Patterns of Mental Health and Resilience Among Nurses and Physicians Throughout the COVID-19 Pandemic

We conducted a longitudinal analysis to explore changes in depression and anxiety symptoms relative to resilience within a large healthcare system in New York City. Results highlight the importance of developing and implementing interventions that foster support and resilience while also ensuring that healthcare workers have access to mental health care.

Abstract

Objective

The COVID-19 pandemic is associated with increased difficulties in emotional well-being among healthcare workers. The current investigation assesses how changes in depression and anxiety symptoms fluctuate relative to resilience across three time points: baseline, 1-year, and 2-year follow-up among a cohort of physicians and nurses (n = 728).

Methods

Generalized linear mixed modeling for repeated-measures was used to assess the association between resilience and mental health symptoms using validated measures for all variables.

Results

The odds of probable anxiety were reduced by 81% for each one-unit increase in resilience (odds ratio = 0.19, 95% confidence interval [0.14, 0.26]). The odds of probable depression were reduced by 88% for each one-unit increase in BRS (odds ratio = 0.12, 95% confidence interval [0.07, 0.2]).

Conclusions

Results highlight the importance of developing and implementing interventions that foster resilience among healthcare workers.

LEARNING OUTCOMES

• Describe the impact of COVID-19 on healthcare workers mental health.

• Identify changes in mental health outcomes longitudinally from initial outbreak of COVID-19 to 2 years following.

• Appraise the role of resilience in mental health symptoms among healthcare workers.

In 2020, the World Health Organization declared the COVID-19 pandemic a national public health emergency.¹ With nearly 7 million deaths worldwide,² COVID-19 has reshaped healthcare and our understanding of the impact of occupational stress on healthcare workers (HCWs). Now, over 3 years later, the negative impact remains. Although HCWs’ well-being has been a significant focus in understanding the nuances of occupational health for years now,³ the pandemic has called for a greater evaluation of mental health difficulties among this group.⁴ COVID-19 abruptly shifted daily routines across the globe and especially within healthcare systems. In addition to increased exposure to direct contact with COVID-19 patients, HCWs were faced with an increased risk of developing mental health symptoms as a result of demanding workloads, lack of proper training, physical and emotional exhaustion, and unsafe conditions overall.^5–7

During the first wave, HCWs reported anxiety, depression, insomnia, posttraumatic stress, and burnout.^8–10 At least one in five HCWs reported symptoms of anxiety and depression, with the highest rates being among females and nursing staff.¹¹ Both nurses and physicians reported experiencing more workplace stress during COVID-19 than ever before, with over a third of them expressing intentions to quit.¹² Most notably, physicians and nurses also die by suicide at a higher rate than that of the general population.^13–15 Given the elevated occupational burden during the pandemic, HCWs are facing an impaired ability to effectively manage additional stressors within the workplace. It is clear that fostering resiliency and providing psychological support to them is critical for both their own well-being and that of patients.

Building HCW resilience has been an important aim across healthcare systems since the start of the pandemic.^16–20 Resilience is defined as one’s ability to positively adapt to adverse or traumatic experiences.²¹ Specifically, within the context of healthcare professionals, resilience has been described as multifaceted, encompassing both personal traits that lead to said positive adaptation as well as occupational factors, such as organizational support and resources during hardships.²² HCWs’ resiliency represents an important factor not only for their own mental health but also in providing satisfactory patient care. In fact, the Systems Model of Clinician Burnout and Professional Wellbeing, a theoretical framework aimed at improving patient care by supporting clinician well-being, indicates that the consequences of HCW burnout extend beyond one’s own risk of suicide and development of mental health symptoms to that of quality of patient care, safety, and satisfaction.²³ This framework suggests that burnout and professional well-being may be mediated by individual and work factors, including that of resilience, coping strategies, and social support. The Systems Model further posits that when resilience is overpowered by stress, individuals face significant health disturbances.²³

Several studies highlighting the importance of resilience and its connection to mental health within samples of nurses indicated that greater resilience is associated with lower posttraumatic stress disorder, anxiety, depression, burnout, and overall emotional exhaustion, as well as a higher sense of accomplishment, increased job satisfaction, improved work relationships, and increased overall well-being.^24–28 Subsequent meta-analyses have also concluded that resilience is negatively correlated to depression, anxiety, and negative affect, and positively correlated to life satisfaction and positive affect.²⁹ More recently, during the initial peak of COVID-19, moderate levels of resilience were reported for both physicians and nurses.^30,31 Although COVID-19’s relative incidence within healthcare systems has become more normalized, HCWs are still reporting high levels of burnout that are persisting.^32,33 The Systems Model indicates that in order to promote the professional and overall well-being of HCWs, systems would benefit by targeting resilience at both the individual and organizational level.²³

While the majority of the current literature explores the short-term psychological effects of COVID-19 on HCWs, primarily at the initial outbreak,^34,35 there is some research that extends into the following years.^36,37 Throughout the first year of the pandemic, psychological symptoms remained mostly unchanged, with notable increases in emotional exhaustion, stress, depression, and state anger.³⁶ Similarly, for physicians and nurses, psychological effects were still prominent 2 years following the first wave, with nurses demonstrating an increase in symptoms of anxiety and depression as well as reporting worsened sleep over time.³⁷ Despite the growing body of literature, within the longitudinal studies that have been conducted, few address patterns of mental health difficulties or examine them relative to resiliency in HCWs.

The present study aims to address the gap by exploring how mental health, specifically anxiety and depression symptoms, change across three time points among both physicians and nurses, beginning with the first wave of the pandemic (baseline) and again at 1-year and two-year follow-up time points. The current investigation assesses how changes in these mental health symptoms fluctuate relative to resilience over time. We hypothesize that both nurses and physicians will demonstrate fluctuations in anxiety and depression symptoms as associated with resilience such that increases in symptoms will be associated with lower levels of resilience. Understanding the longer-term psychological effects of COVID-19 within the healthcare system and the role of resilience allows for potential recommendations regarding implementing interventions and prevention of worsening outcomes in future outbreaks, especially at a systemic level.

METHODS

Participants

The data utilized in this study were derived from a preexisting longitudinal registry that was established to assess HCWs’ mental health and well-being throughout the course of the COVID-19 pandemic. Physicians and nurses, who were employed directly with a large integrated healthcare system in NY, served as two cohorts that voluntarily participated in the well-being survey. While physicians comprised attendings, residents, and fellows, this study only utilized data relative to attendings, as the residency and fellowship data would not extend to future timepoints.

Procedure

To evaluate changes over time, the registry consists of the baseline assessment and a follow-up every 6 months thereafter. For the purpose of this study, data were analyzed from baseline assessment, and 1- and 2-year follow-up time points among both cohorts. For the nursing cohort specifically, the assessments were conducted between the following timepoints: baseline between September 24 and November 25, 2020, 1-year follow-up data between September 27, 2021 and January 24, 2022, and 2-year follow-up data between September 26, 2022 and January 26, 2023. As for the physician cohort, baseline data were collected between June 18 and August 21, 2020, 1-year follow-up between June 28 and October 20, 2021, and 2-year follow-up between June 20 and October 22, 2022.

The study was reviewed and deemed exempt by the Institutional Review Board (IRB#20-0510-OTH). Informed consent and data were obtained through an electronic health questionnaire on REDCap, a Health Insurance Portability and Accountability Act–compliant database. This manuscript adhered to STROBE guidelines (Supplemental Digital Content, http://links.lww.com/JOM/B785).³⁸

Measures

A demographics questionnaire was utilized to collect information regarding age, gender, race, Hispanic ethnicity, marital status, and occupational descriptors such as work setting, position, and whether or not they provided direct COVID-19 care. HCWs’ probable anxiety and depression symptoms were assessed using the Patient Health Questionnaire (PHQ-4), a brief self-report measure. The PHQ-4 consists of four questions, two of which comprise the anxiety subscale and two the depression subscale. Scores greater than or equal to 3 indicate probable anxiety or depression.³⁹ It is considered a reliable and valid measure of anxiety and depression, demonstrating adequate internal consistency with a Cronbach α of 0.65 to 0.81 for depression and 0.74 to 0.84 for anxiety.⁴⁰ Individual resilience was also measured using the Brief Resilience Scale (BRS), which assesses one’s ability to recover and adapt from stressful experiences.⁴¹ The BRS consists of six items with responses ranging from 1–5 (1 = strongly disagree, 5 = strongly agree). Items are aggregated, and higher scores are indicative of higher levels of resilience. The BRS demonstrates sound psychometric properties, with an internal consistency Cronbach α of 0.70.⁴² Both the PHQ-4 and BRS were administered at baseline, 1-year follow-up, and 2-year follow-up.

Data Analysis

All descriptive statistics, including frequency and proportion for categorical variables, and means, standard deviations, median, and interquartile range for continuous variables, were computed. Select demographic factors (age, gender, race, and ethnicity) were held using their baseline status. Whether the nurse or physician cared for COVID patients directly at the time the baseline survey was disseminated was also included in the analyses. Nurse or physician cohort status was included as a covariate. Secondary analyses looking at the nursing cohort stratified into nurse versus nursing staff were also performed. Since 12% of the participants had their marital/partner status change over time, marital status was analyzed using participant’s status at each survey time point.

Longitudinal analyses were conducted utilizing generalized linear mixed modeling. Due to the inherent correlation among responses within subjects, generalized linear mixed modeling for repeated-measures was performed to determine the association between resilience and the odds of probable anxiety or probable depression over time (baseline, 1-year follow-up, and 2-year follow-up). Each endpoint was analyzed separately as a binary outcome using logistic regression. The additional variables in each multivariable model were decided a priori, and included responder cohort (nurse vs attending physician), age at baseline (analyzed as a continuous variable, in years), gender (male or female), race (White, Black/African American, Asian, other/multiple races), ethnicity (Hispanic Latino or not Hispanic/not Latino), marital status at each time point (single/separated/divorced vs married/engaged/in a committed relationship), and direct COVID care status at baseline (yes vs no). Potential interactions between resilience (BRS) and time were also assessed by including a time × BRS term in the model. Multiple imputation was used to impute missing demographic, BRS, or PHQ-4 values, specifically using the fully conditional specification method. Level of significance was set at α = 0.05 level. All analyses were conducted utilizing SAS software Version 9.4 (SAS Institute, Cary, NC). The datasets produced for the current study are available upon reasonable request.

RESULTS

Of the total 34,980 eligible participants sent the questionnaire, 1491 completed the baseline survey. Nurses accounted for 871 participants, whereas physicians accounted for 620 participants. Exclusions were then made based on missing baseline data, having survey responses for only one time point, and removal of residents and fellows as they did not complete data for multiple time points. The pattern of missingness was evaluated using the PROC MI procedure in SAS and deemed arbitrary (nonmonotone). Therefore, the fully conditional specification method was used to impute missing values for all variables. Of the remaining 738 subjects, five subjects who self-reported their gender status as nonbinary or preferred not to answer and six with a partner status of “other” were also excluded. After all exclusions were applied, the total number of participants was 728, corresponding to 495 nurses and 233 physicians, with a mean age of 48.8 ± 11.9 years. The majority were White (78.9%), non-Hispanic (88.6%), and female (79.7%) (see Table 1).

TABLE 1.

Sample Characteristics

	Baseline				Year 1 Follow-up				Year 2 Follow-up
	Anxiety		Depression		Anxiety		Depression		Anxiety		Depression
Variable	Freq	%	Freq	%	Freq	%	Freq	%	Freq	%	Freq	%
Cohort
Nursing	90	19.6	59	12.9	69	18.5	52	13.9	32	12.4	33	12.8
Physician	39	17.2	22	9.7	28	15.7	10	5.6	13	8.2	5	3.2
Gender
Woman	107	19.8	65	12.1	78	18.3	52	12.1	34	10.7	31	9.7
Man	21	14.7	16	11.3	19	15.6	10	8.3	11	11.3	7	7.3
Race
White	95	18.2	60	11.5	81	18.8	51	11.8	32	9.9	29	8.9
Black/African American	9	23.1	5	12.8	3	9.7	5	16.1	3	16.7	3	16.7
Asian	10	15.4	5	7.6	5	12.2	3	7.3	6	14.3	4	9.5
Other/multiple races	10	27.8	7	19.4	5	16.7	2	6.7	4	20.0	2	10.0
Hispanic ethnicity
Yes	23	29.5	13	16.9	8	13.6	5	8.3	8	18.2	6	13.3
No	103	17.1	68	11.3	87	17.9	56	11.5	36	9.8	31	8.5
Partner status
Single/separated/divorced/widowed	45	24.6	28	15.2	19	16.2	16	13.8	19	22.6	14	16.7
Married/engaged/committed relationship	84	16.7	53	10.6	78	18.0	46	10.6	25	7.6	24	7.3
Direct COVID care
Yes	111	21.0	69	13.1	81	18.9	53	12.4	39	12.2	33	10.3
No	18	11.4	12	7.6	16	12.9	9	7.3	6	6.1	5	5.2
	M	SD	M	SD	M	SD	M	SD	M	SD	M	SD
Age in years	45.1	11.5	46.9	11.3	43.7	12.8	45.2	12.3	43.2	11.0	44.4	11.3
Brief Resilience Score	3.2	0.8	3.1	0.8	3.1	0.8	2.9	0.8	3.0	0.8	2.9	0.7

Note: Of the 728 participants with completed baseline data, 37.5% had only the 1-year follow-up, 14.1% had only the 2-year follow-up, and 48.3% had all three time points.

In the fully adjusted models, when evaluating the change in mental health symptoms among HCWs longitudinally, the odds of probable anxiety decreased significantly over time (P = 0.014); however, probable depression did not significantly change over time (P = 0.46). Higher BRS scores and older age were each associated with a significantly decreased odds of probable anxiety status over time, after covariate adjustment (P < 0.0001 for both). The odds of probable anxiety were reduced by 81% for each one-unit increase in BRS score (odds ratio [OR] = 0.19, 95% confidence interval [CI] [0.14, 0.26]), and there was evidence of 4% reduced odds for each year increase in age (OR = 0.96, 95% CI [0.94, 0.98]). Higher BRS scores were also associated with a significantly decreased odds of probable depression status over time, after covariate adjustment (P < 0.0001). The odds of probable depression were reduced by 88% for each one-unit increase in BRS score (OR = 0.12, 95% CI [0.07, 0.20]). See Figures 1 and 2 for adjusted odds ratios from multivariable analysis of anxiety and depression over time. There was not enough evidence to suggest that the odds of probable anxiety nor the odds of probable depression differed between physicians and nurses (P = 0.19 and P = 0.15, respectively). None of the other covariates (gender, race, ethnicity, or direct COVID care status) were significantly associated with the odds of probable anxiety or probable depression. Age was significantly associated with an increased odds of probable anxiety, though not significantly associated with the odds of probable depression.

FIGURE 1.

Forest plot showing adjusted odds ratios from multivariable analysis of anxiety over time throughout course of the COVID-19 pandemic.

FIGURE 2.

Forest plot showing adjusted odds ratios from multivariable analysis of depression over time throughout course of the COVID-19 pandemic.

Additionally, when assessing for interaction effects, there was no significant time by BRS interaction (P = 0.53 and P = 0.057), indicating that the effect of resilience score on the odds of probable anxiety and probable depression, respectively, was not dependent on time point, and thus not included in the final models.

DISCUSSION

Our study findings supported our hypotheses in that increased personal resilience was associated with lower odds of probable depression and anxiety symptoms over time. This study explored how probable anxiety and depression fluctuated relative to resilience over time within a sample of physicians and nurses. Our findings suggest that resilience does indeed play a role in mental health symptoms among HCWs, as higher levels of resilience were significantly associated with reduced odds of both probable anxiety and depression. Specifically, a one-unit increase in BRS score was associated with 81.1% and 90.8% reductions in odds of anxiety and depression respectively, demonstrating the role of resilience as a potentially protective factor among developing negative mental health outcomes. Age was also associated with a significantly decreased odds of probable anxiety status, suggesting that increases in age were associated with decreases in reported anxiety symptoms. A possible explanation for this association could be that older age may demonstrate more experience and thus less anxiety related to workplace stressors and COVID-19 care. These results are consistent with previous research that indicated that individuals ages 18 to 34 were most vulnerable during the pandemic, reporting the highest levels of mental distress of all age groups assessed.⁴³

When evaluating changes in the mental health outcomes longitudinally, nurses and physicians demonstrated statistically significant changes in probable anxiety over time, specifically a decrease in self-reported anxiety, whereas odds of probable depression did not differ significantly over time. One possible explanation for the decrease in probable anxiety may be due to increased comfort in caring for COVID, with increases in exposure to treating COVID-19 patients potentially decreasing the anxiety surrounding doing so, as previous research indicates that at the onset of the pandemic, the most frequently reported occupational stressor was fears regarding contracting COVID-19.⁴⁴ Not only did HCWs fear exposing themselves to COVID-19, but they also displayed worries about exposing their family members. In the initial months of COVID-19, many HCWs were at an even higher risk of experiencing moral injury than usual, especially regarding making decisions in times of ventilator shortages.⁴⁵ Since then, there has been significant research conducted on COVID-19, increasing scientific knowledge about the virus and thus increasing treatment and prevention initiatives (ie, medication and vaccines). Additionally, fewer patients require hospitalization due to COVID-19, and the strains have been less severe overall, decreasing mortality rates. Protocols have been put forth, and adequate personal protective equipment is more readily available. All of these changes may serve as contributors to the decrease in HCW anxiety we observed over time. While the notable decreases in anxiety over time may be attributed to these factors, the maintenance of depression symptoms may be due to factors outside of the workplace. External stressors aside from occupational ones may include the financial strains faced as a result of the pandemic as well as complicated grief.^46,47 Nonetheless, the emotional exhaustion associated with burnout, hopelessness with regard to death of their patients, and significantly understaffed hospitals may all be contributing factors to depression remaining relatively stable over time within the HCW population as it relates to COVID-19.⁴⁸

Results should be interpreted in light of several limitations. Given the longitudinal nature of the study, attrition reduced the sample size since not every participant responded to all three surveys thus jeopardizing statistical power. Likewise, generalizability to physicians and nurses is limited due to the likely presence of some bias (specifically nonresponse, survivorship sampling, volunteer, and/or response biases) so that the participants included in these analyses may not accurately represent the target population of HCWs. Moreover, this study does not include HCWs beyond those that work as physicians or within the nursing department, and thus future research among varying groups of HCWs is warranted. Due to the study’s nature, there are some unmeasured confounding variables, such as the number of years of work experience for each participant. Despite having sent our questionnaire to a large, diverse population (34,980 eligible participants), our sample was primarily represented by females and White individuals. Future studies are needed to explore experiences of racially and ethnically marginalized HCWs specifically. The American Medical Association indicated that Black and Asian physicians experienced increased racism during the COVID-19 pandemic, and thus mental health outcomes may differ among these groups.⁴⁹ It is also important to note that there were differences among the time frame of the sample, such that physicians and nurses were not assessed at the exact same time-points; however, even after controlling for cohort, there were no differences between the groups. Additionally, as mental health outcomes were measured through a brief self-report, future research would benefit from utilizing standardized clinical interviews to more comprehensively assess symptoms of anxiety and depression. While these findings pertain to a healthcare system during the course of the COVID-19 pandemic specifically, future studies will be needed to assess trends beyond a global crisis, as HCWs are repeatedly faced with a multitude of increased stressors aside from outbreaks.

Fostering resilience may serve as an important preventative strategy for decreasing the risk of developing negative mental health symptoms during global health crises and in response to the stressful nature of healthcare work. While this study assessed resilience at the individual level, it is evident that organizational initiatives aimed at strengthening both individual and organizational-level resilience may also benefit HCWs. In fact, Shanafelt and colleagues^50(p365) argue that “greater personal resilience is not the solution” but rather a starting point, and they call for change at the highest levels of an organization, suggesting that healthcare systems need to shift the organizational culture to one that models and bolsters resilience beyond the personal level. This is further supported by the Systems Model, which recommends healthcare systems develop, pilot, and implement organization-wide initiatives aimed at promoting the positive emotional well-being and resilience of HCWs that also shift organizational culture while building agility, adaptability, and self-sufficiency at the organizational level.^23,51 At this level, an organization ideally demonstrates the ability to withstand changing environments and disruptive events and responds effectively to them. ⁵¹ Organizational resilience has been associated with positive outcomes for employees, including enhanced self-efficacy, proficiency, and overall work performance. ⁵² Thus, future research could assess organizational resilience as an additional way of supporting mental health among HCWs.

Currently, several resilience building programs exist such as Stress First Aid (SFA). SFA is a peer support and self-care model based on evidence-based elements originally developed for individuals in high-risk occupations, now being adapted for use among HCWs.^53,54 Adapted use of SFA for HCWs has been found to increase self-efficacy, resource awareness, and resilience rates among this population as well as decrease burnout.⁵⁴ Similarly, Promoting Resilience in Stress Management within the workplace (PRISM at Work) is another program that has demonstrated improvements in mental health outcomes through resilience coaching in this population.⁵⁵ Through systematic programs like SFA or PRISM, HCWs can become readily able to identify and address early signs of stress reactions and promote resilience and coping. These programs target both personal and organizational resilience and may reduce HCWs’ vulnerability to developing worsened symptoms of depression and anxiety, consequently resulting in providing increased satisfactory care to patients.

In sum, HCWs have demonstrated increased distress in the workplace, with the COVID-19 pandemic exacerbating their difficulties and contributing to development of psychological symptoms.^8–10,12 The well-being of both nurses and physicians is impacted by several occupational burdens, and this population exhibits higher rates of suicide than that of the general population.^13–15 Our results highlight the importance of developing and implementing interventions that foster support and resilience at an institutional level while also ensuring that HCWs have access to mental health care without barriers.