Psychological safety in medical education: A scoping review and synthesis of the literature

Abstract Purpose Psychological safety (PS) is the belief that the environment is safe for risk taking. Available data point to a lack of PS in medical education. Based on literature in other fields, PS in clinical learning environments (CLEs) could support trainee well-being, belonging, and learning. However, the literature on PS in medical education has not been broadly assessed. Materials and methods In 2020, authors searched PubMed, Web of Science, CINAHL, Scopus, ERIC, PsycInfo, and JSTOR for articles published prior to January 2020. Authors screened all search results for eligibility using specific criteria. Data were extracted and thematic analysis performed. Results Fifty-two articles met criteria. The majority focused on graduate medical education (45%), and 42% of studies took place within a CLE. Articles addressed organizational and team level constructs (58%), with fewer descriptions of specific behaviors of team members that promote or hinder safety. The impacts of safe environments for trainees and patients are areas in need of more exploration. Discussion Future research should focus on defining specific organizational and interpersonal leader behaviors that promote PS, seek to understand how PS is determined by individual trainees, and measure the impact of PS on learners, learning, and patient care outcomes.

Psychological safety (PS) is increasingly recognized as a prerequisite condition for optimal learning in medical education (Nembhard and Edmondson 2006;Caverzagie et al. 2019;Kim et al. 2020).It is the perception that an environment is safe for interpersonal risk taking, exposing vulnerability, and contributing perspectives without fear of negative consequences (Edmondson 2018).In the organizational psychology literature, PS has been shown to increase worker engagement, job satisfaction, productivity, and team effectiveness (Edmondson 2002;Kessel et al. 2012;Singh et al. 2013;Edmondson and Lei 2014;Newman et al. 2017).In the patient safety and quality improvement literature, PS has been correlated with better communication on interdisciplinary teams, improved patient care outcomes, and reduced turnover in health care teams (Edmondson 2003(Edmondson , 2004;;Nembhard and Edmondson 2006;Havyer et al. 2014;O'Donovan and McAuliffe 2020).In medical education environments, PS frees learners from image management and reduces barriers to asking questions (Torralba et al. 2016;Torralba and Puder 2017;Hylton et al. 2019;Saddawi-Konefka and Scott-Vernaglia 2019;Tsuei et al. 2019;Johnson et al. 2020;McClintock et al. 2021).PS may help learners to disclose gaps in knowledge (Johnson et al. 2020) and support credibility and acceptance of feedback

Practice points
The presence of psychological safety (PS) supports learning by allowing students to be present in the moment, focus on the task at hand, and ask questions without fear of negative interpersonal consequences.Psychological safety has also been tied to physician well being, thriving in stressful situations, and a sense of belonging.Psychological safety is notably lacking in most clinical learning environments, largely due to the systems and structures of clinical teaching settings (e.g.hierarchy and power differentials within teams, frequent team member turnover, and constant evaluation of trainees).Psychological safety can be supported or threatened at each layer of an organization (organizational, team, interpersonal or based on preexisting personal attitudes and beliefs).Team leaders can create PS by setting expectations, supporting relationship building with teams, providing appropriate levels of autonomy, and emphasizing learning when errors occur.from educators (Eva et al. 2012).PS has also been increasingly tied to physician well being (Duggan and Clark 2022), thriving in stressful situations (Dewey et al. 2022), and a sense of belonging (Appelbaum et al. 2022;McClintock et al. 2022).When learners feel psychologically safe, they are able to engage fully, be present in the moment, and concentrate on the task at hand, rather than be consumed by constant self-censoring and/or competition with peers (Tsuei et al. 2019;McClintock et al. 2022).
Despite these many benefits, available data point to a lack of PS in clinical learning environments (CLEs) within the field of medical education.Threats to a learner's PS in these environments exist on a spectrum, from internal, consuming concern about grading on one end (Bullock et al. 2019;AAMC 2020), and outright external, supervisor hostility on the other (Gadson 2018;Osseo-Asare et al. 2018).According to responses to the Medical School Learning Environment Survey (MSLES) administered to 4262 medical students (who matriculated in 2010 and 2011), perceptions of the learning environment worsen as medical school progresses, with the most precipitous drop occurring in the 3rd year of medical school, as students transition into a CLE (Dunham et al. 2017).The 2019 American Association of Medical Colleges (AAMC) Medical Student Graduation Questionnaire found that 40% of learners were 'publicly embarrassed' at least once in medical school, most commonly by clerkship faculty or residents/interns (All Schools Summary Report [date unknown]).Following an event, 76% of students did not report the incident, most commonly citing fear of reprisal or belief that nothing would be done (All Schools Summary Report [date unknown]).These are not new problems (Sheehan et al. 1990), but we have yet to achieve system-wide solutions to trainee mistreatment and fear of retribution in CLEs, and our profession continues to struggle with high rates of burnout, drop out, and career regret related to negative experiences in the learning environment as early as the preclinical years (L.Dyrbye et al. 2021;L. N. Dyrbye et al. 2021).Psychological safety may represent the missing piece in the solution to learner mistreatment, burnout, and wellbeing.
Given the clear benefits of PS across multiple sectors, there is an urgent need to intentionally create and cultivate more psychologically safe learning environments that support learners' education, well-being, and sense of belonging (McClintock et al. 2021).Prior efforts to synthesize the literature related to PS have either been overly broad, so as to be difficult to translate and apply to the medical education context (Newman et al. 2017), focused on patient safety without attention to medical education (Turner and Harder 2018;O'Donovan and McAuliffe 2020), or solely focused on simulation (Dieckmann and Krage 2013;Lackie et al. 2023) and debriefing (Allen et al. 2018), without additional attention to the complexities of real-life CLEs.As a result, the literature on PS in the medical education learning environment has not been systematically examined or synthesized to understand how PS is created or destroyed at the personal, interpersonal, and organizational level.We do not have a synthesized understanding of the impact of PS (or lack thereof) on learners at varying levels of training.Without this careful mapping, it will remain difficult to create a coherent and focused research agenda that can delineate, and support implementation of, evidence informed practices for promoting PS in medical teaching environments.Here, we describe a scoping review to answer the intentionally broad question 'What has been published on PS in medical education?' with specific attention to (1) population: learner stages, (2) context: educational contexts/learning environments, (3) decision making: how decisions about PS are made, (4) description of outcomes, and (5) gaps in existing research that may guide future research agendas.

Methods
We followed a rigorous scoping methodology first outlined by Arksey and O'Malley (2005) and later refined by Levac et al. (2010), the six phases of this scoping review included: (1) identifying the research question; (2) identifying relevant studies; (3) selecting the studies; (4) charting the data using predetermined variables; (5) organizing, summarizing, and finalizing the results and (6) consulting with key stakeholders.
We began by establishing our research group with in-depth knowledge of PS within the medical and organizational psychology literature.We defined our search question broadly by consensus and engaged a health sciences librarian and group expertise to assist with identification of databases and search terms to be used.

Research question
This scoping review was to define 'what has been published about PS in medical education?'This question was intentionally broad, in order to capture a large breadth of literature in our initial search.We focused on medical education by only including studies relevant to any stage of physician training or practice.Though we were collecting data about the type of study and available evidence, we did not specify or evaluate the quality of the evidence in our review of the literature.

Data sources and search strategy
The search was initiated on 13 January 2020, starting with PubMed.Six other databases were also searched, including Web of Science, CINAHL, Scopus, ERIC, PsycInfo, and JSTOR.We enlisted the help of a health sciences librarian to develop the search terms through an iterative process.Within PubMed, the search terms used were as follows: ('Psychological safety') AND ('Education, Medical' [Mesh] OR medical education or clinical learning environment or medical training or undergraduate medical education or internship or residency or fellowship).Within the other databases, the following search terms were used: ('psychological safety' or wellness) and (undergraduate medical education or graduate medical education or clinical learning environment or medical training or internship or residency or fellowship).The search was limited to articles published up until 13 January 2020 and written in English.We excluded non-English language papers due to limited resources for formal translation of non-English manuscripts.We excluded literature that did not pertain to physician training, as the author's believed this to be a unique social environment, and one that all authors understood well, having been medical trainees at one time.
The initial search query resulted in 214 records (Table 1).From these records, 19 duplicates were identified and removed, yielding 195 articles.Paperpile was used to import all of the citations from the databases and journals.

Screening, selection, and data extraction
Two investigators (AHM, JJ) reviewed the title and abstract of all 195 records to determine relevance to the research question.Authors read the first 10 studies and then conferred to define which studies were not relevant to the search question and further clarify and delineate inclusion and exclusion criteria.A Google form was used to capture each study, and two investigators independently determined whether each study was to be included or excluded, which generated a spreadsheet showing agreement and any discordant determinations, which were discussed until consensus was reached.Studies that were excluded as not relevant to the research question were about quality and safety or quality improvement alone, without a mention of education, learning environment, or physicians.Studies that had no mention of medical education were also excluded.
A total of 148 studies were excluded, leaving 47 articles.Following full-text review of all 47 manuscripts, two additional studies were eliminated (one not relating to the research question and the other did not have the full text available in English).Of the remaining 45 articles, several key manuscripts were identified and their references were reviewed to find records not previously found in the initial search.This yielded seven additional records for a final total of 52 included articles (Figure 1).

Charting the data
Once the complete list of manuscripts was agreed upon, the first 10 papers were read in full by JJ and AHM to define and edit the data-charting form.Google Sheets was used to design a spreadsheet for data collection.The following variables were included in the data-charting form: Title, author, year of publication, country of origin of the manuscript, type of article (original research, perspective, review), context (theoretical, simulation, class-room, or clinical teaching environment), learner stage (undergraduate medical education (UME), graduate medical education (GME), continuing medical education (CME), or undefined), and research method used (see Supplemental Material for charting form).The country of origin of the manuscript was based on where the study took place, or the country of origin of the 1st author if no intervention or study took place.No critical appraisal of the quality of the evidence was undertaken.One investigator (KB) read all papers in full and extracted data.Extracted data were checked by a second investigator for accuracy (AHM).The data extraction phase of the scoping review was completed 30 January 2021.

Organization, thematic analysis, summary, and reporting
Based on previous work describing the social and structural supports for PS (Tsuei et al. 2019), and our interest in learner decision making and outcomes, we used an objectivist deductive approach to thematic analysis, based on two existing theoretical frameworks: the social ecological model and a behaviorist model of decision making from social cognitive theory (Bandura and Walters 1977;Bandura 1986).See Figure 2 for a conceptual model of how these two frameworks were blended.After initial data extraction, AHM applied the theoretical frameworks to the extracted data and identified subthemes within the literature.Findings from thematic analysis and creation and classification of subthemes were discussed as a group, and consensus taken between AHM, JJ, and TF where discrepancies or uncertainties arose.Table 2 describes these themes and subthemes and their prevalence in the literature.
The social ecological model (Bronfenbrenner 1974) was used to organize studies based on the level of influence described or addressed within the study design or reporting.This was chosen because of the inherent nesting and interaction of individuals, team members, and organizational systems within medical education (Crowe et al. 2017;Vanstone andGrierson 2019, 2022).For our purposes, these layers included individual, interpersonal, team, and organization.The individual from the social ecological model was 'pulled out' as its own entity, reflecting our belief that the person and their environment interact with each other.For our purposes, the levels of the social ecological model were defined as individual trainee knowledge, skills, attributes, or beliefs (individual), observed interpersonal behaviors that promote or hinder PS (interpersonal), team culture and beliefs of team members or leadership structures within the team (team), and the organizational influences external to team, relating to larger institutional structures, schedules, or competing demands placed on physicians (organization).
Additionally, we sought to investigate how PS is assessed and acted upon by team members.As PS is a cognitive decision made quickly by an individual team member at a discrete point in time (McClintock et al. 2022), we used a behaviorist model of decision making that applied constructs from social cognitive theory (Bandura and Walters 1977;Bandura 1986) to classify social and structural behaviors and dynamics present (stimulus), cognitive processes taking place within individuals (mediational processes), response to stimuli (response), and educational or learning outcomes described.Our conceptual model also highlights the way that stimuli from other team members and the organization both impact, and are impacted by, the individuals in a bidirectional manner as part of the mediational processes.

Consulting with key stakeholders
After organizing, summarizing, finalizing our results, and drafting an initial manuscript, we consulted with experts in the field, chosen based on their number of publications included in our reviews.We solicited feedback on the extent to which our scoping review and summary of the literature captured the existing literature and appropriately summarized the main themes in the literature (Levac et al. 2010).In response to feedback, we took the additional step of analyzing the included articles for the measurement tools, if any, that were used to measure PS in each study.

Article characteristics
We included 52 articles in this scoping review (see supplemental material for a summary of studies), and Figure 1 for a PRISMA-ScR (Tricco et al. 2018) flowchart of the scoping review.Table 3 reports characteristics of the included articles.
Table 3 reports descriptive characteristics of the included articles.The majority of articles were published in the last 5 years, between 2016 and 2020 (n ¼ 38; 72%), with only 11% published in the decade between 2001 and 2010 (Table 3), demonstrating significant increase in attention to the topic in recent years.Thirty-two of the 52 articles (60%) included in the scoping review originated from the United States.There were seven articles (11%) from Canada, and five or fewer articles from each of the following countries: Australia, Austria, Ghana, Ireland, Netherlands, Singapore, Switzerland, and the United Kingdom (Table 3).Most articles were original research (n ¼ 35; 66%), followed by those categorized as perspectives, opinions, or letters to the editor (n ¼ 11; 21%), and seven articles (13%) were structured reviews.The majority of studies (57%) used some form of qualitative methodology, with quantitative and mixed methods less commonly used (19% and 15%, respectively) (Table 3).Five of the 52 articles (9%) were commentary or opinion pieces (Table 3).Data were collected in several different ways in the 52 studies, occurring most commonly via surveys (n ¼ 21) and interviews (n ¼ 12) (Table 3).Many studies employed more than one method or data collection tool.

Populations: learner stages studied
Nearly, all of the 52 included articles (94%) specified a learner stage, with the 45% of articles focused on residents and fellows from GME programs (n ¼ 24) (Table 3).Articles focused on medical students in UME programs accounted for 17% of articles (n ¼ 9), and 11% focused on faculty undergoing CME programs (n ¼ 6) (Table 3).Eleven articles (21%) included a combination of two or three learner groups, and three articles (6%) did not specify a learner stage (Table 3).

Educational contexts
The learning environments of the 52 included articles were categorized as: classroom, simulation, clinical, and theoretical (Table 3).Classroom environments were involved in the fewest number of articles (n ¼ 5; 9%), while simulated learning environments that strive to replicate clinical settings using manikins and medical equipment in order for learners to practice their clinical skills, were assessed in 21% of included articles (n ¼ 11).Clinical learning environments represented less than half of the articles: 22 (42%).Fifteen of the 52 included articles (28%) articles were categorized as having a theoretical learning environment, as they were written as commentary, perspective, or review articles that did not specify a learning environment.

Measures used
Twenty-one studies used survey scales to measure various aspects of the learning environment or to evaluate novel tools to improve PS.Of these, the most commonly used scale (nine studies) was Edmondson's seven-item validated scale (A.Edmondson 1999), or specific questions from the scale, without the use of the entire measure.Other commonly used scales were perceived organizational support (Shore and Tetrick 1991;Hutchison 1997) (three studies), and the Veterans Affairs Learners' Perceptions Survey (Keitz et al. 2003) (two studies).Less commonly used measures included the Speaking Up About Patient Safety Questionnaire (Richard et al. 2021), Team Learning and Beliefs Questionnaire (Van den Bossche et al. 2006), and The Quality Improvement Team Questionnaire (Albritton et al. 2019).Several of these survey tools contain one or more of Edmondson's original survey, as mentioned above.
Five studies designed their own survey method or used a survey but did not attempt to measure PS directly.

Thematic analysis and synthesis
Stimuli: facilitators and inhibitors of psychological safety Themes related to organizational factors and team culture were most commonly described, with relatively fewer studies focused on team structure, observed interpersonal behaviors that could be taught to faculty, or pre-existing antecedents in trainees themselves (Table 2).For a full summary of barriers and facilitators, see Table 4.
Organizational and team culture attributes that promoted PS overwhelmingly described concepts related to interpersonal communication and positive relationships, High productivity pressure, highly complex tasks in medicine Team Team leadership structure without associated social hierarchy Hierarchical structure with Large power distance between members though few studies described specific behaviors that could create these conditions.Organizational attributes included perceived organizational support for work-life balance and an emphasis on team-work, time allotted to support relationship building and social cohesion, protected time for teaching and learning, and lack of formal assessment.Cultural aspects of psychologically safe teams included team leader enthusiasm for the profession and teaching, patient care and learning as core team goals, coaching-oriented feedback, a culture of accountability, and a social contract of respect among team members.Observable behaviors that were named included explicit statements of objectives or expectations of team members, providing an explicit invitation of dialogue and speaking up, treating team members with respect and downplaying of power differentials (i.e.knowing names and faces, addressing learners directly when speaking), asking open ended questions and allowing time for learners to answer, disclosing their own learning process and challenges, and providing autonomy.Individual level factors for learners themselves included affective attributes (comfort with uncertainty, growth mindset), behavioral coping strategies, and nonmodifiable attributes that impacted place in a social hierarchy such as physical attributes, gender, and learner stage.
Conditions that detract from PS were most often related to low perceived organizational support, poor coping with stress in leaders, disrespectful interpersonal communication that reinforced social hierarchy within the team, low quality relationships, and a culture of blame.

Mediating processes
Only one study asked learners directly to describe their decision making process to determine the safety of an environment.Authors outlined a process of primary appraisal to determine if the situation is stressful or not, followed by a secondary appraisal to determine whether the stress is harm or threat (negative) or positive challenge, which further dictated their self-described ability to learn new skills in the operating room.

Learner responses
When perceived to be safe, learners described speaking up, asking questions, disclosing knowledge or skill gaps, embracing challenges, providing more honest feedback, and accepting feedback that was given.Learners additionally described reduced self-monitoring, an ability to focus on learning, and a reduced focus on evaluation and grading.Finally, trainees in psychologically safe environments were more likely to want to stay within their institution after residency.When perceived to be unsafe, studies described fear, silence, hesitation to speak up as well as a focus on evaluation and use of self-monitoring processes to manage image.

Learning or educational outcomes
The majority of studies sought to describe positive outcomes associated with the presence of PS.These included engagement and implementation of QI initiatives in team members, positive ratings on surveys from the US Accreditation Council for Graduate Medical Education (ACGME), higher rated satisfaction from the CLE and more effective debriefings that allowed for effective learning and emotional support following stressful situations.Measured outcomes from negative learning environments were less often reported, with burnout and failure to report unsafe clinical practices being the only measured outcomes.Despite fewer studies seeking to measure negative outcomes, many studies demonstrating a positive outcome with PS described a less desirable outcome in comparison groups where PS was reduced or absent relative to comparison group.

Discussion
This scoping review and synthesis demonstrates that while PS is a concept of increasing interest in medical education, significant gaps do exist in the current evidence-base that present opportunities for focused investigation.In particular, additional research in the UME setting is warranted.The gap in UME related literature is especially notable, as medical students are a unique and particularly vulnerable group to feeling unsafe in the learning environment because of the component of grading and evaluation that is nearly always present (Bynum and Haque 2016), and their traditionally lower status in medical team hierarchies (Vanstone andGrierson 2019, 2022).Other areas where further investigation is warranted include studies that more specifically describe team leader behaviors that create desired organizational and cultural change, studies that foster a deeper understanding of how learners' pre-existing internal attitudes and beliefs impact determinations of safety, and studies that describe how learners mediate the various inputs to make determinations of safe or unsafe.Lastly, further investigation is needed into whether the presence or absence of PS impacts larger outcomes of interest in medical education, such as for measures of learning, burnout, drop-out, belonging in the learning climate, and patient care.
Hierarchy and power distance were highly prevalent themes in the literature, and concepts related to team leaders and leadership behaviors were commonly highlighted as areas to focus change efforts.Rigid social hierarchy was nearly universally described as negatively impacting PS, whereas attempts to decrease power distance and foster team member inclusion were tied to increasing PS in team members.This highlights a major cultural shift facing medicine learning communities: that being a team leader is a role within team structure, but is a problematic negative influence on the learning climate when conflated with social worth and used to enforce a social hierarchy on 'lower ranking' team members.The role of hierarchy also highlights important implications for planning and studying potential interventions, as studies clearly point to the role of leaders in fostering culture change from the 'top down' (Rider et al. 2018;Umoren et al. 2021) is likely to be more successful than attempting to do so from the 'bottom-up' (Edmondson 2003;Calhoun et al. 2013Calhoun et al. , 2014;;Bould et al. 2015;Crowe et al. 2017).The highly important role of leadership in creating PS should not come as a surprise, given that the concept is derived from the organizational literature and that the framework itself describes key 'leadership tasks' that create PS.
The vast majority of existing research lacks sufficient granularity to fully understand how culture change and PS are created or operationalized in the learning climate.While a large portion of the studies are original research, 40% used surveys to gather data, this method cannot provide insights that promote an understanding of individual experiences in the medical learning climate.This is problematic since we know PS is an individual perception of the learning environment (Appelbaum et al. 2016).Similarly, while over half of the literature contains recommendations for team structure or culture, they are often reported as larger cultural shifts in medical education culture or general strategies for debriefing, rather than deliberate interpersonal or specific programmatic structural changes that can be implemented.
The finding of primary and secondary appraisal of the learning environments as a decision making tool for determining safe or unsafe, and the impact it may have on learning, is well supported by literature in human stress and trauma (Van der Kolk 2014; Brown et al. 2021).A 'quick' path through the amygdala signals stress and triggers fight or flight responses, whereas a 'long path' can access the frontal cortex when threat is not detected, which allows for more complex mental processing and likely learning.This model of neural pathways and their impact on learning also aligns with studies describing PS as a mediator of effective learning, with some studies referring to it as 'the engine, not the fuel' (Caverzagie et al. 2019;Kim et al. 2020;Lateef 2020).In this way, this engine (PS) is considered an integral part of the climate, but does not fully encompass or define the whole learning environment.When PS is present, there are fewer perceived risks associated with learning (Bynum and Haque 2016;Edmondson 2002).In psychologically safe environments, learners can focus on the task at hand, ask questions, acknowledge gaps, and are free from the cognitive load related to the stress of image management and evaluation (Tsuei et al. 2019;Johnson et al. 2020;McClintock et al. 2021).Psychologically safe work environments have also been described as a precursor to physician engagement (Duggan and Clark 2022), thriving (Dewey et al. 2022), and a sense of belonging in trainees that are underrepresented in medicine (Singh et al. 2013;Razack and Philibert 2019;Mateo and Williams 2020).It is also not surprising then, that many best practices in medical education (Harden and Laidlaw 2020) contain skills that map to a PS framework, and that the presence of PS may facilitate learning by supporting a 'rest and digest' mental state and allowing access to the frontal cortex.
This study is limited by its focus on physician training only, and valuable studies from the fields of nursing or training of advanced practice providers may have been excluded.This was an intentional choice because the writers of this paper are all physicians and understood the physician culture and training process, helping us to interpret findings in context.Additionally, while we recognize there is a broad body of literature on the learning environment generally speaking which may be relevant to the creation of PS, we did not include studies that did not explicitly name or mention PS, as we were interested specifically in how the concept of PS has been studied and reported.This choice was made because PS is a discrete concept with a clear definition of its own, whereas learning environment or climate alone may contain studies that lack or fail to consider and address all components of PS, and thus cannot be considered to be studying or evaluating comparable concepts.While this choice may have sacrificed our ability to capture potentially relevant studies, this was an intentional choice to allow us to focus on the specific concept of interest.This study, and the applicability of findings, is also limited by a dearth of primary evidence and a current predominance of expert opinion and perspective pieces.As our study method did not include an appraisal of the evidence, these results are presented in parallel with any primary or objective data, but may not ultimately represent the same level of rigor behind all reported findings.Lastly, we did not evaluate the quality of the data in each study, and bias may be present in the larger literature, which our study was not designed to detect.
Psychological safety in medical education must be a priority if we are to optimize learning and wellbeing for trainees, and address the existing challenges in the learning climate.Existing evidence points to the important role of team and organizational leaders in establishing safety through interpersonal behaviors that promote a sense of safety in team members, and the importance of institutional systems and structures that facilitate and support team leaders to do so.Currently, the literature is largely focused on GME and CME, and lacks specific evidence-informed strategies to improve PS or demonstrate its utility in larger outcomes of interest.Future research efforts may be scaffolded and directed by the social ecological model and cognitive theory findings highlighted in this manuscript.Research to define the specific organizational structures and interpersonal leader behaviors that promote PS, understand how PS is determined, and measure the impact of PS on learners, learning, and patient care outcomes is needed.Focusing first on the 'engine' that drives medical education forward is paramount so that the 'fuel' of teaching skills and patient exposure can more effectively reach our learners.

Figure 1 .
Figure 1.Flow diagram of scoping review process.

Figure 2 .
Figure 2. Synthesis of findings.Table 2. Tabulation of concepts reported in studies in the 53 articles included in the scoping review.Theme Subtheme Number % Organizational factors Perceived organizational support, organizational factors impacting PS 18 34 Team Team leader behaviors, hierarchy/power imbalance, collaboration, leader mindset, speaking up about patient safety.Description of leadership structure or imposed/flattened social hierarchy, naming of specific observed behaviors that promote or hinder PS 31 58

Table 1 .
Sources of evidence.

Table 3 .
Characteristics of the 53 articles included in the scoping review.

Table 4 .
Summary and mapping of 'stimuli' that promote or inhibit psychological safety within the social ecological model.