Choosing Your Words Carefully: How Physicians Would Disclose Harmful Medical Errors to Patients

Background  A gap exists between patients' desire to be told about medical errors and present practice. Little is known about how physicians approach disclosure. The objective of the present study was to describe how physicians disclose errors to patients.

Methods  Mailed survey of 2637 medical and surgical physicians in the United States (Missouri and Washington) and Canada (national sample). Participants received 1 of 4 scenarios depicting serious errors that varied by specialty (medical and surgical scenarios) and by how obvious the error would be to the patient if not disclosed (more apparent vs less apparent). Five questions measured what respondents would disclose using scripted statements.

Results  Wide variation existed regarding what information respondents would disclose. Of the respondents, 56% chose statements that mentioned the adverse event but not the error, while 42% would explicitly state that an error occurred. Some physicians disclosed little information: 19% would not volunteer any information about the error's cause, and 63% would not provide specific information about preventing future errors. Disclosure was affected by the nature of the error and physician specialty. Of the respondents, 51% who received the more apparent errors explicitly mentioned the error, compared with 32% who received the less apparent errors (P<.001); 58% of medical specialists explicitly mentioned the error, compared with 19% of surgical specialists (P<.001). Respondents disclosed more information if they had positive disclosure attitudes, felt responsible for the error, had prior positive disclosure experiences, and were Canadian.

Conclusions  Physicians vary widely in how they would disclose errors to patients. Disclosure standards and training are necessary to meet public expectations and promote professional responsibility following errors.

A sizable gap exists between patients' desire to be told about medical errors and current practice, with studies^1-5 suggesting that less than half of harmful errors are disclosed to patients. This failure to disclose errors could diminish patient trust and satisfaction and may increase the likelihood of malpractice claims.^6-11

The causes of this “disclosure gap,” especially the contribution of physicians' attitudes and behaviors, are poorly understood. Prior studies^12-16 suggest that while physicians generally support disclosure, multiple barriers inhibit them from actually talking to patients about errors, such as fear of lawsuits, shame, and lack of disclosure training. Physicians are also uncertain about the content of disclosure (ie, what words to choose when discussing errors with patients).¹⁷ Following errors, patients want an explicit statement that an error has occurred, information about why the error happened, how recurrences will be prevented, and an apology.^18-21 Yet, in prior qualitative research,¹⁷ physicians were often cautious in discussing errors with patients, such as mentioning the adverse event (the harm caused by medical care) but avoiding use of the word “error.” Prior work¹⁷ also suggested that the nature of the error may affect disclosure. Physicians told us they would be less inclined to disclose errors that would not be apparent to the patient unless health care workers pointed out the mistake. Yet, patients want to be informed about all harmful errors, not just errors that are obvious to them.^7,20-22 To our knowledge, no quantitative data exist regarding what information physicians would disclose and whether disclosure varies for errors that would not be apparent to patients.

Physicians' specialty may also influence how they approach disclosing errors. Surgeons routinely talk with patients about potential adverse events during the informed consent process and with colleagues during morbidity and mortality conferences.^23-25 Medical specialists, however, may have less experience discussing adverse events with patients and colleagues.²⁶ In addition, surgeons traditionally espouse a “captain of the ship” mentality in which the captain is considered responsible for errors made by anyone on the team, a belief evolving with changes in surgical education and legal doctrine.^23,25,27 Understanding specialty differences in disclosure could allow one group of physicians to learn from another. However, to our knowledge, no prior studies have explored whether medical and surgical physicians approach disclosure similarly.

Programs to enhance disclosure of errors to patients are proliferating.^28,29 The success of these efforts hinges on better understanding how physicians in different specialties think these difficult conversations should be conducted. Therefore, we surveyed physicians in the United States and Canada to determine the following: (1) how physicians would disclose hypothetical harmful medical errors to patients, (2) whether disclosure differs when an error might be less apparent to a patient and between medical and surgical specialists, and (3) the factors associated with disclosure.

We mailed surveys about error disclosure and patient safety to 4193 medical and surgical physicians in academic and private practice in the United States and Canada. A detailed description of the survey methods is provided elsewhere.³⁰

The survey used the US Institute of Medicine's definitions of adverse event, medical error, and near miss.³¹ We developed pilot tested our own definitions of serious error and minor error (Table 1). General questions solicited respondents' attitudes about key patient safety concepts, malpractice, whether errors should be disclosed, and barriers to disclosure.

To explore how differences in errors might affect disclosure, we created 4 scenarios depicting serious errors (errors that caused permanent harm or transient but potentially life-threatening injury). Widespread consensus supports disclosing such errors to patients.^32-34 Using scenarios allowed us to study how physicians would disclose errors in standardized situations and explore the variation in disclosure.³⁵ These 4 errors were designed to be comparable in severity and to vary along 2 primary axes: specialty (medical and surgical scenarios) and how apparent the error would likely be to the patient (more and less apparent).

Respondents randomly received 1 of 2 scenarios specific to medicine or surgery (Table 2). The more apparent medical error was an insulin overdose due to the physician's handwritten order for “10 U” of insulin being misinterpreted as “100 U,” resulting in severe hypoglycemia. The less apparent medical error was a hyperkalemic dysrhythmia due to failure to check the results of a potassium level ordered after starting a medicine known to cause hyperkalemia, an error that the patient would likely be unaware of unless the physician brought the overlooked potassium result to the patient's attention. The more apparent surgical error involved a retained surgical sponge, while the less apparent surgical error was bile duct injury during a laparoscopic cholecystectomy caused by the surgeon's incorrect use of a new surgical tool. This later scenario was considered less apparent because the patient would be unlikely to suspect that the surgeon's lack of familiarity with the new surgical tool caused the bile duct injury.

For each scenario, respondents received a parallel series of questions, including how serious the error was, how responsible the respondent was for the error, how likely the respondent was to be sued, and how likely the respondent would be to disclose this error to the patient. Five disclosure content questions measured what information physicians would volunteer to the patient about the error. These questions included the following: (1) “What would you most likely say about what happened?” (2) “How much detail would you most likely give the patient about the error?” (3) “What most closely resembles what you would say about the cause of the error?” (4) “What would you most likely say regarding an apology?” and (5) “What would you most likely say about how the error would be prevented in the future?” For each question, 3 response items that represented increasing amounts of information disclosed (no disclosure, partial disclosure, and full disclosure) were provided. The full text of these responses is provided in Table 3. Several rounds of pilot testing, including cognitive interviews with practicing physicians, were conducted to ensure that the survey questions were clear, the scenarios were realistic, and the disclosure options were plausible.

To explore potential differences in responses to the scenarios, we created several composite result categories, including all 4 scenarios combined (“overall”), the 2 more apparent errors compared with the 2 less apparent errors, and the 2 medical scenarios compared with the 2 surgical scenarios. Then we used χ² analyses to determine whether responses to the scenarios differed based on (1) whether the error would be more or less apparent to the patient and (2) whether the error was medical or surgical. Finally, we created medical and surgical multivariate models to explore the predictors of how much information respondents would disclose. Because we assumed a priori that a single factor could represent the 5 disclosure content questions and factor analyses supported a single “significant” factor (with an eigenvalue >1), the dependent variable for the multivariate analyses was a specialty-specific scaled disclosure score created via factor analysis of the 5 disclosure content questions. Creating the disclosure score this way yields a continuously measured score with the advantage over simple summation of not assuming equal item weighting. General linear models were used to determine the effects of key covariates on the medical and surgical disclosure scores, including attitudes toward patient safety, malpractice, and disclosure, scenario-specific attitudes, prior experience with disclosure, and sociodemographics. Full models were reduced using backward deletion with a P<.05 criterion for deletion. Analyses were conducted using SAS statistical software, version 8.2 (SAS Institute Inc, Cary, NC).

It was recently found that US and Canadian physicians' basic attitudes regarding patient safety and error disclosure were largely similar.³⁰ Therefore, for clarity of presentation, we have combined the US and Canadian data while retaining country as a covariate in the multivariate models.

Surveys were completed by 2637 (62.9%) of 4193 eligible physicians (Table 4). The characteristics of these respondents are described in detail elsewhere.³⁰ We highlight the composite responses to all 4 scenarios and compare the more apparent with the less apparent errors and the medical with the surgical scenarios. Responses to each scenario are available in Table 5 and Table 6.

Overall, 85% of physicians thought the error was serious and 81% believed the physician was very or extremely responsible for the error. Of the respondents, 39% were very or extremely concerned that their reputation would be damaged because of the error and 47% thought a lawsuit was somewhat or very likely (Table 5). Of the physicians, 65% said they would “definitely disclose” the error to the patient, 29% would “probably disclose the error,” 4% would “disclose only if asked by the patient,” and 1% would “definitely not disclose” the error.

Of the respondents, 90% considered the more apparent errors to be serious errors, compared with 81% for the less apparent errors (P<.001). Respondents who received the more apparent errors were more likely to report that they would definitely disclose the error compared with respondents who received the less apparent errors (81% more apparent errors and 50% less apparent errors, P<.001) (Table 5).

Medical specialists were more likely to rate the error as serious (90% medical scenarios and 79% surgical scenarios, P<.001), and surgeons thought a lawsuit was more likely (57% surgical scenarios and 40% medical scenarios, P<.001). Despite this increased malpractice concern, surgeons were more likely to report that they would definitely disclose the error (81% surgical scenarios and 54% medical scenarios, P<.001) (Table 5).

Physicians varied widely in the information they reported they would disclose. Of the physicians, 42% would use the word “error,” while 56% mentioned the adverse event but not the error; 50% would disclose specific information about what the error was, while 13% would not volunteer any details unless asked by the patient (Table 6). Similarly, 52% would disclose specific information about why the error happened, but 19% would volunteer no such information. Overall, 94% would apologize, with 61% offering an expression of regret (“I am sorry about what happened”) and 33% choosing an explicit apology (“I am so sorry that you were harmed by this error”). Regarding error prevention, 9% would provide no information unless asked, 54% chose a general pledge to prevent recurrences, and 37% would discuss detailed plans for preventing recurrences (Table 6).

Physicians who received the less apparent errors would disclose less information to the patient than respondents who received the more apparent errors. Respondents to the less apparent errors were less likely to use the word “error” (32% less apparent errors and 51% more apparent errors, P<.001), and were also less likely to choose an explicit apology (28% less apparent errors and 37% more apparent errors, P<.001). Of the respondents, 19% responding to less apparent errors would volunteer no details about what the error was, compared with 8% of those responding to more apparent errors (P<.001) (Table 6). Overall, this pattern also extended to what physicians would tell patients about the error's cause and prevention plans. Respondents who received the less apparent errors were twice as likely to volunteer no information about the error's cause (26% less apparent errors and 13% more apparent errors, P<.001) and to volunteer no information about how recurrences of the error would be prevented (13% less apparent errors and 6% more apparent errors, P<.001) (Table 6).

Despite expressing greater intention to disclose than medical specialists, surgical specialists disclosed less information. For example, 19% of surgeons would use the word “error,” compared with 58% of medical specialists (P<.001); 35% of surgeons would disclose specific details about the error compared with 61% of medical specialists (P<.001). Surgeons chose an explicit apology half as often as medical specialists did (21% surgical scenarios and 41% medical scenarios, P<.001). Surgeons also were one third as likely to provide details about preventing error recurrences (16% surgical scenarios and 52% medical scenarios, P<.001) (Table 6).

Table 7 presents the multivariate medical and surgical models exploring predictors of the amount of disclosure. In the medical and surgical scenarios, independent predictors of disclosing more information included respondents believing the error was serious, respondents feeling more responsible for the error, respondents reporting they would definitely disclose the error to the patient, respondents not agreeing that they would be less likely to disclose an error the patient was unaware of, respondents experiencing relief the last time they disclosed an error, and being Canadian. For the medical scenarios, attitudes about the relationship between disclosure and malpractice were also independent predictors of the content of disclosure. Medical specialists provided more information if they thought disclosing a serious error reduced the likelihood of malpractice and if they disagreed that a potential lawsuit would reduce their willingness to disclose. Medical specialists also disclosed more information if they were not in private practice.

Calls are increasing to fully disclose adverse events and medical errors to patients, but little is known about how physicians approach disclosure.^33,36-38 Our study of more than 2500 practicing physicians in the United States and Canada reveals wide variation in how physicians would disclose harmful errors. For example, 42% of physicians would disclose the fact that the adverse event was due to an error, while 56% would mention the adverse event but not the error. Such variation likely reflects the competing pressures physicians face regarding disclosure, because ethicists and patient advocates promote full disclosure while risk managers and malpractice insurers often urge restraint.^29,39 Standards offer little guidance, requiring disclosure of “unanticipated outcomes” but remaining silent regarding what information should be disclosed.⁴⁰

While all the scenarios involved clear-cut serious errors, many physicians would not explicitly apologize. For example, only 33% would explicitly apologize, while 61% would simply express regret. Studies^41-43 in medicine and other fields suggest that individuals strongly prefer explicit apologies and that such apologies, while not a panacea, may prevent lawsuits and promote faster and smaller settlements when lawsuits are filed. Progressive malpractice companies, like COPIC in Denver, Colo, encourage physicians to apologize when clear-cut serious errors have harmed patients.⁴⁴ The lack of consensus about whether and how to apologize following errors makes it likely that patient expectations are not being met.

We found that physicians in both specialties were less likely to disclose an error that might not be apparent to the patient. Some dimensions of errors might justify disclosing less information, such as if the error caused only trivial harm. However, physicians agreed that all the scenarios represented serious errors. Basing disclosure decisions on whether the patient was aware of the error is not ethically defensible or consistent with standards such as those from the Joint Commission on Accreditation of Health Care Organizations.⁴⁰

This study highlights differences in how medical and surgical physicians disclose errors. Surgeons reported greater intention to disclose errors than medical specialists, but disclosed less information, especially regarding using the word “error” and communicating about error prevention. Perhaps a disclosure style has developed within the specialty of surgery that focuses more on the adverse event itself than on whether the adverse event was due to an error.⁴⁵ While many surgical adverse events are not due to errors, some surgical adverse events are caused by clear-cut errors.²⁵ In the bile duct injury scenario, an error caused by the surgeon's lack of familiarity with a new device, only 21% would disclose the fact that the injury was caused by an error. The patient might logically assume that this adverse outcome was merely an unavoidable complication. Yet, if this patient subsequently learned from a source other than the surgeon how the bile duct injury occurred, the patient's trust and satisfaction would likely be reduced.

Examining the predictors of disclosure in this study raises important questions for further research. Many have suggested that the external malpractice climate is a key determinant of disclosure.^33,46,47 Canadian physicians, who practice in a much less hostile malpractice environment than their US counterparts, indeed would disclose more information than US physicians in this study.^48-50 Yet, in a recent survey,³ US and Canadian patients were equally likely to report that they had not been told about a medical mistake. Furthermore, while medical specialists' malpractice attitudes were associated with disclosure, surgeons' malpractice attitudes were not independent predictors of disclosure. Additional research should explore the complex interrelationships between malpractice issues, physician specialty, and disclosure. Finally, several factors beyond malpractice also influenced how physicians approached disclosure and, therefore, merit further exploration, including physicians' rating of an error's severity, perceived responsibility for the error, prior disclosure experience, and general disclosure attitudes.

This study has several limitations. The scenarios were all hypothetical, and it is not known how physicians would actually behave. Social desirability bias might cause our results to overestimate physicians' willingness to disclose errors. Also, the error scenarios, although designed using a theory-driven approach and carefully pilot tested, did differ in ways other than whether they were surgical or medical errors and how apparent the error might be to the patient. These differences might complicate comparisons between scenarios. In addition, respondents received specialty-specific scenarios. We do not know how surgeons would have responded to the medical scenarios or vice versa. While our response rate was robust, nonresponse bias could have affected our results. In addition, the US physicians we surveyed were located in only 2 states. However, our large sample size of more than 2500 medical and surgical physicians in academic and private practice and in rural and urban settings supports the generalizability of our findings.

Being more open with patients about errors represents a paradigm shift for the medical profession, a process that is just beginning. Additional research should study how disclosure affects litigation to address this real barrier to disclosure. Ethicists should consider whether different dimensions of errors justify disclosing more or less information to patients about errors. By integrating empirical research and normative analyses, the medical profession can develop guidelines for what information patients can expect from their physicians following errors. Ideally, if these guidelines can help physicians choose their words following errors in closer alignment with patients' pREFERENCES, including apologizing and providing information about preventing recurrences, such disclosure could enhance patients' confidence in the honesty and integrity of the health care system.

Correspondence: Thomas H. Gallagher, MD, University of Washington School of Medicine, 4311 11th Ave NE, Suite 230, Seattle, WA 98105-4608 (thomasg@u.washington.edu).

Accepted for Publication: May 11, 2006.

Author Contributions: Dr Gallagher had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Financial Disclosure: None reported.

Funding/Support: This study was supported by grants 1U18HS1189801 and 1K08HS01401201 from the Agency for Healthcare Research and Quality and by the Greenwall Foundation Faculty Scholars Program.

Role of the Sponsor: The funding bodies had no role in data extraction and analyses, in the writing of the manuscript, or in the decision to submit the manuscript for publication.

Acknowledgment: We thank Alison Ebers for her tireless work throughout this project; Kerry Bommarito, MPH, Melissa Krauss, MPH, and Irene Fischer, MPH, for collecting and processing the survey data; and Mary Lucas, RN, MA, for her assistance with manuscript preparation.