A Longitudinal Cohort Study of Factors Impacting Healthcare Worker Burnout in New York City During the COVID-19 Pandemic

Abstract

Objective 

This study aimed to longitudinally examine the prevalence and correlates of burnout in frontline healthcare workers (FHCWs) during COVID-19 in New York City.

Methods 

A prospective cohort study of 786 FHCWs at Mount Sinai Hospital was conducted during the initial COVID surge in April to May 2020 (T1) and November 2020 to January 2021 (T2) to assess factors impacting burnout.

Results 

Burnout increased from 38.9% to 44.8% (P = 0.002); 222 FHCWs (28.3%) had persistent burnout, 82 (10.5%) had early burnout, and 129 (16.5%) had delayed burnout. Relative to FHCWs with no burnout (n = 350; 44.7%), those with persistent burnout reported more prepandemic burnout (relative risk [RR], 6.67), less value by supervisors (RR, 1.79), and lower optimism (RR, 0.82), whereas FHCWs with delayed burnout reported more prepandemic burnout (RR, 1.75) and caring for patients who died (RR, 3.12).

Conclusion 

FHCW burnout may be mitigated through increasing their sense of value, support, and optimism; treating mental health symptoms; and counseling regarding workplace distress.

In Brief

During year one of the COVID-19 pandemic, frontline healthcare workers in New York City experienced high rates of persistent and later onset burnout potentially driven by pre-pandemic burnout symptoms, lower sense of value by supervisors, lower dispositional optimism and having cared for patients who died from COVID.

CME Learning Objectives

After completing this enduring educational activity, the learner will be better able to:

• Examine longitudinally the prevalence and correlate of burnout in frontline healthcare workers (FHCWs) during COVID-19 in New York City
• Identify the factors impacting the different courses of burnout among FHCWs
• Outline ways to mitigate and reduce burnout in (FHCWs)

Burnout is an occupational syndrome characterized by emotional exhaustion, depersonalization, and a lack of perceived effectiveness.¹ It is highly prevalent among healthcare workers (HCWs), with 35% to 54% of physicians and nurses experiencing burnout in the pre–COVID-19 pandemic era.^2–5 Key occupational factors contributing to burnout in HCWs include caring for sicker patients, longer work hours, sleep deprivation, excessive clerical tasks, and imbalance between job demands and resources.^2,3,6–8

Burnout adversely impacts personal well-being of HCWs, resulting in lower job satisfaction, early retirement, and increased risk of depression.⁹ Consequently, HCW burnout adversely impacts the healthcare system and society at large. Among nurses, burnout is associated with increased rates of patient mortality and hospital-transmitted infections.^10,11 Burnout in physicians is associated with poorer quality of care and lower patient satisfaction, increased medical errors, and the likelihood of experiencing a malpractice suit.^12,13 Annual societal cost attributable to physician burnout is estimated to be more than $4.6 billion because of its link to physician turnover and reduced clinical hours.^14–16

Since the beginning of the COVID-19 pandemic, frontline healthcare workers (FHCWs) caring for COVID-19 patients have experienced significant distress and a substantial health burden.^17–19 For example, a recent study in Spain found a positive association between perceived threat of COVID-19 and burnout in nurses, and a study of intensive care unit professionals in the Netherlands reported higher rates of burnout in healthcare professionals working overtime and providing care to COVID-19 patients.^7,20 As a result of these stressors, HCWs are experiencing elevated rates of burnout symptoms in the wake of the COVID-19 pandemic, with prevalence as high as 49% to 67%.^21–23

Although a large body of literature has characterized the prevalence, risk factors, and consequences of burnout in HCWs related to the COVID-19 pandemic, little is known about the longitudinal course of burnout in FHCWs during the pandemic in the United States. From our previous work, we found that psychological distress in healthcare workers declined over this time; however, some studies suggest that burnout may be on the rise.¹⁷ To date, only two known longitudinal studies examined burnout rates in FHCWs before and during the COVID-19 pandemic, both of which reported a significant—approximately 22%—increase in burnout rates because of the pandemic.^20,24 On the other hand, another study recently demonstrated that physician burnout has been stable, if not slightly decreasing, from 2017 (43%) to 2020 (38%).²⁵ Although these studies provide important insights into the pandemic's impact on HCW burnout rates, they did not explore predominant courses of burnout throughout the pandemic (ie, early, persistent, and delayed burnout). Furthermore, scarce research has examined factors that may influence these burnout courses in FHCWs during the pandemic.

To address this gap, we examined the prevalence and risk factors associated with symptomatic courses of burnout among FHCWs in New York City (NYC) during the initial pandemic surge and again 7 months later. We also considered the role of potentially protective factors—such as social support, restorative behaviors (i., sleep hours) and coping strategies—which are linked to reduced risk of mental health symptoms and higher resilience.^17,26–28 We hypothesized that burnout symptoms would increase over time, and that history of mental illness and burnout, pandemic-related stressors, and lower perceived social support would portend a greater risk of persistent and delayed burnout.

METHODS

Participants and Timeline

Data were collected via two anonymous surveys of FHCWs working at Mount Sinai Hospital (MSH), an urban tertiary care hospital in NYC. The first survey was administered during the middle and downward slope of our initial pandemic peak in April to May 2020 (time 1 [T1]), and the second survey was administered at 7-month follow-up between November 2020 and January 2021 (time 2 [T2]), corresponding with a subsequent rise and plateau of the second pandemic surge in NYC as indicated by inpatient census data.

Surveys were administered using the Research Electronic Data Capture platform, and links were emailed to eligible participants. At T2, we sent a follow-up email to the entire T1 sample inviting them to complete the second assessment. Undelivered email invitations indicated that a participant no longer worked at MSH and were thus excluded. Participants self-generated research codes to preserve anonymity, and T1 and T2 surveys were linked using approximate deterministic linkage methods. Linked surveys were considered to be those with exact code matches and those with codes within one generalized Levenshtein edit distance²⁹ in addition to having matches on four of five demographic variables.^30,31 Participants were eligible to receive prizes via raffle by filling out a separate unlinked form.

Individuals eligible to participate in both surveys were MSH frontline HCWs, who were defined as HCWs who directly cared for COVID-19 patients during T1 as part of their standard scope of practice or as a redeployment assignment. The institutional review board at the Icahn School of Medicine at Mount Sinai approved this study.

Measures

Supplemental Table 1, https://links.lww.com/JOM/B258, describes survey items assessed at T1 and T2. Survey items assessed demographic/occupational characteristics, personal health (medical risk and history of mental illness), prepandemic burnout, COVID-19–related personal/occupational variables (preparedness, stressors, infection-related concerns, family-related concerns, work competency concerns), psychosocial characteristics (resilience, positive dispositional traits, work-related inspiration, value/support at work, social support), restorative behaviors, and coping strategies.^1,32–37 Prepandemic burnout was assessed with the two-item Maslach Burnout Inventory (MBI) and asked participants to recall their sense of burnout in the past year before the pandemic onset (Cronbach α = 0.783).^1,32,33

Outcome Measure

Current burnout was assessed using the Mini-Z Burnout Assessment, which is rated on a 5-point scale. The presence of burnout was indicated by a rating of 3 (“I am definitely burning out and have one or more symptoms of burnout, i.e., emotional exhaustion”), 4 (“The symptoms of burnout that I am experiencing won't go away. I think about work frustrations a lot”), or 5 (“I feel completely burned out. I am at the point where I may need to seek help”). The Mini-Z has been validated against the emotional exhaustion subscale of the MBI.³⁸ The Mini-Z burnout measure was selected for the periodic longitudinal assessments to capture changes in levels of burnout across a 7-month time frame (from baseline to follow-up), as it asks about burnout levels over the prior month (as opposed to the MBI, which refers to frequency of burnout symptoms in the past year).

Data Analysis

Data analyses proceeded in the following steps. First, we evaluated the representativeness of our sample.³⁹ Next, we computed descriptive statistics to summarize sample characteristics and the prevalence of burnout over time. Four groups were created based on presence/absence of burnout at T1, T2, or both: no burnout, no burnout at T1 or T2; early burnout, burnout at T1 but not T2; persistent burnout, burnout at T1 and T2; and delayed burnout, no burnout at T1 and burnout at T2. Exploratory factor analyses were then conducted to generate factor scores (0 = mean, 1 = SD) of measures of acute (T1) and postacute (T2) COVID-related stressors, and acute (T1) infection-, family-, and work-related concerns and protective psychosocial characteristics (ie, resilience, optimism; Supplemental Table 1, https://links.lww.com/JOM/B258). χ² Analyses and analyses of variance were conducted with Bonferroni-corrected pairwise contrasts to determine between-group differences in sample characteristics. A multinomial logistic regression analysis was subsequently conducted to identify factors associated with symptomatic burnout courses; only variables associated with burnout in bivariate analyses (P < 0.05) were included in this analysis. We conducted planned post-hoc analyses of multi-indicator variables (ie, COVID-19 stressors) to identify component variables associated with the symptomatic courses of burnout as well as the distribution of support by close supervisor by profession using χ² analysis. Finally, relative importance analyses^40,41 were conducted to identify the relative proportion of variance in symptomatic burnout courses explained by each of the significant independent factors in the multinomial analysis. Using chained equations, missing data (<5%) were handled with multiple imputation method.⁴² Analyses were conducted using IBM SPSS Statistics for Windows, version 28.0 (IBM Corp, Armonk, NY); relative importance analyses were conducted using R statistical software package relaimpo (https://cran.r-project.org/web/packages/relaimpo/relaimpo.pdf).⁴⁰

RESULTS

Response Rate and Participant Characteristics

A total of 3360 of the 6026 invited HCWs completed the T1 survey (55.8%), of which 2579 (76.8%) endorsed frontline responsibilities (FHCWs). Seven hundred eighty-six FHCWs (30.5%) analyzed at T1 completed the T2 survey. Of these 786 FHCWs, 783 had complete data for T1 and T2 burnout and all of the variables included in our analyses. Distributions of age, sex, and profession between the T1 and T2 samples did not differ (all χ² > 1.52, all P > 0.22), nor did demographics differ among T2 completers and noncompleters or when compared with MSH Human Resources data.³⁹

Table 1 shows characteristics of the full sample and stratified by longitudinal courses of burnout. Burnout groups differed on all the variables assessed except race/ethnicity, current psychiatric treatment, and infection-related concerns. Table 2 shows results of a multinomial logistic regression analysis of factors associated with symptomatic courses of burnout. Profession was included in the model but was not retained.

TABLE 1 - Characteristics of the Sample and Longitudinal Courses of Burnout in Health Care Workers on the Frontlines of the COVID-19 Pandemic

	Full Sample (N = 783)	No Burnout (1) (n = 350; 44.7%)	Early (2) (n = 82; 10.5%)	Delayed (3) (n = 129; 16.5%)	Persistent (4) (n = 222; 28.3%)
	n (%) or Mean (SD)	n (%) or Mean (SD)	n (%) or Mean (SD)	n (%) or Mean (SD)	n (%) or Mean (SD)	Test of Difference	Pairwise Contrasts
Demographic and occupational characteristics
Age						34.05***
<35	462 (59.0%)	173 (49.4%)	51 (62.2%)	74 (57.4%)	164 (73.9%)		1, 3 > 4
≥35	321 (41.0%)	177 (50.6%)	31 (37.8%)	55 (42.6%)	58 (26.1%)
Sex						19.47***
Female	569 (72.7%)	229 (65.4%)	61 (74.4%)	97 (75.2%)	182 (82.0%)		4 > 1
Male	214 (27.3%)	121 (34.6%)	21 (25.6%)	32 (24.8%)	40 (18.0%)
Race/ethnicity						11.25	—
White, non-Hispanic	415 (53.0%)	186 (53.1%)	42 (51.2%)	73 (56.6%)	114 (51.4%)
Black, non-Hispanic	39 (5.0%)	17 (4.9%)	3 (3.7%)	2 (1.6%)	17 (7.7%)
Hispanic	46 (5.9%)	19 (5.4%)	7 (8.5%)	6 (4.7%)	14 (6.3%)
Other, mixed race	30 (3.8%)	13 (3.7%)	3 (3.7%)	5 (3.9%)	9 (4.1%)
Prefer not to say	54 (6.9%)	25 (7.1%)	4 (4.9%)	7 (5.4%)	18 (8.1%)
Relationship status						8.83*
Single/divorced/widowed	217 (27.7%)	82 (23.4%)	21 (25.6%)	37 (28.7%)	77 (34.7%)
Married/partnered	566 (72.3%)	268 (76.6%)	61 (74.4%)	92 (71.3%	145 (65.3%)		1 > 4
Living with children	233 (29.8%)	126 (36.0%)	25 (30.5%)	42 (32.6%)	40 (18.0%)	21.67***	1, 3 > 4
Profession						39.82***
Registered nurse	265 (33.8%)	95 (27.1%)	29 (35.4%)	38 (29.5%)	103 (46.4%)		4 > 1, 3
Residents/fellows	183 (23.4%)	95 (27.1%)	16 (19.5%)	29 (22.5%)	43 (19.4%)		NS
Attending MD/DO	183 (23.4%)	86 (24.6%)	20 (24.4%)	41 (31.8%)	36 (16.2%)		3 > 4
Physician's assistant/nurse practitioner	114 (14.6%)	58 (16.6%)	14 (17.1%)	19 (14.7%)	23 (10.4%)		NS
Other	39 (4.9%)	15 (4.6%)	3 (3.7%)	2 (1.6%)	17 (7.7%)		NS
Years in practice	8.3 (8.6)	10.1 (10.1)	7.9 (7.4)	7.1 (7.5)	6.2 (6.2)	10.81***	1 > 3, 4
History of mental disorder	159 (20.3%)	61 (17.4%)	11 (13.4%)	23 (17.8%)	64 (28.8%)	14.65**	4 > 1, 2
Past-year burnout	300 (38.4%)	67 (19.2%)	47 (57.3%)	40 (31.0%)	146 (65.8%)	140.12***	4, 2 > 3 > 1
Current psychiatric treatment	116 (14.9%)	44 (12.7%)	9 (11.1%)	19 (14.7%)	44 (19.9%)	6.58	—
COVID-19 pandemic-related variables
Perceived preparedness	2.8 (1.1)	3.0 (1.0)	2.6 (1.1)	3.0 (1.0)	2.5 (1.1)	12.42***	1 > 2, 4
Acute surge stressors	0 (1.0)	−0.3 (0.9)	0.1 (0.9)	0.0 (0.9)	0.2 (1.0)	15.73***	1 > 2, 3, 4
Postacute surge stressors	0 (1.0)	−0.2 (0.8)	−0.1 (0.9)	0.2 (1.0)	0.3 (1.2)	12.78***	3, 4 > 1
Infection-related concerns	0 (1.0)	−0.2 (0.9)	−0.1 (1.0)	−0.1 (1.0)	0.0 (1.0)	2.54	NS
Family-related concerns	0 (1.0)	−0.3 (1.0)	0.3 (0.9)	−0.2 (1.0)	0.1 (1.0)	10.59***	2, 4 > 1
Work competency concerns	0 (1.0)	−0.2 (0.9)	0.1 (1.0)	0.1 (1.0)	0.2 (1.0)	9.69***	2, 3, 4 > 1
Protective psychosocial variables
Resilience	6.5 (1.3)	6.8 (1.3)	6.2 (1.2)	6.6 (1.3)	6.2 (1.3)	10.75***	1 > 2, 4
Positive dispositional characteristics	0 (1.0)	0.2 (1.0)	0.0 (0.9)	0.0 (0.8)	−0.4 (1.1)	18.17***	1, 3 > 4
Work-related pride, meaning, inspiration	0 (1.0)	0.2 (0.9)	−0.2 (1.0)	0.1 (0.9)	−0.2 (1.1)	9.93***	1, 3 > 4
Feel valued/supported at work	0 (1.0)	0.4 (0.9)	−0.4 (0.9)	0.1 (0.9)	−0.5 (1.0)	49.61***	1 > 2, 4
Perceived social support	12.3 (2.9)	12.8 (2.6)	12.4 (2.8)	12.1 (3.0)	11.7 (3.0)	6.37***	1 > 4
Restorative behaviors
Sleep hours	6.5 (1.1)	6.6 (1.1)	6.3 (1.3)	6.5 (1.1)	6.3 (1.1)	4.91**	1 > 4
Physical exercise	2.2 (1.9)	2.4 (2.0)	2.2 (1.9)	2.5 (2.1)	1.8 (1.7)	4.44**	1, 3 > 4
Coping strategies
Self-sufficient coping	1.4 (1.1)	1.7 (1.1)	1.0 (0.9)	1.5 (1.1)	1.1 (1.0)	19.35***	1 > 2, 4
Socially supported coping	1.0 (0.7)	0.9 (0.7)	1.2 (0.7)	0.9 (0.7)	1.1 (0.7)	7.50***	2, 4 > 1
Avoidant coping	0.8 (0.6)	0.7 (0.6)	0.9 (0.7)	0.8 (0.6)	0.9 (0.6)	9.25***	4 > 1

Pairwise contrasts represent Bonferroni-corrected comparisons between burnout courses; 1 refers to no burnout; 2, early burnout; 3, delayed burnout; and 4, persistent burnout.

DO, doctor of osteopathic medicine; MD, doctor of medicine; NS, not significant; SD, standard deviation.

Statistically significant association: *P < 0.05; **P < 0.01; ***P < 0.001.

TABLE 2 - Results of Multinomial Logistic Regression Analysis Predicting Symptomatic Courses of Burnout in Healthcare Workers on the Frontlines of the COVID-19 Pandemic

	Persistent vs Early Burnout, RRR (95% CI)	Persistent vs No/Low Burnout, RRR (95% CI)	Delayed vs No/Low Burnout, RRR (95% CI)
Demographic and occupational characteristics
Female sex	1.81 (0.96–3.42)	2.42 (1.44–4.06)***	2.02 (1.23–3.32)**
Years in practice	0.98 (0.94–1.02)	0.96 (0.93–0.99)*	0.96 (0.93–0.99)**
Past-year burnout	1.49 (0.87–2.55)	6.67 (4.26–10.43)***	1.75 (1.08–2.83)*
Living with children	0.79 (0.39–1.60)	0.52 (0.29–0.93)*	1.34 (0.78–2.29)
Prepandemic mental disorder	2.27 (1.10–4.69)*	1.51 (0.90–2.51)	0.83 (0.48–1.46)
COVID-19–related variables
Acute surge stressors	1.04 (0.75–1.44)	1.50 (1.14–1.96)**	1.50 (1.14–1.97)**
Postacute surge stressors	1.41 (1.06–1.89)*	1.43 (1.13–1.82)**	1.44 (1.13–1.83)**
Family-related concerns	0.83 (0.60–1.15)	1.24 (0.95–1.61)	0.92 (0.71–1.20)
Protective psychosocial variables
Positive dispositional characteristics	0.70 (0.52–0.95)*	0.70 (0.56–0.88)**	0.80 (0.63–1.02)
Work-related pride, meaning, inspiration	1.12 (0.86–1.45)	0.78 (0.62–0.98)*	0.99 (0.77–1.26)
Feel valued/supported at work	0.85 (0.64–1.13)	0.49 (0.39–0.63)***	0.86 (0.67–1.11)
Coping strategies
Socially supported coping	0.84 (0.56–1.25)	1.52 (1.11–2.10)*	0.99 (0.72–1.35)
Avoidant coping	0.94 (0.59–1.47)	1.61 (1.11–2.31)*	1.10 (0.76–1.60)

RRR, relative risk ratio.

Statistically significant association: *P < 0.05; **P < 0.01; ***P < 0.001.

Prevalence of Longitudinal Courses of Burnout

The prevalence of burnout increased from T1 (n = 306 [38.9%]) to T2 (n = 351 [44.8%]; P = 0.002); 222 FHCWs (28.3%) had persistent burnout (at T1 and T2), 129 (16.5%) had delayed burnout (at T2, not T1), 82 (10.5%) had early burnout (at T1, not T2), and 350 (44.7%) had no burnout (not at T1 or T2).

Persistent Versus Early Burnout

Relative to FHCWs with early burnout, those with persistent burnout were more likely to have prepandemic mental illness, reported greater severity of postacute pandemic-related stressors, and scored lower on measures of protective psychosocial characteristics (ie, optimism; Supplemental Table, https://links.lww.com/JOM/B258).

Post-hoc analyses revealed that lower optimism during the acute surge, making a difficult decision in prioritizing health/survival of COVID-19 patients, caring for patients who died of COVID-19, knowing a friend or colleague who was hospitalized for COVID, and more hours worked were independently associated with a persistent versus early burnout course (Table 3). None of the interactions among significant main effect variables were significant (all P > 0.22).

TABLE 3 - Post-Hoc Analysis of Acute and Postacute Surge Variables Associated With Symptomatic Burnout Courses

Persistent vs Early Burnout	RRR (95% CI)
Acute surge variables
Optimism	0.82 (0.67–0.99)
Made difficult decisions prioritizing COVID-19 patients	4.57 (1.12–18.59)
Know a friend or colleague with COVID-19–related hospitalization	2.51 (1.12–5.65)
No. hours worked	1.02 (1.01–1.04)
Postacute surge variables
Cared for patients who died of COVID-19 complications	2.89 (1.03–8.06)

Persistent vs No Burnout	RRR (95% CI)
Acute surge variables
Cared for patients who died of COVID-19 complications	1.97 (1.22–3.17)
Optimism	0.82 (0.70–0.96)
Derive meaning from work	0.70 (0.51–0.95)
Feel valued by one's close supervisor	0.56 (0.38–0.81)
Coping by venting	1.83 (1.12–3.01)
Postacute surge variables
Cared for patients who died of COVID-19 complications	2.02 (1.02–4.00)

Delayed vs No Burnout	RRR (95% CI)
Acute surge variables
Cared for patients who died of COVID-19 complications	1.66 (1.03–2.67)
Postacute surge variables
Cared for patients who died of COVID-19 complications	3.12 (1.70–5.74)

95% CI = 95% confidence interval; RRR = relative risk ratio.

As shown in Figure 1, making a difficult decision in prioritizing health/survival of COVID-19 patients (36.5%), prepandemic mental illness (21.9%), and lower optimism (15.0%) explained the majority of variance within the multivariable model in persistent versus early burnout. Knowing a friend or colleague with COVID-19–related hospitalization (12.9%), caring for patients who died of COVID-19 during the postacute surge (9.8%), and more hours worked (3.9%) explained the remaining variance.

FIGURE 1: Results of relative importance analysis of predictors of persistent versus remitted burnout. Note: error bars represent 95% confidence intervals.

Persistent Versus No Burnout

Relative to FHCWs with no burnout, those with persistent burnout were more likely to be female and to have endorsed prepandemic burnout, less likely to be living with children, and reported fewer years in practice. They also reported greater severity of acute and postacute surge stressors; scored lower on measures of protective psychosocial characteristics, work-related pride, meaning and inspiration, and feeling valued/supported at work; and were more likely to endorse using socially supported (ie, coping with venting) and avoidance (ie, substance use) among their most common coping strategies.

Post-hoc analyses revealed that caring for patients who died of COVID-19 during the acute and postacute surge, lower optimism, deriving less meaning from clinical work, feeling less valued by one's close supervisor, and engagement in venting and behavioral disengagement coping were independently associated with persistent burnout (Table 3). None of the interactions among significant main effect variables were significant (all P > 0.82).

As shown in Figure 2, prepandemic burnout (30.5%), feeling less valued by one's close supervisor (26.9%), and lower optimism (7.9%) explained the majority of variance in persistent burnout. Female sex (7.0%), fewer years in practice (6.0%), behavioral disengagement (4.7%), venting (4.7%), meaning in clinical work (4.5%), not living with children (3.3%), and caring for patients who died of COVID-19 during the acute (2.9%) and postacute (1.6%) surges explained the remaining variance.

FIGURE 2: Results of relative importance analysis of predictors of persistent versus no burnout. Note: error bars represent 95% confidence intervals.

Value by Supervisor by Profession (Post-Hoc Analysis)

FHCW sense of value by supervisor by profession is shown in Figure 3. After correction for multiple comparisons, registered nurses and other professions were significantly more likely to endorse feeling not valued at all by close supervisors relative to housestaff, attending physicians, and advanced practice providers (27.0% and 21.1% vs 7.1%, 10.9%, 10.5%, respectively; P < 0.001). Attending physicians reported the highest levels of feeling very much valued, which was significantly greater than the proportion of registered nurses who endorsed this level of support (19.7% vs 9.4%, P < 0.001).

FIGURE 3: Sense of value by close supervisor by profession. APP, advanced practice provider; NP, nurse practitioner; PA, physician assistant; RN, registered nurse.

Delayed Versus No Burnout

Relative to FHCWs with no burnout, those with delayed burnout were more likely to be female and to screen positive for prepandemic burnout, and had fewer years in practice and greater severity of acute and postacute surge stressors.

Post-hoc analyses revealed that caring for patients who died of COVID-19 during the acute and postacute pandemic surge was independently associated with delayed burnout (Table 3). None of the interactions among significant main effect variables were significant (all P > 0.12).

As shown in Figure 4, caring for patients who died of COVID-19 complications during the postacute surge (43.5%), fewer years in practice (20.3%), and prepandemic burnout (15.0%) explained the remaining variance in our multivariable models. Female sex (14.5%) and caring for patients who died of COVID-19 complications during the acute surge (6.7%) explained the remaining variance.

FIGURE 4: Results of relative importance analysis of predictors of new-onset versus no burnout. Note: error bars represent 95% confidence intervals.

DISCUSSION

This prospective cohort study found that the prevalence of burnout in FHCWs increased from 38.9% to 44.8% between the acute initial surge and 7 months beyond the acute phase of the COVID-19 pandemic in NYC. This finding contrasts with a decline in psychological distress (depression, anxiety, posttraumatic stress symptoms, or a combination of the 3) observed in this same cohort.³⁹ Results revealed that 28.3% of FHCWs had persistent burnout, 16.5% developed delayed burnout, and 10.5% experienced remission of burnout over the study period.

Burnout was most strongly associated with stressors related to direct patient care, consistent with the international recognition of burnout as an occupational phenomenon and safety hazard.^9,43 In particular, FHCWs who reported having to make decisions prioritizing the health/survival of COVID-19 patients were less likely to recover from burnout, and FHCWs who cared for patients who died of COVID-19–related complications during the postacute surge were more likely to develop delayed burnout. The COVID-19 pandemic exacerbated the volume of death and dying and created unprecedented work conditions for FHCWs who cared for patients who died without loved ones present, risked infecting themselves and their families, and made exceptionally difficult decisions about patient care, with limited resources.⁴⁴ These experiences can be described as moral distress, a consequence of working within an environment under constraints that conflict with one's personal ethics and values.⁴⁵ In previous studies of FHCWs, morally distressing experiences were linked to higher odds of burnout.^46,47 These findings support the need to expand services that address moral distress in FHCWs, such as the implementation of trauma-informed guilt reduction therapy, and the organization of senior clinician teams and ethics committees that can rapidly respond to support those practicing in ethically distressing circumstances.^48,49

Prepandemic burnout and history of mental illness were also linked to current burnout. Prepandemic burnout was a strong predictor of persistent and delayed burnout, whereas prepandemic mental illness was associated with persistent burnout. These findings underscore the importance of identifying at-risk FCHWs and targeting interventions to FHCWs with preexisting histories of burnout and mental illness, both during and preceding crises. In May 2020, our institution expanded mental health supports for FHCWs, including mobile app self-screening for psychological distress, crisis response lines, peer discussion groups, counseling services, and psychiatric care.^50–52 Although these initiatives were crucial in proactively responding to the high demand for mental health resources, healthcare systems should also aim to prevent and treat burnout distinctly and in addition to mental health concerns during crises.⁵³ Indeed, meta-analytic studies have revealed that both individual- and systems-level initiatives can reduce overall burnout and subdomain scores, such as emotional exhaustion and depersonalization,⁵⁴ and that systems-level interventions (ie, scheduling changes, work hour restrictions) were more effective than person-oriented programming (mindfulness, coping strategies).⁵⁵ Nonetheless, studies of systems-level interventions are sparse, highlighting the need for further work on interventions that address burnout during crisis, particularly among vulnerable workers.

FHCWs who felt less valued by their immediate supervisors were more likely to have persistent burnout relative to those with no/low burnout. This finding is consistent with prior work demonstrating that low support from direct supervisors is associated with higher rates of employee burnout.^20,56–58 A longitudinal study of more than 3600 physicians revealed that physicians with immediate supervisors with higher baseline leadership scores were less likely to report burnout.⁵⁹ Another study of more than 57,000 HCWs (excluding physicians) found that each 1-point increase in an immediate supervisors' leadership score was associated with a 7% reduced likelihood of burnout in HCWs.⁶⁰ Furthermore, our prior study found that feeling less valued by hospital leadership FHCWs was linked to persistent psychological distress during the COVID-19 pandemic.³⁹ Importantly, we found that nurses felt overall less valued by their supervisors than physicians, trainees, and advanced practice providers. Taken together, these findings suggest that efforts to promote appreciation and communication behaviors from hospital leadership, and regular authentic support from direct supervisors, with a focus on nurses and those who feel less supported, may help mitigate employee burnout.^61–63

Lastly, we found that relative to FHCWs with no/low burnout, those with delayed and persistent burnout were more likely to be female and have fewer years in practice. This is consistent with prior work^9,64 and parallels findings in our previous study, in which female sex and fewer years in practice were associated with delayed and persistent psychological distress.³⁹ Some authors have speculated that women are at a higher risk of burnout because they face systemic challenges including, but not limited to, lower occupational roles, underemployment, greater caregiving responsibilities, and more negative interpersonal interactions.⁶⁵ Indeed, an analysis of our T1 FHCW cohort revealed that sex differences in psychological distress were no longer observed after accounting for prepandemic and peripandemic characteristics and stressors.⁶⁶ Informal caregiving demands were particularly heightened during the COVID-19 pandemic when children were remote schooling and traditional childcare options were unavailable,⁶⁷ whereas we found that living with children was protective against burnout, pointing to a more complex relationship. Nonetheless, given that women make up greater than 75% of the US healthcare workforce,⁶⁸ there is a critical need to bolster support for early-career, female FHCWs. For example, increasing access to workplace resources, mentorship, and coaching programs have shown promise in reducing burnout of physicians and physician-scientists.^69,70

Limitations of this study must be noted. First, we used the single-item Mini-Z Burnout Assessment, which has been validated in physicians against the emotional exhaustion subscale of the MBI but is not considered the criterion standard for assessing burnout.³⁸ Second, retrospectively assessing burnout (ie, early burnout) may have sensitized some FHCWs to reporting a symptomatic burnout course during the pandemic; however, we have differentiated prepandemic and longitudinal burnout at T1–2 by using distinct assessment measures for each (ie, MBI-2 and Mini-Z, respectively). Third, although FHCWs who completed the T2 assessment did not differ on major demographic characteristics, sampling bias may have occurred. Fourth, our study focused on a single institution in one geographic region, potentially limiting generalizability of our findings. Finally, our modest response rate may limit power and ability to detect differences across burnout groups.

CONCLUSIONS

Results of this longitudinal study suggest that the prevalence of burnout in FHCWs increased from the acute to 7 months after the acute phase of the COVID-19 pandemic, with 45% screening positive for burnout at T2. FHCWs experiencing care-related moral distress and who had preexisting mental health disorders were less likely to recover from burnout. In addition, FHCWs with early burnout and who felt less valued by their direct supervisors were more likely to develop persistent burnout. Given the current staffing crisis in healthcare, ongoing research is needed to evaluate the effectiveness of both individual and organizational prevention strategies and treatment efforts to help mitigate the burden and adverse consequences associated with burnout in FHCWs and other at-risk populations.