Prevalence of intimidation, harassment, and discrimination among resident physicians: a systematic review and meta-analysis

Background The aim of this study was to establish the prevalence of intimidation, harassment, and discrimination (IHD) reported by resident physicians during their training, to identify factors associated with reported IHD, and to identify adverse sequalae associated with IHD. Methods This review followed the PRISMA guidelines. Eight electronic databases were searched for cross-sectional studies reporting the prevalence of IHD among resident physicians. Prevalence estimates were pooledacross studies using random-effects meta-analysis, with variance stabilization using Tukey double arcsine transformation. Heterogeneity was assessed with forest plots, the I2 statistic, subgroup analyses, and multivariate meta-regression. Results 52 cross-sectional studies were included in the meta-analysis. The overall pooled prevalence of IHD was 64.1% (95% confidence interval [CI], 51.0-77.1). Verbal, physical, and sexual IHD were the most common forms of IHD reported by residents. Training status (55.5%), gender (41.7%), and ethnicity (20.6%) were the most commonly cited risk factors for IHD. The most common sources of IHD were relatives/friends of patients, nurses, and patients (cited by 50.7%, 47.8, and 41.7%, respectively). Conclusions The prevalence of IHD among resident physicians is high and associated with multiple negative outcomes, including burnout. Despite the availability of multiple anti-IHD interventions, reports of IHD appear to be rising in many residency programs.


Background
The Canadian human rights commission defines intimidation, harassment, and discrimination (IHD) as unwanted physical or verbal behaviours that are offensive or humiliating, which can occur on the basis of race, religion, sex, age, disability, or other grounds. 1 Specifically, intimidation refers to the use of authority to inappropriately influence behaviour. 2 Harassment is defined as unwelcome or vexatious conduct that occurs on the basis of the perceived status of the target, be it ethnicity, gender, sexual orientation, age status, or other attributes. 3,4 Discrimination denotes to the unjust or prejudicial treatment of different categories of people. 5 In learning environments, IHD often induces fear or anxiety in the learner, causing generally detrimental effects on the learner's ability to succeed. 6 In recent years, the deleterious impacts of IHD on medical trainees across all stages of training have been recognized internationally. 7 The psychiatric sequelae of exposure to IHD have been particularly well studied, with a recent systematic review finding that IHD increased the risk anxiety disorders, sleep disorders, eating disorders, posttraumatic stress disorder, and suicide attempts by three to 16-fold -regardless of sex or age. 8 Therefore, IHD appears to be associated with an increased prevalence of psychopathology as well as specific patterns of psychopathology. 9,10 IHD also has negative impacts on learning and educational outcomes among medical trainees, with recent studies showing that trainees are less likely to pursue a medical speciality that they perceive to be particularly hostile. 2 Exposure to IHD during training also influences a trainee's academic trajectory because it affects their ability to communicate, concentrate, and collaborate. 11,12 Medical students and resident physicians are especially vulnerable to IHD for a number of reasons. 13 As trainees are learners who are dependent on their supervisors and senior residents for promotion through the medical education system, this creates a power gradient, which sets the stage for IHD. 14 In medicine, IHD behaviours are often rationalized as 'rites of passage' or 'beneficial to training' 15 -not unlike the role of distorted cognition in initiating and maintaining cycles of sexual abuse and assault. 16 Similarly, a number of cultural factors within medicine that discourage reporting of IHD and encourage victimization have contributed to the e99 phenomenon of "whistleblowing" within medicine. 14,[17][18][19][20][21] In response to mounting criticism, 14 10,30,33 Despite these efforts, there is evidence that IHD continues, and may even be on the rise within medicine. 6 National surveys conducted by Resident Doctors of Canada in 2012 and 2018 identified that 73% and 78.2% of Canadian residents, respectively, reported at least one instance of IHD during residency, showing a 5% increase over this six-year span. 34,35 While these estimates may also reflect decreased barriers to reporting, particularly as many surveys are kept anonymous, the effectiveness of anti-IHD infrastructure is unclear. 14,18,[22][23][24] Previous reviews have described IHD across different Canadian medical education settings. In 2014, Karim and Ducherer reported that between 45% and 93% of residents reported at least one instance of IHD during residency. 6,36 In another study, two Canadian family medicine programs reported that 33.7% of trainees had experienced IHD during residency. 2 In another survey, 36% of neurosurgery residents reported IHD during their training. 37 Internationally, nearly 50% of surveyed medical students, residents, and fellows report experiencing or witnessing at least one form of IHD during their training in the United Kingdom, United States, Japan, Nigeria, and Canada. 6,12,36,38 A prior, comprehensive systematic review conducted in 2014 by Fnais and colleagues 36  The authors reported that residents cited gender discrimination as the most common form of abuse (n = 3 studies, 1,315 residents, prevalence: 66.6%, 95% CI: 58.7%-74.5%), followed by verbal harassment (n = 12 studies, 9,867 residents, prevalence: 58.2%, 95% CI: 45.5%-70.9%). Among residents, the least common form of IHD was racial discrimination (n = 3 studies, 3,261 trainees, prevalence: 26.3%, 95% CI: 24.2%-28.3%). Heterogeneity was significant across these studies.
Allied health professionals, including nurses, 39,40 physician assistants, 41 home health aides, 42 and social workers, 43 have also experienced high rates of IHD. Converging sources of data suggest that IHD is not limited to North America or to medical trainees specifically, but may represent a larger, and more systemic issue that affects medicine on the whole.
Several factors related to IHD research make it an intrinsically difficult topic to study by way of metaanalysis. The overreliance on cross-sectional surveys (prone to recall bias) and the subjective nature of IHD terminology (often utilizing overlapping definitions) increase heterogeneity between studies, limiting the extent to which results can be pooled. Still, several previous reviews have pooled the available IHD literature 2,6,8,33,36,38 by collapsing multiple forms of IHD, several groups of trainees, and countries or regions, often due to the limited number of studies identified by any individual review. While the review done by Fnais and colleagues 36 is the most comprehensive to date, there are opportunities to expand their findings by identifying more residentspecific studies. This may in turn enable additional subgroup analyses, greater description of betweenstudy heterogeneity, and an improved understanding of the contextual factors involved in IHD among residents.
Therefore, the objective of the study was to establish the prevalence of IHD reported by resident physicians during their training, to identify factors associated with reported IHD, and to identify adverse sequalae associated with IHD. e100

Ethics
Research ethics board approval and consent procedures were waived as this study was a metaanalysis of publicly available studies.

Search strategy
A systematic review protocol was developed with the support of an experienced research librarian using the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines. 44 Medical Subject Headings (MeSH) and free-text searches related to IHD among resident physicians were used to search the following seven electronic databases: PubMed, MEDLINE, EMBASE, PsycINFO, CINAHL, Allied and Complementary Medicine (AMED), and the Cochrane Library. The following MeSH terms were used when searching MEDLINE: "Internship and Residency," "Resident," "Medical Residency," "Intimidation and Harassment." We searched each database from inception to August 1, 2018, with an updated search conducted on May 28, 2019. Specific definitions and examples of IHD are described in Appendix A, while the detailed search strategies for each database are described in Appendix B. The reference lists of included studies and reviews were scanned for additional articles. The ProQuest database of dissertations and theses was also searched for relevant grey literature to supplement findings from published studies. 45,46

Eligibility criteria
English-language studies reporting the prevalence of self-reported IHD among resident physicians-or those where the prevalence could be computed using raw data reported by the studies-were eligible for inclusion in this review. If studies reported data on other groups (such as staff physicians, medical students, or allied health practitioners), residentspecific data were extracted. No restrictions were placed on geographic location, stage of training, type of residency program, date of dissemination, or subtype of IHD.

Outcome measures
Outcome measures were defined a priori for consistency with previous reviews of IHD. 6, 36 The primary outcome was the prevalence of IHD. Secondary outcomes included: • Sources of IHD (staff physicians, residents, medical students, patients, relatives of patients, nurses, and other staff) • Risk factors for IHD (gender, training status, sexual orientation, ethnicity, culture, language, and other factors) • Reporting of IHD (awareness, reporting rates, perceived barriers to reporting) • Impacts of IHD (general satisfaction, quality of life, self-rated mental health, mental health screening) • Proposed resources or solutions to IHD (education, training, policies, infrastructure, supports, wellness, access to an ombudsperson, access to a physician or counsellor, and career or other forms of advice) Crude prevalence estimates were determined by dividing the total number of residents reporting IHD-overall or by subtype-by the total number of survey respondents. For example, if 10 from a total of 50 survey respondents reported sexual harassment, the prevalence of sexual harassment from that particular study was calculated as 20%.

Selection of studies
Both authors screened all articles for inclusion using a two-stage process, supported by Rayyan, a webbased systematic review software. 47 During the first stage, articles were excluded on the basis of title and abstract. Articles deemed relevant by either author progressed to the second stage, where full text versions of all articles were screened against the eligibility criteria. All disagreements were resolved by consensus.

Data extraction and management
A data collection sheet was developed in Microsoft Excel; study coding variables are described in Appendix C. Both authors independently collected data, and discrepancies were resolved by consensus. If there were multiple companion publications reporting on data from the same population, only the most recent analysis was considered. Across studies, prevalence was usually reported as percentages or proportions. Missing data were not included. e101

Assessment of risk of bias in included studies
To assess study quality, the Risk of Bias tool (RoBT) for prevalence studies developed by Hoy and colleagues 48 was used because of its use in other meta-analyses of prevalence studies, 49-52 its high interrater reliability, 53,54 previous validation, 55,56 and simplicity. 55,56 This RoBT consists of ten items addressing four domains of bias and an eleventh summary risk of bias item (described in Appendices F and G). The four domains of bias assess external and internal validity using forced choice responses (yes/no). Each "yes" received a score of 1, while "no" responses received a scored of 0: the total possible score was ten. "Low" risk of bias was defined as scoring 0 to 2 points, "moderate" was defined as 3 or 4 points, while "high" was defined as 5 or more points. [49][50][51][52] Both authors independently scored studies using the RoBT; all disagreements were resolved by consensus. Inter-rater agreement was quantified with the kappa coefficient. 57 Kappa was 0.83, indicating moderate-high agreement across raters.

Statistical methods
All analyses were conducted using the Open Meta Analyst. 58 IHD prevalence estimates were pooled using a random-effects model, generating an overall prevalence and accompanying 95% confidence intervals [CIs]. Statistical heterogeneity was assessed using tau 2 , Q, and I 2 and with forest plots. 59 To stabilize variance across proportions and percentages, the arcsine transformation was applied, which allowed the sampling distribution to better approximate a normal distribution. [60][61][62] Leave-oneout meta-analysis was applied as a method of sensitivity testing to measure the robustness of the results. [63][64][65] Results

Results of the search
The literature search yielded 2,941 unique citations ( Figure 1). From these, 2,876 were excluded because they did not include resident physicians (n = 1,582), did not report IHD (n = 786), did not provide primary data (n = 213), were not published in English (n = 25), or were not eligible study designs (n = 270). The remaining 65 records were obtained and reviewed in full. Reasons for exclusion at the full-text review stage included that the study did not provide primary data (n = 5), did not report IHD (n = 3), did not report prevalence or risk factors of IHD (n = 4), or did not include resident physicians (n = 1). A total of 52 crosssectional studies fulfilled the inclusion criteria for this meta-analysis.

Methodological quality
Among the 52 studies, 37 (71%) recruited a nationally representative sample population. However, only ten (19%) used a random sample to obtain a truly representative sample of the average resident physician. Eighteen studies (35%) used a validated survey instrument. All but two studies used the same mode of data collection for all study respondents. 66,67 Forty-seven studies (90%) provided definitions and example of IHD for respondents; the remaining five studies intentionally excluded IHD definitions to promote completion of the survey without restraint. Forty-one studies (79%) had a survey response rate greater than 50%.
The majority of studies did not provide full demographic descriptions of their resident populations. For example, 13 (25%) studies did not report the sex distribution of respondents, 29 (56%) did not report the whereabouts of the residents' training, 20 (39%) did not report the residents' year of training, and 37 (71%) did not report the residents' age distribution. None of the studies controlled for age or gender to improve the comparability of results across studies.

Post hoc subgroup analyses
Post hoc subgroup analyses were conducted to identify trends in the prevalence of IHD. There were no statistically significant differences in the prevalence of IHD in studies using national, local, or regional samples of residents, or in studies that provided definitions of IHD (relative to those that did not). However, IHD was more prevalent among female residents compared to male residents (p < 0.05) and amongst residents of visible minorities (including Asian, Middle Eastern, or Black residents) relative to Caucasian residents (p < 0.05). However, neither the proportion of female respondents nor the response rate to the survey were significantly associated with the overall prevalence of IHD in metaregression.

Reporting of IHD
The pooled prevalence of reporting awareness of how to report IHD incidents was 41.0% (N = 2311/4416; n = 7 studies; 95% CI = 20.8-61.1%; I 2 = 99.548%), while the pooled rate of reporting of IHD incidents was only 26.9% (N = 2080/9155; n = 9 studies; 95% CI = 16.2-35.2%; I 2 = 99.747%). The top barriers to reporting were fear of retaliation, that residents believed they could handle the incident on their own, that they believed the IHD incident was not significant enough to warrant reporting, that reporting would not improve the situation, or a lack of awareness of reporting infrastructure.

Impact of IHD
Twenty-four studies (46%) reported the impact of IHD on residents, including self-reported psychological sequelae (n = 13), positive screening for common mental health disorders (n = 4), measures of overall life satisfaction (n = 4), self-report of subjective feelings of safety at work (n = 2), and overall disruption in residents' ability to work (n = 2). The majority of psychological sequelae of IHD reported by residents was negative and harmful, including perceived hostility at work, increasing emotionality, anger, frustration, burnout, diminished resilience, increased substance abuse (as a maladaptive coping mechanism for stress), anxiety, depression, fear, and feelings of inadequacy. 75

Sensitivity analysis
Heterogeneity was significant across these studies; however, the prevalence estimates across metaanalyses were robust to the leave-out-one test of significance. e107

Summary
The present study provides an updated systematic review and meta-analysis on the prevalence of IHD among resident physicians. To the best of our knowledge, this review is the largest to specifically focuses on IHD among resident physicians, including 63,378 trainees in total.
Residents described several negative effects of IHDincluding poorer overall mental health, lower quality of life, and decreased satisfaction with work. Residents frequently screened positive for multiple psychiatric disorders, including depression, anxiety, substance abuse, and suicidality. However, residents also proposed several potential resources and solutions for addressing IHD, including education about IHD, decreasing barriers to reporting IHD, and increasing access to supports.

IHD among psychiatry residents
Although the overall prevalence of IHD among psychiatry residents (59.5%) was similar to the overall mean (64.1%), the prevalence of sexual IHD (39.5%) and physical IHD (46.0%) were significantly higher (p < 0.05). Although the latter may be due to the overrepresentation of violence-focused studies among psychiatry trainees, the former may relate to the unique characteristics of the psychiatric discipline, which we attempt to explain here.
Several studies have shown that the ways in which residents and medical students experience psychiatry training is different than in other specialties, which may extend to experiences of IHD during training. [68][69][70] However, the available literature exploring the nature of the learning and training environment in psychiatry is controversial, with one study suggesting that perceptions and personal experiences of IHD within the psychiatric learning environment are low. 71 Previous studies have consistently demonstrated elevated rates of IHD among psychiatry trainees relative to other trainee groups. 36,[72][73][74] This is likely related to unique power differentials, 24 which may be driven by prejudicial views toward the discipline of psychiatry that disproportionately sensitize trainees to IHD. 75 Factors related to compassion fatigue and burnout may also be more common among psychiatry trainees given the empathic demands of their work and vicarious exposure to IHD. 69 Psychiatrists are also paid the least among medical specialties-a systematic factor that is outside of the immediate control of most physicians. [76][77][78] As IHD is considered a major risk factor for burnout, with a 2016 national survey of Canadian psychiatry residents finding that 21% reported symptoms of burnout, IHD encounters appear to play a critical role in psychiatry resident experiences of burnout. 79

Heterogeneity of results
A major source of heterogeneity in our meta-analysis arose from variations in the type of IHD terminology-including specific definitions or examples-used across studies. Although the majority of studies (90.4%) provided definitions, five [80][81][82][83][84] intentionally did not in order to limit the restraint on respondents. However, the presence of IHD definitions was not associated with IHD prevalence. Among the studies incorporating IHD definitions, the majority used terminology that were based on how the respondent perceived IHD, rather than objective criteria. For example, most studies included "uncomfortable propositions of a sexual nature" and "unwanted sexual banter" under "sexual harassment", so respondents that perceived sexually themed interactions as unwanted would respond positively to such questions.

Significance of findings
This review indicates that IHD is a highly prevalent phenomenon among residents of most specialties. Specific risk factors that may increase vulnerability to IHD among resident physicians are not consistently defined in the academic literature. However, this review found that female residents and those belonging to visible minorities were at greater risk.
While the culture of medicine and residents' lack of control over their schedules have been previously cited as the biggest barriers to seeking mental health e108 care, 85 it remains unclear if these factors extend to vulnerability to IHD. However, studies of IHD in other professions may be relevant in defining the vulnerability profile of IHD within medicine. Among army soldiers, personality traits, such as "negative femininity"-which reflects extreme passivity-and "negative masculinity"-which includes antisocial characteristics-were both positively correlated with unwanted sexual experiences among male and female soldiers. 86 Similarly, students who underestimated their own likelihood of being sexually assaulted were at greater risk compared to their peers. 87 Still, vulnerable targets may be less obvious in the medical workplace. Medical students, residents, fellows, postdoctoral fellows and junior faculty are dependent on their superiors for recommendation letters, evaluations, opportunities, mentorship, among other reasons. 23 Residents who have experienced IHD during their training frequently report disappointment with the effectiveness of existing anti-IHD infrastructure, 67 fear of reprisal and retaliation, and identify numerous barriers to reporting IHD. 33

Comparison to other reviews
The prevalence estimates of the present metaanalysis are consistent with prior IHD reviews. 36,88,89 Fnais and colleagues found that 63.4% of residents had experienced at least one form IHD during their training. 36 Karim and Duchcherer found that the prevalence of IHD in residency varied between 45% and 93%. 6 Huang and colleagues found that discrimination occurred in 22.4% of surgeons and surgery trainees [95% CI = 14.0-33.9%]; however, as this was a pooled estimate of medical students, residents, fellows, and staff, it is not directly comparable to our study's estimate, which is exclusive to residents. Interestingly, one of the component studies in the Huang meta-analysis found that among surgical residents only, rates of bullying varied from 11.5-82.5%.
Although the general findings of our study are consistent with IHD literature, 14,18,[22][23][24] IHD prevalence appears to be higher among residents than staff physicians. However, among staff physicians, IHD prevalence is highly variable. For example, in a recent survey of radiologists, only 10% of respondents reported sexual harassment, 90 while a recent survey of all Australasian surgeons found that 49.2% reported experiencing bullying, discrimination and sexual harassment behaviours. 91 Our study's findings are also consistent with reports of high prevalence of IHD in non-medical institutions of higher learning. For example, nearly 30% of college students have previously reported sexual harassment during their education. 22 This finding lends support to the trans-institutional nature of academic IHD. [92][93][94][95] Limitations As our study is a meta-analysis of cross-sectional surveys, issues of selection and information bias are especially relevant. For example, given that the overall response rate was only about 50%, there is a high possibility of selection bias if the non-responders were significantly different than the responders with respect to their experience of IHD. Furthermore, as surveys were based on self-report and required the respondent to recall the duration of their residency experience, this also introduces a high chance of recall bias.
Most surveys included in the meta-analysis provided definitions of IHD. While this may improve the consistency of reporting within a study, definitions were inconsistently used across studies, and may have restricted respondents in how they interpreted their personal experiences of IHD; however, we did not identify a significant difference in IHD prevalence between studies that used definitions and those that did not. Given the subjective nature of IHD, additional factors, such as social desirability, confidentiality, and fear of reprisal, may have influenced the ability of respondents to truthfully complete the surveys.
Although we included all types and kinds of IHD, there were a limited number of studies reporting certain risk factors for IHD, such as training status, which was only reported by one study. However, this one risk factor was found to be highly cited (by 55.5% of respondents), suggesting training status is a significant risk factor for IHD.
Despite inclusion of a grey literature source, the majority of eligible studies identified by this review were published, English-language studies of resident physicians based in the United States, Canada, and the United Kingdom. Therefore, the generalizability of our findings may be limited to these populations.

e109
While all postgraduate year residents were included, 71% were in their first two years, and only a minority of respondents from postgraduate year three or onwards were represented. This is potentially significant as the experiences of more senior residents may be different from more junior trainees.
As our meta-analysis synthesized surveys across decades and across countries, this may have introduced additional heterogeneity in the results. Literature on the reporting of various forms of IHD has suggested that rates of reporting of IHD are on the rise. 10

Strengths
Our study has a number of strengths. First, our metaanalysis is up to date, including several recent studies from 2018 and onwards. Second, our study is comprehensive, synthesizing 52 studies, 63,378 respondents, and multiple medical specialties. Third, our inclusion of all types, sources, and risk factors for IHD enables a rich and thorough discussion about trends in IHD among residents. Fourth, our focus on diverse secondary outcomes including reporting measures, impact measures, and potential solutions to IHD, allows our study to unique contribute to the available literature on IHD among resident physicians, and IHD in medicine.

Future directions
While IHD continues to be a highly prevalent and serious issue for resident physicians, there is hope that a future without IHD in medicine is possible. [104][105][106][107] Future research should explore the efficacy of anti-IHD interventions, such as education or policy change, on the overall prevalence of IHD.

Conclusion
Our study achieved its proposed aim of establishing the prevalence of IHD among resident physicians, IHD risk factors, and potential sources and solutions to IHD. Despite growing recognition of IHD in medical education, the responses of the medical and medical education systems and organizations to IHD has been inadequate. Given the high rates and severe consequences of IHD, it is disappointing that the situation remains unchanged after many years.
However, we end our study with a call to action-that future researchers identify effective intervention and prevention strategies to reduce IHD and its sequelae Unjust assignment of duties Overloading someone with work Education/service imbalance e.g. contractual infractions, inadequate supervision, excessive service load or service assignment without educational merit Academic Being asked to carry out some personal services unrelated to patient care or educational activities Questions/queries were intentionally not answered You were threatened with failure or giving poor evaluations for reasons unrelated to your academic performance

Loss of Privileges or Opportunities
Privileges/opportunities taken away unfairly or in ways that are not related to resident's performance

Retaliation/Recrimination
Reprisal or threat of reprisal for negative feedback of staff, program or service, including the lodging of a complaint or grievance Other Economic abuse Repeated Situations where residents perceive more than one instance of intimidation, harassment or discrimination

Risk Factors/Basis for IHD Examples Gender
Comments about one's gender identity or gender expression Sexist teaching materials Punitive measures and favoritism based on gender

Training Status
Intimidation, harassment or discrimination on the basis of the resident's rank (e.g., postgraduate year 1 versus a postgraduate year 5), relying on intrinsic hierarchical systems

Sexual Orientation
Homophobic remarks Assumptions on the basis of the residents' perceived sexual orientation Ethnicity Racial epithets or slurs Negative stereotypes about a particular ethnic group Culture Disparagement of someone's cultural or religious devotions Language Rude or disparaging remarks on the basis of someone's first language or perceived difficulty with the native language Other Physical appearance, location of training, region of training