SUBJECTS AND METHODS

Subjects

In May 1989 we mailed a questionnaire to 254 house officers in three internal medicine training programs associated with medical schools. Programs were located at large (>500 beds) academic tertiary care hospitals.

Procedures

Questionnaires were filled out anonymously to assure confidentiality. House officers were asked to return a postcard indicating either that they had mailed the completed questionnaire or that they did not wish to participate in the study. If the postcard was not returned, house officers received two additional mailings and a personal reminder from one of the authors. Approval for the study was obtained from institutional review boards at all three institutions.

Questionnaire

The questionnaire was developed after a review of the literature^7,10–19 and two stages of pretesting. Subjects were asked to describe their most significant medical mistake in the last year, their response to it, and the events that followed. A mistake was defined as an act or omission for which the house officer felt responsible that had serious or potentially serious consequences for the patient and that would have been judged wrong by knowledgeable peers at the time it occurred.

Respondents first wrote a paragraph about the mistake and then answered questions about the age and prognosis of the affected patient, adverse patient outcomes, and perceived causes of the mistake.

In describing responses to the mistake, house officers answered questions about the degree to which they accepted responsibility for the mistake, their emotional response to the mistake, discussions about the mistake with others, the institutional response to the mistake, and changes in practice due to making the mistake.

Questions used four-point Likert-type and categorical response formats. Respondents were also encouraged to write comments at the end of the questionnaire.

Scales

We grouped items on the questionnaire into scales representing meaningful concepts on the basis of factor analysis and consensus of the authors’ judgment. Each scale score was created by summing the responses to the items it included.

Causes of the mistake were described by three scales: inexperience (three items), job overload (two items), and case complexity (four items). Responsibility for the mistake was measured with three items from the “accepting responsibility” subscale of the Ways of Coping Scale developed by Folkman and Lazarus.²⁰Emotional distress in response to the mistake was measured with four items. The extent to which the institutional response was judgmental was measured with two items. The extent of discussion was measured by summing affirmative responses to items that asked whether the physician discussed the mistake with the supervising attending physician, another medical person, the patient or family, or at a conference.

Learning from the mistake was measured by two scales that asked house officers how they changed their practices due to the mistake. A scale of constructive changes in practice contained nine items. A scale of defensive changes contained two items. It should be noted that constructive and defensive changes measure separate concepts rather than polar opposites of the same scale. Therefore, a house officer might report both constructive and defensive changes in practice after making a mistake.

Means, SDs, and internal consistency reliability coefficients (Cronbach’s α) for each of these scales are shown in table 1. Relatively large SDs for the overload, judgmental, and defensive scales reflect skewed score distributions. The non-normal distributions make the α coefficient difficult to interpret. To facilitate comparison of the different scale scores, scores were transformed linearly to a scale of 0 through 100, with 0 indicating the lowest and 100 indicating the highest possible score.

Analysis

Analysis was conducted in two stages. In the first stage, two-sample t tests and one-way analyses of variance were used to test the relationship between the dependent variables (constructive change and defensive change) and categorical independent variables (house officer gender and year of residency training; institution and setting of the mistake; patient age group, previous functioning, and life expectancy; whether or not there was a serious outcome; and extent of discussion). Simple correlations were used to evaluate the relationship between the dependent variables and continuous independent variables (scales for causes of the mistake, accepting responsibility for the mistake, and institutional response to the mistake).

In the second stage, variables that had been found to be related to the dependent variables at p<0.15 were included in two multiple linear regression equations to test their independent relationship to (1) constructive changes in practice and (2) defensive changes in practice.

RESULTS

Characteristics of respondents

Of the 254 residents surveyed, 114 (45%) responded by reporting a mistake and completing the questionnaire. An additional 56 residents (22%) returned a postcard acknowledging receipt but declining to complete the questionnaire. The remaining 33% did not respond.

Our study group comprised the 114 respondents who completed the questionnaire. Because the results did not differ by site, we present only aggregated results. 33% of the subjects were women. 36% of the respondents were interns, 32% were junior residents, and 32% were senior residents. The distributions of gender and year of training were similar among respondents and non-respondents.

Types of mistakes

Types and frequency of mistakes are summarized in table 2. The most frequently reported type of mistake was a missed diagnosis (33%). In one typical case, a house officer failed to recognize congestive heart failure in a patient with human immunodeficiency virus disease with severe dyspnea.

Errors in evaluation and treatment were reported in 21% of cases. For example, one resident noted but failed to treat profound hypoglycemia in a patient with the acquired immunodeficiency syndrome admitted with neutropenia and presumed sepsis. The patient had a seizure and died soon thereafter.

House officers reported errors in prescribing and dosing of drugs in 29% of cases. One resident missed an intern’s drug dosing error in an elderly woman with congestive heart failure who was well known to him from previous admissions. “I approved the intern’s admission orders without noting a significant error,” in which an 80 mg dose of a cardiac medication was transcribed as 180 mg. The patient was found dead 2 hours after her first dose.

Errors ascribed to faulty communication were described in 5% of cases. In one such case, a resident accepted misinformation from the emergency department physician that a patient being admitted was not to be resuscitated. “I subsequently found out from the patient’s family and personal physician that the patient was not a ‘no code.’ At that point in time the patient had not been treated aggressively and died 24 hours later.”

Examples of procedural complications, described in 11% of cases, and other types of mistakes are given in table 2; a brief summary of all of the mistakes is presented in table 3.

Outcomes of mistakes

In response to the question “What adverse effects did the mistake have for the patient?”, 90% of residents reported that patients had significant adverse outcomes following mistakes. These included physical discomfort (32%), emotional distress (27%), additional therapy (25%), additional procedure (13%), prolonged hospital stay (24%), and death (31%). Mistakes often had multiple adverse outcomes. For 10% of patients, no adverse outcome was attributed to the mistake. A brief summary of the reported outcomes of the mistakes is included in table 3.

Causes of mistakes

The causes of mistakes reported by house officers varied (table 4). House officers usually attributed mistakes to more than one cause: 54% reported that mistakes were caused in part because they did not know information they should have known (e.g. being unaware of the significance of a prolonged episode of ventricular tachycardia); 51% reported “too many other tasks” (e.g. one resident neglected to continue to administer a required medication, being “too busy with other sick patients and supervising interns and students”); 41% reported fatigue (e.g. after inadvertently ordering potassium replacement as a bolus, one resident commented, “It was 3 am and I’m not sure I was completely awake”).

Circumstances of mistakes

The mistakes occurred during medical school in 3% of cases, during the first year of residency in 53% of cases, during the second year of residency in 36% of cases, and during the third year of residency in 9% of cases. The mistakes happened with inpatients in 77% of cases, emergency department patients in 14% of cases, and outpatients in 9% of cases. The patients involved in the mistakes were less than 18 years old in 1% of cases, 18–64 years in 60% of cases, and 65 years or older in 39% of cases. House officers estimated the life expectancy of patients to be less than 1 month in 10% of the cases, 1–6 months in 22% of the cases, 6–12 months in 18% of the cases, and greater than 12 months in 50% of the cases.

House officers’ responses to mistakes

House officers reported discussing the mistake with the supervising attending physician in only 54% of cases. However, 88% of house officers discussed the mistake with another physician who was not in a supervisory capacity. House officers discussed the mistake with the patient or patient’s family in only 24% of cases; 58% of house officers reported talking to a non-medical person about the mistake. Only 5% of house officers did not tell anyone about the mistake. On a scale ranging from 0 to 100 for extent of discussion, the mean score was 52.5 (SD 22.8). On average, house officers discussed the mistake with two of the following: their supervising attending physician, another medical person, the patient or family, or at a conference.

Most house officers were willing to accept responsibility for their mistakes. Subjects’ responses included “promising to do things differently the next time” in 76% of cases, “criticizing or lecturing oneself” in 62% of cases, and “apologizing or doing something to make up” in 21% of cases. On a scale ranging from 0 to 100 for accepting responsibility, the mean score was 54.5 (SD 22.3).

House officers experienced emotional distress in reaction to the mistakes. After a fatal mistake involving a young patient, one house officer wrote: “This event has been the greatest challenge to me in my training.” They felt remorseful in 81% of cases, angry at themselves in 79% of cases, guilty in 72% of cases, and inadequate in 60% of cases. On a scale that ranged from 0 to 100, the mean level of distress was 71.3 (SD 23.7). The correlation between distress and accepting responsibility was 0.58 (p<0.0001). 28% of house officers feared negative repercussions from the mistake.

A few house officers reported persistently negative psychological impact of mistakes. After a mistake caused the death of a patient, one house officer commented, “This case has made me very nervous about clinical medicine. I worry now about all febrile patients since they may be on the verge of sepsis.” For another house officer, a missed diagnosis made him reject a career in subspecialties that involve “a lot of data collection and uncertainty.”

Institutional responses to mistakes

Mistakes were discussed in attending rounds in 57% of cases and at the morning report or morbidity and mortality conference in 31% of cases. However, house officers stated that, in about half of these conferences (48%), “the tough issues were not addressed.” One house officer believed “the key issues were ignored by the morbidity and mortality committee, i.e. being overworked, having too many patients to care for at one time.”

House officers felt that the hospital atmosphere inhibited them from talking about the mistakes in 27% of cases and that the administration was judgmental about the mistakes in 20%. One house officer felt that public discussion is counterproductive: “Training programs do not sympathize or help one learn from one’s mistakes. Instead, the administration is usually critical and often ostracizes the individual.” In contrast, although another house officer was initially reluctant, she found discussing her mistake to be a positive experience: “Presenting this case at intern’s report was difficult—I felt under a lot of scrutiny from my peers. In the end, I felt as though I had gotten more respect from presenting this kind of case rather than one where I had made a great diagnosis.”

Changes in practice

Almost all residents (98%) reported some change in practice in response to their mistakes. The most frequently reported changes were paying more attention to detail (82%), confirming clinical data personally (72%), and seeking advice (62%). Most residents (98%) reported at least one constructive change. Only 18% reported one or more defensive changes. A summary of constructive and defensive changes reported by house officers is shown in table 5. In addition, 26% of respondents described ordering more tests as a result of their mistakes. In review, the authors believe that ordering more tests might have prevented the mistake in most cases. Thus, we did not group this item with defensive changes.

Factors relating to reported changes in practice

We examined how predictor variables—physician characteristics, patient characteristics, type and seriousness of the mistake, causes of the mistake, and responses to the mistake by the physician and the institution—were related to reported constructive and defensive changes in practice.

In univariate analysis, constructive changes in practice were significantly associated (p<0.05) with female gender, serious outcome, inexperience, or case complexity as causes of the mistake, accepting responsibility for the mistake, and extent of discussion of the mistake. Defensive changes in practice were significantly associated with house officers’ perceptions of job overload as a cause of the mistake and perceptions that the institution responded judgmentally. Changes in practice were not significantly related to age, functional level, or prognosis or to physician year of training or institution.

In multivariate analysis, reported constructive changes in practice were associated with several independent predictors (table 6). Residents were more likely to report constructive changes if the mistake was caused by faulty judgment in a complex case or by inexperience, but they were less likely to do so if they perceived that the mistake was caused by job overload. Physicians who responded to the mistake with greater acceptance of responsibility and more discussion were also more likely to report constructive changes. The independent variables shown in table 6 were associated with 44% of the variance in constructive changes. Constructive change is reported on a scale of 0 to 100, with 33 equivalent to an average response of “disagree somewhat” and 67 equivalent to an average response of “agree somewhat.” The independent effect of a predictor variable on constructive change can be calculated by multiplying the β coefficient by the difference in score or category for that predictor variable, as noted in table 6.

Defensive changes in practice were more likely if there was a judgmental institutional response to the mistake (β=0.37, p<0.001). In multivariate analysis, the model was associated with 29% of the variance in defensive changes. However, the small number of respondents reporting defensive changes gave this analysis relatively little power to detect significant predictors.

COMMENT

Mistakes are inevitable in clinical medicine, given its inherent uncertainty and complexity and the need to make decisions despite limited information. Because house officers are taking on new clinical responsibilities, they may be particularly likely to make mistakes.

This study suggests several ways to help residents learn from their mistakes and institute constructive changes in practice. First, house officers should be encouraged to accept responsibility for their mistakes. In our study, residents who reported accepting responsibility reported constructive changes in practice more often than residents who did not accept responsibility. However, accepting responsibility for mistakes was also strongly associated with emotional distress. For example, one resident described persistent feelings of guilt and shame after inappropriate management of a diabetic foot ulcer led to an amputation. Thus, supervising physicians who encourage house officers to accept responsibility for their mistakes need to respond sensitively to the distress those house officers may experience.

Second, house officers should be encouraged to discuss their mistakes with attending physicians. While house officers candidly described their mistakes in the questionnaire, barely half had told their attending physicians about them, although the attending physician is legally and ethically responsible for patient care. Several house officers expressed the desire for helpful discussion. One resident wanted more discussion so that “some of the unsaid horrors of our experiences can be discussed and dealt with.” Another wrote, “I was very disturbed that there was never really an opportunity to discuss the mistake … I was also very frightened by the impact that carelessness or ignorance on my part could have on someone else’s life.” In training programs, mistakes are traditionally discussed at conferences and rounds. In this study, however, when their mistakes had been discussed in a conference, half of the house officers said that the “tough issues were not addressed.” In non-medical specialties, avoidance of important issues may be a common response to mistakes. For example, in psychiatry, suicide review conferences often transform “negative evidence into a positive display of an attending’s skill.”²¹ In surgery, a morbidity and mortality conference consists of “ceremonial apologies” by attending physicians.⁶ The limited role of residents in these proceedings may preclude useful discussion. Future studies should explore why house officers are reluctant to tell their supervisors about their mistakes and how to encourage fruitful discussion.

Because mistakes may have harmful consequences for patients, it is important to try to reduce their frequency and severity. Our findings regarding the reported causes of mistakes suggest specific strategies for preventing mistakes. First, more active supervision may prevent some mistakes or mitigate their adverse effects. Senior physicians should be more available for critical decisions about patient care, especially in complex cases that require more mature clinical judgment. One officer complained, “As an intern, I couldn’t—and didn’t—know enough to manage the case.” Another speculated, “If I had had more attending support all along with this patient, the diagnosis would have been made much sooner and the patient might have survived.”

Attention must be given to house officer work load. McCue²² has suggested that sleep deprivation during training may teach house officers to tolerate and rationalize unnecessary errors. In our study, house officers reported that job overload played a part in 65% of mistakes. Moreover, house officers who reported being fatigued or having too many tasks to perform were less likely to seek information following a mistake. Such information seeking might help prevent future mistakes.²³

Disclosure of mistakes to patients or their families is a difficult issue. In our study, such disclosure was reported by fewer than one quarter of house officers. This finding is consistent with reports suggesting that physicians are reluctant to tell patients about mistakes.^5,24,25 Legal and ethical experts, however, suggest that a patient generally should be told about a mistake.^4,26–28 Disclosure of a mistake may also foster learning by compelling the physician to acknowledge it truthfully. Indeed, our study suggests that accepting responsibility may precede learning from a mistake. Finally, Hilfiker⁸ argues that disclosing a mistake to the patient may be the only way for the physician to achieve a sense of absolution. However, telling patients about mistakes may be difficult because there are no guidelines about how to do so. One way might be for the attending physician and house officer to inform the patient of the mistake together. Such joint discussions might benefit house officers by providing emotional support and role modeling.

Our findings may be limited in several important ways. First, since accounts of mistakes and changes in practice were anonymous, we have no external confirmation of the data. Some residents may have exaggerated the impact of their mistakes. Many patients were terminally ill and medically unstable, and the mistakes might not have caused the adverse outcomes. Second, the limited response rate, the relatively small sample size, and the sample of internal medicine residents at large teaching hospitals limit the generalizability of our findings. It is likely that non-respondents felt more defensive than respondents. If so, the actual severity of outcomes might be worse than we reported, and the proportion of mistakes that are discussed might be less than our findings indicate. Finally, some associations we found may be due to unmeasured confounding variables rather than cause-and-effect relationships. For example, unmeasured personality characteristics of house officers might cause them both to discuss mistakes with others and to make constructive changes in practice.

Medical training and patient care will benefit from an environment that allows house officers to learn constructively from their mistakes. Supervising physicians need to encourage house officers to accept responsibility for their mistakes and need to provide opportunities for discussing mistakes. Directors of training programs should resolve problems in staffing and scheduling that may contribute to mistakes and impede learning. Physicians can learn from their mistakes even as they strive to minimize their occurrence.