Introduction
In mental health settings, various restrictive practices (RPs) such as physical or mechanical restraint, seclusion, chemical restraint, compulsory admission and informal coercion are commonly employed in response to behaviours perceived as posing a risk of harm to oneself or others (World Health Organization, 2019). RPs are broadly defined as making someone do something they don’t want to do or stopping someone from doing something they want to do (Britain, Reference Paterson, Bennet and Bradley2014). The use of RPs limits the dignity, autonomy, and freedom of movement of individuals and has the potential to cause serious physical and psychological harm (Chieze et al., Reference Chieze, Hurst, Kaiser and Sentissi2019).
Reducing RPs has been one of the forefront issues in international mental health reform over the past two decades. They are considered to be used only as a last resort (Power et al., Reference Power, Baker and Jackson2020). A range of policy reforms and evidence-based frameworks has been introduced to support this goal, including the Open Wards Policy (Jungfer et al., Reference Jungfer, Schneeberger, Borgwardt, Walter, Vogel, Gairing, Lang and Huber2014), the Six Core Strategies (Riahi et al., Reference Riahi, Dawe, Stuckey and Klassen2016), the Safewards model (Maguire et al., Reference Maguire, Ryan, Fullam and McKenna2018) and trauma-informed care (Forkey et al., Reference Forkey, Szilagyi, Kelly and Duffee2021). A key strategy across these frameworks is the integration of reduction strategies into routine clinical care and the replacement of RPs with less restrictive alternatives (Muluneh et al., Reference Muluneh, Chavulak, Lee, Petrakis and Haines2025).
The effective translation of these principles into practice requires meaningful engagement with a diverse range of mental health stakeholders, including service users and family caregivers (Bennetts et al., Reference Bennetts, Pepin, Moylan, Carolin and Lucas2024). However, inconsistent definitions of RPs (Muluneh et al., Reference Muluneh, Chavulak, Lee, Petrakis and Haines2025) and differing stakeholder perspectives hinder shared decision-making, complicating both the practical implementation of policy reforms and the successful reduction of RPs. Healthcare professionals often view RPs as necessary for safety and ward management, whereas many service users often perceive them as harmful and as violations of human rights (Mayers et al., Reference Mayers, Keet, Winkler and Flisher2010). Whether or not specific actions are interpreted as RPs is highly influenced by the divergent perspectives of various stakeholders. This challenges the ability to reach a consensus on what constitutes acceptable routine clinical interventions versus RPs that should be avoided (Murphy et al., Reference Murphy, Qureshi, Endale, Esponda, Pathare, Eaton, De Silva and Ryan2021). Thus, inconsistencies in stakeholder perspectives may be a powerful yet under-recognized factor perpetuating unnecessary reliance on RPs, even in systems formally committed to their reduction.
Several factors including cultural perspectives, legal frameworks, professional training, individual values, shared assumptions, epistemic disparities and lived experiences can affect individuals’ views towards RPs (McSherry, Reference McSherry2013). Additionally, variations in the context in which an action occurs can lead to different interpretations of the same practice. Factors such as the presence or absence of consent, the healthcare professional’s intent, the degree of restriction, the duration of the action, whether doors are locked, whether staff or security initiated the intervention, and whether the action is actually implemented or merely threatened are all likely to influence its classification (Duffy et al., Reference Duffy, Boyle and Del Villar2023). Despite this, few studies have systematically examined how contextual factors affect RP classification across diverse stakeholder groups or explored differences in perspectives for various types of RPs.
This study sought to address two aims related to how different mental health stakeholders conceptualize and classify RPs. First, it sought to characterize the types of scenarios that stakeholders consider should be classified as RPs. Second, it sought to identify the contextual factors that were influential in determining whether a scenario should be considered an RP and to compare these factors between stakeholder groups.
Methods
Design and participants
This study was an international cross-sectional survey involving clinicians, service users, caregivers and researchers. Eligibility criteria varied by participant group. Clinicians included nurses, psychiatrists, psychologists, social workers, medical practitioners, occupational therapists and other trained staff currently working in adult mental health inpatient settings. Service users were individuals with lived experience in such settings and having been discharged six months prior to data collection. Family caregivers were individuals who provided unpaid support to a relative, friend or significant other admitted to an adult mental health inpatient facility. Individuals under 18 years of age or those receiving or providing inpatient care during the survey administration were excluded. The researchers’ group included those who self-identified as researchers with expertise in mental health inpatient care and/or RPs.
Development of survey strategies and instruments
The development of the overall instruments and data collection strategies for this survey followed a co-designed approach, informed by a series of systematic reviews (Belayneh et al., Reference Belayneh, Chavulak, Lee, Petrakis and Haines2024, Reference Belayneh, Lee, Haines, Aluh, Onu, Newton‐Howes, Masters, Kohn, Sin and Goulet2025b; Muluneh et al., Reference Muluneh, Chavulak, Lee, Petrakis and Haines2025). The co-design process involved 33 mental health stakeholders, including clinicians, service users, caregivers, and researchers, representing 17 countries across Africa, Asia, North America, Europe, and Australia (Belayneh et al., Reference Belayneh, Chavulak, Lee, Petrakis and Haines2025a). A detailed methodological description of the survey is available in the published protocol (Belayneh et al., Reference Belayneh, Lee, Haines, Aluh, Onu, Newton‐Howes, Masters, Kohn, Sin and Goulet2025b).
Two rounds of online panel meetings were held (Figure 1). The first round aimed to explore contextual elements influencing how individuals classify an action as a RP, resulting in the identification of 23 key contextual elements that guided the development of case scenario descriptions. These included whether a door was locked or unlocked, the duration of the intervention, the identity of the person administering it, the presence or absence of a risk assessment, attempts at less restrictive alternatives, legal approval, and documented evidence prompting the intervention.
Figure 1. Structure and overall step-by-step process of the co-design work.
In the second round, panel members drafted 81 case scenarios reflecting various contexts to illustrate real-world situations (Supplementary File 1). From these, 44 scenarios were prioritized for survey data collection, covering a range of RPs: 21 mechanical restraint, 8 manual restraint, 16 chemical restraint, 17 seclusion, 4 compulsory admission, 4 informal coercion and 11 involving combined methods. The full list of scenario descriptions is available in Supplementary File 2. Several comparator scenarios were developed within each context, featuring identical core descriptions but differing by a single situational element (see underlined statements in the examples below). For instance, in the door-locking context, two comparator scenarios differ only in which door is locked:
1. A nurse locks the individual person’s room door while the person is inside to prevent escape from the hospital.
2. A nurse locks the main ward door, leaving the individual person’s room door open to prevent escape from the hospital.
This design was chosen to isolate and assess the influence of specific contextual elements on the participants’ decisions regarding whether or not an action constituted an RP.
Recruitment and outcome measurement
Participants were recruited online using the Qualtrics platform, enabling respondents from any country worldwide. Qualtrics is a widely used survey tool offering secure data collection, customizable question formats, and automated data management (Carter and Ponte, Reference Carter and Ponte2022). Its user-friendly interface, multi-device compatibility, and robust features for handling complex survey logic ensured accurate and efficient data collection for this international study (Boas et al., Reference Boas, Christenson and Glick2020; Molnar, Reference Molnar2019). An anonymous survey link and a flyer with a QR code were widely distributed through institutional websites and social media channels, including X, Facebook, Instagram and LinkedIn. Various institutions, professional associations and mental health advocacy groups supported recruitment by hosting the survey, sharing it via mailing lists, and featuring it in newsletters. Additionally, a snowball sampling technique was employed, with international research team members actively promoting the survey through their professional and personal networks and acting as regional contacts for data collection.
The survey questionnaire was professionally translated and expert-reviewed in English, Amharic, Dutch, French, Hebrew, and Greek, allowing participants to select their preferred language via the multilingual survey link.
Outcome data were collected by inviting survey respondents to review narrative descriptions of 44 case scenarios and respond to a closed-ended question asking whether they believed the action depicted constituted an RP. Responses were recorded using a four-point Likert scale: Definitely yes, Probably yes, Probably no or Definitely no.
Data analysis
Descriptive statistics, including frequencies and percentages, summarized participant characteristics and the proportion of participants who classified each scenario as a restrictive practice or not. The study employed ordered logistic regression to examine whether the classification of scenarios as RPs was influenced by differences in the specific contextual categories embedded within each scenario, and to identify which contextual factors contributed to discrepancies in scenario classification both within and between stakeholder groups. These models accounted for the ordinal nature of the 4-point Likert scale responses (‘Definitely yes’, ‘Probably yes’, ‘Probably no’ and ‘Definitely no’). The model was adjusted for respondent ID, with clustered robust standard errors applied to account for repeated measures within participants.
Separate regression analyses were conducted for responses received from clinicians, researchers, service users, and family caregivers to examine within-group differences in scenario classification. A pairwise comparison approach was also utilized, in which each pair of comparator scenarios shared an identical core contextual description but differed in one specific situational detail. A total of 22 paired comparisons were conducted to examine classification patterns both within and across stakeholder groups. For example, in the context of handholding, two scenarios were compared: one involved a nurse firmly holding a person’s arms to prevent self-harm, while the other involved a security staff member performing the same action. Additionally, interaction effect analyses were performed to determine whether specific contextual elements of the scenarios had varying effects across stakeholder groups (clinicians, researchers, service users and family caregivers). Comparisons of other stakeholders’ classifications were made with reference to clinicians’ classifications, given the larger number of clinicians who participated in the study, which can maximize the statistical power of interaction analyses. Additionally, clinicians are the stakeholder group primarily responsible for recording instances of RPs (Grace-Martin, Reference Grace-Martin2020). This approach allowed us to examine how clinicians’ perspectives align with or differ from those of other stakeholders. These models revealed four patterns of scenarios: (1) consistent differences in classification across all stakeholder groups, (2) variation specific to certain groups but not others, (3) no significant effect in any stakeholder group and (4) significant variation within stakeholder groups, but not between them.
All statistical analyses were performed using STATA version 17, with statistical significance set at p < 0.05 and a 95% confidence level.
Results
Characteristics of study participants
A total of 864 participants from 47 countries completed the core set of scenario items in this study. The regional representation includes Africa (39.26%), Europe (34.18%), Asia (8.59%), North America (3.52%) and Australasia (14.71%). The sample included 517 clinicians, 165 researchers, 80 service users and 89 family caregivers.
Classification of RP scenarios
Participants’ responses varied considerably in how they classified the same potential RP scenarios, both between and within stakeholder groups. The service user group more frequently classified scenarios as RPs, providing the highest number of ‘Definitely yes’ responses for 42 out of 44 scenarios (95.45%) compared to classifications made by clinicians, family caregivers and researchers. In contrast, family caregivers recorded the lowest proportion of ‘Definitely yes’ responses for 31 out of 44 scenarios (70.45%).
Overall, the proportion of ‘Definitely yes’ responses for classifying each scenario as an RP was relatively low across stakeholder groups. The highest proportions of ‘Definitely yes’ responses for specific scenarios were as follows: 86.44% among service users for Scenario 22 (restraint without prior risk assessment), 72.38% among clinicians and 66.00% among family caregivers for Scenario 27 (mechanical restraint of both wrists and ankles), and 60.9% among researchers for Scenario 14 (restraint without consent).
Detailed data on the frequency and proportion of participants’ responses for the classification of each of the 44 case scenarios, indicating whether or not they were identified as RPs are provided in Supplementary File 2. The data are presented as cross-tabulations comparing responses across stakeholder groups.
Scenario contexts influencing scenario classification within each stakeholder group
Nine of the 22 contextual elements of RP examined by our comparative scenarios did not influence any of the stakeholder groups’ responses regarding whether the scenarios described were an RP or not (Tables 1 and 2). These elements included: the time of day when the action took place (daytime or nighttime), type of shift (morning or afternoon), whether less restrictive alternatives were attempted before the action, the type of activities the person was prevented from engaging in during the episode (e.g., receiving visitors or leaving designated areas), and the availability of utilities in the seclusion room (empty or furnished). These contexts therefore appear to be less important for inclusion in future work seeking to improve the classification and recording of RPs.
Table 1. Comparison of the influence of variations in specific contextual factors/definitional elements on scenario classification where or not as restricitve practice among individuals within each stakeholder group
* Indicates statistically significant differences in the classification between paired scenarios that shared identical core contexts among individuals within each participant stakeholder group.
Table 2. Interpretation of findings for comparisons of individuals’ classification reposes within each stakeholder group
Thirteen contextual elements showed statistically significant differences in scenario classification in at least one participant group. These were: severity of anticipated harm (moderate vs. severe risk to be prevented), type of professional implementing the action (nurse vs. security staff), door lock status (door locked vs. unlocked), type of locked door (individual room door vs. only main ward door), consent status (family approached and consent given vs. family not approached), duration of the intervention episode (short vs. long), nature of intention (actual implementation of mechanical restraint vs. only threatening with mechanical restraint), risk assessment (conducted and indicating high risk vs. no risk assessment conducted), number of restraint points applied to the person’s body parts (multiple vs. a single point), seclusion room conditions (empty vs. furnished), legal status of medication (medication legally approved vs. not legally approved), medication dosage (dose in line with hospital protocol vs. increased dose beyond protocol) and presence of others during seclusion (person secluded alone vs. with a staff member present). Among these, door lock status, type of locked door, and nature of the intervention showed statistically significant variations within each of the stakeholder groups. The remaining 10 comparisons demonstrated group-specific effects, with significant differences observed only within certain stakeholder groups.
For example, the presence of another person during seclusion influenced the classification of scenarios as an RP for clinicians and researchers, but not for service users or family caregivers. Similarly, variations in duration of intervention and medication dosage showed significant differences for clinicians and service users only. Consent status and severity of harm anticipated to be prevented affected the classification patterns only among service user participants. Family caregivers were the only stakeholder group whose classification responses were significantly influenced by the number of restraint points used.
Individuals within service user groups were generally more sensitive in identifying many scenarios, including subtle forms such as verbal threats, lack of the person’s consent, and the constant presence of staff that were overlooked by other stakeholder groups. However, their scenario classification decisions shifted notably with even minor changes in scenario context descriptions (for example, locked doors versus unlocked doors, action carried out with consent versus without consent). Across all comparisons, service users’ classifications demonstrated statistically significant differences in 10 of the 22 paired scenarios comparisons, followed by clinicians and family caregivers with eight each, and researchers with seven (Tables 1 and 2).
Scenario contexts influencing scenario classification between stakeholder groups
Contrasts in how different stakeholder groups classified the comparative case scenarios examining key contextual elements of RPs revealed that clinicians often held divergent views from service users and family caregivers and occasionally differed from researchers (Table 3).
Table 3. Comparisons of influence of variations in specific contextual factors/definitional elements on scenario classification across stakeholder groups: Results of interaction effect analysis
* Indicates statistically significant variations in the influence of contextual factors/definitional elements on the differences observed in the classification of paired scenarios among researchers, service usersand family caregiver stakeholders compared to clinicians.
Statistically significant variations between stakeholder groups were observed in comparisons involving key contextual elements, including the severity of harm, consent status, duration of the intervention, nature of the intervention, presence of a formal risk assessment, use of less restrictive alternatives before applying RPs, number of restraint points used and legal framing of the intervention.
Service users’ responses exhibited the most frequent divergence from clinicians, with statistically significant differences identified across five contextual comparisons: severity of harm, consent status, duration of the intervention, nature of the intervention, and the use of less restrictive alternatives. Family caregivers demonstrated statistically significant differences from clinicians in three contexts (the nature of the intervention, the number of restraint points applied and the legal framing of the action). Researchers, by contrast, differed significantly from clinicians’ classifications in two contextual areas: the nature of the intervention and the presence or absence of a formal risk assessment.
Discussion
This study revealed divergent perspectives within and between clinicians, service users, family caregivers, and researchers on how interventions are conceptualized and classified as RPs. These classification decisions were influenced more by the context in which the practice was applied, the professional role of the assessor, and their underlying epistemological orientation, rather than solely by the coercive nature of the action itself. Such inconsistencies in perspectives on RPs hinder collaborative efforts among stakeholders to implement reduction strategies effectively (Bennetts et al., Reference Bennetts, Pepin, Moylan, Carolin and Lucas2024) and undermine the consistency of rights monitoring and accountability mechanisms. This ultimately hinders the shift towards more recovery-oriented and evidence-based approaches to care (De Cuyper et al., Reference De Cuyper, Vanlinthout, Vanhoof, van Achterberg, Opgenhaffen, Nijs, Peeters, Put, Maes and Van Audenhove2023), which require the active involvement of service users and their family caregivers in care planning (Slade, Reference Slade2017).
A common pattern across all groups was the tendency to associate restrictiveness primarily with visible physical or environmental features like mechanical restraints or locked doors, while often overlooking more subtle actions (Series, Reference Series2015). A clear example of this is the context of door locking status: a substantial proportion of participants across all groups classified a seclusion scenario involving locked doors as an RP. However, this proportion declined markedly when the same scenario was presented without locked doors, despite the core coercive nature of the context remaining unchanged. In reality, individuals are often compelled to remain in a single room not necessarily because of physical barriers, but also due to implied negative consequences if they attempt to leave the room on their own (Klingemann et al., Reference Klingemann, Świtaj, Lasalvia and Priebe2022). Such forms of restriction are likely to be overlooked and left unreported, even though they can seriously limit personal freedom and carry ethical and psychological consequences comparable to more overt forms of restriction (Muluneh et al., Reference Muluneh, Chavulak, Lee, Petrakis and Haines2025).
The issue of locked doors in psychiatric settings raises critical human rights concerns due to the restriction of individual liberty, as emphasized by the Universal Declaration of Human Rights (Szmukler et al., Reference Szmukler, Daw and Callard2014). The World Health Organization and the Office of the United Nations High Commissioner for Human Rights promote open-door policies aimed at reducing coercion by balancing safety with autonomy (World Health Organization, 2023). However, the findings of this study challenge the assumption that open-door policies alone reduce coercion. Such policies may remove visible physical barriers, while subtler informal restrictions can persist if left unrecognized. This highlights that policies, classification frameworks, and clinical guidelines need to move beyond visible physical measures and give greater attention to less overt forms of restriction, ensuring that the full spectrum of RPs is accurately recognized and addressed (Beeri et al., Reference Beeri, Baumberger, Zwakhalen and Hahn2025).
This study highlights the stakeholder-specific nature of classification decisions, demonstrating how differences in professional roles, lived experiences, and shared assumptions shape individual perspectives on RPs (Butterworth et al., Reference Butterworth, Wood and Rowe2022). Service users, in particular, showed the greatest sensitivity to contextual factors related to power, autonomy, and psychological safety, whereas most other groups relied on policies and protocols (Bennetts et al., Reference Bennetts, Pepin, Moylan, Carolin and Lucas2024). In contrast, clinicians’ classifications were more influenced by legal and procedural factors such as formal risk assessments, intervention duration, and medication dosage, often shaped by institutional protocols and medico-legal frameworks. These findings suggest that policies and protocols are likely influenced by clinician needs and preferences. This underscores the importance of valuing lived experiences and integrating them with clinical judgments to promote inpatient practices that prioritize safety, trust, and empowerment across all stakeholder groups (Forkey et al., Reference Forkey, Szilagyi, Kelly and Duffee2021).
Family caregivers generally aligned with clinicians and researchers, viewing interventions through the lens of risk mitigation rather than autonomy. This supports prior findings that family members’ views often differ from service users’ and align more with clinical staff due to safety concerns (Hupé et al., Reference Hupé, Larue and Contandriopoulos2024). These differences highlight the importance of integrating perspectives across diverse stakeholders in policy development and the implementation of reduction strategies (Molloy et al., Reference Molloy, Kilcoyne, Belcher and Wykes2025).
Researchers fell between other groups, demonstrating more context sensitivity than family caregivers but less than clinicians and service users, who are often the subjects of their studies. Their classification judgments tended to rely more heavily on legal definitions and procedural guidelines, likely reflecting their familiarity with formal policy frameworks and research ethics standards, while placing less emphasis on psychological dimensions (Hupé et al., Reference Hupé, Larue and Contandriopoulos2024). Importantly, some contextual elements identified in recent systematic reviews (Hupé et al., Reference Hupé, Larue and Contandriopoulos2024; Muluneh et al., Reference Muluneh, Chavulak, Lee, Petrakis and Haines2025), as core conceptual themes for defining RPs, such as the time of day, the activity the person was engaged in during the episode, and whether less restrictive alternatives were attempted did not influence how RPs were actually interpreted by any of the stakeholder groups. The disconnect between academic theory and real-world decision-making highlights the limitations of relying solely on theory-derived definitions and classification frameworks (Belayneh et al., Reference Belayneh, Abebaw, Amare, Haile and Abebe2019), which may not fully capture how RPs are perceived and experienced in clinical practice (Hupé et al., Reference Hupé, Larue and Contandriopoulos2024). This underscores the need to empirically validate such systems in real-world settings to ensure their effectiveness and applicability.
Implications
Diverging views on what constitutes an RP among stakeholders complicate collaborative decision-making and hinder the consistent application of least-restrictive alternatives (Rose et al., Reference Rose, Perry, Rae and Good2017). These discrepancies affect consistency in the implementation, monitoring and evaluation (Janssen et al., Reference Janssen, van de Sande, Noorthoorn, Nijman, Bowers, Mulder, Smit, Widdershoven and Steinert2011) of reduction strategies in mental health care and the integration of least restrictive alternatives in clinical practice (Muluneh et al., Reference Muluneh, Chavulak, Lee, Petrakis and Haines2025). Practices that one healthcare professional considers least restrictive may be perceived as restrictive or even harmful by others (Bennetts et al., Reference Bennetts, Pepin, Moylan, Carolin and Lucas2024) or experienced as intrusive by service users. Therefore, clinicians are required to thoughtfully balance or integrate different viewpoints with their clinical judgment to reach a workable approach when implementing RP reduction strategies and selecting the least restrictive alternatives (Duffy et al., Reference Duffy, Boyle and Del Villar2023). Such inclusive approaches ensure practices are conceptually coherent, ethically grounded, and practically applicable across diverse settings (Rose et al., Reference Rose, Perry, Rae and Good2017).
Inconsistencies in how RPs are defined also affect the validity and comparability of research. For example, researchers analyzing hospital records may use criteria that differ from those applied in clinical incident reporting systems. Similarly, when data are collected by asking service users if they can recall whether and how frequently they experienced RPs during their hospital stay, their responses may be distorted by differences in how they interpret what constitutes RPs. Such variabilities can introduce systematic bias, ultimately limiting the utility of research findings for benchmarking, evaluating trends, and guiding quality improvement efforts (Haines et al., Reference Haines, Cornwell, Fleming, Varghese and Gray2008). Researchers are encouraged to triangulate data from multiple sources, such as medical records, incident reports, and routine surveys of staff and service users. This can help lead to a more comprehensive and reliable evidence base, reducing underreporting and capturing informal or undocumented practices that might otherwise be overlooked by using individual methods (Hallett and McLaughlin, Reference Hallett and McLaughlin2022).
The overall findings underscore the need for harmonized definitions and classification systems for RPs. Achieving consensus across stakeholders is a long-term process and may be challenging. Even providing a written definition for healthcare practices does not necessarily mean stakeholder agreement on the definition and classification will improve (Haines et al., Reference Haines, Massey, Varghese, Fleming and Gray2009). Variations in perspectives on RPs between stakeholders and across regions, informed by cultural differences, educational backgrounds, and local policy frameworks, are likely to pose challenges to the harmonization process. These efforts may be more effective if supported by intermediate and complementary strategies such as the development of robust monitoring systems and educational programs (Janssen et al., Reference Janssen, van de Sande, Noorthoorn, Nijman, Bowers, Mulder, Smit, Widdershoven and Steinert2011). Ongoing capacity-building initiatives, such as therapeutic patient education and rights-based programs like the WHO’s Quality Rights initiative (World Health Organization, 2019) can further support progress towards harmonization (Haines et al., Reference Haines, Massey, Varghese, Fleming and Gray2009; Lickiewicz et al., Reference Lickiewicz, Efkemann, Husum, Lantta, Pingani and Whittington2024).
Limitations
The use of hypothetical scenarios may not adequately capture individuals’ interpretations of whether an action constitutes an RP in clinical settings. In reality, such judgments are often shaped by contextual pressures, emotional responses, institutional dynamics, and organizational stress. These factors may not be fully simulated through scenario vignettes. However, the use of standardized cases facilitated meaningful comparisons across participants and regions while also strengthening internal validity. Moreover, the engagement of diverse mental health stakeholders in the co-design of these scenarios enriched the study by incorporating a wide range of perspectives and contexts.
Participants in this study were drawn from a variety of professional backgrounds across 46 countries worldwide. However, some countries and regions contributed larger numbers of responses than others, with a small number of responses from North and South America. This may bias our overall results towards the responses of regions with more respondents. The legal frameworks and cultural factors may influence the classification and reporting of RPs in these regions in particular and may lead to different results compared to what we have observed in this study. Additionally, a relatively small sample of service users (n = 80) compared to clinicians (n = 517) may have reduced statistical power in some subgroup comparisons, potentially limiting the robustness of inferences drawn from service user perspectives. The purposive and snowball sampling techniques may also have introduced selection bias, as participants were likely those with a particular interest or experience in RPs, which could affect the representativeness of the findings.
The varying perspectives of mental health stakeholder groups towards RPs may influence their interpretation of case scenarios. These perspectives can be shaped by differences in training levels, cultural norms across regions, local policies and laws, disciplinary backgrounds, and prior exposure to RPs and mental health care. This highlights the need for future research to explore whether and how these factors influence stakeholders’ views regarding whether or not a certain action constitutes RP. Employing mixed-methods designs or subgroup analyses could help better understand these influences, thereby strengthening the evidence base to endorse and apply for internationally acceptable, stakeholder-informed RP frameworks.
Conclusions
This study underscores the need for internationally harmonized definitions and standardized classification systems for RPs, co-designed with diverse stakeholders, including clinicians, service users, family caregivers and researchers. Without such frameworks, efforts to measure, monitor and reduce RPs risk being undermined by inconsistent understanding and documentation. Embedding these definitions into training, reporting systems and research tools will support valid comparisons across settings, foster ethical and rights-based care, and ensure that reform efforts proceed with clarity, accountability and meaningful impact.
Supplementary material
The supplementary material for this article can be found at https://doi.org/10.1017/S2045796025100322.
Availability of data and materials
The de-identified raw data for this manuscript are available and can be accessed from the corresponding author upon request at zelalem.muluneh@monash.edu or zelalembe45@gmail.com.
Acknowledgements
The author team gratefully acknowledges the valuable contributions of all co-design panel members, whose diverse perspectives, rooted in both professional expertise and lived experience were essential in informing the study design. Their input played a crucial role in shaping case scenarios that authentically reflect real-world experiences with the implementation of RPs in adult mental health inpatient settings. Additionally, they contributed significantly to the refinement, validation and pretesting of the survey strategies and data collection instruments, which were translated into multiple languages. We also extend our sincere appreciation to the survey participants, whose responses form the foundation of the evidence presented in this manuscript, as well as to the institutions and individuals who supported the dissemination of the survey.
Author contributions
Zelalem Belayneh, Den-Ching A. Lee and Melissa Petrakis conceived the study, designed the methodology, developed the instrumentation, led data collection and contributed to manuscript writing. Zelalem Belayneh and Terry Haines managed the data, conducted the analysis and interpreted the results. Zelalem also drafted the manuscript, revised it through several rounds and handled correspondence with the journal for publication. Deborah Oyine Aluh, Justus Uchenna Onu, Giles Newton-Howes, Kim Masters, Yoav Kohn, Jacqueline Sin, Marie-Hélène Goulet, Tonje Lossius Husum, Eleni Jelastopulu, Maria Bakola, Sau Fong Leung, Kathleen De Cuyper, Eimear Muir-Cochrane, Yana Canteloupe, Emer Diviney, Lesley Barr, Jim Ridley and Didier Demassosso contributed to the study design, instrumentation, data collection and critically reviewed the manuscript.
Financial support
This research received no specific grant from any funding agency, commercial or not-for-profit sectors.
Competing interests
None.
Ethical Standards
This research received ethics approval from the Monash University Human Research Ethics Committee (MUHREC) (Project ID: 3642). The authors assert that all procedures contributing to this work comply with the ethical standards of the relevant national and institutional committees on human experimentation and with the Helsinki Declaration of 1975, as revised in 2000. Participation in the study was entirely voluntary. Before accessing the survey, participants were presented with an online explanatory statement outlining the study’s aims, scope, and inclusion and exclusion criteria. To proceed, participants had to read this information and respond to a consent statement. Only those who provided informed consent could access the survey questions; the survey automatically terminated for those who declined.
Open access
