Health Plan Members' Views about Disclosure of Medical Errors

Since the publication of the Institute of Medicine's report, “To Err Is Human” (1), the problem of medical errors has gained increased recognition by the general public (2). Researchers and policymakers have begun to focus serious efforts on improving patient safety (3-6), but the complexity and uncertainty inherent in the practice of medicine make it unlikely that medical errors will ever be eliminated completely. When medical errors occur, effective physician–patient communication is critical. Ethical and professional guidelines emphasize that physicians have a responsibility to disclose medical errors (7-9), and recent standards link disclosure of adverse events to hospital accreditation (10). There may also be pragmatic reasons for disclosure—patients who pursue litigation may be motivated by a desire for a full explanation of the adverse event (11-13), and patients may be less likely to sue physicians who provide full and honest explanations (14).

Although the preponderance of opinion is that disclosure is the right thing to do when a medical error has resulted in patient injury, disclosure often does not occur (15-20). Several articles (21-33) and continuing medical education materials developed by national organizations (34) offer guidance, but there is limited empirical literature related to disclosure of medical errors to patients or families (11-20, 35-41). Of these articles, most focused on whether disclosure occurred or would occur; few examined the disclosure process or the consequences of disclosure in any way. Thus, the lack of empirical research in this area may be a barrier to disclosure.

We therefore undertook a study to assess patients' views of disclosure. We surveyed a representative sample of members of a large health plan to measure responses to a set of 8 vignettes, each of which described a medical error, the clinical outcome of the error, and a physician–patient dialogue about the event. The goals of the study were to examine the relationship of the error, severity of clinical outcome, and the level of disclosure to 5 dependent variables: likelihood of changing physicians, likelihood of seeking legal advice, patient satisfaction, trust, and emotional response. We also sought to determine patients' attitudes, beliefs, and preferences about medical errors and disclosure in general.

Methods

Sample

We conducted a mailed questionnaire survey of 1500 members of a New England–based health maintenance organization. Member names were randomly sampled from a master list of all health plan members 21 to 85 years of age (n = 130 009). Selected members were sent a letter indicating that they would be receiving the questionnaire, followed approximately 1 week later by the questionnaire. A $5 cash incentive was enclosed with the questionnaire. Nonrespondents were sent a reminder postcard 1 week later and a second copy of the questionnaire 1 week after. The response rate was calculated as the number of usable completed surveys returned (958 surveys) divided by the number of mailed surveys (1500 surveys) minus the number of surveys returned by the mail service as undeliverable (47 surveys).

The study protocol was approved by the institutional review boards of the participating institutions.

Questionnaire Design and Administration

The questionnaire was developed on the basis of a systematic review of the literature (11-20, 35-41) and the results of 2 focus groups comprising 14 patient volunteers. The questionnaire was refined after pretesting for length and comprehensibility.

The final questionnaire, which is available from the first author, consisted of 4 sections: 1) a vignette describing a medication error and the clinical outcome of the error, followed by a physician–patient dialogue about the error; 2) items assessing participants' response to the scenario and dialogue; 3) items assessing attitudes, beliefs, and preferences about medical error and disclosure in general; and 4) background and demographic items. Each section is described in detail in this paper.

Vignettes and Dialogues

Eight versions of the questionnaire were developed to produce a completely crossed factorial design with 2 levels each of the following variables: type of medication error (inadequate monitoring of antiepileptic medication vs. prescribing penicillin when penicillin allergy is documented), clinical outcome of the error (life-threatening outcome vs. less serious outcome), and level of physician disclosure about the error (nondisclosure with limited information and responsibility vs. full disclosure with high information, responsibility, and promise to prevent recurrences). Figure 1 summarizes the vignette and dialogue combinations comprising each condition, and the Appendix presents them in detail. Each participant received only 1 version; that is, a single error and outcome vignette with a corresponding dialogue wherein the physician and patient or family member discussed the error and its outcome. Distribution of versions was randomized.

• Download figure
• Download PowerPoint

Although most medical errors are considered systems errors (42) rather than the responsibility of individual providers, the actions of the treating physician may contribute to, or mitigate the results of, an error. Furthermore, the physician will often be considered—by the health care team and by the patient—to be the most appropriate person to speak to the patient or family after an error has occurred (43-45). We therefore chose vignettes in which the physician had direct involvement in the error and restricted our focus to the physician as communicator. These decisions were not intended to contradict the importance of a systems approach to patient safety or perpetuate a culture of blame, but rather to provide a starting point for study in this important area.

Three items were included to assess whether the vignettes were effective in communicating as intended; these items queried whether respondents perceived that an error had been made in each scenario and dialogue combination, whether the physician was perceived as responsible, and whether the error was perceived as preventable. Five response options were provided: “strongly disagree,” “disagree,” “agree,” “strongly agree,” and “don't know.”

Responses to the Vignette and Dialogue

Respondents were instructed to imagine themselves as the patient or family member in the vignette and dialogue and to respond accordingly. Two items, developed for this study, focused on behavior. Respondents were asked to indicate whether they would change physicians or seek legal advice. The same 5 response options mentioned earlier were provided. For the analysis, responses were recoded to be dichotomous, with “agree” or “strongly agree” responses compared with all others. Patient satisfaction was assessed by using 5 items drawn from the American Board of Internal Medicine Patient Satisfaction Questionnaire (46). Trust in the physician was assessed by using 8 items drawn from the Trust in Physician Scale (47) plus 1 supplemental item (“This doctor seemed to be withholding important information.”). Item wording was modified slightly to refer to the vignette rather than a real encounter. Five semantic differential items were developed for this study to assess emotional response: “angry” or “calm,” “frightened” or “reassured,” “worried” or “relieved,” “wronged” or “respected,” and “dissatisfied” or “satisfied.” All were on a 5-point scale; the most positive response option (for example, “reassured”) was assigned a value of 5, and the most negative response option (for example, “worried”) was assigned a value of 1. The variable emotional response was computed as the mean across the 5 semantic differential items, with higher scores indicating a more positive emotional response.

Attitudes, Beliefs, and Preferences about Medical Error and Disclosure

A set of 9 items was developed on the basis of the focus group interviews and a review of the literature to measure attitudes and beliefs about medical errors and disclosure in general. Five response options were provided: “strongly disagree,” “disagree,” “neutral,” “agree,” and “strongly agree”. Nine items were developed to assess respondents' preferences for action after a medical error. These items were also based on the review of the literature on disclosure and were intended to directly assess preferences for commonly recommended responses to error (for example, “I would want to know the full details of how the error occurred.”). Again, 5 response options were provided: “strongly disagree,” “disagree,” “neutral,” “agree,” and “strongly agree.”

Background and Demographic Items

Final items queried personal experience with medical error or injury, history of having filed a lawsuit against a physician or other health care provider, sex, education, age, and race or ethnicity.

Statistical Analysis

We compared background characteristics of respondents and nonrespondents by using a chi-square test for sex and a t-test for age. Associations between the 3 design variables (type of medication error, severity of clinical outcome, and level of disclosure) and the 2 dichotomously coded variables (likelihood of changing physicians and likelihood of seeking legal advice) were modeled by using logistic regression analysis. Odds ratios and 95% CIs were estimated. The association between design variables and patient satisfaction, trust, and emotional response was modeled by using linear regression analysis. Linear regression coefficients and 95% CIs were estimated. All regression analyses were conducted by evaluating 3 models: a main-effects model that included only the 3 design variables; an interaction model that included the 3 design variables plus all 2-way and 3-way interactions (where terms for 3-way interactions were not statistically significant, further analyses were conducted to test for statistically significant 2-way interactions); and a full model that included adjustment for age, sex, education, race or ethnicity, personal experience with errors and personal suffering due to errors in addition to the 3 design variables and any statistically significant design variable interactions determined by the interaction model.

Model-based means and percentages were estimated by setting other variables in the model to their mean values.

Chi-square tests were used to assess whether each of the design variables (type of error, severity of clinical outcome, and level of disclosure) influenced responses to the items intended to measure general attitudes, beliefs, and perceptions related to medical error, using a criterion value of a P value less than 0.01.

Statistical analyses were done with Stata software, version 8 (Stata Corp., College Station, Texas), and SPSS software, version 11.5 (SPSS, Inc., Chicago, Illinois).

Role of the Funding Source

The funding source had no role in the collection, analysis, or interpretation of the data or in the decision to submit the manuscript for publication.

Results

A total of 958 completed usable questionnaires were returned, for a response rate of 66%. Table 1 presents the characteristics of survey respondents. Compared with nonrespondents, respondents were more likely to be female (57% of respondents were female, compared with 44.7% of nonrespondents; Pearson chi-square, 21.2; P < 0.001). Respondents were also older than nonrespondents (mean age for respondents, 51.9 years; nonrespondents, 45.0 years; t = 8.29; P < 0.001). Most respondents (91.6%) reported their race or ethnicity as white, which is consistent with the U.S. census statistics for this area for the year 2000, in which 89.6% of respondents identified themselves as white (48). With respect to education, 28.5% of respondents 25 years of age or older reported earning a 4-year college degree or higher compared with 27% of census respondents 25 years of age or older (48).

Table 1. Characteristics of Respondents

Almost one third of the sample (31.4%) reported personal experience of a medical error in their own treatment or that of a close family member, 12.1% reported having personally suffered an injury or harm as the result of a medical error, and 1.4% reported having filed a claim or lawsuit against a physician or other health care provider (Table 1). This rate of reporting personal experience with error is slightly lower than the rate reported by a national survey in which 42% of the public and 35% of physicians reported experience with an error in their own care or that of a family member (19).

The randomization of questionnaire versions resulted in groups that were balanced in terms of age, sex, education, race or ethnicity, and personal experience with errors and personal suffering due to errors (data not shown).

Validation of Error Vignettes

In response to the validation items, most respondents in all conditions agreed that the physician had made an error (95.8%), that the physician was at fault (88.8%), and that the incident could have been prevented (94.8%).

Likelihood of Changing Physicians or Seeking Legal Advice

Figure 2 presents the observed data and Table 2 summarizes the results of the logistic regression analyses for patients' likelihood of changing physicians. All 3 independent variables, type of medication error (inadequate monitoring of antiepileptic medication vs. prescribing penicillin when penicillin allergy is documented), clinical outcome of the error (life-threatening outcome vs. less serious outcome), and level of physician disclosure about the error (nondisclosure with limited information and responsibility vs. full disclosure with high information, responsibility, and promise to prevent recurrences) were related to the likelihood of changing physicians. Nondisclosure, life-threatening outcome, and the inadequate monitoring error all increased the likelihood of changing physicians (Table 2). However, interpretation of the odds ratios for this model is complicated by the statistically significant 2-way interaction (error situation by clinical outcome, likelihood ratio test; P = 0.03). This suggests that the effect of the clinical outcome may vary depending on the error; we therefore constructed separate models for each error situation (Table 2). The results of these additional analyses suggest that while the effect of disclosure was similar in the 2 error situations studied here, clinical outcome had a somewhat greater effect in the missed allergy error situation compared with the monitoring error situation (odds ratios, 3.086 [95% CI, 2.10 to 4.53] vs. 1.724 [CI, 1.18 to 2.51], respectively).

• Download figure
• Download PowerPoint

Table 2. Predictors of Changing Physicians and Seeking Legal Advice

Figure 2 reports the observed data and Table 2 reports the results of the logistic regression analyses for patients' likelihood of seeking legal advice. Overall, nondisclosure, life-threatening outcome, and the inadequate monitoring error were all associated with a higher likelihood of seeking legal advice (Table 2). However, a statistically significant 3-way interaction was found (likelihood ratio test vs. full model; P = 0.04), again complicating the interpretation of the odds ratios for the main-effects model and suggesting that the effect of disclosure may vary depending on the clinical outcome and the type of error. We therefore constructed 4 separate regression models, 1 for each error and outcome combination (Table 2). The results of these additional analyses suggest that the level of disclosure had little effect on the likelihood of seeking legal advice in 3 of the 4 error-and-outcome situations. However, for the missed allergy and serious outcome situation, those in the nondisclosure situation were much more likely to indicate that they would seek legal advice than those in the full disclosure situation (odds ratio, 6.593 [CI, 2.17 to 19.96]). Figure 2 illustrates these relationships and highlights the relatively high percentage of respondents who indicated that they would probably seek legal advice under some conditions.

Inclusion of age, sex, education, race or ethnicity, personal experience with errors, and personal suffering because of errors resulted in less than a 5% change in the odds ratios reported in Table 2. However, older patients were less likely to indicate that they would probably change physicians (odds ratio, 0.88 [CI, 0.80 to 0.96]; P = 0.003; for each 10-year change in age) or seek legal advice (odds ratio, 0.81 [CI, 0.73 to 0.91]; P = 0.001; for each 10-year change in age).

Satisfaction, Trust, and Emotional Response

Figure 2 presents the observed means and Table 3 presents the results of the linear regression analyses predicting patient satisfaction, trust in the physician, and patient's emotional response. Nondisclosure was associated with lower ratings on all 3 dependent measures: lower patient satisfaction, less trust in the physician, and a more negative emotional response. Similarly, the life-threatening clinical outcome was associated with less positive ratings on all 3 dependent measures than the less serious clinical outcome. Error type had a less consistent effect. The inadequate monitoring error was associated with lower trust and more negative emotional response than the missed allergy information error, but there was no difference in patient satisfaction in the 2 error situations. No statistically significant interactions were found for any of the 3 outcomes, suggesting that the effect of each independent variable was consistent regardless of the level of the other independent variables, at least for the conditions studied in this paper.

Table 3. Predictors of Satisfaction, Trust, and Emotional Response

Inclusion of sex, education, race or ethnicity, personal experience with errors, and personal suffering because of errors in the models resulted in less than a 5% change in the regression coefficients (differences) reported in Table 3. There was a small effect of education on satisfaction, with the college-educated respondents reporting lower satisfaction (β = −0.1 [CI, −0.2 to 0.0]; P = 0.05). There was also a small effect of race or ethnicity, with nonwhite respondents reporting greater satisfaction (β = 0.3 [CI, 0.05 to 0.49]; P = 0.02). Trust decreased with increasing education (college vs. less than college education; β = −0.15 [CI, −0.25 to −0.05]; P = 0.002) and reporting of personal experience with medical errors (β = −0.14 [CI, −0.24 to −0.05]; P = 0.004). Emotional response became more positive with increasing age (β = 0.08 [CI, 0.04 to 0.12]; P = 0.001; age coded in 10-year units) and decreased with reporting of personal experience with medical errors (β = −0.21 [CI, −0.34 to −0.08]; P = 0.002).

Attitudes, Beliefs, and Preferences about Medical Error and Disclosure in General

Table 4 presents the percentage of respondents agreeing with statements intended to measure beliefs and attitudes about medical errors and disclosure in general. Of interest, almost half of all respondents (46.5%) expressed agreement with 2 apparently incompatible statements (“It is realistic to expect that doctors will make errors” and “Patients have a right to expect that their doctors will not make errors”). Table 4 also presents the percentage of respondents favoring various responses to a medical error, which are ordered from most desirable to least.

Table 4. General Opinions and Preferences for Action

Discussion

Our findings suggest that full disclosure after a medical error reduces the likelihood that patients will change physicians, improves patient satisfaction, increases trust in the physician, and results in a more positive emotional response. Full disclosure may also reduce the likelihood that patients will seek legal advice under some, but not all, circumstances. Considering the error or outcome situations studied here, we found that full disclosure had a statistically significant effect on the likelihood of seeking legal advice in only 1 situation (missed allergy error and serious clinical outcome).

Of the 3 independent variables in this study (type of error, severity of clinical outcome, and level of disclosure), only level of disclosure is controlled by the physician once an error is discovered. Our results suggest that physicians can influence the consequences of the disclosure process in a positive way if they disclose fully by explaining what occurred, acknowledging responsibility, apologizing, and promising to work to ensure that the error does not recur. These findings are consistent with the results of a recently published study that used focus groups (36). However, our results also suggest that physicians should not assume that full disclosure guarantees a positive response; patients are also influenced by the clinical outcome and the specifics of the error situation. Again, this is consistent with reports that the characteristics of the injury are influential in determining whether patients pursue litigation (49, 50). Thus, providers should be aware that, under some circumstances, full disclosure could have little effect. The percentage of respondents who agreed that they would seek legal advice was relatively high under some circumstances, which is clear from Table 2 and Figure 2. It is noteworthy, however, that complete disclosure did not increase the likelihood of patients agreeing that they would seek legal advice under any circumstance studied here.

Our findings confirmed that patients want to be told of medical errors, even if there is nothing that can be done about them. Few respondents agreed that some circumstances might justify nondisclosure. These observations are consistent with results of previous studies, in which most patients endorsed disclosure of medical errors (14, 19, 36, 37). Our findings also suggest that financial issues should be addressed when an error results in injury. Most patients would want medical fees associated with an error to be waived, and most favored financial compensation if an injury occurred. Relatively few patients favored compensation when no injury occurred. While we did not directly examine the effect of compensation on the likelihood of seeking legal advice, others have found that patients are more likely to consider legal action if the physician pursues payment when an injury has occurred (49) and that financial pressures (including medical bills associated with the adverse event) may motivate legal action (51).

The finding that almost half of the respondents expressed agreement with 2 seemingly contradictory statements—that “It is realistic to expect that doctors will make errors” and “Patients have a right to expect that their doctors will not make errors”—deserves comment. Although agreement with both statements may seem logically inconsistent, we feel that both beliefs are central to the public's attitudes about medical errors and that neither should be discounted. Gallagher and colleagues (36) reported similar attitudes among focus group participants who expressed a belief that medical errors are inevitable and that the possibility of experiencing an error in one's own care was frightening. Similarly, Brodsky and colleagues (52) reported that approximately 50% of survey respondents believed that it was not rare for physicians to make mistakes, but almost all respondents (99.3%) agreed that “People should be able to trust their doctors to give them the right care.”

Our findings make clear that when errors occur, most patients want detailed explanations, a sincere apology, and assurances that steps will be taken to prevent recurrences. At least in the abstract, safety is more important to patients than compensation or punishment. However, almost 4 in 10 respondents favored punishing the physician and an even greater number (> 6 in 10 respondents) favored reprimanding the physician. It is possible that the impersonal nature of the survey, which did not refer to the respondent's own physician, was responsible for these relatively high rates of endorsing negative consequences.

Our study has several limitations. Our use of vignettes and dialogues allowed us to systematically vary factors in ways that would be unethical to manipulate in practice. However, this approach assumes that responses to hypothetical situations are predictive of responses to real situations.

Because we drew our sample from a single geographic area, we cannot assess whether our conclusions are valid for other populations. The fact that the study was conducted in the context of a managed care plan may also have affected patients' perceptions and responses, especially about questions related to cost considerations.

Our findings add to the current knowledge of patients' beliefs, preferences, and attitudes about medical errors and disclosure but also highlight the need for additional research. Among the many unanswered questions is whether patients perceive errors as attributable to systems and how this relates to the primacy of the physician–patient relationship. The error events in the vignettes in this study were systems errors (that is, there were several points where systems changes could have prevented the adverse event), but we did not make this explicit to participants. Patients' responses would probably change if the vignettes stressed the complexity of the system. Future research efforts should consider the effect of such variations on patients' responses. Consideration of the systems perspective and its implications is also relevant to the disclosure dialogue. We chose to portray the physician as sole communicator in the dialogue with the patient or family member, but such conversations could include people other than or in addition to the physician; patient preferences and responses to such variations should be investigated.

There are several dimensions of the clinical outcome of the error that were not explicitly studied but seem to be influential. Future research should investigate the effect of variables such as the pain associated with the outcome; whether hospitalization, surgery, or therapy is required; whether the injury is temporary or permanent; and the patient's status before the injury. Several variables probably influence the disclosure process and patients' responses, including who initiates the conversation, timing of the conversation (relative both to the occurrence of the adverse event and to determination of causality), existence of a previous physician–patient relationship, and financial considerations.

Conventional wisdom suggests that when a patient is harmed by medical error, full disclosure—including acceptance of responsibility, an apology, and an explanation—will result in the best outcomes for both patient and physician. We found that full disclosure incorporating these elements and assurance of efforts to prevent recurrence resulted in more positive outcomes in terms of patient satisfaction, trust, and emotional response and decreased the likelihood of changing physicians. The effect of disclosure on the likelihood of seeking legal advice was more complex, suggesting strong situational influences in this area. Both the clinical outcome of the error and the specifics of the error situation influenced how people respond to medical errors. We conclude that full disclosure fulfills patients' expectations and may help sustain or strengthen the patient–physician relationship, but it may not prevent litigation under some circumstances.