Burnout and resilience among Canadian palliative care physicians

Physicians and resident physicians who were active members of Canadian Society of Palliative Care Physicians (CSPCP) or Société Québécoise des Médecins de Soins Palliatifs (SQMDSP) as of March 18, 2019 were invited via email to participate. Members who were non-physicians, retired physicians, and physicians working in countries outside of Canada, and physicians and resident physicians from CSPCP who had not authorized research participation requests were excluded from the invitation email. Details of the survey’s sampling frame are presented in a supplementary figure (Additional file 1), which presents the derivation of the total of 569 physicians from the respective societies’ membership lists to whom invitation emails were sent: 479 members in the CSPCP and 90 members in the SQMDSP.

Over the course of 5 weeks (between March 18th, 2019 and April 23rd, 2019), administrators of CSPCP and SQMDSP distributed an initial invitation email (followed by two reminders), with embedded and secure SurveyMonkey® electronic links to both English and French versions of the survey.

The survey, inclusive of questionnaires, consisted of 57 questions in total, and of these a set of 10 questions concerned participant demographics: clinical practice, clinical settings, training and remuneration arrangements. These demographic questions were developed by the research team in English and translated to French by a certified translator. The last question of the survey, ‘Thank you for your participation. Leave a comment if you wish”, provided participants with the option of inserting informal free text comment.

Burnout was measured using the Maslach Burnout Inventory-Human Services Survey for Medical Professionals [MBI-HSS (MP)], a validated 22-item questionnaire considered to be the gold standard tool for measuring professional burnout [3]. The questionnaire has 3 subscales to evaluate each domain of burnout, including emotional exhaustion (EE; 9 items), depersonalization (DP; 5 items), and personal accomplishment (PA; 8 items). Each item is rated using a seven-point Likert scale grading the extent each item is experienced by the respondent, ranging from never (0) to everyday (6). Additive scores from items under each subscale represent the total score for that subscale and are used in determining the level of risk for burnout. Based on normative data in medical professionals, EE was scored as high for total score ≥ 27; DP scored as high for total score ≥ 10; PA scored as low for a total score ≤ 33 [2]. High burnout was defined as having a high score on EE (≥27) and/or DP (≥10) subscale, in line with other studies measuring burnout in medical professionals [28, 29, 38, 41]. Resilience was measured using the 25-item Connor-Davidson Resilience Scale (CD-RISC) [31], a widely accepted measure of resilience [42]. Respondents rate their level of agreement with each scale item on a five-point Likert scale ranging from not true at all (0) to true nearly all the time (4). Possible scores range from 0 to 100, with higher total scores reflecting greater resilience. English and French-Canadian versions of the MBI-HSS (MP) and CD-RISC tools were used.

All open text comments were carefully reviewed and then imported into NVivo 12, a qualitative data analysis software program [43]. Drawing on an inductive approach, wherein each participants’ comments were reviewed as a specific unit of text [44], two authors, one a qualitatively trained researcher, coded the open-text into initial codes, capturing the content, meaning and context of each participant response [45, 46]. after moving between and across initial codes and exploring patterns and responses across participants, a more situated, and nuanced thematic coding process was developed [44‐47]. Moving between and across initial codes, and focusing on the overall study’s research question in exploring the relationship between burnout, resilience, demographic and job factors, we began to explore patterns across the initial codes, thereby developing thematic codes which offered further insights, understanding and context into the quantitative data collected. In dialogue with the larger research team, our open-text thematic analysis was then situated in conversation with the emerging conceptual understanding of how physicians across disciplines experience burnout, resilience and wellness [48‐52].

Demographic and job characteristics of respondents are shown in Table 1. Almost half (49%) of the respondents were aged > 50 years and over two thirds (70%) of respondents were female. A majority of respondents (53%) worked 41–60 h per week. Most respondents worked predominantly (> 50% of clinical time) in palliative care (79%), and in a variety of settings. Approximately half of the respondents (51%) spent ≥20% of work time on non-clinical activities. Most respondents trained in family medicine/ general practice (68%). Just under half of the respondents (47%) had confirmed additional formal training in palliative care. The most frequent remuneration arrangement was a clinical or mixed clinical-academic salaried position in 44% of respondents.

The Maslach Burnout Inventory-Human Services Survey for Medical Professionals MBI-HSS (MP) includes 3 subscales, emotional exhaustion (EE), depersonalization (DP), and personal accomplishment (PA).

Table 2 shows respondents who scored low, moderate or high on EE, DP and PA based on tertiles derived from normative data. Overall, 36.4% of respondents had high EE, 15.1% had high DP, and 7.9% had low PA. We found that 38.2% of respondents had one manifestation of high burnout based on having high EE and/or high DP.

The Connor-Davidson Resilience Scale (CD-RISC).

Table 2 shows CD-RISC scores stratified into quartiles. Median CD-RISC score of respondents was 74, which was marginally above the 25th percentile of a US normative population [53].

Table 3 shows bivariable associations between demographic and job variables listed and manifestation of burnout. Older physicians (> 60 years of age) were at much lower risk for burnout as compared to younger physicians (p = 0.0003). Similarly, there was a trend toward decreased burnout in increased years worked (> 30 years) and increased length of time spent in palliative care (> 20 years) with no statistical significance. There was a trend toward decreased burnout with working ≤40 h per week. Burnout was not significantly different for those whose clinical practice was predominantly in palliative care, nor for those who spend ≥20% of work time on non-clinical activities. Prevalence of burnout was higher for those on salary as compared to fee for service (FFS)/combination of salary and FFS (p = 0.04). There was no gender difference in burnout rate.

Table 4 shows bivariable associations between demographic and job variables listed and mean CD-RISC score. Older physicians (> 60 years of age) reported significantly higher resilience than younger physicians (p = 0.03). There was no gender difference in reported resilience. While not statistically significant, there was a trend toward increased level of resilience with years in practice, increased hours worked per week and having a clinical practice predominantly focussed on palliative care.

The multivariable logistic regression analysis examined the level of independent association of demographic and workplace variables with burnout (Table 5). Of the included variables, only age > 60, (OR = 0.05; 95% CI, 0.01–0.38), compared to age ≤ 40, had an independent association with lower burnout. Multivariable linear regression was conducted to examine differences in mean CD-RISC resilience scores in association with the same demographic and workplace categorical variables (Table 6). Of the included variables, there were increased CD-RISC scores of 7.77 (95% CI, 1.97–13.57), 5.54 (CI, 0.81–10.28), and 8.26 (CI, 1.96–14.57) in association with age > 60, a predominantly palliative care clinical practice (> 50% of clinical time), > 60 h worked per week as compared to ≤40 h worked, respectively. In an alternate model that included years in practice instead of age, there was a trend [Mean CD-RISC score difference of 4.59 (CI, − 0.96-10.14)] toward an increase in resilience level with > 30 years in practice as compared to < 10 years (Additional file 2).

Twenty-five (15.15%) participants included additional comments at the end of the survey. Comments ranged from a few words, to a few sentences, to paragraphs and longer narratives. Comments were carefully reviewed and coded into initial codes and then thematic codes [44‐47]. Thematic codes offered important insights on: the unique “double-edged sword” context of palliative care provision- meaningful, sustaining and impactful work with patients while also needing to advocate and fight for administrative, organizational, and political support to be able to provide optimal treatments, services and supports. A number of participants noted their exhaustion in having to continually push against myths and reluctance that other care givers/providers have about the value and importance of palliative care. Lastly, some participants shared their own personal and professional experiences with burnout, the support that was missing, and the lessons they had learned in building resilience and taking care of themselves. Comments about the concepts, instruments and application to palliative care were also carefully reviewed (Table 7).

One limitation of this study is the response rate (29%). However, our response rate is higher than in comparison with other national surveys of burnout across all physician groups (17.1% in the United States [10] and 8.5% in Canada [36]). It is also possible that some palliative care physicians are non-members of either society and therefore would not have been invited to participate, introducing sampling bias. Furthermore, we did not have access to the demographics of either membership rosters for comparison with our study sample, further limiting the study’s generalizability. The cross-sectional study design precludes a determination of causality or direction of effect for the associations observed. We were also unable to capture variability in experienced burnout over time, including past experiences with burnout. Variations in MBI-HSS (MP) cut-off scores used to define burnout is another limitation in this area of research: our study results were only comparable to other studies with similar MBI-HSS (MP) cut-offs. A systematic review on the prevalence of burnout among physicians found that of studies using MBI-HSS (MP), there were at least 47 distinct definitions of burnout, highlighting the importance of developing a consensus definition of burnout [67]. We recognize the limitations that exist around how and in what way open text comments can be analysed, particularly when these were not framed a priori, specific questions were not asked directly, and open-text items in surveys have both explicit and implicit constraints in really understanding participants’ expereinces [62, 63]. Despite these cautions, participants who did provide open-text responses offered very important, rich and fulsome accounts of their perceptions and experiences, and we believe these reflection offer important insights which offer further support to our quantitative findings, while also offering direction for future research areas to explore.