Initial self-reported data on sleep and burnout in pulmonary, critical care and sleep medicine: an initiative from the Assembly on Sleep and Respiratory Neurobiology of the American Thoracic Society

Abstract

Rationale

Health worker burnout has reached crisis proportions, threatening workforce sustainability. Limited data exist on the burden of burnout in pulmonary, critical care and sleep medicine (PCCSM), a high-demand and strained specialty.

Objective

At the Assembly of Sleep and Respiratory Neurobiology of the American Thoracic Society, we aimed to gather exploratory data on burnout in this group.

Methods

During a dedicated series of five virtual town halls (THs), we polled the audience regarding self-reported burnout. Topics included the scope of the problem, role of sleep, impact on clinical and academic operations, contributors in vulnerable groups, and mitigation strategies.

Results

A high proportion experienced burnout (45%) and 58% considered premature retirement. Insufficient sleep (53%) was common, most often due to excessive workload (57%) curtailing sleep through early morning meetings and electronic medical record (EMR) documentation. 36% also reported having a sleep disorder. Sleepiness (69%) and fatigue (58%) impaired work performance and patient care, and 54% reported a fatigue-related, personal-safety incident. Contributors to burnout in vulnerable communities included bias/discrimination (81%), harassment (44%) and assault (12%). Respondents predominantly endorsed organizational mitigating strategies: promoting a culture of “recovery time” (96%) and healthy sleep (86%), and periodic evaluation and accountability of leadership (86%).

Conclusions

In this convenience sample of participants in a TH series regarding burnout in PCCSM, self-reported burnout was common. Sleep disturbance is a prevalent, under-recognized, but potentially modifiable contributor. The high reported rates of discrimination and harassment suggest that vulnerable groups may be at particular risk. To reduce burnout, system-level interventions aimed at transforming organizational culture and promoting leadership accountability were strongly endorsed.

Introduction

Burnout encompasses three domains: emotional exhaustion, depersonalization or cynicism, and lack of personal fulfillment [1]. Health worker burnout has reached crisis proportions, with a steep rise in its prevalence from 45–55% to 61% [2] precipitated by the COVID-19 pandemic [3]. Importantly, health worker burnout has been associated with patient safety incidents [4] and premature exit from the workforce [5], exacerbating already-existing strains on remaining health workers. The U.S. Surgeon General recently designated health worker burnout as a public health threat [6, 7] and the National Academy of Medicine (NAM) declared identifying and tackling the key drivers of burnout among U.S. health workers a top priority [8]. Both entities emphasized the urgent need for effective, system-based interventions. Most recently, a Senate resolution passed unanimously to expand national awareness of health worker burnout, and to support the well-being of health workers [9].

Pulmonary, critical care and sleep medicine (PCCSM) workers face high work demands, thus potential for high risk of burnout [10, 11]. Indeed, burnout within this group increased during versus prior to the COVID-19 pandemic (68% vs. 42%) [12]. In addition, a recent American Academy of Sleep Medicine (AASM) statement posits that disturbed sleep and sleep disorders may worsen burnout among health workers, based on data from physicians-in-training and nurses [13], and called for an acceleration of research on this topic. Moreover, a consensus paper from 17 professional societies, including those in PCCSM, emphasized that societies must play a leading role in addressing burnout and support wellbeing, by providing education and awareness, and advocating for mitigation strategies [11]. Data regarding the burden, contributors, and impact of burnout specifically in PCCSM specialties, however, remain sparse. Moreover, the key role of sleep health as a potential mediator of burnout remains unknown, despite the high prevalence of sleep deprivation, long work hours, and high occupational, environmental, and personal demands which compete with sleep opportunity.

Recognizing the escalating crisis of burnout and the pressing need for solutions, the Assembly on Sleep and Respiratory Neurobiology of the American Thoracic Society (ATS-SRN) undertook a grassroots effort by organizing a series of town halls (THs), to explore: (1) the scope of the problem; (2) the role of sleep; (3) the impact on clinical, research and educational activities; (4) burden in vulnerable communities, and; (5) potential mitigation strategies. Each town hall encompassed a live audience poll, the results of which we summarize in this report. An initial report of this work was published in abstract form [14].

Methods

Town hall format and participant engagement

The virtual TH series included five interactive sessions. Table 1 shows their titles and a priori established educational objectives. The series was hosted by ATS staff via Zoom (Zoom Video Communications, Inc., San Jose, California) between April 2022 and March 2023. ATS staff tallied the number of attendees at the start of each TH and poll. The University of Pennsylvania Institutional Review Board approved the study, and waived documentation of consent (protocol # 854649; approval date October 25, 2023).

Table 1 Topics of the five town hall sessions and their educational objectives

Each of the five successive THs addressed a key topic: (1) scope and contributors, (2) the role of sleep, (3) impact on research and clinical operations, (4) contributors and impact in vulnerable communities, and (5) the role of organizational structures and responses in mitigating this epidemic. Each session started with brief presentations by 2–3 evidence experts that began by defining burnout [1], before further addressing the specific objectives of the session. An interactive, live audience poll followed. Subsequently, the hosts, speakers, attendees, and an invited panel of 6–12 experts representing multiple stakeholders (government agencies, health care systems, professional societies, educational/licensing boards) engaged in a discussion. Panelists were invited to attend more than one TH and had the option to answer poll questions or not, as they wished.

Invitations to the THs were announced in the weekly ATS bulletin for several weeks before each event date. Membership in the ATS was not required to register or participate. The announcements were emailed by the ATS office to their confidential ATS membership list, along with a registration link. Following registration, interested members were emailed a unique link to connect to the virtual session. There were no additional restrictions on registration or participation. Responses were not tracked at the individual level, to maintain privacy and confidentiality.

Poll content

The poll questions were created by consensus among the organizers (IG, SP, MT), to align with the topic of each TH session, and were limited to 7–11 questions per TH, with additional questions on demographics (age, gender, and self-identified race). To enhance content validity, questions and response choices of each TH were thoughtfully designed after careful consideration of published data on the topic. Response options to the questions were either multiple choice (with single or multiple choices allowed), ranking of multiple choices, or binary (yes/no). ATS staff administered polls using the built-in Zoom tool. Each attendee could answer each question only once. Participants could choose to respond to some or all of the poll questions. ATS staff collected the responses, collated them, and submitted de-identified data to the co-organizers for analysis. Apart from age, gender and race, all participant details (including names, professional title) were anonymized, so responses within or between sessions could not be linked to any participant.

Data analysis

Data were analyzed using JMP Pro 16^® (SAS Institute, Cary, NC) statistical software. Descriptive analyses were conducted, but inferential statistics were not performed due to the small sample sizes. For each TH, to account for attendees who exited or joined the TH prior to the start of poll, we tallied the number of attendees at the beginning of the TH session and separately, at the start of the survey. For each TH, we determined the total number of participants who responded to each question, then averaged them across all questions in the session to obtain the number of survey respondents for that TH. The overall response rate per each TH was calculated as the mean number of respondents to the poll divided by either the number of attendees at the beginning of TH session or at the start of the survey. We summarized data with means and standard deviations (±SD) or number (proportion of selected choices relative to the number of respondents to each question), as applicable. Each question along with all of its choice of response options offered are shown in the Tables and Figures.

Results

Response rate and sample characteristics

Table 2 presents attendee numbers and response rates, at the beginning of each TH and at the start of each poll, first averaged across all THs, then by each individual TH.

Table 2 Response rates and respondent characteristics across all five townhalls and for each individual townhall

Across all THs, the numbers of attendees at the beginning of each TH ranged from 23 to 45 and at the start of the survey poll ranged from 23 to 40. The mean number of respondents to poll questions ranged from 12 to 31, yielding a mean (±SD) response rate across THs of  65(±17)% when based on the number of attendees at the beginning of the session and of  71(±11)% when based on the number of attendees at the start of the poll.

More women than men completed the poll in TH2 to TH5 (54 to 68% women vs. 32 to 46% men), whereas slightly more men than women competed the poll for TH1 (53 men vs. 47% women). Across all THs, the reported race of respondents was 65% White, 22% Asian, 5% Black, 3% chose not to respond and the remainder were of other racial backgrounds.

Overall burden and impact of burnout on worker retention

Almost half of respondents (45%) self-reported experiencing burnout (Fig. 1; Panel A). 43% of the sample reported ‘probably or definitely’ with an additional 15% contemplating (‘not sure’) (58% total) for leaving the profession before planned retirement because of burnout (Fig. 1; Panel B). Moreover, 57% reported that work obligations often/always conflicted with personal responsibilities, such as childcare, eldercare, or household/personal care activities (Fig. 1, Panel C).

Fig. 1

Burden of Burnout and Impact on PCCSM Workforce Retention. Attendees were asked to choose only one among the options presented for each question: (A) experiencing burnout; (B) leaving the profession early (represents combined results from polls asking the same question in two different townhalls: TH3 and TH4); (C) work obligations conflicting with personal responsibilities. The question in Panel B was asked in both Town Hall 3 and 4 and therefore, the shown data reflect the combined results of the two questions. Number of participants who responded and % of responses relative to this number are shown for each question

Contributors to burnout

The factors identified as most strongly contributing to burnout were workload (76%) followed by unreimbursed activity (14%), rather than illness in self or family members (7%) or mental health (3%) (Table 3).

Table 3 Contributors to burnout among survey respondents

The most frequently cited contributors to burnout in clinical care included: (1) administrative staff exited, quit or were out sick (74%); (2) inefficient workflows for in-person patient care (74%), and (3) electronic medical record (EMR) inefficiencies (70%) (Table 3). The most often reported contributors in research operations were competing clinical demands (81%) and insufficient protected time (72%) (Table 3).

Sleep disturbances and specific impacts on workplace performance and personal safety

Figure 2, Panel A shows that only 46% of respondents endorsed sufficient sleep opportunity often or always, whereas only 38% ‘sometimes’ and 15% ‘rarely or never’. Figure 2, Panel B shows that excessive workload and family obligations (57% each) were the most frequently cited reasons for insufficient sleep opportunity, with high stress (50%) and misaligned work schedules (36%) also common, while personal physical and mental health were least-often reported (21%). Early morning meetings (38%), EMR documentation and research obligations were each (23%) identified as the most important factors contributing to workload-related impacts on sleep (Fig. 2, Panel C). 36% reported that they suspected or knew they had a sleep disorder with an additional 10% being unsure (Panel D).

Fig. 2

Sleep Disturbances and Determinants. The role of sleep in self-reported burnout: (A) sleep opportunity; (B) factors impacting getting sufficient sleep; (C) work-related factors impacting sleep; (D) suspected sleep disorders. Questions in Panel A, C and D permitted only 1 response selection per question and number of participants who responded and % of responses relative to this number are shown, for each question. For the question in Panel B, attendees were permitted to choose as many responses as needed; responses are displayed in order of most to least frequently selected; number of participants who responded to each question and % of responses per each response choice in that question, relative to this number are shown

Abbreviations: EMR-electronic medical record

Regarding work and personal safety, over two thirds of respondents (69%) endorsed adverse effects of sleepiness on performance at work ‘sometimes’ (Fig. 3, Panel A) and none ‘often or always’. 58% reported that fatigue played a role in their ability to provide best patient care a few times a month (Fig. 3, Panel B) and none ‘all the time’. Additionally, 39% responded ‘yes’ and another 15% reported ‘maybe’ (54% total) (Fig. 3, Panel C) to having experienced a fatigue-related personal safety incident since completing training, such as needle-stick injury and error or driving-related accident.

Fig. 3

Impact of Sleepiness and Fatigue on Performance and Safety. Self-reported impact of (A) sleepiness and (B) fatigue on work, and (C) of fatigue on personal safety. Each question in Panel A-C allowed only one response among the multiple choices offered. Number of participants who responded and % of responses per question relative to this number are shown

Bias, harassment and diversity in the workplace

Most individuals reported having faced challenges with discrimination related to unequal task assignments (91%) or salary, employment opportunities and promotion (82%) (Fig. 4, Panel A). Personal incidents related to discrimination or harassment were also common: bias and discrimination at the workplace (for gender, race, sexual orientation, religion, disability, or other reason) (81%), and harassment and hostility from patients (56%) and from superiors (44%). 31% experienced unwanted sexual conduct, 25% experienced harassment and hostility from peers and 12% reported having been assaulted.

Fig. 4

Burnout in Vulnerable Communities. Data on self-reported challenges: (A) related to work-related bias; (B) discrimination and harassment in the workplace; (C) relevant mitigating strategies offered by institutions. Multiple response choices per question were permitted for all questions. Number of participants who responded to each question and % of responses per each response choice in that question, relative to this number are shown

Among the numerous response options offered to address diversity, equity, and inclusion (DEI) issues in the workplace (Fig. 4, Panel C), the most often reported were through paid external consultants or institutional DEI leader (57%), meditation and yoga, and volunteer, unpaid groups addressing DEI issues (50% each). There were no reports of increased administrative support implemented at the workplace, to address burnout.

Effectiveness of current mitigation measures

As shown in Table 4, respondents identified employers (72%) as the most important entity in offering interventions to address burnout, much more often than payers (12%), government agencies (6%), licensing boards (3%), professional societies (3%) or research funding agencies (3%). While 62% reported that their institutions were addressing burnout, a vast majority (82%) felt that the institutional response was insufficient. Most institutional countermeasures targeted individual (48%) rather than institutional (13%) factors. Individual countermeasures were perceived as insufficient (67%) to address the crisis. Only 35% reported that their institution was targeting both individual and institutional measures.

Table 4 Effectiveness of current mitigating measures among TH5 survey respondents

Respondents’ endorsed solutions and mitigation strategies for burnout

As shown in Fig. 5, Panel A the most important organizational mitigating strategies identified focused on: periodic evaluation and accountability of leadership across all levels (86%), promoting a culture of ‘recovery time’ by reducing the need for after-hours work (96%), mental health (by destigmatizing care, increasing access to care, and protecting confidentiality during licensure and accreditation/credentialing processes) (85%), and a culture of healthy sleep through education about sleep, destigmatizing the need for sleep (86%) and implementation of rest and naps opportunities (78%).

Fig. 5

Strategies to Mitigate Health Worker Burnout. Self-endorsed mitigating strategies for burnout, addressing: (A) organizational leadership and culture; (B) clinical operations; (C) research, education and regulatory requirements. Participants were permitted to select multiple choices for each of the questions in this Figure. Responses are displayed in order of most to least frequently selected; numbers of participants who responded to each question and % of responses per each response choice in that question, relative to this number are shown

In clinical operations, the following measures were identified as strategically important: optimizing workload distribution by having all workers perform at the top of their skill level (89%), optimizing EMR for efficient use (87%) (Fig. 5, Panel B), and limiting the amount of documentation required for regulatory compliance (billing, coding, risk management) (68%) (Fig. 5, Panel C). In academic, education and research-related environments, offering realistic protected time (88%) and reforming salary structure (83%) were identified of top importance.

Discussion

Our grassroots, exploratory poll generated key considerations for the PCCSM workforce. Firstly, 45% of the TH respondents self-reported experiencing burnout, a rate that matches the crisis proportions cited in multiple systematic surveys. Second, burnout may also drive a substantial proportion of participants to leave the profession prematurely (58%), straining an already-understaffed workforce. Third, insufficient sleep and sleep disorders may be novel, under-recognized contributors, and may affect both patient and health worker safety. Fourth, previously-identified risk factors for burnout that concerned EMR inefficiencies [15] and the lack of administrative support [11, 16] continued to be experienced by a high proportion in this group. Finally, the most-endorsed mitigation strategies concerned system-based solutions by employers focused on novel intervention targets, such as promoting a culture of recovery time, optimizing sleep, and ensuring leadership accountability. Such measures would be critically important, as they represent deviations from currently offered strategies, which predominantly target individual-level factors, such as meditation and yoga that were also uniformly perceived to be insufficient.

Our polls’ target audience included workers in PCCSM, who face high work demands, and bore considerable front-line responsibilities during the COVID-19 pandemic [12]. A pre-pandemic cross-sectional, mixed-methods survey study in a convenience sample of ATS 2018 conference attendees, reported a lower prevalence (38%) of respondents who met burnout criteria, while workload and the burden of EMR documentation were also the most frequent contributors [10]. In particular, issues with administration and management, loss of autonomy and workplace-related conflicts were also important contributors identified in this earlier study [10]. Amid the acute shortage of health workers following the COVID-19 pandemic [7], our results generate questions about worker retention, due to potential premature exit from the profession endorsed by 43% of respondents, with an additional 14% contemplating it. This proportion (57% total) is larger than that of a 2022 poll in nurses, in which 49% responded “maybe or yes” when asked if they intended to leave direct patient care in the next 6 months due to burnout [17]. If this finding is confirmed in larger surveys, additional strain may be looming for this already limited but essential specialty workforce.

Our work not only aligns with previously identified, potential targets for intervention [18], but also suggests novel ones, such as sleep health, which are currently not addressed after training [19]. Sleep deprivation and maladaptive behaviors around sleep were identified as often overlooked keycontributors to burnout [13]. Also, our findings resemble data reported by faculty members in a teaching hospital [20], who reported a similarly high prevalence (29%) of undiagnosed sleep disorders, such as insomnia, obstructive sleep apnea, insufficient sleep and circadian rhythm misalignment, and restless legs syndrome, thus suggesting a role of screening health workers for sleep disorders. No biological basis exists to suggest that sleep need and the necessity to safeguard sleep health decline after trainees become practitioners, especially as work demands and stress may grow considerably when reaching that level; additionally, that sleep disturbances increasingly acquired during training [21] likely do not cease upon entering the attending workforce, and may have long-lasting consequences. Changing the organizational culture regarding sleep health by providing rest breaks, systemic education on healthy sleep habits, reducing early-morning meetings and overall, promoting a culture of ‘recovery time’ outside of the work were identified as key interventions. Aligning work schedules to circadian patterns by offering autonomy in scheduling [22] may yet be another important intervention. These measures may indeed be critical, to address the reduced quality in patient care that respondents attributed to sleepiness (69%) or fatigue (58%), and since 39% experienced personal safety incidents due to fatigue. Our data mirror past literature in medical residents, in whom needle stick injuries (43%) and drowsy-driving accidents (46%) were associated with overnight- and extended-duration shifts [23,24,25,26]. While mental health and burnout may be reciprocally linked, the fact that majority of participants (85%) endorsed mental health as a target for interventionconveys concern about the potential increased vulnerability to developing or perpetuating burnout among those with mental health issues. This aligns with the declaration of the World Health Organization that burnout is an occupational phenomenon rather than a medical condition [27], assigning it a specific International Classification of Diseases-11 diagnostic code distinct from depression and other mental health conditions, within its 2019 ICD-11 release [28]. This implies that workers suffering from burnout require different interventions than psychotropic medications and psychotherapy. Moreover, in the recently released report, the National Academy of Sciences prioritized mental health as a key intervention in curbing burnout and promoting well-being of health workers [18].

Among vulnerable communities, in descending order, a high prevalence of bias/discrimination, harassment by superiors, unwanted sexual conduct, harassment/hostility from peers, and assault were perceived to impact burnout. We could not link these reports to the race or gender of respondents, since responses were anonymous. However, the assertion that higher rates of burnout may be experienced by vulnerable groups is supported by recent documentation of gender disparities in PCCSM [29]. Moreover, prior data indicate that race/ethnicity may magnify PCCSM disparities, aside from gender and other factors such as sexual orientation and disabilities [29].

Most respondents emphasized the value of systemic over individual interventions, highlighting the importance of periodic evaluation and accountability of leadership. These respondents’ needs are echoed by measures proposed by national organizations [30, 31], which furthermore offer practical implementation toolkits for organizations to use [32]. Moreover, a taskforce of professional societies with growing awareness of the burden of burnout in their fields, similarly, highlighted the importance of addressing health system factors and recognized the critical role they themselves could play in advocacy, to mitigate contributing factors and promote their members’ wellbeing [11]. Specifically, they could encourage prioritizing burnout initiatives through collaboration, education, and promoting future research, while supporting both organizational and individual mitigation strategies [11]. Last, some have suggested that leadership term limits may be an underutilized method of boosting accountability and driving a transformational, adaptable and diverse leadership culture, which may improve quality of patient care and fiscal margins [33]. The National Institutes of Health and other institutions have adapted and enacted this strategy [34].

Strengths of our work include a structured format with expert presentations to inform attendees of definitions and scientific evidence, before requesting responses to poll questions. Our work was timely, coinciding with the publication of two major reports by the National Academy of Medicine and the U.S. Surgeon General that endorse some of the same strategies chosen by our respondents [6, 8]. Polls included broad content: prevalence, contributors, personal and organizational impacts, and intervention targets, which included unaddressed topics, such as sleep health and experiences of vulnerable groups. The high proportion of positive responses across THs serves as an indicator that the participants found the content matter relevant to their individual experiences and is unlikely participants could have influenced each other in their responses, since each virtually connected to the sessions via unique links. While some of our findings align with prior data, novel knowledge on the role of sleep, impact on vulnerable groups and mitigating strategies emerged, to help guide targeted interventions.

While our polls drew reasonably high response rates (71%) and the estimates were similar to prior reports, our study has several limitations which limit generalizability to the larger PCCSM workforce. The small sample size (ranging from 12 to 31 survey respondents/ TH) is a significant limitation, yielding reduced power to detect meaningful differences and increase in the margin of error around estimates. Second, the additional requirement of protecting anonymity of respondents precluded linking individual characteristics with their responses, thereby, subgroup analyses (by demographics, professional role and other covariates) and a deeper understanding of impact on vulnerable groups. Additionally, since participants could attend more than one TH, we could not gauge the number of unique participants across all TH. , Third our poll questions were not previously standardized or validated to assess burnout, highlighting that these are preliminary results, primarily intended for hypothesis generation. Fourth, while burnout was defined in the expert lectures preceding the polls, participants were asked to self-report it, which may reflect burnout-like symptoms rather than an objective diagnosis. Finally, due to the lack of defined eligibility criteria, this is an exploratory study in a convenience sample, and therefore open to participation bias resulting from their own experiences with burnout or other relevant characteristics. For example, workers with high levels of burnout may have been either more interested to participate and learn more about the topic, while others may have chosen not to attend due to cynicism and loss of hope in addressing this issue. This vulnerability to selection bias also limits the generalizability of these exploratory data to the broader PCCSM community. Thus, our findings in this small convenience sample, merely begin to, but do not fully address the gap in data on this topic. However, the results are provocative enough to provide some initial perspective to stakeholders, and prompt larger studies with well-characterized populations, including professional role (physician, nurse practitioner, physician assistant, etc.) and career stage, and utilizing objective physiologic measures of stress, and assessments of sleep, sleepiness, vigilance and performance.

In summary, this exploratory, hypothesis-generating initiative on burnout leveraged a targeted, convenience sample attending a series of THs in PCCSM, a specialty facing high workloads and sustainability concerns. Given high rates of response positivity at each TH, the poll questions tested here deserve further evaluation in intervention studies among larger and more diverse groups, which are representative of the PCCSM workforce. Additionally, future studies should consider evaluating burnout as it pertains to profession and career stage. As other national organizations have reiterated [6,7,8], to ‘turn the corner’ and retain and recruit talent in this specialized workforce, the time to act is now. Our exploratory poll data also indicate that specific, system-level interventions , such as addressing sleep health, increasing leadership accountability, enhancing diversity and inclusion should be considered. In the best interest of our patients and heavily taxed workforce, follow-up action in response to these exploratory and intriguing results is not optional, but imperative.