incentives, and reporting culture. The operational definition of improvement should be negotiated openly, with explicit safeguards against lowering recorded harm by excluding difficult patients or discouraging complaints. An ethically credible indicator system invites scrutiny from the people whose experiences it claims to represent WHO & OHCHR (2023) Crawford et al. (2016).

Open Disclosure, Complaint Resolution, and Learning After Harm

When patients experience an adverse event, clinical obligations extend to recognition, timely explanation, access to appropriate follow-up, and correction of processes that remain hazardous. Open disclosure is a communication practice and an accountability process: it concerns what is known, what remains uncertain, what support can be offered, and how the patient can participate in subsequent review. Gallagher and colleagues' work on the attitudes of patients and physicians toward error disclosure documents the central relevance of patients' expectations and professional barriers. The study establishes reported attitudes in a particular research context; specific disclosure duties remain shaped by local law and professional policy Gallagher et al. (2003).

In mental health care, an explanation must take account of the patient's emotional state, possible previous experiences of involuntary treatment, privacy needs, and available supporters. A purely procedural apology may cause further injury when it minimizes the person's account or provides no route to correct the record. Equally, a declaration that an error occurred before facts have been established can misrepresent causal uncertainty. A responsible response distinguishes confirmed information from provisional findings, communicates in accessible language, and gives clear routes for further questions and independent review NHS England (2022/2025) WHO & OHCHR (2023).

Incident learning must extend beyond communication to evaluation of completed corrective actions. Organizations can monitor the timeliness and comprehensibility of disclosures, reported resolution, recurrence of analogous events, and whether patients perceive that their input informed improvement. Complaints are an important source of detection, though reliance on complaints alone selectively captures people able and willing to navigate formal processes. Accessible reporting and independent advocacy may broaden visibility. The strongest ethical repair combines honest explanation with demonstrable changes to the conditions that contributed to harm NHS England (2022/2025) Vincent et al. (1998).

Safety at the Boundaries of Services and Jurisdictions

Many harms arise between organizational units rather than within a single therapeutic technique. Emergency departments, inpatient teams, community clinicians, primary care, social services, pharmacies, and digital platforms may hold different portions of an individual's history. Transitions introduce opportunities for lost test results, conflicting treatment plans, unclear responsibility, and failures to reach people after acute episodes. Studying one setting without its connections obscures the actual causal pathway. A safe system therefore requires continuity of relevant information, agreed responsibilities, accessible reassessment, and effective follow-up National Academies (2015) Vincent et al. (1998).

Jurisdiction adds another dimension. Standards for consent, involuntary care, confidentiality, duty to protect, and access to records vary among countries and may change over time. International WHO guidance articulates rights-based directions for policy, while local legal obligations must be verified from applicable legislation and regulatory materials before clinical implementation. This dossier treats legal differences as an object of comparative investigation; it does not supply legal advice or assert that a single rule governs all readers. Scientific conclusions about harms can travel across borders only when populations, intervention definitions, and institutional conditions are sufficiently comparable WHO & OHCHR (2023) World Health Organization (2021b).

Cross-system evaluation asks whether a safety intervention in one institution shifts people or costs to a different provider and whether vulnerable groups lose access when services narrow eligibility. Measures of continuity should include successful follow-up contact, availability of understandable care plans, patient experience of transitions, and major outcomes over a consistent period. The pathway from better handoffs to reduced injury needs empirical evaluation rather than assumption. A systems model that reaches across institutional boundaries also offers a practical basis for investigating clinical care delivered through emerging human–artificial arrangements World Health Organization (2021a) NHS England (2022/2025).

Equity, Cultural Context, and Differential Exposure to Harm

Clinical harm is unequally distributed when barriers to care, discrimination, language differences, service geography, disability access, and economic insecurity change the conditions of assessment and treatment. A service may follow the same formal procedure for everyone while offering materially different opportunities to understand decisions or obtain follow-up. Equity research therefore asks which groups encounter particular interventions, how harm is detected, whether patients' accounts are credited, and who has meaningful access to alternatives. These questions require stratified analyses and attention to the mechanisms linking social position, institutional rules, and patient outcomes WHO & OHCHR (2023) World Health Organization (2021b).

Claims of differential treatment must be evaluated with appropriate denominators and potential confounders. Higher use of compulsory pathways in a group does not alone identify the full causal contribution of clinician prejudice, legal criteria, crisis-service access, differences in presentation, or prior care experiences. Nevertheless, unequal patterns signal a safety and justice question requiring investigation. Research can combine linked population data, qualitative accounts, chart reviews, and comparisons across otherwise similar services. Restricting inquiry to recorded clinical severity risks overlooking how exclusion and delayed help shape the severity observed at admission Corderoy et al. (2025) Donabedian (1988).

Patient-defined measures must be culturally interpretable. A questionnaire developed for one service population may fail to capture experiences of language minorities, neurodivergent persons, or groups whose expectations about care differ. Translation, cognitive interviewing, participation of service users, and tests of measurement invariance can strengthen comparisons. Clinical ethics requires that care remain intelligible and responsive while evidence about a specific intervention is still accumulating. In operational terms, safety is assessed not just by the average effect but by which subgroups receive benefit, who bears disproportionate burdens, and whether proposed improvements reduce or widen disparities World Health Organization (2021b) World Health Organization (2021a).

Developmental and Family Contexts of Clinical Safety

Children, adolescents, and older adults may face distinct configurations of decision-making authority, communication support, developmental vulnerability, and dependence on caregivers. These contexts affect how a service establishes consent or assent, assesses capacity, responds to safety concerns, and balances confidentiality with involvement of significant others. A developmental account avoids treating chronological age as a universal substitute for the capacity to understand a particular decision. It evaluates the person's actual abilities and needs together with applicable legal rules and the effects of family and institutional arrangements Appelbaum (2007) WHO & OHCHR (2023).

For younger patients, adverse consequences of labeling or poorly explained compulsory intervention may extend through education, peer relationships, family trust, and later willingness to seek help. For older adults, sensory impairments, medical comorbidity, polypharmacy, and access barriers can complicate differential diagnosis and communication. Empirical safety studies should report age groups, family and caregiver involvement, settings, and whether the intervention was developmentally adapted. It is unsafe to assume that a result obtained from one age group predicts benefit or harm in another without appropriate evidence National Academies (2015) World Health Organization (2021b).

Family members and caregivers may provide essential longitudinal information and practical support; they may also hold views that conflict with the patient's wishes or well-being. A structured care process identifies their role, manages confidentiality, and considers the patient's preferences and legal entitlements. The relevant safety outcomes include the person's functioning, experienced coercion, understanding of plans, family burden, and continuity of support. These are interdependent observations that require careful causal interpretation. A service is more likely to learn from adverse events when the voices of affected individuals, including those requiring communication accommodations, are substantively represented Elwyn et al. (2012) WHO & OHCHR (2023).

Digital Mental Health and Human–Artificial Clinical Systems

Digital clinical tools can support appointment access, symptom reporting, record review, structured assessment, and communication, while introducing vulnerabilities associated with privacy, incomplete context, automated classification, and difficult-to-contest recommendations. The clinical ethics of these tools depends on their actual function and deployment. A reminder service, decision-support application, automated psychotherapy interface, and generative conversational system present different risks and different evidence requirements. Claims of clinical efficacy, safety, or diagnostic reliability must therefore refer to the evaluated system, intended users, comparator, and specific version rather than to artificial intelligence as an undifferentiated category National Academies (2015) World Health Organization (2021a).

Human–artificial interaction changes responsibility distribution when users attribute professional authority to generated responses or clinicians rely on automated summaries in high-stakes choices. A system may appear coherent despite missing patient history, producing plausible but inaccurate explanations that are difficult to detect without source validation. Risk can also arise through escalating emotional reliance or unauthorized storage of intimate disclosures. These pathways require direct empirical measurement in clinical populations, including prospective monitoring of errors, crisis escalation, informed-choice comprehension, confidentiality incidents, and user-reported adverse experiences. Technical performance alone is a limited proxy for patient safety NHS England (2022/2025) National Academies (2015).

In the Artificial Era, clinical governance must make the respective responsibilities of human professionals, institutions, and artificial systems explicit. Conceptual work on artificial agency or human–artificial relationships can inform hypotheses, but philosophical attribution does not establish therapeutic competence or a right to delegate irreversible decisions. The clinical test is whether deployment produces demonstrable benefits without unacceptable physical, psychological, informational, or rights-related harm, under independent evaluation and meaningful patient choice. Audit trails, human review, version-specific monitoring, and withdrawal mechanisms are plausible safeguards whose actual effectiveness must be tested rather than presumed World Health Organization (2021a) WHO & OHCHR (2023).

Ethical Conflicts, Proportionality, and Competing Outcomes

Clinical safety decisions commonly involve several legitimate objectives that cannot be reduced to a single metric. Reducing immediate physical danger, respecting liberty, supporting long-term recovery, preserving trust, and ensuring fairness may demand different priorities at different moments. Ethical deliberation must identify the decision actually at issue, feasible alternatives, the evidence supporting expected benefit, the burdens imposed on the person, and the opportunities for reassessment. It should also make explicit who is authorized to decide and by which legal process, since technical expertise does not by itself confer unlimited authority over a person's life Appelbaum (2007) WHO & OHCHR (2023).

Proportionality evaluates the relationship between an intervention's expected protective effect and the severity of its restrictions or adverse consequences. Its application becomes especially demanding when a severe outcome is rare and prediction is weak. For example, classifying many people as at high risk can expose a large number of individuals to restrictive interventions while offering uncertain precision about whom the intervention helps. Carter and colleagues' meta-analysis illustrates the predictive-value problem, while NICE NG225 directs clinical services away from using global risk strata as gatekeepers to treatment or discharge Carter et al. (2017) NICE (2022).

A transparent ethical analysis must describe the time horizon as well as the immediate action. A measure that apparently limits exposure for several hours could undermine future help-seeking if its implementation is humiliating or frightening. Conversely, a collaborative approach may require substantial resources and careful support to be safe during an acute crisis. Policy choices should be evaluated across clinical outcomes, patient experience, social consequences, distributional effects, and feasible alternatives. This is an argument for richer evidence and accountable decision processes rather than for a universal algorithm that settles every case Corderoy et al. (2025) World Health Organization (2021b).

Critical Appraisal of the Current Evidence Base

The empirical foundation is uneven across outcomes and settings. General medical record-review programs supply established methods for identifying care-related injury, while mental health safety studies frequently emphasize coercion, adverse psychological experience, diagnostic uncertainty, and continuity. Randomized psychotherapy trials offer potential causal evidence for intervention effects but often report deterioration and adverse events inconsistently. Studies comparing compulsory and voluntary admission are vulnerable to confounding by severity and availability of alternatives. Incident reports underdetect outcomes that are difficult to recognize, and patient surveys face attribution and response bias. These limitations are research findings in their own right because they determine what can reasonably be concluded Brennan et al. (1991) Honkalampi et al. (2025) Corderoy et al. (2025) Crawford et al. (2016).

Psychiatry-specific reviews make this evidence structure more concrete. Thibaut and colleagues' 2019 systematic review included 364 publications across 31 countries, with information from over 150,000 participants. It identified ten inpatient safety domains including violence, coercive interventions, environmental hazards, medication safety, and clinical decision-making. Study quality and outcome definitions varied substantially, and heterogeneity limited meaningful pooling. These findings establish a broad research map, yet they do not produce a single reliable rate of all psychiatric harm. The review also excluded various sources of routine quality-improvement data and many ambulatory contexts, reinforcing the need to specify what a particular evidence synthesis was designed to capture Thibaut et al. (2019).

Russotto and colleagues' systematic review published in January 2026 synthesized 92 studies of inpatient psychiatric patient-safety incidents using a WHO classification framework. Commonly reported domains included behavior-related incidents, medication events, and falls; contributing conditions included patient factors, communication and staff issues, organizational practices, and environmental hazards. Only a minority of reviewed studies reported quantitative outcomes of prevention measures. This is evidence that the description of risks has outpaced credible evaluation of many mitigation strategies, particularly in settings poorly represented in the literature. The review's hospital-focused sampling must not be generalized without evidence to outpatient clinics, community support, or digital mental health systems Russotto et al. (2026).

Some apparent numerical conflicts result from incompatible estimands rather than inconsistent data. The WHO's widely cited global patient-safety burden estimates concern health care broadly, not mental health care alone. The 5.08% estimated frequency of outpatient diagnostic error applies to a synthesis of selected U.S. adult outpatient datasets, while the 5.2% negative-effects figure from Crawford and colleagues measures self-reports among survey respondents receiving psychological treatment in England and Wales. These percentages describe different events, populations, methods, and denominators. Neither can be used to estimate the proportion of all psychiatric patients harmed by treatment World Health Organization (2023a) Singh et al. (2014) Crawford et al. (2016).

Contested findings require equal attention to the quality of positive and negative evidence. An intervention may improve average scores while causing serious adverse effects in a subgroup; it may reduce an event recorded by staff while increasing unreported distress. Strong safety claims should address plausible alternatives, measurement properties, missing outcomes, baseline differences, and external validity. The relevant synthesis is therefore conditional: some organizational interventions have credible mechanisms and policy support, but their clinical benefit cannot be assumed without an evaluation suited to the outcomes and setting. Research designs should be assessed by what they establish, not by whether they confirm a preferred ethical position Pedersen et al. (2024) Panagioti et al. (2019) Rozental & Řiháček (2025).

A Research Agenda for Safer Mental Health Care

Future work should establish a shared but multidimensional taxonomy of mental health care-related adverse events, including physical injury, psychological deterioration, patient-defined coercion, informational harm, loss of access, and longer-term effects on trust and functioning. Such a taxonomy must specify temporal windows, severity, causal attribution, and reporting channels. It should be tested across countries and service types with direct patient participation. Prospective cohorts that integrate patient reports, validated outcomes, clinical documentation, and service-use data would provide a more complete picture than any single existing source. Independent replication and transparent methods are needed to support cumulative inference Rozental & Řiháček (2025) NHS England (2022/2025) World Health Organization (2021a).

Comparative research is especially needed on alternatives to restrictive admission, supported decision-making, safer crisis transitions, complaint-responsive care, and interventions to reduce diagnostic error. Where randomization is ethically or practically difficult, carefully designed longitudinal and quasi-experimental studies can still improve on uncontrolled institutional before–after accounts. Studies should evaluate whether benefits persist, whether harmful practices are displaced, and whether groups with limited access experience different effects. Economic and resource analyses matter because implementing safer alternatives often requires investment, but resource costs should be interpreted alongside dignity, rights, and social participation Corderoy et al. (2025) Pedersen et al. (2024) WHO & OHCHR (2023).

Digital mental health research should preregister clinically meaningful outcomes, report system versions and data dependencies, evaluate model-supported decisions against credible alternatives, and monitor unforeseen harms over time. Institutional studies should test whether incident response policies actually lead to verifiable improvements rather than merely increasing paperwork. The broader scientific challenge is to connect ethically justified standards with independently measured outcomes through falsifiable causal mechanisms. A system becomes credibly safer when the conditions causing harm change and those changes are evident to patients and researchers across successive encounters NHS England (2022/2025) National Academies (2015) World Health Organization (2021a).

Conclusion

Clinical ethics, iatrogenic harm, and patient safety designate an integrated field of inquiry into how health systems acting in the name of care can also produce injury, restriction, uncertainty, and exclusion. Historical record-review research established methods for identifying treatment-related injury; systems theory located errors within organizations and care pathways; contemporary psychiatric research and human rights guidance extended the relevant outcomes to coercion, dignity, patient authority, and longitudinal participation in care. The strongest synthesis joins these traditions through observable processes, clear ethical criteria, and evidence appropriate to each question Brennan et al. (1991) Reason (2000) WHO & OHCHR (2023).

The evidence supports several conclusions. Harm is multidimensional and requires both clinical and patient-defined measurement. Diagnostic error and suicide-risk prediction should be treated as problems of uncertainty and follow-up, with risk scores evaluated within their actual predictive limits. Involuntary admission is associated with a complex pattern of potential benefit and harm, and observational comparisons cannot fully resolve causality. Psychological treatment can be accompanied by adverse experiences, yet survey frequencies of perceived harm are not interchangeable with causal estimates. Safety interventions must address organizational mechanisms and demonstrate outcomes across time and populations National Academies (2015) Carter et al. (2017) Corderoy et al. (2025) Crawford et al. (2016).

An adequate scientific and ethical standard for mental health care therefore rests on four commitments: reliable detection of harm, transparent causal reasoning, respect for the person's agency and rights, and sustained organizational learning. These commitments create a common basis for examining traditional clinical institutions and emerging human–artificial systems without assuming that an innovation is safe because it is new or beneficial because its intention is therapeutic. The research horizon is an empirically accountable model of care in which improvements are established through observable reductions in avoidable harm and improvements in the person's continuing opportunity for informed, dignified support Donabedian (1988) World Health Organization (2021a) World Health Organization (2021b).

References

  • Appelbaum, P. S., & Grisso, T. (1988). Assessing patients' capacities to consent to treatment. New England Journal of Medicine, 319(25), 1635–1638. https://doi.org/10.1056/NEJM198812223192504
  • Appelbaum, P. S. (2007). Assessment of patients' competence to consent to treatment. New England Journal of Medicine, 357(18), 1834–1840. https://doi.org/10.1056/NEJMcp074045
  • Brennan, T. A., Leape, L. L., Laird, N. M., Hebert, L., Localio, A. R., Lawthers, A. G., Newhouse, J. P., Weiler, P. C., & Hiatt, H. H. (1991). Incidence of adverse events and negligence in hospitalized patients: Results of the Harvard Medical Practice Study I. New England Journal of Medicine, 324(6), 370–376. https://doi.org/10.1056/NEJM199102073240604
  • Carter, G., Milner, A., McGill, K., Pirkis, J., Kapur, N., & Spittal, M. J. (2017). Predicting suicidal behaviours using clinical instruments: Systematic review and meta-analysis of positive predictive values for risk scales. British Journal of Psychiatry, 210(6), 387–395. https://doi.org/10.1192/bjp.bp.116.182717
  • Corderoy, A., Kisely, S., Zirnsak, T., & Ryan, C. J. (2025; published online 2024). The benefits and harms of inpatient involuntary psychiatric treatment: A scoping review. Psychiatry, Psychology and Law, 32(5), 734–781. https://doi.org/10.1080/13218719.2024.2346734
  • Crawford, M. J., Thana, L., Farquharson, L., Palmer, L., Hancock, E., Bassett, P., Clarke, J., & Parry, G. D. (2016). Patient experience of negative effects of psychological treatment: Results of a national survey. British Journal of Psychiatry, 208(3), 260–265. https://doi.org/10.1192/bjp.bp.114.162628
  • Donabedian, A. (1988). The quality of care: How can it be assessed? JAMA, 260(12), 1743–1748. https://doi.org/10.1001/jama.260.12.1743
  • Elwyn, G., Frosch, D., Thomson, R., Joseph-Williams, N., Lloyd, A., Kinnersley, P., Cording, E., Tomson, D., Dodd, C., Rollnick, S., Edwards, A., & Barry, M. (2012). Shared decision making: A model for clinical practice. Journal of General Internal Medicine, 27(10), 1361–1367. https://doi.org/10.1007/s11606-012-2077-6
  • Gallagher, T. H., Waterman, A. D., Ebers, A. G., Fraser, V. J., & Levinson, W. (2003). Patients' and physicians' attitudes regarding the disclosure of medical errors. JAMA, 289(8), 1001–1007. https://doi.org/10.1001/jama.289.8.1001
  • Honkalampi, K., Urhonen, H.-R., & Virtanen, M. (2025). Negative effects in randomized controlled trials of psychotherapies and psychological interventions: A systematic review. Psychotherapy Research, 35(1), 100–111. https://doi.org/10.1080/10503307.2024.2301972
  • Institute of Medicine. (2000). To err is human: Building a safer health system (L. T. Kohn, J. M. Corrigan, & M. S. Donaldson, Eds.). National Academies Press. https://doi.org/10.17226/9728
  • Institute of Medicine. (2001). Crossing the quality chasm: A new health system for the 21st century. National Academies Press. https://doi.org/10.17226/10027
  • Leape, L. L., Brennan, T. A., Laird, N., Lawthers, A. G., Localio, A. R., Barnes, B. A., Hebert, L., Newhouse, J. P., Weiler, P. C., & Hiatt, H. H. (1991). The nature of adverse events in hospitalized patients: Results of the Harvard Medical Practice Study II. New England Journal of Medicine, 324(6), 377–384. https://doi.org/10.1056/NEJM199102073240605
  • Lilienfeld, S. O. (2007). Psychological treatments that cause harm. Perspectives on Psychological Science, 2(1), 53–70. https://doi.org/10.1111/j.1745-6916.2007.00029.x
  • National Academies of Sciences, Engineering, and Medicine. (2015). Improving diagnosis in health care (E. P. Balogh, B. T. Miller, & J. R. Ball, Eds.). National Academies Press. https://doi.org/10.17226/21794
  • National Institute for Health and Care Excellence. (2022). Self-harm: Assessment, management and preventing recurrence (NICE Guideline NG225). https://www.nice.org.uk/guidance/ng225/chapter/recommendations
  • NHS England. (2022; updated 2025). Patient safety incident response framework. https://www.england.nhs.uk/long-read/patient-safety-incident-response-framework/
  • 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
  • Pedersen, M. L., Gildberg, F. A., Baker, J., & Tingleff, E. B. (2024). A systematic review of interventions to reduce mechanical restraint in adult mental health inpatient settings. International Journal of Mental Health Nursing, 33(3), 505–522. https://doi.org/10.1111/inm.13267
  • Reason, J. (2000). Human error: Models and management. BMJ, 320(7237), 768–770. https://doi.org/10.1136/bmj.320.7237.768
  • Rozental, A., Kottorp, A., Boettcher, J., Andersson, G., & Carlbring, P. (2016). Negative effects of psychological treatments: An exploratory factor analysis of the Negative Effects Questionnaire for monitoring and reporting adverse and unwanted events. PLOS ONE, 11(6), e0157503. https://doi.org/10.1371/journal.pone.0157503
  • Rozental, A., & Řiháček, T. (2025). How to measure negative effects of psychological interventions. Current Opinion in Psychology, 65, 102084. https://doi.org/10.1016/j.copsyc.2025.102084
  • Russotto, S., Conti, A., Masini, A., Tempia Valenta, S., Vanhaecht, K., Mira, J. J., & Panella, M. (2026). Patient safety incidents in the psychiatric inpatient setting: Determinants, consequences, and strategies. A systematic review. Frontiers in Psychiatry, 16, 1703768. https://doi.org/10.3389/fpsyt.2025.1703768
  • Singh, H., Meyer, A. N. D., & Thomas, E. J. (2014). The frequency of diagnostic errors in outpatient care: Estimations from three large observational studies involving US adult populations. BMJ Quality & Safety, 23(9), 727–731. https://doi.org/10.1136/bmjqs-2013-002627
  • Strauss, B., Rosendahl, J., & Klatte, R. (2026). Adverse experiences in psychological treatments: Where do we stand? Current Opinion in Psychology, 67, 102161. https://doi.org/10.1016/j.copsyc.2025.102161
  • Thibaut, B., Dewa, L. H., Ramtale, S. C., D'Lima, D., Adam, S., Ashrafian, H., Darzi, A., & Archer, S. (2019). Patient safety in inpatient mental health settings: A systematic review. BMJ Open, 9(12), e030230. https://doi.org/10.1136/bmjopen-2019-030230
  • Vincent, C., Taylor-Adams, S., & Stanhope, N. (1998). Framework for analysing risk and safety in clinical medicine. BMJ, 316(7138), 1154–1157. https://doi.org/10.1136/bmj.316.7138.1154
  • World Health Organization & Office of the United Nations High Commissioner for Human Rights. (2023). Mental health, human rights and legislation: Guidance and practice. https://www.who.int/publications/i/item/9789240080737
  • World Health Organization. (2021). Global patient safety action plan 2021–2030: Towards eliminating avoidable harm in health care. https://www.who.int/publications/i/item/9789240032705
  • World Health Organization. (2021). Guidance on community mental health services: Promoting person-centred and rights-based approaches. https://www.who.int/publications/i/item/9789240025707
  • World Health Organization. (2023). Patient safety [Fact sheet, September 11]. https://www.who.int/news-room/fact-sheets/detail/patient-safety