Participants

Procedures

We obtained informed consent from all participants before data collection. Qualitative data were gathered with a semistructured interview that allowed doctors to tell their individual stories about caring for a patient who died. At the end of the interview doctors were asked six questions that summarised the interview material (see table 1). Doctors then completed self administered questionnaires. We have focused on the quantitative data, including the six questions administered at the end of the interview and self administered questionnaires.

Measures

We grouped the measures under one of three headings: doctors' experiences in providing patient care, doctors' emotional reactions to the death of the patient, and doctors' use of coping resources after the patient's death (table 1). We used three measures that show good reliability and validity in general populations: the grief reaction scale,9 a short version of the impact of events scale,10 and the brief COPE.11 The distributions for these scales were positively skewed in our sample, indicating that most doctors endorsed either “never or rarely” or “occasionally” having a given symptom or using a particular coping strategy. Therefore we dichotomised the scale items into endorsed/not endorsed and summed the rescaled items. The resulting distributions met normality assumptions necessary for conducting analyses.

We used literature review and discussions with experts to generate questions relevant to doctors' experiences in caring for the dying. We piloted and revised the questions to enhance their content validity. Questions generated by the study included: “How emotionally powerful was this death?” and “How satisfying was it to participate in the care of this patient?” Doctors' responses to the questions fell within a normal distribution.

Data analyses

Table 1 shows the outcome variables, including doctors' experiences, emotional reactions, and use of coping resources. The key predictor variable of interest was level of training (intern, resident, or attending physician); other predictors were variables related to doctors and patients, such as sex, marital status, duration of care (table 2). To reduce the probability of type I errors from multiple testing, we set a more stringent predetermined α of 0.01 (two tailed), and we have reported only those variables that were consistently related to outcome variables (that is, a given variable was significantly related to at least three outcome variables).

We report univariate statistics as means and standard deviations for continuous variables and percentages for categorical variables. We tested bivariate associations using χ² tests to compare categorical variables, correlation coefficients to measure linear associations, independent groups t tests to compare means, and analysis of variance to compare means of more than two groups.

The analysis of variance models to test the association between level of training and the outcome variables needed to account for doctors clustering within shared index patient cases. We used the CROSSTAB procedure (SUDAAN, Research Triangle Institute, NC) and SURVEYMEANS and SURVEYREG (SAS Institute, Cary, NC) procedures to compute the model estimates.12 These programs adjust for covariance related to cases among doctors' scores so that the three analysis of variance groups were independent of one another.

Doctors' experiences in providing patient care

Most of the doctors (154, 82%) had not cared for the patient before the final admission to hospital. During that admission, doctors knew the patient for only a short period of time: 68 (36%) knew the patient for three days or less, another 68 (36%) knew the patient for four to seven days, and only 53 (28%) reported caring for the patient longer than a week. In total 117 doctors (62%) classified their relationship to the patient as not close (for example, ≤ 3 on a 0-10 point scale) (table 3). On the other hand, 21 (11%) reported feeling very close to the patient, and this was associated with providing patient care for a longer period of time (r = 0.38, P < 0.01).

One hundred and thirty nine (74%) doctors thought that taking care of the patient was a satisfying or very satisfying experience (table 3). Longer durations of providing care were associated with more satisfying experiences (r = 0.20, P < 0.01). Few doctors reported conflict among family and members of the healthcare team regarding the goals of patient treatment (32, 17%), poor symptom management (8, 4%), or patient suffering (24, 13%).

Doctors' emotional reactions to the patient's death

Descriptive analyses indicated that many doctors reported moderate emotional impact from a patient's death, though 58 (31%) doctors rated the death as having strong emotional impact (table 3). Many doctors (103, 55%) reported that the patient's death disturbed them very little, but 43 (23%) reported that the patient's death was very disturbing to them. On average, doctors reported experiencing two (out of a possible 14) symptoms of grief after the death. The most commonly reported symptoms were “feeling upset when thinking about the patient” (88, 47%) and feeling “numb” (45, 24%). Doctors reported around one symptom (out of six) on the impact of events scale, with “Avoid getting upset when thinking about the patient” being the most commonly endorsed item (70, 37%).

The differences in emotional reactions by level of training (from analysis of variance models) were non-significant. Although interns reported more symptoms of grief than attending physicians (mean 2.9 (SD 2.9) v 1.7 (2.5), P < 0.03), this finding was not significant. Longer durations of providing care were consistently associated with stronger emotional reactions: r = 0.21 (P < 0.01) for emotional impact, r = 0.19 (P < 0.01) for disturbing, r = 0.26, (P < 0.01) for grief, and r = 0.20 (P < 0.01) for the impact of events scale. Female doctors reported more symptoms of grief than male doctors (3.3 (3.2 v 1.8 (2.1), P < 0.01).

Doctors' use of coping resources after the patient's death

Doctors reported using an average of two coping strategies (out of five) to manage the emotions associated with an index patient's death. The most commonly endorsed were “getting emotional support from others” (86, 46%), “trying to see the death in a different light to make it seem more positive” (68, 36%), and “turning to work or other activities to take your mind off of it” (68, 36%). Female doctors used more coping behaviours than male doctors (2.2 (1.5) v 1.4 (1.4), P < 0.01), and residents used more coping behaviours than either interns or attending physicians (2.1 (1.5) v 1.6 (1.4) and 1.2 (1.4), P < 0.01).

Most doctors (156, 83%) reported needing little emotional support from their colleagues (table 3). Interns reported needing more emotional support than attending physicians (2.3 (2.2) v 1.1 (1.7), P < 0.01). Female doctors reported needing more emotional support than male doctors (2.5 (2.4) v 1.3 (1.7), P < 0.01). Although most doctors in need of support felt they had received it from their colleagues, 66 (35%) felt their needs had gone unmet.

Most residents (42, 74%) and interns (43, 68%) spoke to the attending physicians about the death of the patient (table 4). However, residents (48, 84%) and interns (56, 89%) relied more on talking with each other or another resident about their experiences with the death (table 4). Less than a quarter of residents (14, 24%) and interns (13, 21%) identified an attending physicians as the most useful person in helping them cope with the patient's death (table 4).

Potential risk factors

Our data show that while most deaths do not elicit a strong emotional response in doctors, we can identify those situations in which greater needs for emotional support may be present. These include the presence of a long doctor-patient relationship or the involvement of a female doctor or an intern. Being able to identify deaths in which the clinicians may be “at risk” of higher levels of emotional distress provides educators with some guidance about situations in which extra attention to these issues is essential. Unfortunately, in our current environment, these needs are likely to be unaddressed. Research has shown that senior physicians in charge of training house officers tend to underestimate emotional distress in junior doctors.19 20 Furthermore, the culture of medicine, through its emphasis on biomedical issues and its lack of emotional support,21 may also contribute to trainees' reluctance to report needs for emotional support. Although interns and residents were usually successful in identifying someone among their peers to talk to about the death, 46% (36) of attending physicians found that no one was helpful. What is the cumulative impact of emotionally draining deaths on attending physicians who “do not need” emotional support?

Limitations

Our study had several limitations. While we made every effort to contact respondents within two weeks of the patient's death, interviews were sometimes delayed, leading to potential recall bias. Although we used standardised instruments whenever possible, few standardised measures exist that specifically assess doctors' emotions. The quantitative measures used may have lacked the sensitivity to measure subtle and highly individual reactions to death that may be better elucidated through qualitative methods. We will report on our qualitative findings in subsequent papers.

Conclusions

Our research provides new insights into the effect of patients' deaths on doctors and raises some questions about current medical training in the United States. Doctors are moved by the deaths of the strangers for whom they care, and they are often powerfully affected by the deaths of patients with whom they have forged close relationships. The attending physicians in charge of the learning of their interns and residents do not often discuss these strong emotional responses. This conveys a message about how death is to be handled and potentially isolates learners who could benefit from having an opportunity to receive a seasoned perspective on what it is like to care for a patient who dies. A conspiracy of silence toward emotions can potentially cause trainees to develop maladaptive coping patterns that lead to burnout and other forms of emotional distress.22 The finding that a substantial number of trainees did not talk to their attending physician, and that most did not find attending physicians to be the most helpful resource in dealing with the death, points to a major gap in the clinical education of interns and residents and an important opportunity for attending physicians to improve their clinical training skills. As the clinical leader and role model for the team, it is incumbent on the attending doctor to teach and model both excellent patient care and appropriate self care in managing challenging patients. Senior physicians have a major role in legitimising discussions of these issues and in helping doctors in training to construct their losses in a meaningful and adaptive manner. Further research should investigate both how the attending physicians respond to distress in junior doctors as well as studies of how junior doctors' opinions regarding how attending physicians can be supportive.