Effect of Marriage on Burnout among Healthcare Workers during the COVID-19 Pandemic

Since the onset of the COVID-19 pandemic, burnout symptoms have been prevalent among healthcare workers. Living with spouses can be complex and was associated with an increased burnout risk during the COVID-19 pandemic. This study investigated the relationship between living with spouses and burnout among healthcare workers during the COVID-19 pandemic. We distributed questionnaires to participants working in a hospital affiliated with a medical university in Taiwan. The questionnaires were the Copenhagen Burnout Inventory, which comprises personal burnout (PB), work-related burnout (WB), and client burnout subscales; the Nordic Musculoskeletal Questionnaire; and information on basic demographic variables, family factors, living habits, work-related factors, and physical health factors. Multiple linear regression and mediation analysis were used. We obtained 1615 (63.81%) valid questionnaires. After analysis revealed that marriage was an independent risk factor for PB; however, the effect of marriage on WB was nonsignificant after controlling for risk factors. Parenthood, less alcohol use, reported sleep duration less than six hours, less overtime, less shift work, and participation in leisure activities with family and friends were found to be mediators between marriage and a lower WB level. In addition, chronic diseases, frequent neck pain, and shoulder pain were suppression factors. In summary, marriage was associated with an increased risk of PB. Married individuals sustain a high WB level because of changes in family roles, living conditions, and work conditions. Overall, helping healthcare workers to maintain well-being in marriage or family living may be effective in decreasing burnout during the COVID-19 pandemic.


Introduction
Burnout was first described in 1974 by clinical psychologist Herbert Freudenberger [1]. The Maslach Burnout Inventory [2] and the Copenhagen Burnout Inventory (CBI) [3] have been demonstrated to be appropriate and effective burnout measurement tools in healthy populations [4]. Burnout was recognized in the International Classification of Diseases, 11th Revision, as an "occupational phenomenon" [5]. It was defined as a state of physical, emotional, and mental exhaustion that results from long-term involvement in work situations that are emotionally demanding [6]. Clinically, severe burnout can present as emotional exhaustion, physical fatigue, cognitive impairment, disturbed sleep, and functional impairment [7,8]. The stressors often contribute to clinical burnout [9].

Methods
This cross-sectional observational study was conducted during the COVID-19 pandemic. We distributed questionnaires to 2531 employees from a hospital affiliated with a medical university in Taichung, Taiwan, from March to April 2021. Of the 1633 (64.52%) responses received, 1615 (63.81%) were deemed valid. This survey comprised the Nordic Musculoskeletal Questionnaire, the CBI, and information on basic variables for demographic. The professional fields of 1615 healthcare workers were classified in Supplementary Information Table S1 and reclassified as physicians, nurses, professional and technical personnel, and administration staff.
We assessed the participants' education level ("high school or lower", "Bachelor's degree", "Master's degree", or "PhD and above"), marital status ("Married" or "Other"), number of children ("zero", "one", "two", "three", or "more than three"), with a nonzero number of children reclassified as the new variable "parenthood." Relationships with family and friends were evaluated by asking if the participants engaged in leisure activities with family or friends (LAFF) during vacation time. The response options for a suitable parametric test 5-point Likert scale method [34] were "always", "often", "sometimes", "rarely", and "never", with corresponding scores of 100, 75,50,25, and 0 points, respectively. The presence of one or more chronic diseases was also noted as "yes". Alcohol use in the past month was rated as "always", "often", "sometimes", "rarely", and "never" corresponding to 100, 75,50,25, and 0 points, respectively. Responses for sleep duration were <5, 5-6, 6-7, 7-8, or >8 h per day, which were reclassified as <6 h and SLD >6 h per day. Responses for exercise habit were "at least once a day", "at least once a week", "at least once a month", "less than once a month", or "never", and exercising at least once a day or week was reclassified as "regular weekly exercise." The response options for overtime were "rarely", "less than 45 h per month", "45-80 h per month", and "more than 80 h per month." Responses other than "rarely" were reclassified as "experiencing overtime work." The responses to the question on the shift schedule were "day shift work", "night shift work", "irregular shift work", and "regular shift work." The professional fields were classified as nurses, administration staff, physicians, including attending physicians and residents, and professional and technical personnel.
The Chinese version of the CBI is reliable and valid for the assessment of burnout problems [35,36]. The scales are listed in Supplementary Information Table S2. The response options-"always", "often", "sometimes", "rarely", and "never/almost never"-were scored as 100, 75, 50, 25, and 0 points, respectively, except for item C13, which was inversely scored (i.e., 0, 25, 50, 75, and 100 points, respectively); the calculated mean values indicated the PB and WB level for the participants.
MS pain was considered a confounder of burnout due to the close correlation between musculoskeletal pain and burnout [15,37]. Therefore, MS pain must be one of the adjusted variables in a multiple linear regression model on burnout. We adopted the Nordic Musculoskeletal Questionnaire, modified and translated by the Taiwan Institute of Occupational Safety and Health [38], to determine if the presence of pain was attributable to work-related factors in the preceding year. The response options on pain frequency are: every day, once a week, once a month, once every half year, or at least once every half year, corresponding to 100, 80, 60, 40, and 20 points, respectively. We used factor analysis [39] to determine new underlying variables to effectively explain the questionnaire. The results and calculation process are presented in Supplementary Information Table S3. The redefined underlying variables were neck and bilateral shoulder pain, both ankle pain, and bilateral knee pain, respectively.
The t test and one-way ANOVA were adopted for testing the difference on the dependent variable among two or more independent variables. The univariate and multiple linear regression models were adopted for the model. Whether a mediation effect existed among independent variables, the dependent variable, and the mediation factor was determined using the strategy proposed by Baron and Kenny [40], in which, (1) for the first-stage effect, the independent variable significantly affects the mediator, (2) the independent variable significantly affects the dependent variable in the absence of the mediator, (3) for the second-stage effect, the mediator exerts a significant unique effect on the dependent variable, and (4) the effect of the independent variable on the dependent variable weakens upon the addition of a mediator to the model.
A mediation model suitable for the combination of categorical and continuous variables was developed by Iacobucci (2012) [41]. The formulas are listed in Supplementary Information Table S4. The traditional mediation effect is generally recognized as a, b, c, and c in the same direction. A study indicated that when a × b has the opposite sign to that of c , a × b is an estimate of the suppression effect [42], the statistical significance of which can be verified using the Sobel test.

Results
Tables 1 and 2 primarily identified risk factors for PB and WB. These risk factors for PB and WB would be added as controlled variables in multiple linear regression models. According to the strategy proposed by Baron and Kenny [40], the mediation factor must satisfy the condition that an independent variable is significantly associated with the mediation factor. Therefore, Tables 3 and 4 mainly determined if the association between marriage state (an independent variable) and a possible mediation factor (the risk factor of PB or WB) were significant in statistics. Table 5 presents the multiple linear regression models of the marriage effect on PB and WB in the absence and presence of confounders.
These models could determine if marriage states independently affect PB and WB. The mediation model was established in Table 6, which could find the mediation factors of the marriage effect on burnout.      −5.20 *** 0.14 ** 0.05 6.53 *** 0.41 2.76 ** * p < 0.05; ** p < 0.01; *** p < 0.001; c', direct effect; a, the linear or logistic regression coefficient of marriage against the mediation factor (first-stage effect); s a , the standard error of a; b, the linear regression coefficient of mediation factor against WB in the presence of the marriage factor (second-stage effect); s b , the standard error of b; 1 , shift work includes irregular and regular shift work.

To Establish the Linear Regression Models of Marriage Effect on Burnout
Three models-M 0 , M 1 , and M 2 -were established ( Table 5). M 0 was a univariate linear regression for PB/WB (dependent variable) and married (independent variable). M 1 and M 2 were multiple linear regression models for PB/WB (dependent variable) and married (independent variable), with control for the variables affecting PB/WB in Tables 1 and 2. In the M 2 model for WB, two dummy variables-married and parenthood-could create multicollinearity in the multiple regression because 93.3% of the married individuals were also parents ( Table 3). As a result, two dummy variables would not contribute to the M 2 model together.
In the M 0 model for PB and WB, marriage was not associated with a lower PB level (B = −1.06, p = 0.240); however, marriage was significantly associated with a lower WB level (B = −4.30, p < 0.0001). In M 1 and M 2 for PB and WB, marriage was significantly associated with a lower PB level (B = 2.12, p = 0.018) and not associated with a lower WB level (B = −1.05, p = 0.201). Marriage was an independent risk factor for PB, and the effect of marriage on WB can be explained by other risk factors for WB (Table 5).

To Establish the Mediation Models of Marriage and Their Effect on Burnout
The effect of marriage on WB could be explained by other risk factor effects on WB (Table 5). To further explore, we modeled marriage, the mediation factor, and WB. The mediation models used the risk factors for WB that were significantly associated with marital status as possible candidates for mediation factors. Parenthood, exercise habits, alcohol use, sleep duration < 6 h, experience with overtime, irregular work shifts, LAFF, and neck and bilateral shoulder pain were added to the candidate list. Table 6 summarizes the results of the mediation models for the candidate variables.

The Marriage Effect on Burnout
We found that marriage was an independent risk factor for PB; however, the marriage effect on WB could be explained by other risk factors. Notably, parenthood, ever alcohol use, sleep duration less than 6 h per day, experience of overtime, shift work, and LAFF were all found to be mediators of the marriage effect on WB.

Discussion
Through multiple linear regression, we explored the relationship between marriage and burnout during the COVID-19 pandemic and discovered that marriage was associated with increased PB. Married individuals reported low WB, which may be attributed to the parent's role, reduced alcohol use, frequent sleep duration of more than 6 h, rare working overtime, rare having shift work, and frequent participation in leisure activities with family and friends when on vacation.
Individuals engaging in regular exercise had low levels of PB and WB (all p < 0.001; Table 1). The physiological changes due to physical activity may reduce physiological sensitivity to chronic stress, which could lead to faster physical recovery from stressful situations and a reduced burnout risk [43][44][45][46].
The frequent alcohol consumers had higher levels of PB and WB than those who rarely consumed alcohol in a month (Tables 1 and 2). Whether alcohol use reduces stress is debatable [47,48]. In addition, burnout was strongly associated with alcohol abuse or dependence [49,50].
Individuals with a sleep duration of <6 h and those who work overtime or in shifts had a higher risk of PB/WB than others (Tables 1 and 2). Sleep duration of <6 h was correlated with self-reported stress [51] and higher risks of burnout [14]. Overtime and shift work are related to short or disturbed sleep, which could also cause high levels of burnout [52][53][54][55][56][57][58].
Studies have identified that work-related stressors, high workload, and work-home conflict are associated with burnout among physicians [59][60][61][62]. Notably, burnout remains markedly more prevalent among practicing physicians than among individuals in other fields after adjustment for work hours and other factors [63,64]. According to our data, physicians have higher PB and WB levels than nurses, professionals, technical personnel, and administration staff (Tables 1 and 2).
Burnout was associated with an increased risk of musculoskeletal pain and can predict the onset of regional neck/shoulder and lower back pain [37]. We discovered that frequent neck and bilateral shoulder pain was associated with an increased risk of PB and WB (both p < 0.001; Tables 1 and 2).
Family members and the support of friends and colleagues play a vital role in alleviating burnout [17,65]. In addition, LAFF relieves stress, helps individuals cope with emotions, and moderately maintains physical and mental health [66,67]. Positive engagement in LAFF was associated with a low level of PB and WB among our participants (Table 4).
Marriage was significantly correlated with a high level of PB in the presence of adjusted variables (Table 5; M1; B = 2.12, p < 0.05). PB is the degree of physical and psychological fatigue and exhaustion [3], which excludes factors from the workplace and the client. In our cohort, 93.31% of married individuals had ≥1 children (Table 3). This implies that the effect of marriage on PB can be explained by the effect of parenthood. Parenthood is complex and stressful, which exposes parents to chronic stress known as "parental burnout" and it has been reported that there are significant and positive correlations between parental burnout and professional burnout [68]. Therefore, raising children in marriage relationships plays a vital role in worsening PB. Our study was conducted during the COVID-19 pandemic. Studies have demonstrated that in addition to lower PB levels among single people during the COVID-19 pandemic than among married people, disturbances in home and family lives were correlated with burnout [17,33]. A study from Taiwan during the COVID-19 pandemic revealed the following fears among medical staff: (1) passing on COVID-19 to relatives and friends; (2) being separated from family; and (3) the inconvenience of taking care of children or family members [69]. These concerns may also explain the high levels of PB among married individuals in our cohort.
Individuals with married status would consume alcohol in moderation [70]. The married co-twins consumed fewer alcoholic beverages than did their single or divorced co-twins [71]. On the basis of the association between marriage and decreased alcohol use (Table 3), we used the mediation model in Table 5 to determine that decreased alcohol use is a reason that married individuals have a low WB level (Z = −2.12, p < 0.05).
Studies have shown that unmarried individuals reported sleep duration that was 1.53 times shorter than that of married individuals, an average of 4.8 min less [72,73]. Our data indicate that unmarried individuals reported sleep durations that were 1.35 times shorter than those of married individuals (Table 6). In addition, the mediation model demonstrated that a low proportion of sleep duration < 6 h is a reason why married individuals have lower WB levels than those with "other" marital status (Z = −2.76, p < 0.01).
Although marriage is associated with longer lifespans [74], our data indicate that married individuals exhibited a higher proportion (52.66% vs. 45.34%) of chronic disease than did those with the "other" marital status. The presence of chronic disease suppressed the relationship between marriage and decreased WB (Z = 2.42, p < 0.01; Table 6).
Participation in family leisure led to improved interactions and cohesion within families and improved parents' and children's well-being, which is integral to well-functioning marriages [75][76][77]. Our mediation model (Table 6) indicated that the increased frequency of leisure activities with family and friends was also a reason for the marriage effect on decreased WB (Z = −4.74, p < 0.001).
Married workers have a higher risk of musculoskeletal pain than unmarried, divorced, widowed, and separated workers [78]. Married individuals reported a higher frequency of neck and bilateral shoulder pain than those with the "other" marital status (Table 4). However, the mediation model (Table 6) revealed that sustained neck and bilateral shoulder pain was a suppression factor, and married individuals have low levels of WB. Low mood or stress at baseline is more likely to result in neck and shoulder pain, indicating that married individuals may experience more stress than those with the "other" marital status [79].
We discovered that marriage affected PB and WB differently during the pandemic. Specifically, marriage and family living were attributed to severe PB. This may be because, during the pandemic, married healthcare workers may additionally worry about their family or children [69]. However, according to mediation models, marriage and living with a spouse alleviated WB during the pandemic. According to the theory of work-family enrichment [80], participation in one role may enrich the quality of life in the other. The previous research had determined that work and family identities and support from the family were one of the individual, work, and family antecedents of enrichment [81]. In addition, participation in family leisure could improve interactions, family cohesion, functioning, and satisfaction [75,82], which are also integral to well-functioning marriages [77]. The mediation models showed that raising children, less alcohol use, longer sleep duration, less overtime, less shift work, and being positively engaged in LAFF are the main reasons that married individuals sustain low WB. We believe these mediation factors hint that individuals were positively trying to play a successful family role, which could further enhance the quality of life of another role played, such as staff. This could explain why married individuals sustain lower WB than those who are not married.
Therefore, strong relationships with spouses and family members are necessary to alleviate burnout, especially the burnout that results from work-related factors. We suggest that well-being after marriage and positively taking a competent role (wife or husband) in a marriage are reasons that the effects of marriage on burnout differ. This is a key finding that has been rarely discussed in other studies.
We did not collect data on the age of the participants' children, which can affect individuals' stress levels. To overcome this limitation, we controlled for the participants' age. We also did not collect data on parental and marital burnout, stress, or satisfaction; these factors may affect the relationship between marital status and burnout. Work-family conflict results from a psychological phenomenon of imbalance between work and home life, which is an important model for exploring the relationship between family and work. Since the present study did not collect related data using the Measure of Work-Family Conflict Questionnaire, we are unable to further explore the marriage effect on burnout using the work-family conflict model. In addition, multigenerational families and taking care of elderly parents are common in Taiwanese society, but this study did not explore these factors.  [83]. In addition, the measured means of PB and WB in India [84] and Lithuania [85] [86]. Obviously, different countries and regions may experience varying levels of overall burnout among due to differences in medical culture and environment. Therefore, the same study should be conducted in different countries for further comparison and to explore the reasons for differences in burnout. Although people who complete the the burnout questionnaire are not necessarily at risk for clinical burnout [8], people with short-term stress have higher levels of burnout [87], shorter recovery duration, and a more favorable prognosis compared to those with clinical burnout [88]. Therefore, CBI is still an effective tool for early warning of clinical burnout and stress.

Conclusions
In conclusion, we found that marriage was an independent risk factor for PB during the COVID-19 pandemic. In addition, the protective effect of marriage on WB is mainly attributed to the parental role that emerges after marriage, changes in living habits and health, decreasing amounts of overtime and shift work, and positive engagement in LAFF. Overall, maintaining marital well-being and healthy relationships with their family is the most effective strategy for decreasing burnout during pandemics and other high-workload situations. Specifically, an employee health promotion program should assist employees in breaking bad habits by elastically adjusting their work schedule, encouraging family leisure activities, and positively improving their relationships with their spouses and children.
Supplementary Materials: The following supporting information can be downloaded at: https:// www.mdpi.com/article/10.3390/ijerph192315811/s1, Table S1: The professional field of participants;  Informed Consent Statement: Informed consent was obtained from all subjects involved in the study. Written informed consent has been obtained also from the participants to publish this paper.

Conflicts of Interest:
The authors declare no conflict of interest.