Article Text

Primary care physicians’ willingness to disclose oncology errors involving multiple providers to patients
  1. Kathleen Mazor1,2,
  2. Douglas W Roblin3,4,
  3. Sarah M Greene5,
  4. Hassan Fouayzi1,2,
  5. Thomas H Gallagher6
  1. 1Meyers Primary Care Institute, Worcester, Massachusetts, USA
  2. 2Department of Medicine, University of Massachusetts Medical School, Worcester, Massachusetts, USA
  3. 3Center for Health Research, Kaiser Permanente Georgia, Atlanta, Georgia, USA
  4. 4Division of Health Management & Policy, School of Public Health, Georgia State University, Atlanta, Georgia, USA
  5. 5Methods and Infrastructure Program, Patient-Centered Outcomes Research Institute, Washington DC, USA
  6. 6Department of Medicine, University of Washington, Seattle, Washington, USA
  1. Correspondence to Dr Kathleen Mazor, Meyers Primary Care Institute, 630 Plantation Street, Worcester, MA 01605, USA; kathleen.mazor{at}umassmed.edu

Abstract

Background Full disclosure of harmful errors to patients, including a statement of regret, an explanation, acceptance of responsibility and commitment to prevent recurrences is the current standard for physicians in the USA.

Objective To examine the extent to which primary care physicians’ perceptions of event-level, physician-level and organisation-level factors influence intent to disclose a medical error in challenging situations.

Design Cross-sectional survey containing two hypothetical vignettes: (1) delayed diagnosis of breast cancer, and (2) care coordination breakdown causing a delayed response to patient symptoms. In both cases, multiple physicians shared responsibility for the error, and both involved oncology diagnoses.

Setting The study was conducted in the context of the HMO Cancer Research Network Cancer Communication Research Center.

Participants Primary care physicians from three integrated healthcare delivery systems located in Washington, Massachusetts and Georgia; responses from 297 participants were included in these analyses.

Main measures The dependent variable intent to disclose included intent to provide an apology, an explanation, information about the cause and plans for preventing recurrences. Independent variables included event-level factors (responsibility for the event, perceived seriousness of the event, predictions about a lawsuit); physician-level factors (value of patient-centred communication, communication self-efficacy and feelings about practice); organisation-level factors included perceived support for communication and time constraints.

Key results A majority of respondents would not fully disclose in either situation. The strongest predictors of disclosure were perceived personal responsibility, perceived seriousness of the event and perceived value of patient-centred communication. These variables were consistently associated with intent to disclose.

Conclusion To make meaningful progress towards improving disclosure; physicians, risk managers, organisational leaders, professional organisations and accreditation bodies need to understand the factors which influence disclosure. Such an understanding is required to inform institutional policies and provider training.

  • Communication
  • Patient-centred care
  • Diagnostic errors
View Full Text

Statistics from Altmetric.com

Introduction

Full disclosure of harmful errors to patients, including a statement of regret, an explanation of what occurred, acceptance of responsibility and commitment to prevent recurrences, has been a clear-cut standard for nearly a decade.1 Yet, there is continuing evidence that effective disclosure often does not occur.2–4 An absent or poor response by the clinicians involved can make a bad situation much worse, while full disclosure may mitigate the negative impact.5

This persistent disclosure gap may partly reflect an overly simplistic view of the problem. Historically, disclosure has been conceptualised as the individual physician disclosing ‘their error’, and the primary strategy that organisations have used to enhance disclosure is individual physician skills training.6 ,7 Yet the existing literature paints a much more complex reality, suggesting that important features of the event itself, the clinicians involved and their perceptions of responsibility for what occurred, as well as organisational factors such as institutional policies and support for disclosure, may influence whether and how disclosures occur.8

A more complete and nuanced understanding of the factors affecting clinicians’ propensity to disclose will be important to develop effective interventions. Therefore, we created two difficult but realistic hypothetical cases to determine how primary care physicians (PCPs) would respond. The cases depicted two challenging communication situations: (1) a delayed diagnosis of breast cancer and (2) a care coordination breakdown causing a delayed response to patient symptoms. Both cases were expected to be challenging because each involved multiple physicians sharing responsibility for what happened. Further, both were oncology cases, a clinical environment where the consequences of errors can be especially harmful. The purpose of this study was to examine the extent to which PCPs’ perceptions of event-level, physician-level and organisation-level factors influence a physician's intent to disclose a medical error in these stressful situations.

Methods

Setting

This cross-sectional survey was conducted in the HMO Cancer Research Network (CRN) Cancer Communication Research Center. These National Cancer Institute-funded projects involve a consortium of research organisations affiliated with integrated healthcare delivery systems, and work to improve cancer care through a programme of population-based research. Three CRN sites, located in Washington, Massachusetts and Georgia participated. The study protocol was approved by the University of Washington Institutional Review Board (Application #36619), Fallon Community Health Plan Institutional Review Board (Project #1184) and the Kaiser Permanente Georgia Institutional Review Board (Protocol #GA-08JCavi-01).

Study sample

All PCPs were identified at each clinical site by the research teams. Sample selection was limited to those in clinical practice at least 20% of time and seeing adult patients (total n=630). Written informed consent was not required. Completion of the questionnaire was considered indicative of consent.

Data collection

Survey administration

A modified version of the Tailored Design Method was used.9 Eligible participants were mailed a packet containing an invitation letter (endorsed by one or more leaders of the participating healthcare organisations), the questionnaire with a study identifier and a return envelope. The incentive varied across sites to comply with site policies and procedures. Initially participants were given the option of completing the survey via a link to a web-based version or on paper, but a problem with the survey software led to removal of the web-based option. Up to three reminders were sent to non-responders. The survey was conducted between June and September 2011.

Survey instrument

The questionnaire was developed by the research team and pretested with a convenience sample of physicians not included in the target population. Their suggestions were incorporated into the final questionnaire (available on request from the first author).

The questionnaire contained several discrete sections. The sections and corresponding items included in the analyses reported here are described in detail below.

Vignettes

The questionnaire contained two vignettes (see online supplementary appendix 1). The first described a delay in the diagnosis of breast cancer (delayed cancer diagnosis). In this vignette, the respondent is asked to imagine himself/herself as covering for a colleague who is on leave. The patient had previously mentioned a breast lump to the colleague, who noted the lump and recommended re-examination. The covering physician, focused on managing the patient's elevated blood pressure, misses the note about the lump, and is unaware of the need for follow-up until the patient mentions the lump again. A subsequent biopsy reveals cancer, with three positive axillary nodes. The second vignette described a breakdown in care coordination leading to a delayed response to patient symptoms (care coordination breakdown). In this vignette, a patient with a history of hypothyroidism who is being treated for colon cancer calls in several times to the nurse reporting fatigue and diarrhoea. The nurse reassures him that this is normal for patients undergoing chemotherapy, but that if it gets worse he should call back. She also passes on the information to both the oncologist and the PCP, each of whom assumes the other is following the problem. After a third call, the oncologist requests the patient to come in. The patient passes out while driving to the labs; in the emergency room he has a haematocrit of 16 with guaiac-positive stools.

Dependent variables

Each vignette was followed by four items measuring intent to disclose in the situation described. These vignette-specific items were adapted from items used in prior studies10 and asked what the physician would be likely to say with respect to (1) an apology, (2) an explanation, (3) the cause of the event and (4) plans for preventing recurrences. Three to four response options were provided for each item, representing non-disclosure, partial disclosure or full disclosure. Each non-disclosure response was assigned a value of 0, each partial disclosure response was assigned a value of 1 and each full disclosure response was assigned a value of 2 (see table 2). Intent to disclose scores were computed for each vignette using the mean across the four disclosure items (potential range 0 to 2).

Independent variables

Three sets of items were included to define event-level, physician-level and organisation-level factors that might be associated with intent to disclose. Event-level factors were assessed separately for each vignette using questions about the respondent's views on (1) personal responsibility for the event (‘How responsible are you, as the PCP for this event?’) and colleague's responsibility for the event (‘How responsible is your colleague…?’), (2) beliefs about seriousness of the event and (3) predictions as to whether the patient would file a lawsuit as a result of the event. Physician-level factors included: (1) four items assessing value placed on patient-centred communication (eg, ‘It is important that I know my patients as people’), (2) two items assessing self-efficacy with respect to communication (eg, ‘I communicate effectively with my patients who have cancer’) and (3) two items assessing feelings about practice (eg, ‘I often think about leaving practice’).11 Physician perceptions of organisation-level factors were assessed as follows: (1) a single item on organisational support for communication (‘My organisation values good clinician-patient communication’) and (2) two items on time constraints (eg, ‘I don't have enough time to educate patients about treatment options’). Response options for all items assessing physician-level and organisation-level factors were strongly agree, agree, disagree and strongly disagree.

For the factors assessed using multiple items (ie, the physician-level factors and the organisation-level variable time constraints), scale summary scores were computed as the mean of numeric values assigned to each response, and coefficient alpha was calculated. Each of the event-level factors (beliefs about personal responsibility, event seriousness and about the likelihood that the patient would file a suit) and one organisation-level factor (perceptions of organisational support for communication) was measured using a single item. Because responses to the organisation-level item on perceived support for communication were highly skewed, we dichotomised this measure.

Covariates

Physician gender and years in practice were used as covariates in the analyses.

Statistical analysis

Descriptive statistics (eg, frequencies, means and SDs) were computed for each item and variable. Differences in mean intent to disclose scores by respondent characteristics (site, gender, years in practice) were evaluated using analysis of variance or a t test; analyses were conducted separately for each of the two vignettes.

Bivariate relationships among the event-level, physician-level and organisation-level variables and the relationships of these factors with intent to disclose were examined using partial correlation coefficients. We then used generalised estimating equation (GEE) models (in order to take into account that physicians were nested within site), predicting intent to disclose from the event-level, physician-level and organisation-level variables. These multivariate models examined the extent to which all variables (ie, event-level, physician-level and organisation-level variables) were associated with intent to disclose (adjusting for gender and years in practice). Models were evaluated separately for each vignette.

Statistical analyses were conducted with the IBM Statistical Package for Social Sciences (V.22) and SAS package (V.9.2; SAS Institute, Cary, North Carolina, USA).

Role of the funding source

The study was funded by the National Cancer Institute (P20CA137219). The funder had no role in the study's design, conduct or reporting.

Results

A total of 333 PCPs responded to the survey (333 of 630; 53% response rate overall; response rates by site were 58%, 35% and 59% for sites A, B and C, respectively). Of the 333 returned questionnaire, 297 (89% of 333, 47% of 630) are included in the analyses reported here; respondents missing one or more variables key to these analyses were omitted. Respondent characteristics are summarised in table 1. The majority of respondents had been in practice for more than 10 years (71.7%). Over half (55.6%) indicated they often questioned whether the demand of their practice was worth the toll, and over a third (36.7%) often thought about leaving practice.

Table 1

Respondent characteristics

Table 2 presents physicians’ intent to disclose in relation to each vignette. A majority of respondents would not be fully forthcoming in either situation, providing only a limited or no apology, limited or no explanation and limited or no information about the cause. For the delayed cancer diagnosis vignette, a majority would provide limited or no information about efforts to prevent recurrences, though for the care coordination breakdown vignette a majority (57%) would be forthcoming about such efforts. Overall, physicians would be more forthcoming in the context of the care coordination breakdown, compared with the delayed cancer diagnosis (p value for the t statistic associated with paired comparison of mean intent to disclose scores was <0.001); however, for both vignettes, the fact that the average intent to disclose score was close to 1 indicates an intent to partially rather than fully disclose the event. Respondents at one site (A) had higher intent to disclose scores than respondents at another site (B) for the delayed diagnosis vignette (p=0.008), but no other statistically significant differences were associated with site, gender or years in practice (data not shown).

Table 2

Categorisation of level of disclosure, disclosure items, response options and percentage within each category

Table 3 summarises physicians’ responses on the constructed scales and individual items related to event-level variables. In both vignette situations, most physicians would consider both themselves and their colleague as at least partially responsible for the event. Almost three-quarters of physicians (74.4%) assigned at least some responsibility to both self and colleague for the delayed cancer diagnosis vignette. Only 0.3% perceived only self as responsible and 4.0% perceived neither as responsible. We examined whether physicians who perceived shared responsibility for the event (n=221) were more likely to disclose than those who perceived their colleague to be responsible (n=63) using a t test, and found that those who perceived responsibility to be shared were more likely to disclose (intent to disclose scores were 0.91 and 0.57, respectively, p<0.001). For the care coordination breakdown, 90.3% of respondents perceived responsibility to be shared, 1.4% indicated that only the colleague was responsible, none indicated only self was responsible and 8.3% viewed neither self or colleague as responsible. The very small number of respondents perceiving only the colleague as responsible (n=4) precluded a meaningful comparison of intent to disclose for those who perceived shared responsibility versus for those who perceived their colleague as solely responsible for the care coordination breakdown scenario.

Table 3

Responses and partial bivariate correlations (adjusted for site) for event-level responses

Both events were characterised as serious errors by a majority of physicians, though many more considered the delayed cancer diagnosis to be a serious error than considered the care coordination breakdown to be a serious error (83% vs 61% respectively). Consistent with this, approximately 70% of respondents considered it very or somewhat likely that the patient would file a lawsuit in the delayed cancer diagnosis case; this percentage was much less—but still substantial—for the care coordination breakdown (27%). The bivariate correlations suggest relatively strong associations between two of the event-level variables (perceived responsibility for the event and event seriousness) and intent to disclose for both vignettes. The expectation that the patient would file a lawsuit was associated with intent to disclose only in the context of the care coordination breakdown vignette.

Summary statistics for the physician-level and organisation-level variables are presented in table 4, with the bivariate correlations describing the associations between each of the physician-level, and organisation-level variables and intent to disclose. The correlations for the physician-level and organisation-level factors indicate that valuing patient-centred communication was associated with intent to disclose for both vignettes, while negative feelings about practice and perceived time constraints were each associated with intent to disclose for one but not both of the vignette situations.

Table 4

Responses and partial bivariate correlations (adjusted for site) for physician- and organisation-level variables

The multivariate GEE models predicting intent to disclose adjusting for gender and years in practice were also evaluated separately for each vignette. Results were generally consistent with the bivariate analyses (table 5). As in the bivariate analyses, three variables were predictive of intent to disclose in both circumstances: (1) beliefs about the seriousness of the event; (2) perceived responsibility for the event and (3) perceived value of patient-centred communication. In addition, the perception of time constraints was predictive of intent to disclose for the delayed diagnosis vignette, and negative feelings about practice were predictive of intent to disclose for the care coordination breakdown vignette. The perception that one's organisation values communication showed a very small but statistically significant effect for the care coordination breakdown. The expectation that the patient would file a lawsuit and communication self-efficacy was not predictive in either model.

Table 5

Crude and adjusted association of physicians’ perceptions and characteristics with intent to disclose*

Discussion

The importance of effective disclosure of harmful medical errors to patients and families is as clear as the profession's inability to improve the disclosure process. Our study, which used challenging but realistic cases of oncology error involving multiple providers, demonstrates the need to develop a more robust and nuanced understanding of the factors affecting disclosure at the level of the event, the provider and the organisation. Such an understanding will be important to make meaningful headway in meeting patients’ and families’ needs when care has gone wrong.

Our study shows that, despite all the attention over the past 14 years to the importance of open communication with patients following error, full disclosure remains elusive. Most physicians would provide only partial disclosure after either of the events studied here. When asked what they would tell the patient about the cause of the error, 77% of the physicians for the delayed diagnosis of cancer case, and 58% of the respondents for the care coordination breakdown case, would offer either no information or make vague references to ‘miscommunications’. More than half of the physicians in both cases reported they would not volunteer an apology or would offer only a vague expression of regret. This limited approach to disclosure by physicians in this study falls far short of patient expectations and national standards.12–16

The issue of disclosure of delayed diagnosis is particularly important, as such delays have been identified as common underlying cause of malpractice lawsuits.17 We found that about 55% of respondents would provide no explanation or a vague explanation of what happened in the delayed cancer diagnosis vignette. This percentage is similar to the finding from a recent survey of radiologists’ responses to a vignette describing a delayed diagnosis of breast cancer. Most radiologists indicated they would not fully disclose (9%) or would provide information only if asked about the patient (51%) in a case whether a delayed diagnosis of breast cancer was because the radiologist had originally viewed in the wrong order.18 Physicians are likely to find discussion of delays in diagnosis especially challenging, in part due to uncertainty about whether the delay affected the patient's outcome. Compounding this is the additional challenge of communicating this uncertainty to the patient. Because such cases are so difficult, they should prompt disclosure coaches to provide extra support and attention in the planning and execution of these discussions.

The relationship between expectations about malpractice lawsuits and disclosure deserves further study. While we found that many physicians considered it at least somewhat likely that the patient would file a malpractice lawsuit in both vignettes, we did not detect a relationship between this expectation and intent to disclose for the delayed diagnosis case. Prior studies also have not found a consistent relationship between malpractice experiences and intent to disclose.10 ,18 Thus, while physicians appear to be highly cognisant that malpractice lawsuits are a risk, how this affects intent to disclose is not yet clear. This may be attributable to the fact that studies of patients’ attitudes have found that the impact of full disclosure on patients’ likelihood of pursuing a lawsuit appears to be inconsistent,19 ,20 and a systematic review of the impact disclosure found few real-world studies of the impact of disclosure policies.21

These two challenging vignettes, each involving more than one physician, also highlight the critical role that physician perception of responsibility for error may be playing in the disclosure process. While the association between physicians’ perceived responsibility for the event and their intention to disclose was present in both cases, it was a particularly strong relationship in the delayed diagnosis of cancer case, where this factor emerged as the most important single predictor in the multivariate model. Interestingly, for the delayed diagnosis of cancer case, a full 25% of physicians considered themselves not at all responsible for the error, while only 4% considered themselves fully responsible. Further, cross-tabulation of perceptions of personal responsibility by colleague's responsibility revealed that the majority of respondents perceived both physicians to bear some responsibility for the event in both vignettes. Few studies have explored how physicians’ perceptions of responsibility for an error affect physicians’ attitudes and their intent to disclose. One exception is a recent small study of Italian physicians which found that participants were more explicit about errors when the simulation portrayed relatively clear responsibility for an event, compared with the simulation portraying shared responsibility (8 of 10 participants compared with 3 of 10 participants), but were no more or less likely to offer an explicit apology (55% overall).22 These rates are similar rates found for American surgeons in simulated encounters which were not intended to simulate shared responsibility, where 57% used explicit terms to describe an error and 47% offered an explicit apology.23 While more work is needed to better understand how to help physicians disclose effectively when responsibility is shared, these findings suggest that disclosure coaches should ask explicitly about perceptions of responsibility as part of the coaching process to better prepare physicians to have these discussions.

Concepts of responsibility for medical error have varied widely over the course of the patient safety movement, swinging from the historical ‘blame and shame’ approach to a brief emphasis on a ‘blame free environment’, now settling into a focus on ‘Just Cultures’ that tries to distinguish between individual and system error and apportion responsibility accordingly.24–27 Greater perception of personal responsibility for error has also been associated with both greater openness to learning from error and also increased emotional distress.28 Much more work is needed to understand how clinicians accept responsibility for error, and how acceptance of responsibility affects the way they discuss errors with patients.

Our study also offers insight into the role that the factors associated with the environments in which PCPs practice may also influence willingness to disclose. While the effects of physicians’ perceptions of time constraints on the intent to disclose was smaller and less consistent than event-level or physician-level dimensions, this nonetheless appears to be having an independent effect on the way physicians approach disclosure in at least some situations. It may be that physicians who feel most constrained by time pressure in their day-to-day practice experience such pressure as pervasive and are less likely to allot time even for a potentially high stakes conversation such as disclosure of a medical error. This could be short-sighted as non-disclosure is likely to have a negative impact and damage the physician–patient relationship. System-level support may be needed to provide physicians with the skills to have disclosure conversations and to create the time to have such conversations.

Our study has several implications for healthcare providers, institutions and researchers. Virtually all organisations now conduct regular surveys to measure safety climate, often using the Agency for Healthcare Research and Quality instruments.29 Yet these measures shed little light on clinicians’ willingness to be transparent with patients and families following medical error. Organisations should consider using vignette-based surveys to measure how physician-level and organisation-level factors, similar to those we measured and which are measured in surveys of safety climate, might be related to error disclosure.16 Findings from such vignette-based surveys might be used to improve the practice environment, promote proactive approaches to error disclosure and, as literature suggests, mitigate risk of litigation in the event of medical error when disclosure is muted.30 Survey instruments can be developed to include a range of potential disclosure scenarios in order to assist institutions and clinicians to better understand what circumstances pose particularly challenging scenarios for clinicians to disclose errors to patients and family members. Current strategies to promote disclosure frequently rely on creating institutional policies, providing disclosure skills training to clinicians, and on occasion creating disclosure coaching services. While these represent important building blocks, institutions should recognise the much broader range of event-level, physician-level and organisation-level factors that affect physician well-being, and, in turn, influence the way their clinicians communicate with patients when care has not gone well. For example, these findings suggest that supporting physicians in creating sufficient time to disclose may be an important missing element. At the same time, the fact that many physicians in this study would not be forthcoming about these complex errors suggests that studies examining barriers to implementation of disclosure in real-world settings are needed.

There are several important limitations to this study. This was a vignette-based survey, in which the physicians indicate what they might do in response to a scenario; however, this is not necessarily what might happen in practice. One might expect that social desirability bias would lead physicians to be more forthcoming on such a survey, where the potential for any negative consequences are low. We did not assess prior exposure to training in error disclosure; such training would be expected to influence intent to disclose. We are unable to measure causality, so it is unclear whether full organisational support for disclosure would actually impact the disclosure practices of clinicians. The physicians participating in this study may not be broadly representative of physicians in the USA, as their organisational structures and culture may support a more collaborative relationship and greater sharing of responsibility for care. Future research should aim to explore these factors more fully.

While the basic principles underlying effective disclosure of medical errors to patients and families are straightforward, the real-world response to harmful medical error is inevitably complex. To make meaningful progress around actually closing the gap between expectations for disclosure and current practice, healthcare providers, institutions and scientists need to embrace this complexity, continue the effort to unravel the critical factors influencing disclosure and their relationship to one another and then ensure that these learning are fully incorporated into institutional policies and provider training.

References

View Abstract

Supplementary materials

  • Supplementary Data

    This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.

Footnotes

  • Contributors KM: Study planning and design; conduct of the study; data analysis and interpretation; reporting including drafting, critically reviewing and revising manuscript. DWR: Study planning and design; conduct of the study; data review and interpretation; reporting including drafting, critically reviewing and revising manuscript. SMG: Study planning and design; conduct of the study; reporting including critically reviewing and revising manuscript. HF: Data analysis and interpretation; reporting including drafting and critically reviewing and revising manuscript. THG: Study planning and design; conduct of the study; data review and interpretation; reporting including drafting, critically reviewing and revising manuscript.

  • Funding National Cancer Institute (P20CA137219).

  • Competing interests None declared.

  • Ethics approval University of Washington Institutional Review Board (Application #36619); Fallon Community Health Plan Institutional Review Board (Project #1184) and Kaiser Permanente Georgia Institutional Review Board (Protocol #GA-08JCavi-01). Group Health Human Subjects Review Committee HS-08-057.

  • Provenance and peer review Not commissioned; externally peer reviewed.

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.