Physicians’ and Nurse Practitioners’ Level of Pessimism About End-of-Life Care During Training: Does It Change Over Time?

Abstract

Context

An enhanced understanding of trainee attitudes about end-of-life care is needed to inform interventions to improve clinician communication about dying and death.

Objectives

To examine changes in trainee pessimism about end-of-life care over the course of one academic year and to explore predictors of pessimism among residents, fellows, and nurse practitioners.

Methods

We used baseline and follow-up surveys completed by trainees during a randomized controlled trial of an intervention to improve clinician communication skills. Surveys addressed trainee feelings about end-of-life care. Latent variable modeling was used to identify indicators of trainee pessimism, and this pessimism construct was used to assess temporal changes in trainee attitudes about end-of-life care. We also examined predictors of trainee pessimism at baseline and follow-up. Data were available for 383 trainees from two training programs.

Results

There was a significant decrease in pessimism between baseline and follow-up assessments. Age had a significant inverse effect on baseline pessimism, with older trainees being less pessimistic. There was a direct association of race/ethnicity on pessimism at follow-up, with greater pessimism among minority trainees (P=0.028). The model suggests that between baseline and follow-up, pessimism among younger white non-Hispanic trainees decreased, whereas pessimism among younger trainees in racial/ethnic minorities increased over the same time period.

Conclusion

Overall, trainee pessimism about end-of-life care decreases over time. Pessimism about end-of-life care among minority trainees may reflect the influence of culture on clinician attitudes about communication with seriously ill patients. Further research is needed to understand the evolution of trainee attitudes about end-of-life care during clinical training.

Introduction

Although medical trainees feel that learning from dying patients is a valuable experience, engaging in end-of-life care is a daunting endeavor for many (1). Feelings of sadness often accompany the care they provide to patients with serious illness, and the death of a patient may be associated with guilt or a sense of failure (1, 2). Trainees also may worry about the tension between supporting their patients’ hope while still providing realistic and accurate information about their illness (1, 2). A trainee’s attitude about end-of-life care may influence their communication practices (3), and a negative outlook may affect their ability to address difficult topics such as dying and death with seriously ill patients. An understanding of trainee attitudes about end-of-life care is vital to the development of educational interventions to enhance communication skills, yet little is known about the factors that influence trainee attitudes about communication with dying patients.

Using data from a five-year, randomized trial evaluating a training program designed to improve clinician communication skills (4, 5), we examined temporal change in individual medical resident and nurse practitioner attitudes about communication with dying patients. We focused our study on an attitude of pessimism, defined as “an inclination to emphasize adverse aspects, conditions, and possibilities or to expect the worst possible outcome” (6). We felt that this was an important attitude to study, given the potential for pessimism to impact a trainee’s communication with patients who have serious illness. The ability to address negative attitudes about end-of-life care also may promote trainee wellness and attenuate provider burnout (7). Finally, it may be possible to identify a pessimistic attitude toward communication about end-of-life care in the early stages of training and address this with an educational intervention. In addition to changes in trainee attitudes over time, we investigated associations between trainee characteristics and attitudes about communication with dying patients. Direct experiences with dying and death may improve attitudes about end-of-life care (8), thus we hypothesized that pessimism for all trainees would diminish over time and that older trainees would be less pessimistic about talking about end-of-life care.

Methods

Study Design and Participants

Data for this study were drawn from baseline and follow-up surveys completed by internal medicine residents, subspecialty fellows, and nurse practitioners during the Improving Clinician Communication Skills (ICCS) trial (5). Participants were recruited from the University of Washington (UW) and the Medical University of South Carolina (MUSC) between 2007–2012. Eligible trainees included all internal medicine residents, as well as fellows in pulmonary and critical care, oncology, geriatrics, nephrology, and palliative medicine subspecialties. Nurse practitioners and advanced practice nurses were eligible if they were in, or had recently completed, training programs that included care for adults with life-threatening or chronic illnesses.

Prior to the beginning of the academic year, trainees were contacted by mail or email. Study materials included letters from the director of their training program and from the investigators, a consent form, a refusal postcard and a postage-paid return envelope. Three additional reminders were provided to non-respondents using the following approach: 1) a second mailing at two weeks; 2) an email reminder at four weeks; and 3) a phone contact at six weeks. After written consent was obtained, participants were randomized to usual education or an inter-professional simulation-based communication skills workshop. Randomization was stratified by study site, discipline (MD or ARNP/APRN) and year of training. Baseline surveys assessing attitudes towards end-of-life care were sent to participants at the beginning of the academic year and follow-up surveys were sent at the year’s end, after the intervention or control education period. For trainees who completed both pre- and post-intervention questionnaires, the average number of days between completion of the two questionnaires was 268 days (approximately nine months), ranging from 45 days to 420 days. Median elapsed time was 280 days. Human Subjects approval was obtained from the institutional review boards at both institutions.

Developing a Pessimism Construct

To identify indicators of pessimism, we used trainee responses on surveys addressing their feelings about providing end-of-life care. This survey of trainee attitudes has been used in several other studies (2, 9), but has not been validated. Trainees completed these surveys at the time of their enrollment, which coincided with the beginning of the academic year, and then at the end of the academic year during which they participated. The survey of trainee attitudes about providing end-of-life care includes 18 items (Appendix I, available at jpsmjournal.com). Thirteen items are rated on a scale ranging from 1 (“strongly disagree”) to 4 (“strongly agree”) and the remaining items are scored on a scale from 1 (“never”) to 5 (“always”). Study personnel (A.C.L., R.A.E., J.R.C., L.D.) reviewed the 18 items on the survey and identified those most likely to measure pessimism that might affect end-of-life care provided to patients and their family members. The following three items were selected: 1) There is little that can be done to ease the suffering of grief; 2) It is not possible to tell patients the truth about a terminal prognosis and maintain hope; and 3) Talking about death tends to make patients with terminal illnesses more discouraged. Valid comparisons of a latent construct over time require evidence of longitudinal measurement invariance (i.e., that the construct has time-invariant meaning) (10). Thus, our first analysis goal was to establish that either the complete set of three selected indicators, or a subset of two of the indicators, exhibited measurement invariance over time. Each test for longitudinal invariance involved a structural equation model (SEM) based on weighted least squares (WLSMV) estimation, with indicators defined as ordered categorical variables. Each SEM included two correlated time-specific constructs: one representing baseline pessimism and the other representing pessimism at follow-up. The two constructs were measured with parallel sets of time-specific indicators, with indicator loadings and thresholds constrained to equality over the two time periods, with the construct at follow-up regressed on the baseline construct. A construct was judged to exhibit longitudinal measurement invariance if the model, thus constrained, fit the observed data, based on a χ² test of fit with P>0.05. Of the models tested, a two-indicator model exhibited better fit than the model with all three preferred indicators. In addition, the items in the two-indicator model were the most frequently chosen items in our a priori selection of preferred indicators. This two-indicator model was felt to best represent trainee attitudes about communication with patients suffering from serious, life-limiting illness, thus the final resulting pessimism construct (with χ²=1.728, 3 df, P=0.6306) was measured with two indicators: 1) It is not possible to tell patients the truth about a terminal prognosis and maintain hope; and 2) Talking about death tends to make patients with terminal illnesses more discouraged.

Predictor Variables

To examine the association between trainee characteristics and trainee pessimism, we selected several trainee factors obtained from self-report. These included trainee type (physician or nurse practitioner), age, sex, and race/ethnicity (white non-Hispanic or other). We also assessed baseline self-reported personal and professional experience with death prior to enrollment in this study, including the number of patients to whom the trainee personally delivered bad news, the number of times during training when the trainee personally discussed end-of-life treatment options, the number of times he/she observed other clinicians discussing end-of-life treatment issues, and whether he/she had ever experienced the death of a close relative or friend. Data from program offices were used to determine post-graduate training year and recruitment site (UW or MUSC). The trainee’s randomization status (control or intervention group) also was tested as a predictor.

Data Analysis to Identify Predictors of Pessimism

Analysis of the contribution of trainee characteristics to baseline and follow-up pessimism was based on a final causal model that was similar to the model used earlier to identify the time-invariant pessimism construct, in that it included the two time-specific two-indicator pessimism constructs, the time-invariant constraints on indicator loadings and thresholds, and a structural link from baseline pessimism to pessimism at follow-up. However, the final model included exogenous predictors of the two pessimism constructs, resulting from a series of tests involving the following potential predictors: recruitment site, trainee type, age, sex, racial/ethnic minority status, post-graduate training year, the four variables related to trainees’ personal and professional experience with death, and randomization status. Beginning with a model in which the 10 variables other than randomization status were hypothesized as predictors of each of the two pessimism constructs, and randomization status was hypothesized as a predictor of pessimism at follow-up, we removed non-significant predictors sequentially, in order of decreasing P-value, until only the predictors with P<0.05 remained. As with all structural models in the SEM tradition, this model posits a specific hypothesized causal structure and tests how well this model fits the empirical data. Acceptable fit of the model to the data, established with a χ² test of fit having P>0.05, does not prove that the model is the true representation of causal processes in the world, but indicates that it is a plausible representation of such processes.

Results

We approached 1068 eligible trainees, of whom 472 (44%) were randomized. Participation rates were higher for physicians than nurse practitioners (55% vs. 18%; P<0.001). Participation rates were also higher for non-Hispanic whites compared with minorities (61% vs. 52%; P=0.032). This study included the 383 trainees who were randomized, assigned to a workshop, and completed at least one of the surveys. An additional 62 trainees were randomized and assigned to a workshop, but did not complete either survey. The majority of trainees were white, non-Hispanic and located at the UW (Table 1). The sample included slightly more women than men, and most of the trainees were physicians. Average postgraduate training year was 2 (standard deviation [SD] 1.4), with an average trainee age of 30.5 (SD 5.5). At baseline, 6% of trainees agreed and 1% strongly agreed with the statement “It is not possible to tell patients the truth about a terminal prognosis and maintain hope.” Three percent of trainees agreed with this statement at follow-up. Ten percent of trainees agreed and 2% strongly agreed with the statement “Talking about death tends to make patients with terminal illnesses more discouraged” at baseline and 6% agreed and 1% strongly agreed at follow-up (Table 2).

Table 1.

Characteristics of 383 Trainees from the University of Washington and the Medical University of South Carolina

Characteristic	Statistic
University of Washington trainee, n (%)	262 (68)
Female, n (%)	218 (57)
Race, n (%)
White	297 (78)
Asian	62 (16)
African-American	10 (3)
More than one race	9 (2)
Unknown	5 (1)
Ethnicity, n (%)
Non-Hispanic	365 (95)
Hispanic	13 (3)
Unknown	5 (1)
Physician	324 (84.6)
Post-graduate Training Year, mean (SD)	2.0 (1.4)
Age, mean (SD)	30.5 (5.5)

Table 2.

Trainee Responses to Pessimism Indicators

Statement and Responses	Baseline		Follow-up
	n	%	n	%
It is not possible to tell patients the truth about a terminal prognosis and maintain hopea
Strongly disagree	162	47	141	49
Disagree	160	46	135	48
Agree	21	6	9	3
Strongly agree	3	1	0	0
Talking about death tends to make patients with terminal illnesses more discouragedb
Strongly disagree	75	22	78	28
Disagree	228	66	187	66
Agree	34	10	16	6
Strongly agree	7	2	2	1

^aAt baseline 37 trainees had no response, and 98 had no response at follow-up. This includes trainees who did not complete a questionnaire at either time point.

^bAt baseline 39 trainees had no response, and 100 had no response at follow-up. This includes trainees who did not complete a questionnaire at either time point.

Trainee Pessimism Over Time

A structural equation model without exogenous predictors indicated that in the sample as a whole, there was a significant decrease in pessimism between the baseline and follow-up assessments: mean baseline pessimism was constrained to 0.000; intercept at post-intervention=−0.139 (P=0.012). After the variation explained by baseline pessimism was removed, there was still significant variation left in follow-up pessimism (residual variance=0.097, P=0.027), suggesting that factors other than baseline pessimism contributed to trainees’ pessimism at follow-up.

Predictors of Trainee Pessimism

Of the exogenous predictors examined, only two explained significant amounts of variation in pessimism: trainee age and racial/ethnic minority status (Fig. 1). Trainee age had a significant effect on baseline pessimism (older trainees were less pessimistic than younger ones). Racial/ethnic minority status had a close-to-significant effect on baseline pessimism (P=0.058), and an additional significant direct effect on pessimism at follow-up (P=0.028), with minority trainees showing more pessimism than white non-Hispanic trainees at both time points.

Figure 1.

The direct effect of each exogenous predictor represents the portion of the exposure effect not mediated by other predictors, whereas the indirect effect is the portion of the exposure effect that is exerted through other, mediating, variables. The total effect represents both the direct and indirect effects of the exogenous predictors on the outcome. Here, the total effects of both age and race/ethnicity on pessimism at follow-up were statistically significant.

Although the model included no statistical test of whether change from baseline to follow-up differed significantly by race/ethnicity, it provides evidence of differential patterns of change between the two racial/ethnic categories. Estimates computed from coefficients in the model suggest that through about age 35, minority trainees’ pessimism tended to increase between the two assessment points, while white non-Hispanic trainees’ pessimism tended to decrease. Then, by about age 36, both racial/ethnic groups showed decreases in pessimism between baseline and follow-up, but with greater decrease for white non-Hispanics than for minorities (Fig. 2).

Figure 2. Estimated Average Pessimism at Baseline and Follow-up for Selected Age Groups.

Estimates are based on parameter estimates from the structural equation model in Fig 1. Because the model constrained the baseline intercept (representing mean pessimism among non-Hispanics at age zero) to 0.000, and predicted decreased pessimism for each year of age, all estimated pessimism levels at both baseline and follow-up are negative. Shorter bars indicate more pessimism. The selected ages represent the following: 23 (youngest trainee), 29 (sample median), 35 (last age at which estimate for minorities increased between baseline and follow-up), 36 (first age at which estimate for minorities decreased between baseline and follow-up), 63 (oldest trainee). Between the ages of 23 and 35, the estimates suggested baseline-to-follow-up decreases in pessimism among white non-Hispanic trainees and increases among minority trainees. After age 35, estimates for both racial/ethnic groups decreased from baseline to follow-up, but with smaller decreases for minorities than for white, non-Hispanics.

A significant amount of residual variance remained in baseline pessimism after removing the effects of age and race, but residual variance in pessimism at follow-up, unexplained by the joint effects of age, race, and baseline pessimism, was not statistically significant. None of the variables related to trainee personal or professional experience with death explained a significant amount of the remaining variance in either baseline or follow-up pessimism.

Discussion

Using trainee responses on surveys addressing their attitudes about communication with dying patients, we identified a set of indicators of pessimism about end-of-life care. We then used this pessimism construct to assess temporal changes in trainee attitudes about end-of-life care and to examine predictors of trainee pessimism at baseline and follow-up assessments. In the sample as a whole, we found a significant decrease in pessimism over time. However, there were significant effects of trainee age and racial/ethnic minority status on pessimism. The total effect of age on pessimism was statistically significant at both time points, with younger trainees more pessimistic than older ones. The total effect of racial/ethnic minority status was significant only at follow-up, with minority trainees more pessimistic than white non-Hispanics.

For trainees who have yet to accrue significant experience caring for seriously ill patients, attitudes about dying and death are likely to originate from what may be limited exposure to palliative and end-of-life care during medical or nursing school (11, 12). Importantly, these experiences may be influenced by the hidden curriculum, with promotion of negative feelings about the care of dying patients (13). In our study we specifically addressed pessimism about end-of-life care and found an attenuation of pessimism over time as well as less pessimism among older trainees. These findings may reflect the relationship between trainee attitudes and exposure to death, with improvement in attitudes about end-of-life care linked to experiences caring for seriously ill patients during clinical rotations (8). Palliative care experience has been associated with lower death anxiety and higher knowledge about end-of-life care among trainees (14), and practical experience with dying patients can have a lasting effect on trainees by enhancing knowledge about palliative and end-of-life care, influencing assumptions about illness and suffering, and diminishing negative emotional reactions about end-of-life care (8, 15). In our study, we did not identify significant associations between trainee pessimism and self-reported personal or professional experience with death. However, the effects of trainee experiences with dying and death may not be completely captured by the number of encounters a trainee has with seriously ill patients. The quality of the interactions between trainees and patients at the end of life may be more important than the quantity of interactions. Residents who have less than positive experiences during clinical rotations may come away with negative attitudes, whereas those who have positive experiences may find themselves less pessimistic about palliative and end-of-life care. For example, lower levels of empathy have been reported among residents following an inpatient hematology-oncology rotation (16), a setting where dying and death are common. This would suggest that simply placing residents into such environments is not the answer; on the other hand, giving them opportunities to care for seriously ill patients under the guidance of clinicians skilled in palliative care may be beneficial. Palliative care specialists may play an essential role in providing such high-quality experiences, improving trainee understanding of and comfort with palliative care (17).

Although trainee experiences with dying and death seem to promote positive attitudes and diminish pessimism about end-of-life care, the effects of these encounters may not be consistent among all trainees. We found a near-significant effect of minority race/ethnicity on baseline pessimism and significantly higher pessimism among minority trainees at follow-up than was exhibited by white non-Hispanic trainees. Moreover, estimates computed from model coefficients suggested an actual increase in pessimism over time among younger trainees of minority race/ethnicity. By contrast, older trainees in the minority group, like white non-Hispanic trainees of all ages, tended to exhibit decreased pessimism over time, although the average estimated rates of decrease were lower for the minority group than for white non-Hispanics. A potential explanation for racial/ethnic differences in attitudes about end-of-life care may be related to cultural perspectives about dying and death. Patient preferences about end-of-life care differ significantly by race/ethnicity (18–20), and physician preferences tend to mirror those expressed by patients of the same race/ethnicity (21). It is not surprising then, that physician attitudes about communication with seriously ill patients might be influenced by race/ethnicity and the culture in which their attitudes about dying and death developed (22). In our sample, where most of the minority trainees were Asian, specific cultural attitudes about disclosure and information-sharing may offer some insight into our finding of higher levels of pessimism among trainees of minority race/ethnicity.

Our pessimism construct contained two items that focused on discussing death and prognosis with patients, conversations which may not be considered appropriate in non-Western cultures (23). In Asian cultures, family members may be more likely to feel a responsibility to protect the patient from psychological distress or discomfort that might arise from disclosure of information about a terminal diagnosis, and the decision to include the patient in such discussions may be made by the family, not the physician (23). This pattern of disclosure differs from common practice in the U.S., where the desire to protect patient autonomy often prompts full disclosure of diagnosis and prognosis to the patient and family rather than the family alone (24). These cultural differences in disclosure and information-sharing may extend into the communication practices of trainees. A survey of internal medicine residents in Japan and the U.S. found differences in communication strategies when residents were presented with scenarios related to disclosure of a terminal diagnoses (25). Among the residents who supported disclosure to both patient and family, 44% of Japanese residents noted that they would inform the family first, but only 2% of U.S. residents would do the same. Attitudes of Asian-American trainees about communication with seriously ill patients may reflect cultural norms regarding discussions of dying and death, and an attitude that may seem pessimistic may instead be an expression of cultural differences regarding the importance of the family unit in medical decision-making about end-of-life care. Adherence to prevailing attitudes about truth-telling and disclosure may be daunting for trainees whose cultural backgrounds do not support prevailing health care concepts. An improved understanding of the relationship between trainee race/ethnicity and attitudes about end-of-life care is needed to develop appropriate and effective training methods. It is also essential that efforts to enhance clinical education in palliative and end-of-life care incorporate the concept of cultural sensitivity not only for patients and family members, but also for trainees.

This study has several important limitations. First, the survey used to assess trainee attitudes has not been validated. Second, the set of available indicators of pessimism may not have perfectly captured the desired construct. However, using latent variable modeling with the available indicators, we were able to identify a set of indicators of trainee pessimism that exhibited measurement invariance over time and were, therefore, appropriate for assessing change. Third, we had limited information about trainees’ prior experiences with palliative and end-of-life care and all information we obtained was based on self-report. This may have hampered our ability to identify associations between trainee experience and attitudes about end-of-life care. Fourth, not all trainees provided responses for the two items comprising our pessimism construct at baseline or follow-up assessments. However, all data available from each trainee were used to model the constructs. Finally, this study had a limited number of African-American and Hispanic trainees, limiting our ability to assess the attitudes of these underrepresented minority groups.

In conclusion, we found temporal changes in trainee attitudes about communication with dying patients, with a decrease in pessimism over time. We also identified differences in trainee attitudes by race/ethnicity, with pessimism increasing over time for minority trainees. The attenuation of pessimism among the majority of trainees may be related to clinical experiences with dying and death. The important role that positive, instrumental and practical experiences may play in the development of optimistic attitudes towards palliative and end-of-life care should influence the clinical education of trainees. If done well, dedicated time with palliative care providers and participation in the care of seriously ill patients may serve as lasting and influential educational experiences for trainees. Pessimism as captured in our study among minority trainees may be best understood by taking into account the cultural context in which clinicians derive their attitudes and beliefs about communication and decision-making with seriously ill patients. Future efforts to support cultural sensitivity training in end-of-life care should focus both on sensitivity toward the patient and family as well as the perspectives and concerns of the clinician.

APPENDIX I

	Strongly Disagree	Disagree	Agree	Strongly Agree
1. There is little that can be done to ease the suffering of grief	1	2	3	4
2. The physician/nurse practitioner has a responsibility to provide bereavement care to the patient’s family after death	1	2	3	4
3. Physicians have an obligation to tell patients when death is imminent	1	2	3	4
4. It is not possible to tell patients the truth about a terminal prognosis and maintain hope	1	2	3	4
5. Psychological suffering can be as severe as physical suffering	1	2	3	4
6. At their request, patients with terminal illnesses should be given whatever medications are necessary to relieve pain, even if the medications hasten death	1	2	3	4
7. Depression is normal in patients with terminal illness	1	2	3	4
8. Talking about death tends to make patients with terminal illnesses more discouraged	1	2	3	4
9. Depression is not treatable in patients with terminal illnesses	1	2	3	4
10. Family members tend to interfere in the care of patients with terminal illnesses	1	2	3	4
11. The physician’s/nurse practitioner’s responsibility is to the patient; other professionals should deal with the needs of the family	1	2	3	4
12. Physicians/nurse practitioners have a responsibility to help patients prepare for death	1	2	3	4
13. An interdisciplinary team approach to terminal illness treats patients’ medical needs better than conventional care	1	2	3	4
	Never	Rarely	Sometimes	Often	Always
14. Caring for patients who are dying is depressing	1	2	3	4	5
15. I feel guilty after a death	1	2	3	4	5
16. When a patient dies, I feel as if I have failed as a physician/nurse practitioner	1	2	3	4	5
17. I dread having to deal with the emotional distress of family members	1	2	3	4	5
18. When one of my patient dies, I grieve for the patient	1	2	3	4	5