key: cord-0797899-fwz9rolb
authors: Hirten, R. P.; Danieletto, M.; Tomalin, L.; Choi, K. H.; Zweig, M.; Golden, E.; Kaur, S.; Helmus, D.; Biello, A.; Pyzik, R.; Calcogna, C.; Freeman, R.; Sands, B. E.; Charney, D.; Bottinger, E. P.; Keefer, L.; Farinas, M. S.; Nadkarni, G.; Fayad, Z. A.
title: Factors Associated with Longitudinal Psychological and Physiological Stress in Health Care Workers During the COVID-19 Pandemic
date: 2020-12-22
journal: nan
DOI: 10.1101/2020.12.21.20248593
sha: 4ce7a93a72b56f69d0dab050594a4eced6dcb0b8
doc_id: 797899
cord_uid: fwz9rolb

Introduction: The Coronavirus Disease 2019 (COVID-19) pandemic has resulted in psychological distress in health care workers (HCWs). There is a need to characterize which HCWs are at increased risk of psychological sequela from the pandemic. Methods: HCWs across seven hospitals in New York City were prospectively followed in an ongoing observational digital study using the custom Warrior Watch Study App. Participants wore an Apple Watch for the duration of the study measuring HRV throughout the follow up period. Surveys were obtained daily. Results: Three hundred and sixty-one HCWs were enrolled. Multivariable analysis found New York City COVID-19 case count to be significantly associated with increased longitudinal stress (p=0.008). A non-significant decrease in stress (p=0.23) was observed following COVID-19 diagnosis, though there was a borderline significant increase following the 4-week period after a COVID-19 diagnosis via nasal PCR (p=0.05). Baseline emotional support, baseline quality of life and baseline resilience were associated with decreased longitudinal stress (p<0.001). Baseline resilience and emotional support were found to buffer against stressors, with a significant reduction in stress during the 4-week period after COVID-19 diagnosis observed only in participants in the highest tertial of emotional support and resilience (effect estimate -0.97, p=0.03; estimate -1.78, p=0.006). A significant trend between New York City COVID-19 case count and longitudinal stress was observed only in the high tertial emotional support group (estimate 1.22, p=0.005), and was borderline significant in the high and medium resilience tertials (estimate 1.29, p=0.098; estimate 1.14, p=0.09). Participants in the highest tertial of baseline emotional support and resilience had significantly reduced amplitude and acrophase of the circadian pattern of longitudinally collected heart rate variability. Conclusion: Our findings demonstrate that low resilience, emotional support, and quality of life identify HCWs at risk of high perceived longitudinal stress secondary to the COVID-19 pandemic and have a distinct physiological stress profile. Assessment of HCWs for these features can identify and permit allocation of psychological support to these at-risk individuals as the COVID-19 pandemic and its psychological effects continue in this vulnerable population.

Introduction: The Coronavirus Disease 2019 (COVID-19) pandemic has resulted in psychological distress in health care workers (HCWs). There is a need to characterize which HCWs are at increased risk of psychological sequela from the pandemic.

Methods: HCWs across seven hospitals in New York City were prospectively followed in an ongoing observational digital study using the custom Warrior Watch Study App.

Participants wore an Apple Watch for the duration of the study measuring HRV throughout the follow up period. Surveys were obtained daily.

Results: Three hundred and sixty-one HCWs were enrolled. Multivariable analysis found New York City COVID-19 case count to be significantly associated with increased longitudinal stress (p=0.008). A non-significant decrease in stress (p=0.23) was observed following COVID-19 diagnosis, though there was a borderline significant increase following the 4-week period after a COVID-19 diagnosis via nasal PCR (p=0.05). Baseline emotional support, baseline quality of life and baseline resilience were associated with decreased longitudinal stress (p<0.001). Baseline resilience and emotional support were found to buffer against stressors, with a significant reduction in stress during the 4-week period after COVID-19 diagnosis observed only in participants in the highest tertial of emotional support and resilience (effect estimate -0.97, p=0.03; estimate -1.78, p=0.006). A significant trend between New York City COVID-19 case count and longitudinal stress was observed only in the high tertial emotional support group (estimate 1.22, p=0.005), and was borderline significant in the high and medium resilience tertials (estimate 1.29, p=0.098; estimate 1.14, p=0.09). Participants in the highest tertial of baseline emotional support and resilience had significantly reduced amplitude and acrophase of the circadian pattern of longitudinally collected heart rate variability.

Conclusion: Our findings demonstrate that low resilience, emotional support, and quality of life identify HCWs at risk of high perceived longitudinal stress secondary to the COVID-19 pandemic and have a distinct physiological stress profile. Assessment of HCWs for these features can identify and permit allocation of psychological support to these at-risk individuals as the COVID-19 pandemic and its psychological effects continue in this vulnerable population.

Increasing rates of SARS-CoV2 infections and hospitalizations, growing workloads, and concern regarding personal protective equipment have resulted in a large psychological burden on health care workers (HCWs). 1 pandemic amplifies the risk of these adverse outcomes. [1] [2] [3] Cross sectional studies have demonstrated that front line HCWs are at a high risk of depression, anxiety, insomnia and distress compared to the general population. [4] [5] [6] HCWs on wards serving patients with COVID-19 reported higher levels of stress, exhaustion, depressive mood and burnout. 7, 8 However, there is limited longitudinal data on the pandemics psychological impact on this group, limited data across health care occupations, no means to identify which HCWs are at risk of developing psychological sequela over time, and no objective evaluation of the stress response in HCWs. Identification of at risk HCWs will allow for appropriate allocation of mental health resources.

Advances in digital technology provide a means to address these limitations.

Smart phone Apps can administer surveys and integrate wearable devices, such as the Apple Watch, to monitor the autonomic nervous system (ANS), a primary component of the stress response. ANS function can be ascertained through measurement of heart rate variability (HRV), a measure of the parasympathetic and sympathetic nervous systems impact on cardiac contractility through calculation of changes in the beat to beat intervals. 9

This is an observational cohort study. The primary objective of the study was to identify characteristics associated with longitudinal stress in HCWs. The secondary aim was to determine whether changes in HRV associate with features protective against longitudinal stress development. HCWs across 7 hospitals in New York City (NYC) (The . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) preprint CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) preprint

The copyright holder for this this version posted December 22, 2020. ; https://doi.org/10.1101/2020.12.21.20248593 doi: medRxiv preprint differences between heart beats, categorized as the Interbeat Interval. SDNN is a time domain index reflecting sympathetic and parasympathetic nervous system activity. 9 This is recorded by the Apple Watch during approximately 60 second recording periods (ultra-short period). All data generated by the Apple Watch are stored on the iPhone and transferred with completion of App surveys.

To account for gaps created by unanswered weekly surveys and allow comparison for each patient, we created a new chronological variable called a 'period'.

To account for participants having different time windows between each weekly survey, a period was assigned to each weekly survey according to participants' starting and ending date. When a participant's survey was completed less than 7 days from their previous survey date, the day after the previous survey date was regarded as the starting window date for the next period. When a participant's survey was done 7 days or more apart from the previous survey date, the starting window date was set to 6 days prior to the current survey date. To integrate weekly psychological metrics and daily risk/health metrics, results of the daily surveys were summarized by the periods defined by the weekly surveys. Daily survey data was summarized for each period, eg: mean number of risk days per period, mean number of days left home per period, and mean symptom severity per period. To examine associations between the NYC COVID-19 case-count and perceived stress raw NYC case-count data was obtained for modelling and summarized as a mean case-count per period. 16 Occupation Classification

The occupation of each participant was collected at enrollment. However, due to the pandemic, the roles, risks and responsibilities of these occupations may have changed when compared to non-pandemic job descriptions. We therefore created a new occupation metric to identify which participants were seeing patients during the study. Occupation was calculated as follows: 1) Daily clinical occupation was calculated . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) preprint

The copyright holder for this this version posted December 22, 2020. ; https://doi.org/10.1101/2020.12.21.20248593 doi: medRxiv preprint from the daily survey where participants classified the type of patient or non-patient care responsibilities, he or she had that day. Those who reported either (a) exposure to patient areas but without patients diagnosed with COVID-19 or those being evaluated for COVID-19, or (b) exposure to areas with patients confirmed to have COVID-19 or people being investigated for COVID-19 infection, were assigned as clinical for that day.

Those who responded they were at work but not caring for patients or those who were working remotely were classified as non-clinical for that day. Autoregressive correlation structure (over period) was found to be suitable to our data significantly increasing the likelihood function (LRT p<0.001) and leading to minimal Akaiki information criterion (AIC)/Bayesian information criterion (BIC). Model coefficients were estimated using a restricted maximum likelihood approach (REML) method using R's nlme packages. Hypothesis of interest were tested using contrasts through the capabilities of the emmeans package.

Univariate models tested the association of each variable with longitudinal stress and identified associated factors. Variables with p<0.10 in the marginal ANOVA test were considered significant and included in the multivariate analysis. Although in univariate models random effects include only the intercept, in multivariate models, a random effect . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) preprint

The copyright holder for this this version posted December 22, 2020. ; https://doi.org/10.1101/2020.12.21.20248593 doi: medRxiv preprint for the NYC case burden was found to be significant (LRT<0.001, lower AIC/AIC), indicating heterogeneity in the association of this variable with stress across subjects.

Heart Rate Variability Modelling HRV captured from the Apple Watch demonstrated a sparse non-uniform sampling and circadian pattern making it amenable to analysis via a COSINOR model. This approach models the daily HRV circadian rhythm over a period of 24 hours which can be 

. CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) preprint

The copyright holder for this this version posted December 22, 2020. ; https://doi.org/10.1101/2020. 12.21.20248593 doi: medRxiv preprint In response to the COVID-19 pandemic we launched the Warrior Watch Study, comprised of our custom iOS App which integrates survey metrics with physiological signatures acquired from the Apple Watch. Three hundred sixty-one HCWs, characterized as any worker in a health system, were enrolled across seven hospitals in NYC in this ongoing observational study between April 29 th and September 29 th , 2020, when data was censored for analysis ( Table 1) . Participants had a mean age of 37 years, were 69.3% female and were followed for a mean of 60 days (IQR 21-98 days).

Clinical trainees had higher baseline resilience, compared to clinical non-trainees (p=0.03) and staff (p=0.01), higher optimism (p=0.04) and emotional support (p=0.01) compared to staff, and higher emotional support compared to clinical non-trainees (p=0.01) (Supplementary Table 2 ).

Univariate analysis evaluated the relationship between baseline demographics and prospectively collected survey metrics with longitudinal perceived stress (weekly Table 3 ). Baseline factors including resilience, optimism, emotional support, quality of life, male gender, and age were significantly associated with lower longitudinal stress. Baseline anxiety/depression, body mass index (BMI), weight, and asthma were significantly associated with increased longitudinal stress. Participants were stratified into emotional support tertials (low, medium, high). A significant reduction in stress during the 4-week period after COVID-19 diagnosis occurred only in participants in the highest tertial of emotional support (effect estimate -0.97, p=0.03), but not in the medium (effect estimate -0.62, p=0.48), and low tertials (effect estimate 0.08, p=0.93) (Figure 2A) Table 4 ).

Significant reduction in the amplitude and acrophase of the circadian pattern of longitudinal SDNN was observed between participants with high compared to medium (p<0.001; p<0.001) and low (p=.008; p=0.004) baseline emotional support, respectively (Figure 3a and 3b) . Significant changes in the circadian pattern of SDNN was also observed when the cohort was stratified based on baseline resilience (Figure 3c and   3d) . The amplitude and acrophase of the circadian pattern of SDNN was significantly lower in subjects with high resilience compared to those with low (p<0.001; p=0.048) and medium (p<0.001; p<0.001) resilience, respectively (Supplementary Table 5 ).

. CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) preprint

The copyright holder for this this version posted December 22, 2020. ; https://doi.org/10.1101/2020. 12.21.20248593 doi: medRxiv preprint In summary, we conducted the first study to identify HCW characteristics that correlate with longitudinal stress during the COVID-19 pandemic and identify employees at risk of psychological sequela. We found worsening longitudinal stress is associated with the number of COVID-19 cases in the community, highlighting the effect of the environmental stressor. Baseline emotional support, resilience and quality of life defined which HCWs were prone to perceived longitudinal stress, not occupation class, and characterized a unique ANS stress profile.

In line with our findings, prior work shows that emotional support and resilience buffer against stress. 18, 19 Resilience, defined as a reduced vulnerability to environmental stressors and the ability to overcome difficulty, is crucial to establishing social relationships and is tied to social support, which also acts as an environmental protective factor against adversity. [20] [21] [22] In addition to demonstrating their stress protective effect in multivariate analysis, when we further evaluated NYC COVID-19 case count, a factor associated with longitudinal stress over time, we again found that those with lower emotional support or resilience were vulnerable with a dynamic stress response uncoupled from the environmental COVID-19 stressor. Similarly, the transient reduction in stress that occurs after a COVID-19 diagnosis only occurs in those with high emotional support and resilience.

A strength of our study is the objective assessment of this observation through longitudinal HRV measurements. HRV is a marker of the physiological stress response on the ANS. 23 We found that participants with high emotional support or resilience have a physiologically distinct ANS profile demonstrating the impact of these characteristics on physiological metrics of stress. The multiple dimensions in which we reaffirmed the importance of these features substantiates their effect on longitudinal stress in HCWs.

One of these features, resilience, is modifiable through targeted interventions, providing an opportunity to increase it in HCWs with low resilience. Several resilience building interventions have demonstrated to be effective in HCWs, 24, 25 however, our findings linking HRV alterations with degree of resilience, makes HRV focused resilience building exercises an attractive option. 26 Strengths of the study are its multicenter longitudinal study design. Furthermore, the number and type of longitudinal variables we capture allows for a robust multivariate . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) preprint . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this this version posted December 22, 2020. ; https://doi.org/10.1101/2020.12.21.20248593 doi: medRxiv preprint . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) preprint

The copyright holder for this this version posted December 22, 2020. ; https://doi.org/10.1101/2020.12.21.20248593 doi: medRxiv preprint Figure 1 . Multivariate analysis of factors associated with longitudinal stress. The scatter plot shows estimated coefficients (±confidence intervals) for variables used in the multivariate analysis. Stars indicate that variable has significant (p<0.05) association with longitudinal stress while crosses indicate a borderline significant relationship (p<0.10). Positive association is indicated in blue, negative association in red.

. CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) preprint

The copyright holder for this this version posted December 22, 2020. ; https://doi.org/10.1101/2020.12.21.20248593 doi: medRxiv preprint . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) preprint

The copyright holder for this this version posted December 22, 2020. ; https://doi.org/10.1101/2020.12.21.20248593 doi: medRxiv preprint Figure 3 . Exploring the relationship between HRV, emotional support and resilience. Plots (A, C) show mean (± 95% Confidence Intervals) HRV MESOR, Amplitude and Acrophase for participants with low, medium and high emotional support (A) or resilience (C). Stars indicate significant differences between groups. Plots (B, D) show average daily circadian HRV rhythm for participants with low, medium and high emotional support (B) or resilience (D). (+p<0.1, *p<0.05, **p<0.01, ***p<0.001).

. CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) preprint . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) preprint . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) preprint

The copyright holder for this this version posted December 22, 2020. ; https://doi.org/10.1101/2020.12.21.20248593 doi: medRxiv preprint

The Psychological Impact of Epidemic and Pandemic Outbreaks on Healthcare Workers: Rapid Review of the Evidence

Survey of stress reactions among health care workers involved with the SARS outbreak

A Call to Protect the Health Care Workers Fighting COVID-19 in the United States

Factors Associated With Mental Health Outcomes Among Health Care Workers Exposed to Coronavirus Disease

Mental Health in Frontline Medical Workers during the 2019 Novel Coronavirus Disease Epidemic in China: A Comparison with the General Population

Mental Health Outcomes Among Frontline and Second-Line Health Care Workers During the Coronavirus Disease 2019 (COVID-19) Pandemic in Italy

Psychosocial burden of healthcare professionals in times of COVID-19 -a survey conducted at the University Hospital Augsburg

Factors contributing to healthcare professional burnout during the COVID-19 pandemic: A rapid turnaround global survey

An Overview of Heart Rate Variability Metrics and Norms. Front Public Health

A global measure of perceived stress

Development of a new resilience scale: the Connor-Davidson Resilience Scale (CD-RISC)

Measuring social health in the patientreported outcomes measurement information system (PROMIS): item bank development and testing. Qual Life Res

Two-item PROMIS® global physical and mental health scales

Distinguishing optimism from neuroticism (and trait anxiety, self-mastery, and self-esteem): a reevaluation of the Life Orientation Test