Introduction
Crises such as the COVID-19 pandemic can have widespread impacts on the healthcare system, resulting in organizational changes that can negatively impact healthcare worker burnout and turnover. During previous crises, like the 2005 Kashmir earthquake [
1], the 2011 Fukushima earthquake, tsunami, and nuclear incident [
2], and 2017 Harvey and Maria hurricanes [
3], burnout and other mental health issues increased among healthcare, disaster, and related workers. Healthcare worker (HCW; i.e., providers and staff) burnout, primarily characterized by emotional exhaustion and depersonalization, [
4] has been widespread across many specialties, in the US [
5,
6] and globally [
7‐
9], during the COVID-19 pandemic. Turnover intent, or the intent to leave one’s job, among HCWs was also high in the US [
10] and internationally [
11,
12] during the pandemic. Prior to the pandemic, the burnout phenomenon [
13] was associated with both individual and organizational drivers [
14], and linked to numerous negative healthcare consequences [
14‐
16], including increased medical errors, reduced patient satisfaction, poorer quality of care, and increased turnover
. Similarly, previous research has identified turnover as a possible driver of primary care shortages, [
17,
18] and found that it is very expensive, with the estimated costs of replacing a single HCW ranging from $14,000 for a medical assistant [
19] to $1,000,000 for a physician [
20].
The COVID-19 pandemic tested the capacity of healthcare systems to respond to a crisis, requiring healthcare system administrators to adapt organizational structures and processes overnight and changing organizational contexts in ways that may have contributed to higher levels of burnout and turnover among primary care healthcare workers (HCWs). Little is known, however, about how COVID crisis-induced changes may have impacted primary care HCW burnout. Crisis levels of acute and emergent care often required surge staffing, [
21] which may have included temporarily reassigning primary care HCWs into COVID-related screening or inpatient care. These staffing changes may have contributed to increased burnout among the reassigned HCWs exposed to new stressors and among the remaining HCWs with increased workloads. Those HCWs with higher workloads may have experienced increased burnout, as has been documented in research prior to the pandemic [
14]. In addition, the rapid shift towards virtual care (e.g., video- or telephone-based care), [
22] due to COVID-related quarantining and social distancing restrictions, may have increased primary care HCW burnout via changes in work flows, workload, learning how to use the virtual care platforms, and/or educating patients on how to use the modality [
10,
23]. Conversely, virtual care use may have reduced HCW stress and associated burnout by reducing the risk of COVID-19 infection.
Some characteristics of individual HCWs have been shown to be protective against burnout and turnover. Specifically, more engaged employees may be more resilient and less susceptible to burnout. Engagement is characterized by energy, involvement, and effectiveness on the job [
24]. A systematic review and meta-analysis [
25] of burnout and engagement in 37 studies across job fields showed that higher engagement and lower burnout were consistently correlated with fewer health complaints, more job satisfaction, and more organizational commitment. To date, employee engagement has not been well-studied in primary care. Protective factors, like employee engagement and positive working environments (e.g., a workplace with a sense of community), may attenuate increases in HCW burnout [
5] and turnover intentions [
26], but it is unknown whether engagement can buffer the impact of crisis-induced organizational changes on burnout and turnover.
Good leadership and strong workgroups may have also been protective of burnout among individual HCWs during the pandemic. High quality leadership, characterized by values like trust, respect, mentorship, and inspiration, has been consistently associated [
14,
27] with lower burnout among physicians and other HCWs. Coworkers also matter – evidence suggests that in VA primary care, workgroups with good collaboration and communication and high competency [
28‐
30] have members that are less likely to be burned out. Strong leadership and good relationships with coworkers were likely very important during the chaotic times of the early pandemic, and these characteristics may have protected against higher HCW burnout.
Our objectives in this study were to 1) examine changes in reported VA primary care HCW burnout and turnover intent prior to and during the early COVID-19 crisis and 2) examine how individual protective characteristics (engagement, and perceptions of workload, leadership, and workgroups) and COVID-related organizational contextual factors (via COVID-19 burden, virtual care use, and prior year burnout and turnover intent), were associated with burnout and turnover intent during the first six months of the pandemic.
Methods
Data sources
Data sources included survey data from the 2017 to 2020 VA All Employee Surveys (AES) and 2020 administrative data from the VA COVID Shared Data Resource (COVID SDR), the VA Corporate Data Warehouse (CDW), and the VHA Service Support Center (VSSC).
The AES is a yearly survey of VA employee attitudes [
31] and is typically administered in June (see Table
1 for yearly response rates), except in 2020 when COVID delayed administration until September. The AES is anonymous at the individual-level, with identifiers for an individual’s healthcare system. Each annual survey is cross-sectional, and responses are not linked across years. The CDW is a national repository of VA clinical and administrative information that includes clinical, administrative, financial, enrollment, and benefits data [
32]. The VSSC contains information on healthcare system complexity, detailed below. The COVID SDR [
33] was specifically created at the beginning of the pandemic by the VA to pool COVID-19 clinical data from all VA healthcare systems.
Table 1
Average burnout and turnover intent for VA primary care HCWs from 2017–20 AES
2017 | 59.5 | 41.6 | 6170/14825 | 33.0 | 4821/14623 |
2018 | 61.6 | 35.7 | 5781/16201 | 34.5 | 5720/16649 |
2019 | 63.9 | 35.4 | 6191/17504 | 34.3 | 6177/18017 |
2020 | 69.4 | 37.2 | 7320/19680 | 30.6 | 6096/19894 |
Sample
The sample included primary care HCWs who completed the AES surveys from 2017–2020. Data from 2017–19 was also used to describe pre-pandemic trends and to create healthcare system-level predictors for burnout and turnover intent for the year prior to the beginning of the pandemic (2019). Eligible respondents indicated that they worked on a VA primary care team in one of the following professions: primary care provider (PCP: physician [MD/DO], nurse practitioner [NP], and physician assistant [PA]), registered nurse (RN), clinical associate (licensed practical nurse [LPN], licensed vocational nurse [LVN], nursing assistant, intermediate care technician, and health technician), and administrative associate (administrative or clerical employee working in a clinical area, or a general administrative employee).
Measures
Outcomes
The outcomes included individual-level burnout and turnover intent. Burnout was assessed using a single-item measure for emotional exhaustion (“I feel burned out from my work.”) and a single-item measure for depersonalization (“I worry that this job is hardening me emotionally.”). Both were measured on a 7-point frequency scale (never, a few times a year or less, once a month or less, a few times a month, once a week, a few times a week, or every day). For analysis, we created a dichotomous measure of burnout, with 1 = experiencing symptoms once a week or more versus 0 = less than once per week on either the emotional exhaustion or the depersonalization item, as in previous studies [
34,
35]. Similarly, turnover intent was a dichotomous measure, with 1 = a “yes” response (“yes, but taking another job within VA”, “yes, to retire,” “yes, to take another job within the federal government,” “yes, to take another job outside the federal government,” and “yes, other”) to the question of “Are you considering leaving your job in the next year, and if so why?” and 0 = “no”.
Individual-level protective factors
We also include several individual-level factors that that may be protective against burnout and turnover intent.
The first, employee engagement, was a factor that we hypothesize may be protective against crisis-induced organizational stressors, and that previous studies have suggested might be protective against burnout and turnover intent. Individual-level employee engagement was assessed using a 5-point agreement scale ranging from “strongly disagree” to “strongly agree,” ranked from 1 to 5, for four statements: “I recommend my organization as a good place to work.”; “This organization really inspires the very best in me in the way of job performance.”; “I always do more than is actually required.”; and “My job is more than just a paycheck to me.” Respondents were rated as highly engaged if they agreed or strongly agreed with most items in this set (i.e., a score of 18 or above) [
36]. All four items loaded onto one factor with an eigenvalue greater than 1 (Cronbach’s alpha = 0.76).
We also included three other individual-level factors that may be associated with burnout and turnover intent: perceptions of workload, leadership, and workgroup. These individual factors and their corresponding survey items are as follows: reasonable workload (“My workload is reasonable.”), supervisor listening (“My supervisor listens to what I have to say.”), supervisor respect (“My supervisor treats me with respect.”), supervisor trust (“I have trust and confidence in my supervisor.”), workgroup cooperation (“The people I work with cooperate to get the job done.”), workgroup competency (“My work unit has the job-relevant knowledge and skills necessary to accomplish organizational goals.”), and workgroup collaboration (“Workgroups collaborate to accomplish shared objectives.”). All of these factors were assessed using 5-point agreement scales, ranging from “strongly disagree” to “strongly agree,” and ranked from 1 to 5. Respondents who “agreed” or “strongly agreed” with a factor’s survey item were considered to endorse that factor.
Healthcare system-level predictors of burnout and turnover intent
Organizational contextual factors specific to the COVID crisis included healthcare system-level COVID-19 burden, virtual care use, and prior year average HCW burnout and turnover intent. Using data from the COVID SDR, COVID-19 burden was operationalized as the number of COVID-19 tests administered and deaths per 1000 patients from 03/15/20 to 09/15/20 (i.e., the first six months of the pandemic and six months prior to the administration of the AES 2020 survey). These measures were averaged by healthcare system and grouped into quartiles (tests) or terciles (deaths) for analysis, to adjust for uneven distribution of data and ease interpretation of results. Testing rates were grouped into four categories: 9.2 to 38.7 (lowest quartile), 38.9 to 47.6 (2nd quartile), 47.8 to 61.4 (3rd quartile), and 65.2 to 471.7 tests per 1000 unique patients (highest quartile). Death rates were grouped into three categories: 0.0 to 0.27 (lowest tercile), 0.27 to 0.69 (middle tercile), and 0.69 to 3.65 deaths per 1000 unique patients (highest tercile).
The number of primary care visits delivered virtually by patient location (including video, phone and supplementary remote [e.g., store-and-forward video, audio, or image messages, or specialty visits like tele-smoking cessation]) [
22] and in-person from 03/15/20 to 09/15/20 by healthcare system were extracted from the CDW. A measure for healthcare system-level proportion of virtual care use was created by dividing total virtual primary care visits by all primary care visits (in-person + virtual visits) for each healthcare system.
We used 2019 AES data to create a measure of the proportion of primary care HCWs with high emotional exhaustion or depersonalization in 2019 by healthcare system, using the same definition of dichotomous burnout described above. A similar healthcare system-level turnover intent predictor variable was also constructed using 2019 AES data.
Other covariates
Individual-level demographic covariates included professional role (provider, registered nurse, clinical associate, or administrative associate); gender (male or female); race (white, black/African American, Asian, American Indian or Alaskan Native, or Native Hawaiian or other Pacific Islander); ethnicity (non-Hispanic or Hispanic); age (29 and under, 30–49, or 50 +); VA tenure (less than 2 years, between 2 and 10 years, between 10 and 20 years, or more than 20 years); and supervisor status (yes or no).
We also included a healthcare system-level complexity measure using the VHA Complexity Model [
37] categories available from VSSC: group 1 (combining groups 1a, 1b, and 1c; most complex), group 2, and group 3 (least complex). In general, group 1 healthcare systems have medium-to-high patient volumes, medium-to-high risk patients, more complex clinical programs, and have medium-to-large teaching and research programs [
38]. Group 2 and 3 healthcare systems generally have low-to-medium patient volumes, lower risk patients, and less complex clinical programs. Both group 2 nor 3 healthcare systems typically have small or no teaching or research programs.
Statistical analyses
We used descriptive statistics to characterize study outcomes, predictors, and covariates. To compare burnout and turnover intent levels before and during the pandemic crisis period, we examined the prevalence of burnout and turnover intent among VA primary care HCWs from the 2017 to 2020 administrations of the AES.
For multivariable analyses, individual HCW survey responses for 2020 were linked to 2020 and 2019 healthcare system-level (e.g., VA hospital-based medical center and its affiliated clinics) data. We used logistic regression, modeling burnout and turnover separately, and including employee engagement; perceptions of workload, leadership, and workgroup; COVID-19 burden; virtual care use; and prior year burnout as predictors. Models also included demographics and healthcare system complexity as control variables, with cluster adjustments at the healthcare system level.
Discussion
Trends in VA primary care burnout between 2017–2020 varied but hovered around 37%, with a high rate of 42% in 2017, suggesting little impact due to the early COVID-19 pandemic. Thirty-seven percent of VA primary care HCWs reported high burnout and nearly one-third reported turnover intent in September 2020. The lack of large increases in rates of burnout and turnover intent during 2020 may indicate the presence of local supports (e.g., strong leadership or good crisis response) that offset HCW burnout, or persistent non-COVID drivers of both outcomes in VA primary care.
Thirty-seven percent of VA primary care HCWs reported high burnout in 2020, and nearly 31% reported an intent to leave their job. Burnout rates among physicians and other advanced practitioners outside of VA ranged from 38 to 57% in late 2020 according to two national surveys, [
39,
40] indicating that VA experienced less burnout during the early pandemic than community physicians. Turnover intent rates in late 2020 among non-VA physicians and advanced practitioners [
39] were similar to the rates we found in among VA HCWs in our analysis. National data for nurses has not been published, but a state-level survey from New Jersey indicated that 37% of nurses intended to leave their jobs in late 2020, [
41] which is higher in our sample of VA primary care HCWs (31%). In another state-wide analysis in early 2022, intent to leave was even higher among nurses in Michigan at 39% [
42]. These comparator studies contain non-primary care specialties, so this analysis may underestimate the gap in burnout between VA and non-VA primary care HCWs, as burnout in both community and VA primary care is generally higher than in most other specialties [
43,
44]. VA HCWs in primary care may have had even lower rates of burnout and turnover intent than equivalent HCWs in the community than what is suggested here.
We found that highly engaged HCWs were less likely to report either high burnout or their intent to leave practice in 2020. Engagement, commonly conceived as either the opposite of burnout [
45] or an independent positive construct, [
46] is a cultivated, rather than emergent, phenomenon [
47,
48]. The Job Demands-Resources Model formalizes these differences between burnout and engagement, and proposes that burnout is driven by job demands (e.g., workload), whereas engagement is driven by job resources (e.g., high quality supervisors or leaders, social support from colleagues, etc.) [
49]. Cultivating these resources in “normal” times can increase engagement, and potentially protect against a large increase in burnout that results from increases in demands during crises. Engagement has been previously associated with positive organizational culture in a small sample of nurses [
50] (as conceived by the six areas of worklife: workload, control, reward, community, fairness, and values [
26]) and with lower burnout across professions and countries [
5,
26]. In our sample, high engagement was linked to lower burnout during the pandemic, suggesting that engagement may be a bulwark against drivers of burnout external to an organization. Our analyses also showed that healthcare system-level average burnout in 2019 was associated with individual-level burnout and turnover intent in 2020, suggesting the persistent and longitudinal relationship between an organizational culture of burnout and these individual outcomes.
Individual perceptions of reasonable workload, high quality leadership, and good workgroups were linked with lower rates of burnout and turnover intent in our data as well. Higher workload during the COVID-19 pandemic was found to be related to more burnout in a wide range of studies of HCWs conducted across the world [
51]. Workloads pushing HCWs beyond their training or interfering with their personal lives [
52] and a desire to decrease one’s workload, [
7] were found to be particularly related to high burnout. Leadership, overall, [
53] and in terms of communication and support [
54], has also been previously linked to lower burnout among HCWs during the pandemic. The relationships between good workgroups, or teams, and burnout during the pandemic have not been extensively studied in the literature, but there is some evidence that team identification [
55] and support [
56] were related to lower pandemic-era burnout in primary care and emergency medicine. Finally, it is important to note that only workload had a stronger relationship with burnout than engagement in our findings, suggesting that engagement may encompass aspects of the working environment that are more than just the sum of workload, leadership, and workgroups. Perceptions of good leadership and workgroups were less related to decreased burnout, implying that these constructs were less protective than overall engagement among primary care HCWs during the early pandemic.
Our results also show that healthcare system-level COVID-19 burden, shift to virtual care use, and complexity were not associated with either high burnout or turnover intent. While high burnout has been reported among HCWs across specialties and geographies during the pandemic, [
5‐
9] the relationships between primary care burnout or turnover intent and COVID-19 burden has not been widely studied. We found only one study conducted in a Belgian intensive care unit with nurses that found little association between burnout and the proportions of COVID-19 patients or deaths (to total patients or deaths) [
57]. Similarly, the effect of virtual care on healthcare outcomes during the pandemic has been widely studied, [
58] but its relationship with burnout or turnover intent has only been the subject of a few analyses. These previous studies suggest that greater telehealth self-efficacy (i.e., comfort with the use of telehealth or virtual care) [
10] or fewer “difficulties with new technologies” [
23] during the pandemic may be associated with lower burnout or turnover intent. Our results suggest that the COVID-19 crisis and associated virtual care use had no relationship with burnout or turnover intent. It is possible that neither COVID-19 nor virtual care use were as impactful as longer-term organizational contextual factors associated with burnout and turnover, such as staffing shortages and instability [
29,
59].
Our study had a few notable limitations, including the inability to identify individuals in the anonymous AES data and match those data with prior-year individual-level AES data. In addition, our COVID-19 deaths and tests and virtual care data were healthcare system averages across the first six months of the pandemic (03/15/20 to 09/15/20) and may not have adequately captured variations due to local virus surges. Finally, our measures of COVID burden may not have reflected the true impact of the pandemic on primary care, including the uncertainties associated with staff reassignments, patient screening procedures, care management for vulnerable patients, and the sudden shift to virtual modalities.
More research is needed to explore whether interventions to improve job resources and employee engagement can protect HCW from the negative impacts of crises affecting healthcare delivery and other system-level changes that may increase stress among HCW. There is some evidence that evidence-based quality improvement, an implementation strategy for clinical guidelines and care models, may have reduced VA PCP burnout during patient-centered medical home implementation [
60]. This reduction in burnout may have been related to engaging primary care HCWs in participatory decision-making and the empowerment of frontline HCWs to address workplace quality issues. In another study of PCPs and staff outside the VA, participants who did not help design burnout reduction interventions actually experienced an increase in burnout [
61]. Edwards and colleagues examined 715 small-to-medium-size non-VA primary care practices, and found that those with zero burnout had better working environments and used more quality improvement strategies, compared to high burnout practices [
62]. These results suggest that interventions that empower and engage providers and staff may be key to improving organizational climate and reducing burnout.
In conclusion, burnout and turnover intent among VA primary care HCWs was high, but lower than pre-pandemic trends during the first six months of the COVID-19 pandemic, and COVID-specific organizational contextual factors were not associated with high burnout or turnover intent. We found that highly engaged providers and staff were less likely to be burned out or to intend to leave their jobs. Those with reasonable workloads, and positive perceptions of their leadership and workgroups were also less likely to be burned out. Interventions to improve organizational climate, and increase job resources and engagement, should be developed and evaluated, with the intent of buffering the impacts of external and organizational contextual factors on burnout. Improving primary care HCW working conditions now may be the key to protecting against high burnout in the face of future challenges, whether they be pandemics or other healthcare crises.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.