Physicians have a duty to disclose harmful medical errors to patients.1–4 When a trainee contributes to a harmful error, the trainee shares responsibility for disclosure with the attending physician, and it may be most appropriate for the attending physician and trainee to disclose the error together.3,5,6 Responding to errors transparently (i.e., openly and honestly) is a key element of patient safety,7 and many medical centers have implemented disclosure policies aimed at promoting transparency and improving patient care.8,9 However, disclosure remains difficult for physicians, and many errors go undisclosed.10–13
Within teaching hospitals, this difficulty is acutely felt by trainees at the front lines of patient care.10,14 Trainees occupy the bottom of the medical hierarchy and, when deciding whether to disclose an error, face a host of competing considerations.10,15 Few medical schools and residency programs offer formal instruction in error disclosure.15 Thus, trainees learn disclosure skills through direct observation of supervising physicians.16 Medical students and residents are in a formative stage of their professional development, and their attitudes and behaviors regarding disclosure may be influenced by their learning environment, a phenomenon known as the “hidden curriculum.”17 Specifically, role models for responding to errors may impact trainees.18 Thus, we measured trainees’ exposure to negative and positive role-modeling for responding to medical errors, and examined the association between exposure to role-modeling and trainees’ attitudes and behaviors regarding disclosure.
All 435 first-year residents and third-year residents in the general surgery, internal medicine, neurology, neurosurgery, obstetrics–gynecology, orthopedic surgery, otolaryngology, pediatrics, plastic surgery, and urology residency programs at two large U.S. academic medical centers, and all 1,187 fourth-year medical students at seven U.S. medical schools, were eligible to participate. Medical schools were composed of four public and three private institutions representing all four regions of the United States. Participation in the study was voluntary, and consent was implied by survey completion. The study was approved by the institutional review boards at Partners Healthcare and at each study site.
Residents were surveyed between May 2011 and September 2011. Students were surveyed between September 2011 and June 2012. Surveys were administered via e-mail link to an electronic, anonymous questionnaire using REDCap (Research Electronic Data Capture) version 5.0.8 (Vanderbilt University, Nashville, Tennessee).19
The questionnaire used the Institute of Medicine’s definition of a medical error: “the failure of a planned action to be completed as intended or the use of a wrong plan to achieve an aim.”20 For this study, we developed and pilot-tested our own definition of a “harmful” medical error: “a medical error that results in transient or permanent injury to patients (e.g., physical discomfort or disability, need for additional therapy or procedures, emotional distress, or death).”
The questionnaire included items about experience with medical errors, training for responding to errors, behaviors related to disclosure and factors contributing to nondisclosure, frequency of exposure to role-modeling for responding to errors, and attitudes regarding disclosure (see Supplemental Digital Appendix 1, http://links.lww.com/ACADMED/A183). Attitudinal items were based on an empirically derived taxonomy of factors that facilitate or impede disclosure and a review of the literature.4,10,11,15,21–23 Role-modeling items were based on prior qualitative research.18 The questionnaire was pilot-tested for face validity and clarity.
To incorporate multiple attitudinal and role-modeling variables into multivariate analyses, we performed principal component analysis24 on the five role-modeling variables and the nine attitudinal variables to create summary variables. Three of the five role-modeling variables substantially and exclusively loaded to factor one—positive role-modeling scale (score range: 3–12, Cronbach α = 0.91). The remaining two role-modeling variables substantially and exclusively loaded to factor two—negative role-modeling scale (score range: 2–8, Cronbach α = 0.64). All nine attitudinal variables loaded substantially to factor one, and none of the variables meaningfully loaded to factor two. Thus, we used only factor one—disclosure attitudes scale (score range: 9–36, Cronbach α = 0.74). Scale scores were created by summing the four-point Likert scale responses of the variables included in each scale. Higher scores on these scales represent more frequent exposure to that type of role-modeling and more positive attitudes regarding disclosure, respectively.
We created a composite outcome variable—nontransparent behavior in response to a harmful error—composed of all respondents who reported (1) not disclosing a harmful medical error they made to the patient/family, (2) not disclosing a harmful medical error they made to more senior team members, or (3) attempting to evade responsibility (e.g., attempt to conceal the error or shift the blame) for a harmful medical error. Respondents who reported nondisclosure of harmful errors were asked to identify contributing factors.
Adequate training on responding to medical errors was assessed by respondents’ agreement or disagreement with the fol lowing statement: “I have received adequate training on how to respond to medical errors.”
Frequencies and means were used to describe respondents’ characteristics. To simplify the reporting of the results, we dichotomized the four-point Likert scales for attitudinal and role-modeling variables. We used the chi-square test to assess the overall differences in categorical variables and the Mantel–Haenszel chi-square test for trend to analyze binary respondent characteristics across levels of training. We used ANOVA to assess differences in means across levels of training and Pearson product–moment correlations to assess the bivariate relationship between two continuous variables.
The primary outcomes were (1) attitudes regarding disclosure as measured by the disclosure attitudes scale and (2) nontransparent behavior in response to a harmful error. The primary predictors were frequency of exposure to (1) negative role-modeling as measured by the negative role-modeling scale and (2) positive role-modeling as measured by the positive role-modeling scale. Covariates included gender, level of training, specialty/anticipated specialty, study site, prior experience with errors, and training for responding to errors. To assure anonymity of the respondents, we did not query age or race/ethnicity.
To assess independent predictors of attitudes regarding disclosure and nontransparent behavior in response to a harmful error, the primary predictors were entered into backward selection, regression models using a stay criterion of P < .05. To control for potential confounders, all covariates that approached significance (P < .20) on bivariate analyses were included in the models. To provide a sufficient sample size and number of outcome events (i.e., nontransparent behavior in response to a harmful error) for multivariate analysis, interns and residents were collapsed into a single category in multivariate models.
P values of less than .05 were considered significant. Analyses were performed using SAS versions 9.2 (SAS Institute, Inc., Carry, North Carolina).
Surveys were completed by 884 of the 1,622 eligible trainees. The overall response rate was 55%, with subgroup response rates of 53% (631/1,187) for medical students, 53% (119/226) for interns, and 64% (134/209) for residents. Table 1 describes respondents’ characteristics.
Most trainees (54%, 478/884) had observed a harmful medical error. Residents were significantly more likely than medical students (75% [100/134] versus 51% [321/631]; P < .001) and interns (75% [100/134] versus 49% [58/119]; P < .001) to have observed a harmful medical error. Having made a harmful medical error increased significantly by level of training (Table 1; trend P < .001), as did reporting adequate training for responding to errors (Table 1; trend P = .007).
Table 2 describes the frequency of trainees’ exposure to role-modeling for responding to errors. Frequency of exposure to role-modeling did not differ significantly by level of training. More than 77% of trainees were seldom or more often exposed to positive role-modeling across all three items (i.e., observed a more senior team member  accept responsibility for an error,  disclose an error to a patient, or  apologize to a patient for an error), whereas more than 50% of trainees were seldom or more often exposed to negative role-modeling across both items (i.e., observed  colleagues treated harshly [e.g., humiliated or verbally abused] for errors they made or  a more senior team member attempt to evade responsibility [e.g., attempt to conceal the error or shift the blame] for an error).
Table 2 illustrates trainees’ attitudes regarding disclosure. Medical students were more likely than interns (29% [184/631] versus 18% [22/119]; P = .02) or residents (29% [184/631] versus 18% [24/134]; P = .008) to believe that if they acknowledged making a medical error, they would be treated harshly (e.g., humiliated or verbally abused). Students also were more likely than interns (42% [266/631] versus 31% [37/119]; P = .02) or residents (42% [266/631] versus 30% [40/134]; P = .008) to believe that if they acknowledged a medical error, it would negatively affect their career. Students were less likely than interns (59% [374/631] versus 69% [82/119]; P = .05) and residents (59% [374/631] versus 69% [93/134]; P = .03) to feel prepared to have a conversation with a patient in which they disclosed an error they made. Trainees who reported adequate training for responding to errors were more likely to feel prepared to have a disclosure conversation with a patient (86% [329/383] versus 44% [220/501]; P < .001). Feeling free to express concerns to other members of the health care team about medical errors in patient care increased significantly by level of training (Table 2; trend P = .01). Virtually all trainees (>97%) “agreed” that harmful errors should be fully disclosed to patients and that apologizing to patients for medical errors is important; however, only about half “strongly agreed” with these statements (Table 2).
Predictors of attitudes
Table 3 illustrates the results of multivariate analyses, by level of training and overall, of the association between the frequency of exposure to negative and positive role-modeling and attitudes regarding disclosure. Student, intern/resident, and overall models yielded similar results. In the overall model, as in the student and intern/resident models, training on how to respond to errors and positive role-modeling had the largest positive effects on attitudes (standardized effect coefficients, 0.32, P< .001 and 0.26, P < .001, respectively), while negative role-modeling had the largest negative effect on attitudes (standardized effect coefficient, −0.26, P < .001). Male gender was independently and positively associated with attitudes regarding disclosure (standardized effect coefficient, 0.08, P = .007). Two of the nine attitudinal items differed significantly by gender. Women were significantly less likely than men to feel free to express concerns about errors in patient care (62% [284/458] versus 69% [295/426]; P = .03) and to feel prepared to discuss errors with patients (55% [252/458] versus 70% [297/426]; P < .001). Reporting adequate training for responding to errors did not differ significantly by gender (42% [194/458] females versus 44% [189/426] males; P = .46).
Table 4 presents the composite outcome, nontransparent behavior in response to a harmful medical error. Among the 11% (68/631) of students and 36% (92/253) of interns and residents who reported making a harmful medical error, the proportion reporting nontransparent behavior did not differ significantly by level of training (44% [30/68] students versus 32% [29/92] interns and residents; P = .10). Specifically, students did not significantly differ from interns and residents in nondisclosure of harmful errors to patients (34% [23/68] students versus 25% [23/92] interns and residents; P = .22). Trainees of all levels who reported nondisclosure to patients similarly endorsed three factors as contributing to this outcome including the error only causing minor harm (67%, 31/46), trainees’ belief that disclosure was not in the patient’s best interest (44%, 20/46), and reluctance of more senior team members to disclose the error to the patient (30%, 14/46). Students were more likely than interns and residents to report not disclosing a harmful medical error to a more senior member of the team (16% [11/68] students versus 5% [5/92] interns and residents; P = .03). Students who reported nondisclosure to senior team members identified concerns about their evaluations (64%, 7/11) and the error causing only minor harm (64%, 7/11) as the two most common factors contributing to this outcome.
Predictors of behavior
Table 5 presents the results of multivariate analyses, by level of training and overall, of the relationship between frequency of exposure to negative and positive role-modeling and nontransparent behavior in response to a harmful error. More positive attitudes regarding disclosure were independently associated with decreased odds of nontransparent behavior in the intern/resident and overall models. In the overall model, as in the student and intern/resident models, negative role-modeling was independently associated with an increased likelihood of nontransparent behavior (OR 1.37, 95% CI 1.15–1.64, P < .001). Positive role-modeling and adequate training for responding to medical errors were not directly associated with behavior.
Discussion and Conclusions
This study found that exposure to role-modeling predicts both trainees’ attitudes and behavior regarding the disclosure of harmful medical errors. Specifically, we found that trainees reported more frequent exposure to positive role-modeling than negative role-modeling for responding to errors. More frequent exposure to negative role-modeling was independently associated with more negative attitudes regarding disclosure and an increased likelihood of nontransparent behavior in response to a harmful error. In contrast, positive role-modeling and training on how to respond to errors were independently associated with more positive attitudes, but did not directly protect against nontransparent behavior.
These findings indicate that negative role-modeling may be more influential than positive role-modeling for trainees. This mirrors the general psychological finding that negative events have a greater impact on individuals than positive events.25 Although interest in role-modeling is growing, the literature is largely qualitative and descriptive.26–31 This study provides some of the strongest evidence to date on the importance of role-modeling in the professional development of trainees and its implications for patient care. The findings suggest a “hidden curriculum” regarding error disclosure in which the behavior modeled by more senior team members may have a greater impact on subsequent trainee behaviors than the values and expectations espoused by the educational and health care system. A prior study demonstrated that trainees who had observed a variety of unethical behaviors were more likely to behave improperly themselves.32 This study furthers that research by demonstrating the relationship between trainees’ exposure to specific role-modeled behaviors and their own related conduct, and empirically compares and contrasts the relationship of positive and negative role-modeling on trainees’ attitudes and behaviors.
The findings suggest that negative role models for responding to errors may be a significant impediment to disclosure among trainees and thus an important element of organizational safety culture within academic medical centers. To foster a culture of safety and promote professionalism, physicians who act as negative role models, including resident physicians who themselves may be role models for other residents and students, need to be identified and, in a measured way, confronted by senior leadership. When queried, trainees have no trouble identifying negative role models.31 Educational leaders (e.g., clerkship and residency directors) should provide confidential means for trainees to report negative role models without concern for retaliation or adverse impact on their evaluations. It is troubling that nearly a third of students and a quarter of interns and residents in this study feared harsh treatment if they acknowledged making an error. Educational leaders should have a zero tolerance policy when it comes to punishment or retaliation for error disclosure made in good faith. Institutional leadership (e.g., chief medical officers and quality/safety officers) should develop initiatives to restrict the practice of physicians at all levels who, in spite of feedback, persistently exhibit negative behaviors until those behaviors are remediated. Nonpunitive “awareness” feedback and, if needed, “guided authority” interventions, have demonstrated effectiveness in remediating disruptive physicians and may provide a useful template for dealing with negative role models.33,34
Institutional leaders should ensure that all physicians receive sufficient training in patient safety and error disclosure to allow them to function as effective role models for responding to errors.35 Training should include learners across all levels and disciplines (e.g., students, residents, fellows, faculty, and nurses) to ensure that everyone receives the same message and to allow trainees to appropriately integrate the reactions and responses of more senior physicians—their role models—and of their peers in order to normalize discussion of medical errors.36
The best methods for training physicians for responding to errors are unknown and in need of research.8 Thus, we relied on trainees’ own perceptions of the sufficiency of their training and found that less than half of trainees reported adequate training; those who perceived receiving adequate training had more positive attitudes regarding disclosure including a greater sense of preparedness for disclosing errors to patients. This is important given that nearly half of students and a third of interns and residents felt unprepared to have conversations disclosing errors to patients.
Consistent with prior research, we found that most trainees had made or observed a harmful medical error.10,15 Two decades ago, Wu and colleagues10 reported that up to 75% of residents had not disclosed the most significant medical error they made to the patient, and only about half had discussed the error with their attending physician. In our study, we found that over a third of trainees reported nontransparent behavior in response to a harmful medical error they made. This reduced prevalence of nontransparent behavior is consistent with another more recent study on the topic13 and may reflect the tremendous amount of effort that has gone into promoting disclosure over that past two decades.9,37 Additionally, trainees may have underreported their involvement with errors. Prior studies have shown that trainees cope with errors by minimizing or distancing themselves from what happened.5,11 Although this lower prevalence was a welcome finding, it is still much higher than desired and indicative of the need for continued efforts to facilitate greater transparency.
This study unexpectedly found that male gender was independently associated with more positive attitudes regarding disclosure. Given that female trainees were significantly less likely than male trainees to feel free to express concerns about errors in patient care and to feel prepared to discuss errors with patients, this gender difference may be due to differences in empowerment and confidence consistent with prior research.38 More research to confirm and better understand and remedy these differences is needed.
This study has some limitations. First, despite a response rate typical for a study of physicians in training,39,40 nonresponse bias could have affected the results. Given the anonymous nature of this survey and a lack of information about nonrespondents, we are not able to compare respondents with nonrespondents. However, the gender and specialty (surgical versus nonsurgical) distribution of respondents closely resembles that of the entire population of U.S. medical students and residents.41,42 Second, although errors are generally salient events and thus may be recalled more accurately than more routine events, the data are subject to recall bias. Third, although the survey was anonymous, it relies on self-reported behavior and exposure to role-modeling and thus may be subject to social desirability bias. If social desirability bias is present, the results likely underestimate the number of trainees who contributed to errors, exposure to negative role-modeling, and trainees’ nontransparent behavior. Fourth, the study was cross-sectional, limiting the ability to measure the long-term effect of role-modeling on attitudes and behaviors. Fifth, we did not collect data on the type or amount of the training for responding to errors that trainees received. Therefore, we could not assess whether some training strategies may be more or less effective at affecting attitudes and behavior. Lastly, although trainees report disclosing errors to patients,15 there is no consensus regarding their role in disclosure; thus, it is challenging to draw normative conclusions about their behavior.
Despite these limitations, this study highlights an important factor associated with attitudes and behaviors regarding medical error disclosure that had been previously unexamined. As policies promoting disclosure become increasingly common,43,44 it will be critical to ensure that physicians are well prepared to act accordingly. Addressing the problem of negative role-modeling, increasing positive role-modeling, and creating a culture of safety in which physicians at all levels are trained and encouraged to disclose errors is essential to realizing effective and comprehensive implementation of disclosure policies.
Acknowledgments: The authors are indebted to the trainees who participated in the study. They are grateful to Jay Jacobson, MD, for helpful comments on an earlier draft, Ilene N. Moore, MD, JD, for her help implementing the study, and E. Francis Cook, ScD, for his assistance with the study methods.
1. Rosner F, Berger JT, Kark P, Potash J, Bennett AJ. Disclosure and prevention of medical errors. Committee on Bioethical Issues of the Medical Society of the State of New York. Arch Intern Med. 2000;160:2089–2092
2. Council on Ethical and Judicial Affairs. E-8.12 Patient Information. Code of Medical Ethics of the American Medical Association: Current Opinions With Annotations. 2010–2011 ed. 2010 Chicago, Ill American Medical Association
3. Wu AW, Cavanaugh TA, McPhee SJ, Lo B, Micco GP. To tell the truth: Ethical and practical issues in disclosing medical mistakes to patients. J Gen Intern Med. 1997;12:770–775
4. Gallagher TH, Waterman AD, Ebers AG, Fraser VJ, Levinson W. Patients’ and physicians’ attitudes regarding the disclosure of medical errors. JAMA. 2003;289:1001–1007
5. Wu AW, Folkman S, McPhee SJ, Lo B. How house officers cope with their mistakes. West J Med. 1993;159:565–569
6. Lee SK, Cowie SE. MSJAMA: Medical students and remediation of error. JAMA. 2001;286:1082–1083
8. Gallagher TH, Studdert D, Levinson W. Disclosing harmful medical errors to patients. N Engl J Med. 2007;356:2713–2719
9. Gallagher TH, Levinson W. Disclosing harmful medical errors to patients: A time for professional action. Arch Intern Med. 2005;165:1819–1824
10. Wu AW, Folkman S, McPhee SJ, Lo B. Do house officers learn from their mistakes? JAMA. 1991;265:2089–2094
11. Gallagher TH, Waterman AD, Garbutt JM, et al. US and Canadian physicians’ attitudes and experiences regarding disclosing errors to patients. Arch Intern Med. 2006;166:1605–1611
12. Kaldjian LC, Jones EW, Wu BJ, Forman-Hoffman VL, Levi BH, Rosenthal GE. Disclosing medical errors to patients: Attitudes and practices of physicians and trainees. J Gen Intern Med. 2007;22:988–996
13. López L, Weissman JS, Schneider EC, Weingart SN, Cohen AP, Epstein AM. Disclosure of hospital adverse events and its association with patients’ ratings of the quality of care. Arch Intern Med. 2009;169:1888–1894
14. White AA, Bell SK, Krauss MJ, et al. How trainees would disclose medical errors: Educational implications for training programmes. Med Educ. 2011;45:372–380
15. White AA, Gallagher TH, Krauss MJ, et al. The attitudes and experiences of trainees regarding disclosing medical errors to patients. Acad Med. 2008;83:250–256
16. Etchegaray JM, Gallagher TH, Bell SK, Dunlap B, Thomas EJ. Error disclosure: A new domain for safety culture assessment. BMJ Qual Saf. 2012;21:594–599
17. Hafferty FW, Franks R. The hidden curriculum, ethics teaching, and the structure of medical education. Acad Med. 1994;69:861–871
18. Martinez W, Lo B. Medical students’ experiences with medical errors: An analysis of medical student essays. Med Educ. 2008;42:733–741
19. Harris PA, Taylor R, Thielke R, Payne J, Gonzalez N, Conde JG. Research electronic data capture (REDCap)—a metadata-driven methodology and workflow process for providing translational research informatics support. J Biomed Inform. 2009;42:377–381
20. Kohn LT, Corrigan J, Donaldson MS To Err Is Human: Building a Safer Health System. 2000 Washington, DC National Academy Press
21. Kaldjian LC, Jones EW, Rosenthal GE. Facilitating and impeding factors for physicians’ error disclosure: A structured literature review. Jt Comm J Qual Patient Saf. 2006;32:188–198
22. Kaldjian LC, Jones EW, Rosenthal GE, Tripp-Reimer T, Hillis SL. An empirically derived taxonomy of factors affecting physicians’ willingness to disclose medical errors. J Gen Intern Med. 2006;21:942–948
23. Gallagher TH, Garbutt JM, Waterman AD, et al. Choosing your words carefully: How physicians would disclose harmful medical errors to patients. Arch Intern Med. 2006;166:1585–1593
24. Jolliffe I. Principal component analysis. Encyclopedia of Statistics in Behavioral Science. 2005 Chichester, UK: John Wiley & Sons, Ltd. In:
25. Baumeister R, Bratslavsky E, Finkenenauer C, Vohs K. Bad is stronger than good. Rev Gen Psychol. 2001;5:323–370
26. Ambrozy DM, Irby DM, Bowen JL, Burack JH, Carline JD, Stritter FT. Role models’ perceptions of themselves and their influence on students’ specialty choices. Acad Med. 1997;72:1119–1121
27. Côté L, Leclère H. How clinical teachers perceive the doctor–patient relationship and themselves as role models. Acad Med. 2000;75:1117–1124
28. Elzubeir MA, Rizk DE. Identifying characteristics that students, interns and residents look for in their role models. Med Educ. 2001;35:272–277
29. Matthews C. Role modelling: How does it influence teaching in family medicine? Med Educ. 2000;34:443–448
30. Wright S. Examining what residents look for in their role models. Acad Med. 1996;71:290–292
31. Beaudoin C, Maheux B, Côté L, Des Marchais JE, Jean P, Berkson L. Clinical teachers as humanistic caregivers and educators: Perceptions of senior clerks and second-year residents. CMAJ. 1998;159:765–769
32. Feudtner C, Christakis DA, Christakis NA. Do clinical clerks suffer ethical erosion? Students’ perceptions of their ethical environment and personal development. Acad Med. 1994;69:670–679
33. Hickson GB, Moore IN, Pichert JW, Benegas MBerman S. Balancing systems and individual accountability in a safety culture. From Front Office to Front Line. 20112nd ed Chicago, Ill Joint Commission Resources, Inc.:1–32 In:
34. Pichert JW, Hickson GB, Moore IN Using Patient Complaints to Promote Patient Safety: The Patient Advocacy Reporting System (PARS). 2008 Bethesda, Md Agency for Healthcare Research and Quality (AHRQ)
35. Report of the Lucian Leape Institute Roundtable on Reforming Medical Education.Unmet Needs: Teaching Physicians to Provide Safe Patient Care. 2010 Boston, Mass Lucian Leape Institue at the National Patient Safety Foundation
36. Bell SK, Moorman DW, Delbanco T. Improving the patient, family, and clinician experience after harmful events: The “when things go wrong” curriculum. Acad Med. 2010;85:1010–1017
37. Joint Commission. . Standard RI.2.90. 2013 Hospital Accreditation Standards. 2013 Joint Commission Resources, Inc. In:
38. Bartels C, Goetz S, Ward E, Carnes M. Internal medicine residents’ perceived ability to direct patient care: Impact of gender and experience. J Womens Health (Larchmt). 2008;17:1615–1621
39. Agrawal JR, Huebner J, Hedgecock J, Sehgal AR, Jung P, Simon SR. Medical students’ knowledge of the U.S. health care system and their preferences for curricular change: A national survey. Acad Med. 2005;80:484–488
40. Kellerman SE, Herold J. Physician response to surveys. A review of the literature. Am J Prev Med. 2001;20:61–67
43. Kachalia A, Kaufman SR, Boothman R, et al. Liability claims and costs before and after implementation of a medical error disclosure program. Ann Intern Med. 2010;153:213–221
44. Kachalia A, Mello MM. New directions in medical liability reform. N Engl J Med. 2011;364:1564–1572