The relationship between personality, work, and personal factors to burnout among clinical psychologists: exploring gender differences in Sweden

ABSTRACT

This study investigated the effects of gender, personality (prosocialness, relational-interdependent self-construal, and brooding), job demands, affective work rumination, and personal-to-work conflict on burnout (exhaustion and disengagement), among clinical public-health psychologists in Sweden. The participants answered a self-report questionnaire (n = 828). Hierarchical multiple regression analyses showed that affective work-rumination, brooding and personal-to-work conflict were most strongly associated with exhaustion, whereas affective work-rumination, brooding, role conflict, and prosocialness most strongly predicted disengagement. Furthermore, in the full models, quantitative job demands and relational-interdependent self-construal related to exhaustion, whereas emotional demands related to disengagement. Interestingly, role conflict had a positive relation and emotional demands a negative relation to disengagement. Women reported higher exhaustion, but not higher disengagement, than men. Women also reported higher levels on most of the independent variables. In sum, the results show that a broad range of factors influence burnout among clinical psychologists.

KEYWORDS:

• Burnout
• exhaustion
• disengagement
• clinical psychologists
• job demands
• personality
• gender

Burnout has increasing personal and societal costs in modern societies (Eurofound, 2018). Both the type of profession and the national context influence burnout (Innstrand, Langballe, Falkum, & Aasland, 2011; Purvanova & Muros, 2010). Exhaustion, a central dimension of burnout, is especially evident in human service professions, that is, professions where client-worker relations and emotionality is at the core and where the proportion of female workers often is high (Dollard, Dormann, Boyd, Winefield, & Winefield, 2003; Hasenfeld, 2010; Mor Barak, Nissly, & Levin, 2001). Conditions in work and private life differ for men and women, and the results of meta-analyses shows that women often have been found to experience higher levels of burnout than men (e.g., Purvanova & Muros, 2010). However, it has been noted that studies on gender differences and burnout across various professional and national contexts are scarce (Purvanova & Muros, 2010). Illustrative of this, a recent qualitative meta-analysis on burnout in applied psychologists included 29 studies, but the majority of these studies (20) were conducted in the U.S. (McCormack, MacIntyre, O’Shea, Herring, & Campbell, 2018). Moreover, though numerous studies have investigated how a range of different variables relate to burnout (e.g., Hakanen & Bakker, 2017), studies examining the role of personality on burnout are few. For example, only a few of the studies in the McCormack et al. study (McCormack et al., 2018) attended to personality determinants for burnout.

The present study explored the relation between different factors, including personality, job demands, affective work rumination and private life on burnout (i.e., exhaustion and disengagement) among clinical psychologists in Sweden. In this context we also explored gender differences and gender was operationalized as self-reported gender. We see gender as an interaction effect between social, cultural and biological variables (for a detailed discussion of the “gender/sex” conception, see Hyde, Bigler, Joel, Tate, & van Anders, 2019).

It is sometimes assumed that burnout is foremost triggered by high levels of job demands (e.g., Mealer et al., 2016). However, in the present study we take the perspective that not just job demands, but also the individual’s personality, tendency towards work rumination, and events and activities in private life may also contribute to burnout. Our study contributes by being one of the first to study the relation between personality measures and factors outside of work on burnout in clinical psychologists. Gyllensten and Palmer (2005) deplored the lack of research on the influence of personality factors on stress (see also, Molero Jurado, Pérez-Fuentes, Gázquez Linares, & Barragán Martín, 2018).

This study also contributes by investigating burnout among clinical psychologists in Sweden. The training to become a registered psychologist in Sweden includes a completed MSc Psych and a year of internship in a clinical setting. A majority of the registered psychologists work in clinical settings within public sectors. The Swedish social context has long been influenced by political reformist social democratic ideas and has strong laws for the protection of employees’ rights, including parental leave (Duvander, Haas, & Hwang, 2017; Valarino, Duvander, Haas, & Neyer, 2018). This, together with a high standard of living, may help explain why Sweden is among the countries in Europe with the lowest level of burnout (Schaufeli, 2018). Despite these beneficial conditions there has been an increase in number of paid sick-leave cases in Sweden during the last years, which has been more pronounced among women (Swedish Social Insurance Agency, 2016).

Previous research on the investigated variables relating to burnout

As noted above, our study, apart from job demands, also included other types of factors which may help explain the differences found between genders on burnout. By this approach we attempted to achieve a better understanding of the relative importance for burnout of the different types of explored factors. We next review previous research on these factors.

Personality factors

Personality factors may contribute to burnout. An example is brooding, that is, the tendency to engage in negative self-evaluative thought involving a “passive comparison of one’s current situation with some unachieved standard.” (Treynor, Gonzalez, & Nolen-Hoeksema, 2003, p. 256). Previous research reports that women show a greater brooding tendency, compared to men (Treynor et al., 2003), and to be more anxious than men (Hyde, 2014; McLean & Anderson, 2009). Moreover, Matud (2004) found that women reported coping styles characterized by more emotional and avoidance coping and less rational and detachment coping. These findings may, as suggested by Matud (2004), at least partly be explained by that women encounter less controllable and more negative life events and attitudes then men. Related to this, “men are socialized to use more active and instrumental coping behaviors, and women are socialized to use more passive and emotion-focused behaviors” (Matud, 2004, p. 1411). Based on the reviewed research, we expected that higher tendencies to brooding would be associated with higher levels of exhaustion and disengagement, and that women would report higher levels of brooding than men.

A further personality factor that may be related to burnout is sociality. Sociality is of relevance for burnout among clinical psychologists, as it has been argued to be of importance for understanding workers’ engagement and well-being in public service occupations (Bakker, 2015). However, sociality may also lead to more effort being exerted at the job and therefore to a higher risk for exhaustion. With respect to gender, Yarnell et al. (2015) argued that women are often socialized with the norm to prioritize other people’s needs ahead of their own, and that women tend to be more critical of themselves as well as to have lower levels of self-esteem and self-compassion (Yarnell et al., 2015). Moreover, research has found that women tend to be more engaged in their near social environment than men (Caprara, Steca, Zelli, & Capanna, 2005; McDonough & Walters, 2001; Paro et al., 2014). For example, Caprara et al. (2005) concluded that women tend to show more empathy and provide more emotional support to others, whereas men are more likely to engage in immediate and concrete helping actions (see also Diekman & Clark, 2015; Gonzalez-Morales, Peiró, Rodriguez, & Greenglass, 2006). Furthermore, Paro et al. (2014) found that female medical students reported more emphatic and personal distress and emotional exhaustion compared to male students. Other research show that women report more network-related events, that is, stressors experienced by others in their social network (e.g., McDonough & Walters, 2001).

Another aspect of sociality relates to individuals’ self-construction. In brief, interdependent self-construal refers to the tendency to identify oneself with the values prevalent in one’s group, whereas an independent self-construal is the tendency to see one's identity as separated from others in the near social environment. Studies of self-construal in the U.S. have found that women tend to have a more interdependent self-construal and men a more independent self-construal (Cross & Madson, 1997). For clinical psychologists, higher score on the sociality measures (prosociality and relational-interdependent self-construal, henceforth: type of self-construal) might be associated with being more prone to identify aspects of client cases as unfinished and as deserving more work.

Job demands

Many different types of job demands have been identified (e.g., Schaufeli & Taris, 2014). Some typical examples of job demands are quantitative demands, time pressure, role conflict and emotional demands. These may have different effects on the different dimensions of burnout (McCormack et al., 2018). Prior research has found that, in general, job demands relate to burnout (e.g., Bakker & Demerouti, 2014, 2017; Bakker, Demerouti, & Verbeke, 2004) and, specifically, to exhaustion in psychologists (McCormack et al., 2018). Consistent with this, it has been reported that work hours per week is associated with exhaustion and occasionally with depersonalization among clinical psychologists (Rupert & Kent, 2007; Rupert & Morgan, 2005; Rupert, Stevanovic, & Hunley, 2009). However, the relation between job demands and disengagement may not be as strong as for exhaustion. For instance, Innstrand et al. (2011) found that human service professions showed lower disengagement then other professions such as bus-drivers and IT-advertisement workers.

With respect to gender and job demands, some research indicates that women may often have more stressful lives compared to men (Matud, 2004; McDonought & Walters 2001; Nolen-Hoeksema, Larson, & Grayson, 1999). For example, in a study based on a convenience sample of the Spanish population (n = 2816), women were found to experience higher levels of chronic stress and to report more somatic and psychiatric symptoms compared to men (Matud, 2004). Similar results were found in a Canadian national sample (McDonough & Walters, 2001). Given these reports, we expected job demands to be related to higher exhaustion and women to report higher levels of job demands than men.

Affective work rumination

Work rumination is the tendency to ruminate about work during non-work time. Affective work rumination is a specific type of rumination, characterized by repetitive, negative and non-constructive rumination about work and work-related issues during non-work time. Thus, work rumination is triggered by high levels of demands and stress at work and is related to exhaustion, for example by prolonging the negative effect of stressors at work and hinder recovery (Bennett, Bakker, & Field, 2018; Cropley et al., 2017; Sonnentag, 2018). Furthermore, affective work rumination is likely to lead to negative conclusions about work and thereby to disengagement. With respect to gender, research has found that women report higher levels of rumination compared to men (e.g., Hyde, 2014; Jose & Brown, 2008; Nolen-Hoeksema et al., 1999). Furthermore, researchers have argued that gender differences in rumination occur as an effect of women’s greater lack of social control and by women being socialized into social gender stereotypes including expectations of greater emotionality (e.g., Jose & Brown, 2008; Nolen-Hoeksema et al., 1999). However, though women tend to report higher tendencies to engage in ruminative thinking, the association between rumination and health and well-being does not seem to differ between the genders (e.g., Jose & Brown, 2008; Wendsche & Lohmann-Haislah, 2017).

Personal-to-work conflict

Home, family and other private engagements may be demanding and, in turn, influence the individual’s ability to carry out their work and may therefore increase burnout. The present study aspired to catch a broad range of non-work factors relating to demands outside of work. Therefore, we used items from the personal-to-work conflict (PWC) scale by Wilson and Baumann (2015). These items ask about the experienced effect of the respondent’s total private life on work efficiency. Rupert et al. (2009) found that both family-to-work conflict (FWC) and work-to-family conflict (WTF) were related to higher exhaustion and depersonalization among clinical psychologists. In line with this, we expected an effect of PWC on both exhaustion and disengagement.

With respect to gender, the results of a recent meta-analyses found that in general, and in contrast to lay perceptions, men and women do not differ in the report of conflicts between work and family (Shockley, Shen, DeNunzio, Arvan, & Knudsen, 2017). Furthermore, Rupert et al. (2009) found no gender differences in clinical psychologists’ FWC conflict, but they did find a nearly significant effect for WFC conflict, in that men tended to report more WTF conflict. Given this, although research has found that women, also in Sweden, take greater responsibility for family and home matters (see e.g., Bernhardt, Noack, & Hovde Lyngstad, 2008; Öun, 2013), we expected no gender differences for the PWC measure.

Research on gender and burnout

Understanding gender differences in burnout and its determinants is important in order to hinder unwarranted generalizations and may help to guide the design of preventive actions for men and women (e.g., Gonzalez-Morales et al. 2006; Matud, 2004; Purvanova & Muros, 2010). However, Purvanova and Muros (2010) in their meta-analysis (183 studies and 409 effect sizes) noted that “only a handful of authors” (p. 168) have investigated the relation between gender and burnout. An important result in this meta-analysis was that emotional exhaustion was “slightly more” associated with woman, whereas depersonalization was “somewhat more” associated with men (p. 168).

In general, the interaction between biological, social and cultural factors generates social stereotypes about gender, including gender roles. As an illustration of the consequences of such intersectional interactions, Purvanova and Muros (2010) concluded that gender differences relating to exhaustion and disengagement were greater in countries with harsher labor policies, that is, less employee’ benefits (e.g., the U.S), compared to European countries with more progressive labor policies, that is, more generous employee’ benefits (e.g., Sweden). Also of relevance is that Gyllensten and Palmer (2005) in a review found that females, compared with men, were more frequently represented in high-stress groups in professions associated with higher education (e.g., clinical psychologists). Furthermore, Innstrand and colleagues found that gender differences in burnout varied across eight occupational groups (Innstrand et al., 2011). These results indicate the importance of investigating gender differences in specific professions and, in general, to take an intersectional perspective on gender (i.e., approach gender differences in relation to contextual factors such as country, culture, and/or ethnicity).

Results pertaining to gender differences in burnout for psychologists have been somewhat inconclusive. In a recent meta-analysis, McCormack et al. (2018) found that female psychologists experience more exhaustion than their male colleagues, whereas men experience more disengagement than women. However, Rupert and co-workers, across three studies, found no general gender differences for exhaustion among clinical psychologists in the U.S. (Rupert & Kent, 2007; Rupert & Morgan, 2005; Rupert et al., 2009). Nevertheless, occasional gender differences were found within specific clinical work settings, such as agency or independent settings (Rupert & Kent, 2007; Rupert & Morgan, 2005). Also relevant is that Rupert and colleagues found over involvement with clients to be related to exhaustion (Rupert & Kent, 2007; Rupert & Morgan, 2005) and to depersonalization (Rupert & Morgan, 2005).

The present study

Burnout was measured by using the indicators exhaustion and disengagement. Disengagement is of interest in the present context of clinical psychologists, as Delgadillo, Saxon, and Barkham (2018) found that disengagement, but not exhaustion, was related to therapeutic treatment outcome.

Hypotheses. Our hypotheses were mainly based on the research described above.

Hypothesis 1 stated that women would report higher level of brooding, higher level of sociality on both sociality measures, higher levels of job demands, higher affective work rumination and higher exhaustion, compared to men. In addition, men were expected to report higher disengagement than women. The remaining hypotheses concerned the associations between the investigated independent variables and exhaustion and disengagement.

Hypothesis 2 expected higher prosocialness and tendency towards relational self-construal to be related to higher exhaustion. Moreover, Hypothesis 2 expected less disengagement for people with higher sociality in terms of prosociality and interdependent self-construal.

Hypothesis 3 expected brooding to be positively related to both exhaustion and disengagement.

Hypothesis 4 expected that reported job demands would be positively related to exhaustion.

Hypothesis 5 expected affective work rumination to be positively related to both exhaustion and disengagement.

Hypothesis 6 expected personal-to-work conflict (PWC) to be related to exhaustion, since PWC might attenuate the individual’s resources needed to handle stressors at work.

As noted above, there is a lack of theory, as well as of research, on gender interaction effects in research on burnout among clinical psychologists. Therefore, we did not formulate any hypotheses for the interaction effects investigated for gender.

Method

Participants

In all, 857 psychologists responded to the invitation to participate in the study (31 % response-rate). However, 24 participants did not provide information on gender and were therefore not included in the analyses. The gender question provided three alternatives (i.e., “women”, “man”, and “other”). Three participants choose the response alternative “other” and were also excluded since this group was too small for making any statistical inferences. Finally, two additional participants were excluded as they only had answered the gender question. Thus, the final sample included 828 participants (78 % women, mean age = 43 years, SD age = 11 years).

Hence, in terms of age and gender the sample was representative of the Swedish psychologist population (approximately N = 7,900 licensed psychologists, 73 % women, mean age 42 years; Statistics Sweden, 2018). Moreover, most participants (62 %) reported to work full time (31 % worked 70–99 %). The average reported work-hours per week was 38 hours, among both male (M = 38.16, SD = 7.58) and female psychologists (M = 38.13, SD = 8.15).

Procedure

First, contacts were made with human-resource representatives at the three largest Swedish counties (regions). The counties were selected in order to receive contact information to a larger number of clinical psychologists, and to encompass metropolitan areas, small towns, and rural areas. Information about the study was given, along with a request to receive email-addresses to all clinical psychologists employed in each county. Invitations were then sent out to all the 2,763 collected email-addresses, containing information about the study, confidentiality, and contact information to the researchers. The study was approved by the local ethical committee (the Regional Ethical Review-board, Gothenburg secretariat, Sweden, nr. 608–17).

Each email included an individual link to the web-based survey. If participants chose to participate, they gave their informed consent by activating the individual link. The survey was open for participation for three weeks, and two reminders were sent out (one per week) to those participants who at that time had not yet completed the survey. The survey took approximately 20–25 minutes to complete.

Measures

Independent variables

Sociality

Two measures assessed levels of sociality. The Prosocialness Scale for Adults measures prosocial orientation (e.g., helpfulness, empathy) and has been reported to hold high reliability (PSA; Caprara et al., 2005, c.f. α =.91). In the present study the PSA demonstrated similar reliability (α = .88). Example items are “I am pleased to help my friends/colleagues in their activities” and “I share the things that I have with my friends”. Each of the 16 items was answered by use of a five-point Likert scale, ranging from “Never/almost never true” (1) to “Almost always/always true” (5).

Furthermore, the Relational-Interdependent Self-Construal scale measures an individual’s degree of relational-interdependent self-construal (i.e., the extent to which one thinks of oneself in terms of one´s close relations) and demonstrates high internal consistency (RISC; Cross, Bacon, & Morris, 2000; c.f. α = .85–90). In the present study, the RISC scale demonstrated similar reliability (α = .89). The 11 items are rated on a response scale ranging from “Strongly disagree” (0) to “Strongly agree” (6). Example items are: “My close relationships are an important reflection of who I am” and “When I feel very close to someone, it often feels to me like that person is an important part of who I am”.

Brooding

Brooding was measured by the five items on the Brooding subscale of the Ruminative Responses Scale, which has been demonstrated to hold reliability (RRS; Treynor et al., 2003; c.f. α = .77). In the present study, similar reliability was demonstrated for the Brooding subscale of the RRS (α = .79). Example items are “What am I doing to deserve this?” and “Why do I always react this way?” Items are rated on a 4-point scale ranging from “Never” (1) to “Always” (4).

Job demands were measured by four subscales from the validated Swedish medium-length version of the Copenhagen Psychosocial Questionnaire (COPSOQ II; Berthelsen, Hakanen, & Westerlund, 2018; Pejtersen, Kristensen, Borg, & Bjorner, 2010). In line with the previous research, the scales included and assessed in the present study demonstrated high internal consistency: Quantitative demands (QD, α = .87; c.f. α = .83 in Berthelsen et al., 2018), Emotional demands (ED, α = .77; c.f. α = .80 in Berthelsen et al., 2018), Work pace (WP, α = .87; c.f. α = .87 in Berthelsen et al., 2018) and Role conflict (CO, α = .76; c.f. α = .65 in Berthelsen et al., 2018). QD and WP were assessed by three items each, and ED and CO by four items each. QD, ED and WP were rated on a five-point scale ranging from “Never/almost never” (1) to “Always” (5), whereas CO was measured on a five-point scale ranging from “To a very low degree” (1) to “To a very high degree” (5). Item examples are: “Do you get behind with your work?” (QD); “Is your work emotionally demanding” (ED); “Do you need to keep a high work pace throughout the day?” (WP); “Are contradictory demands placed upon you at work?” (CO). The final scores were transformed in line with the principle of the COPSOQ II (i.e., 1 = 0, 2 = 25, 3 = 50, 4 = 75, 5 = 100).

Affective work rumination was measured by the Affective rumination subscale of the Work Rumination scale (WRS; Cropley, Michalianou, Pravettoni, & Millward, 2012). In the present study, high internal consistency was demonstrated for the subscale (α = .88), in line with the reports of previous research (e.g., α = .90, Querstret & Cropley, 2012). The 5 point scale ranges from “Very seldom/never” (1) to “Very often/always” (5). An example item is: “Do you become tense when you think about work-related issues during your free time?”

Personal-to-work conflict (PTW) assesses the extent to which one perceives that one’s private life hinder one’s effectiveness at work. In the present study, PTW was measured with three of the five items (i.e., items 7–9) of PTW subscale by Wilson and Baumann (2015). The items were: “My personal activities produce stress that makes it difficult to concentrate at work”, “My personal activities drain me of energy I need to do my job”, and “I am often too tired to be effective at work because of my involvement in personal activities.” Each item was answered on a scale ranging from “Totally disagree” (1) to “Totally agree” (5). For the three items of the PWC used in the present study, the Cronbach’s alpha (α) was .90 (c.f. α = .93 for the original five-item version of the PWC in Wilson & Baumann, 2015).

Outcome variables

Exhaustion was measured by the validated Swedish version of the Shirom-Melamed Burnout Questionnaire (SMBQ; Lundgren-Nilsson, Jonsdottir, Pallant, & Ahlborg Jr, 2012; Melamed, Kushnir, & Shirom, 1992). The scale has 22 items and includes four subscales: Physical exhaustion (8 items, example item: “I am physically exhausted”), Listlessness (4 items, example item: “I feel alert”; reverse scored), Tension (4 items, example item: “I feel relaxed”; reverse scored), and Cognitive weariness (6 items, example item: “I feel I am not thinking clearly”). Each item is rated on a seven-point scale ranging from 1 “Almost never” to 7 “Almost always”. SMBQ is validated and demonstrated to hold high reliability (e.g., Person Separation Index, PSI = .97: Lundgren-Nilsson et al., 2012), and similar reliability was found in the current study (α = .96). The SMBQ was used in the present study in order to facilitate comparison of the results with other occupations and groups in Sweden, as it is the officially recognized scale for clinical assessments of burnout in Sweden (Eurofound, 2018).

Disengagement was measured by use of the validated Swedish version (Peterson et al., 2011) of the disengagement subscale on the Oldenburg Burnout Inventory (OLBI: Demerouti, Bakker, Vardakou, & Kantas, 2002; Halbesleben & Demerouti, 2005). The disengagement subscale has 8 items, rated on a scale ranging from “Totally disagree” (1) to “Totally agree” (4). Example items are: “It happens more and more often that I talk about my work in a negative way” and “I feel more and more engaged in my work” (reverse scored). In the present study, the disengagement subscale of the OLBI demonstrated similar reliability (α = .80) compared to previous research (c.f. α = .83, Peterson et al., 2011). Scales not already translated to Swedish were translated from English by use of back translation.

Results

Descriptive statistics for all variables in the study are reported in Table 1. To test Hypothesis 1 regarding mean level differences between men and women independent sample t-test statistics were calculated and these are also reported in Table 1. To control for multiple testing, the False Discovery Rate was used and Benjamin-Hochberg’s adjusted p-values were calculated (Glickman, Rao, & Schultz, 2014). The calculations indicated that p < .02 should be considered significant. As can be seen, in line with Hypothesis 1, significantly higher reports on most variables, including exhaustion, were observed for women, compared to men. However, no gender differences were found in the reports of disengagement or personal-to-work conflict.

Table 1. Means (SDs) for women and men for the investigated variables and t-tests of gender differences

	Women		Men		t-value	p-value	Cohen’s d
	Mean (SD)	n	Mean (SD)	n
Age	43.21 (10.96)	650	44.23 (10.97)	178	−1.10	.272	0.09
Prosocialness	56.77 (8.45)	642	55.06 (8.01)	174	2.40	.017	0.20
Relational self-construal	42.09 (9.95)	630	39.24 (10.31)	169	3.28	.001	0.30
Brooding	1.88 (.52)	645	1.75 (.49)	175	3.07	.002	0.26
Quantitative demands	3.17 (.82)	646	2.88 (.86)	178	4.13	<.001	0.35
Time pressure	3.39 (.82)	646	3.15 (.80)	178	3.51	<.001	0.30
Emotional demands	3.93 (.63)	646	3.77 (.62)	177	3.02	.003	0.26
Role conflict	2.83 (.75)	645	2.63 (.77)	176	3.04	.002	0.26
Affective work rumination	2.72 (.82)	646	2.39 (.87)	176	4.72	<.001	0.39
Personal to work conflict	2.21 (1.03)	644	2.26 (1.05)	175	−.640	.522	0.04
Exhaustion	3.56 (1.13)	625	3.31 (1.01)	166	2.58	.010	0.23
Disengagement	1.99 (.48)	625	2.05 (.52)	166	−1.42	.155	0.12

Hierarchical multiple regression analyses

To test hypothesis 2–6 two hierarchical multiple regression analyses were computed, one for exhaustion and one for disengagement. (The bivariate correlations among the study variables can be found in Table 1 in the supplements). Socio-demographic variables were entered first and then variables were entered according to presumed time precedence. Using this approach, the steps were the following. Step 1: gender; Step 2: age; Step 3: sociality (prosocial orientation, type of self-construal); Step 4: brooding; Step 5: the four job demands; Step 6: affective work rumination; Step 7: personal-to-work conflict, and Step 8: interaction-terms with gender. We entered affective work rumination after job demands as it to a large extent can be seen as a reaction to work stressors. The results from the eight steps of the hierarchical multiple regression analyses are reported in Tables 2 and 3, only the new predictors for each step are presented. The complete tables can be found in the supplementary materials. The direct effects of the predictors in the final significant model are presented in text below. To correct for multiple testing the procedure suggested by Cohen, Aiken and West (2014) was used. That is, only predictors in blocks which added significantly to explained variance are interpreted. Analysis was also performed to check for possible curvilinear relationship between the predictors and the outcome variables but no such relations were found.

Table 2. Hierarchical regression results for exhaustion

Variable	b	SE	β	p	95% CI for b		Adj R^2	∆R^2	p
					LL	UL
Step 1							.01	.01	.009
Gender^a	−0.26	0.10	−.09	.009	−0.45	−0.06
Step 2							.04	.04	<.001
Age	−0.02	0.00	−.19	<.001	−0.03	−0.01
Step 3							.04	.00	<.525
Prosocialness	0.00	0.01	.03	.475	−0.01	0.01
Relational Self-construal	0.00	0.00	−.04	.290	−0.01	0.00
Step 4							.27	.23	<.001
Brooding	1.08	0.07	.50	<.001	0.95	1.21
Step 5							.37	.09	<.001
Quantitative demands	0.27	0.05	.21	<.001	0.17	0.37
Time pressure	0.03	0.05	.03	.473	−0.06	0.13
Emotional demands	0.11	0.06	.06	.049	0.00	0.22
Role conflict	0.20	0.05	.13	<.001	0.10	0.29
Step 6							.48	.11	<.001
Affective work-rumination	0.56	0.04	.43	<.001	0.48	0.65
Step 7							.52	.04	<.001
Personal to work conflict	0.24	0.03	.22	<.001	0.18	0.29
Step 8							.52	.01	.208
Gender x Age	0.00	0.01	−0.01	.676	−0.02	0.01
Gender x PSA	−0.01	0.01	−0.04	.221	−0.03	0.01
Gender x RISC	0.00	0.01	0.02	.487	−0.01	0.02
Gender x BR	0.04	0.16	0.01	.806	−0.27	0.35
Gender x QD	0.12	0.11	0.05	.263	−0.09	0.34
Gender x TP	−0.08	0.10	−0.03	.441	−0.28	0.12
Gender x ED	0.01	0.13	0.00	.921	−0.24	0.27
Gender x RC	−0.10	0.11	−0.03	.366	−0.32	0.12
Gender x AWR	−0.20	0.10	−0.07	.055	−0.41	0.00
Gender x PTW	−0.11	0.07	−0.05	.123	−0.24	0.03

The regression coefficients for each step are presented. ^a Women = 0 and Male = 1. N = 786. Relational Self-construal = Relational Interdependent Self-construal. PSA = Prosocialness, RISC = Relational Interdependent Self – construal, BR = Brooding, QD = Quantitative demands, TP = Time pressure, ED = Emotional demands, RC = Role Conflict, AWR = Affective work rumination, PTW = Personal to work conflict.

Table 3. Hierarchical regression results for disengagement

Variable	b	SE	β	p	95% CI for b		Adj R^2	∆R^2	p
					LL	UL
Step 1							.00	.00	.147
Gender^a	0.06	0.04	.05	.147	−0.02	0.15
Step 2							0.4	.04	<.001
Age	−0.01	0.00	−.19	<.001	−0.01	−0.01
Step 3							0.7	.04	<.001
Prosocialness	−0.01	0.00	−.17	<.001	−0.01	−0.01
Relational Self-construal	0.00	0.00	−.05	.179	−0.01	0.00
Step 4							.16	.09	<.001
Brooding	0.30	0.03	.32	<.001	0.24	0.37
Step 5							.22	.06	<.001
Quantitative demands	−0.01	0.02	−.01	.817	−0.05	0.04
Time pressure	0.00	0.02	.00	.917	−0.05	0.04
Emotional demands	−0.04	0.03	−.06	.102	−0.10	0.01
Role conflict	0.17	0.02	.27	<.001	0.13	0.22
Step 6							.27	.05	<.001
Affective work-rumination	0.17	0.02	.30	<.001	0.13	0.22
Step 7							.27	.00	.902
Personal to work conflict	0.00	0.02	.00	.902	−0.03	0.03
Step 8							.28	.02	.091
Gender x Age	0.00	0.00	−0.03	.332	−0.01	0.00
Gender x PSA	0.00	0.00	0.04	.336	0.00	0.01
Gender x RISC	−0.01	0.00	−0.11	.004	−0.02	0.00
Gender x BR	−0.15	0.09	−0.07	.084	−0.32	0.02
Gender x QD	0.01	0.06	0.01	.899	−0.11	0.13
Gender x TP	0.05	0.06	0.04	.401	−0.06	0.15
Gender x ED	−0.04	0.07	−0.02	.572	−0.18	0.10
Gender x RC	0.02	0.06	0.01	.765	−0.10	0.14
Gender x AWR	0.03	0.06	0.03	.591	−0.08	0.14
Gender x PTW	−0.04	0.04	−0.04	.257	−0.12	0.03

Note. The regression coefficients for each step are presented. ^a Women = 0 and Male = 1. N = 786. Relational Self-construal = Relational Interdependent Self-construal. PSA = Prosocialness, RISC = Relational Interdependent Self – construal, BR = Brooding, QD = Quantitative demands, TP = Time pressure, ED = Emotional demands, RC = Role Conflict, AWR = Affective work rumination, PTW = Personal to work conflict.

Most of the steps in the regression analysis for exhaustion significantly added explained variance to the model confirming Hypothesis 3–6. The largest contributions in explained variance were from brooding, job demands and affective work rumination (see Table 2). However, contrary to Hypothesis 2, the two sociality measures did not add significant variance and neither did the interactions.

To interpret the direct (unique) effects of the predictors we used the standardized coefficients from step 7 which explained 52% of the variance. This was the final model before the interaction effects were included. In this model age (β = −.06, p = .033) and type of self-construal (β = −.07, p = .012) were negatively associated with exhaustion when controlling for the other predictors although the associations were small. Brooding (β = .25, p < .001), quantitative demands (β = .12, p < .001), affective-work rumination (β = .43, p < .001) as well as person-to-work conflict (β = .22, p < .001) were all positively related to exhaustion. The association between gender (β = .01, p = .667), prosocialness (β = .01, p = .698), time pressure (β = .02, p < .495), emotional demands (β = .00, p = .895) and role conflict (β = .01, p = .667) with exhaustion were all non-significant.

Step 8 in the regression analysis investigated the interactions between gender and the other variables. Since this step did not add significantly to the variance the interaction terms were not interpreted.

The next regression analysis used disengagement as an outcome variable. Again most steps in this regression analysis significantly added explained variance to the model which confirmed Hypothesis 2, 3 and 5. The largest contribution in variance was from brooding (see Table 3). The steps that did not add significant variance were gender, PWC and the interaction terms.

To interpret the direct (unique) effects of the predictors we used the standardized coefficients from model 6, since model 7 was not significant. Thus step 6 constitutes the final model before the interaction terms were added, which still significantly added explained variance. Model 6 explained 27% of the variance in disengagement. In this model gender (β = .10, p = .001) was positively associated with disengagement and age was negatively associated (β = −.14, p < .001). There were also negative associations between disengagement and prosocialness (β = −.16, p < .001) and type of self-construal (relational-interdependent) (β = −.07, p = .026). Brooding (β = .19, p < .001) was also positively associated with disengagement. Emotional demands (β = −.09, p = .006) were negatively associated with disengagement while role conflict (β = .19, p < .001) was positively related. Finally, affective-work rumination (β = .30, p < .001) was positively related to disengagement. No significant association was found between disengagement and quantitative demands (β = −.07, p = .104), time pressure (β = −.02, p = .622).

Step 8 in the regression analysis investigated the interactions between gender and the other variables. Since this step did not add significantly to the variance the interaction terms were not interpreted.

Discussion

Burnout in clinical psychologists is important to study not only in order to alleviate the suffering of the psychologists themselves, but also because research has reported that therapists with more tendency to burnout tend to be less effective in their treatment outcomes (Delgadillo et al., 2018; McCormack et al., 2018). The present study investigated Swedish clinical psychologists and studied the relation between personality (sociality and brooding), job demands, affective work rumination, private life and gender, on the one hand, and the burnout dimensions exhaustion and disengagement, on the other. Exhaustion is commonly seen as the core dimension of burnout (e.g., Schaufeli, 2018), but the disengagement dimension also adds useful information. In addition, we analyzed gender differences for the independent variables and for the burnout variables.

The level of exhaustion in our sample was comparatively high. For example, the cut-off point for “severe burnout” has been suggested to be a SMBQ score of 4.4 (Lundgren-Nilsson et al., 2012). In our sample 21.6 % had a score above this level. Below, we first deal with our general results for exhaustion and disengagement, and then those for gender differences.

In contrast to our expectations in Hypothesis 2, the results did not support the idea that the aspects of sociality measured in this study were positively related to exhaustion. This is of interest since much of the theorizing reviewed in the Introduction indicates such a relation. In fact, we found a significant negative relation between the tendency to relational self-construal and exhaustion, suggesting that sociality may act as a resource protecting against exhaustion. Given that the relation was rather weak, further research should explore how the protecting effect of sociality against exhaustion among clinical psychologists can be utilized in order to decrease exhaustion in practical work contexts.

We also explored a further aspect of personality, brooding. In support of Hypothesis 3, brooding, a general tendency to engage in negative and self-centered rumination, showed to have the same level of association with exhaustion as personal-to-work conflict and a stronger association than any of the four job demands. This indicates that personality variables are of interest to explore further in order to better understand burnout. Moreover, affective work rumination was the variable strongest related to exhaustion and this variable (r = .42), apart from exhaustion (r = .51), was the variable most strongly correlated with brooding. Thus, although affective work rumination is similar to rumination, a thinking style, it may still be closely functionally related to the personality variable brooding. Accordingly, in brief, our results support a strong relation between personality variables and exhaustion.

As expected (Hypothesis 4), job demands explained substantial variance in exhaustion, although only quantitative demands were related to exhaustion. Thus, role conflict, emotional demands and time pressure showed no relation to exhaustion. This was somewhat unexpected, but might, speculatively, be explained by that the education of clinical psychologists includes some training how to relate to these types of factors which may decrease the extent to which they are associated with burnout.

Moreover, as expected in Hypothesis 5 and 6, affective work rumination and personal-to-work conflict (PWC) were both related to exhaustion. These results are in line with previous research, as is the lack of support for gender differences in reported level of PWT.

On a general level, our model for exhaustion explained 52% of the variance. Thus, the model appears to include important variables related to exhaustion among clinical psychologists.

We next discuss the regression analysis for disengagement. In similarity to exhaustion, brooding and affective work rumination were noticeably related to disengagement, thus supporting Hypothesis 3 and 5. However, these relations were at slightly lower levels than for exhaustion, which may have contributed to the final model explaining less of the total variance in the data, compared to the regression model for exhaustion.

Moreover, the sociality measures, especially prosociality, demonstrated a negative relation to disengagment, providing support for this part of Hypothesis 2. Thus, greater sociality may protect against disengagement. The reason may be that sociality is in itself a motivating factor leading to engagement in one’s work activities, especially in a caring profession such as clinical psychology (see e.g., Bakker, 2015; Innstrand et al., 2011). Previous research suggests that this effect can be facilitated by means of strong identification with one’s work (Innstrand et al., 2011).

In contrast to previous research, certain types of job demands were clearly related to disengagement (emotional demands and, especially, role conflict). With respect to emotional demands, it should be noted that this relation was negative (β-weight = −.09), thus more reported emotional demands were related to less disengagement. This result deserves further research, but, in general, emotional demands are more likely to be perceived by engaged people, than by disengaged. The positive correlations between emotional demands and the sociality measures (especially for prosocialness, r = .22) are in line with this observation. With respect to role conflict, such conflicts may challenge the individual’s identity conception and thereby facilitate withdrawal and disengagement. This, in contrast to experienced high quantitative demands that foremost may be expected to lead to exhaustion. For example, Lorente, Salanova, Martínez, and Schaufeli (2008), using a two-wave design and studying teachers, concluded that quantitative demands were the only type of the five investigated job demands that were related to exhaustion. Finally, given that identity issues are central to the individual, role conflict may increase affective work rumination (these variables were correlated, r = .46) which would contribute to the effect on disengagement.

Of interest is that personal-to-work conflict did not show any relation to disengagement as its β-weight was close to zero. This may be explained by that pressure from life outside of work is not in itself a reason for employees to feel disengagement from work; it may take events and circumstances at the work place to have such an effect.

Our model for disengagement explained 28% of the variance. This suggests that other variables, not part of our study, may be of more importance for disengagement than the variables we explored. For example, lack of job resources is an important factor for explaining disengagement (Bakker et al., 2004; Fernet, Austin, Trépanier, & Dussault, 2013).

In similarity to other studies (e.g., Rupert & Morgan, 2005; Rupert & Kent, 2007; Rupert et al., 2009), our results showed a linear negative relation between age and the two burnout measures. For exhaustion, some of the age effect is accounted for by differences in personal life (PWC) since the beta for age dropped in the final model when PWC was included. However, it continues to be significant even when accounting for PWC. Therefore age explains above and beyond what is captured by stresses in personal life. One additional effect of age might be that with increased work experience clinical psychologists tend to become better at handling both the quantitative and emotional demands at work. Also, it is reasonable to think that with increasing life experience in general one becomes wiser and might find it easier to not become overly involved in work. Accordingly, advice from older clinical psychologists, for example in the context of teaching and supervision, about how to handle one’s work engagement might be helpful to ameliorate the risk for burnout.

We next discuss the results for gender differences. In general, the effect sizes for the gender differences were small, the largest was found for affective work rumination (Cohen’s d = .39). In line with Hypothesis 1 and most previous research, women reported higher exhaustion then men, but in terms of Cohen’s d the effect was small. In general, as no difference between men and women were found with respect to number of hours worked per week, or with respect to PWC conflict, these variables do not help explain women’s higher exhaustion. Sweden is a country renowned for equality between the genders when it comes to wages and household work. However, the genders are still not equal as differences in wages and household choirs remain (Bernhardt et al., 2008; Öun, 2013). Similarly, Rupert et al. (2009) did not find any gender differences for FTW in the U.S., even though the U.S. male psychologists did work more hours per week, compared with women (Rupert & Kent, 2007; Rupert & Morgan, 2005; Rupert et al., 2009).

Moreover, although women reported higher levels on the two measures of sociality (thereby providing support for that part of Hypothesis 2), these measures were not very important in the regression analysis and therefore also may not help explain women’s higher exhaustion level.

However, many of the other differences between men and women found for the independent variables can help to explain this result. For example, women’s higher level of brooding and affective work rumination (in partial support of Hypothesis 1) may have increased their exhaustion. In addition, and in partial support of Hypothesis 1, women reported higher levels on each of the four job demand measures, including quantitative demands which were found to be related to exhaustion. However, the present study used a cross-sectional data set, and conclusions about the causal relations would demand a longitudinal data analysis approach.

In contrast to previous research (McCormack et al., 2018; Purvanova & Muros, 2010), and our Hypothesis 1, we found no gender differences for disengagement. Thus, it seems that women’s higher level in, for example, affective work rumination, brooding and role conflict did not matter in this context. Given that prosociality was negatively related to disengagement, women’s greater prosociality may have helped to protect them against disengagement. For men, the results by Rupert and Kent (2007) suggested greater over involvement with clients in male clinical psychologists but no relation between over involvement and depersonalization. This indicates that males might already be somewhat more involved in their work. However, since Rupert and Morgan (2005) did not find male over involvement but found a relation between over involvement and depersonalization, this issue needs further research.

Finally, the lack of interaction effects with gender in the context of exhaustion and disengagement indicates that the two genders reacted in similar ways to similar levels of the investigated independent variables. This is broadly in line with Hyde et al. (2019) who argued that the genders are much more overlapping with respect to their biological (e.g., brain and neuroendocrinological) and social characteristics than has traditionally been assumed. This was supported by the results from a very large metasynthesis on research on gender differences by Zell, Krizan, and Teeter (2015), showing that of the 106 meta-analyses included, 81% of the effect sizes were between .01 and .30 (measured in Cohen’s d), that is, absent, trivial or small.

Limitations

This research has various limitations. For example, the data was collected by means of an online survey. Such surveys typically limit the degrees of freedom for the respondents’ responses. However, online surveys allow for more respondents. As our research questions did not aim at in-depth analysis of the situation for individual people, but rather at better understanding of the role of different types of variables for burnout in a larger group of clinical psychologists, we found the online survey format appropriate for our aims. Moreover, research has found that the content of people’s responses is not much affected by the format of online surveys (e.g., Gosling & Mason, 2015; Gosling, Vazire, Srivastava, & John, 2004). Moreover, our design was cross-sectional and therefore causal interpretations are risky; a longitudinal design is needed to provide a better basis for causal conclusions. In addition, our study was limited in that it only covered information on participants who self-identified as male or females. The number of participants who identified with other gender categories was too small to be analyzed in this study. In addition, given that we see sex and gender as intertwined, our gender question in this context did not specify whether we asked about biological sex or social gender, thus, the participants were free to use their own interpretation. In future research more attention should be given to which specific definition of gender is used and the participants should be made aware of this. In order to more fully cover the breath of gender self-categorizations, future research should attempt to include a larger number of people with other identifications than male or female. Furthermore, intersectionality issues were not well addressed in this study, for example, issues involving effects of ethnicity or culture in the burnout context and in future research such issues should be better explored (see e.g., Hyde et al., 2019). Moreover, it would have been informative to separate between clinical work demand and administrative job demands as was done successfully in the above cited studies by Rupert and colleagues.

Speculatively, the results from this study suggest the practical importance for clinical psychologists of finding a balance between client involvement and identification with work while not getting so over involved with work that the recovery processes necessary to avoid burnout are hurt. Norms for clinical psychologists’ therapeutic work tend to include awareness of one’s own reactions. Similarly, self-awareness with respect to one’s degree of exhaustion and the extent of one’s affective work rumination during free time might help the clinical psychologist to identify warning signals when one’s work efforts and engagement risk leading to burnout effects and when, thus, more free time and recovery from work are called for. Moreover, many clinical psychologists receive clinical supervision sessions. Such sessions might include efforts to help them to maintain healthy boundaries relating to client involvement and work balance, which in turn may lead to a decreased risk of exhaustion. These issues relate to clinical psychologists’ self-care and we also recommend that they are included in clinical psychologists’ basic and continuing education courses.

Conclusions

In a Swedish sample of clinical psychologists, affective work-rumination in particular, but also brooding and personal-to-work conflict were the strong predictors for exhaustion. For disengagement affective work-rumination and brooding were the strong predictors. These results suggest the importance of considering personality variables and similar variables (e.g., thinking style) in order to better understand burnout among clinical psychologists. With respect to gender differences, female psychologists showed higher exhaustion, but not higher disengagement compared with males. As personal-to-work conflict did not differ between the genders, in line with Sweden’s comparatively high gender quality, it might not have contributed to the gender difference for exhaustion. Also noteworthy is that no significant gender-interaction effects were found. However, the gender differences found in levels of personality and similar characteristics, and in work demands, may help to explain women’s higher exhaustion level.

Disclosure statement

The authors have no potential conflicts of interest to report.

Supplemental materials

Supplemental data for this article can be accessed here.

Additional information

Funding

This research was supported by a AFA Insurance, Sweden to the first author [dnr 160248];

Notes on contributors

Carl Martin Allwood

Carl Martin Allwood is professor of psychology at the Department of Psychology, University of Gothenburg, Sweden. His research interests are in judgment and decision making, social cognitive psychology, culture oriented psychology and burnout. He has edited several books and published more than 90 papers in international scientific journals in different areas of psychology.

Martin Geisler

Martin Geisler is a Postdoctoral Research Fellow at the Department of Psychology, University of Gothenburg, Sweden, affiliated Researcher at the Centre for Work Life and Evaluation studies, Malmö University, Sweden, and Research Manager at the Region Sörmland, Sweden. He conducts research in work and organizational psychology, occupational health and well-being, decision-making, and social cognition.

Sandra Buratti

Sandra Buratti, PhD, is associate professor and licensed psychologist at the Department of Psychology at Gothenburg University. Her research interests include judgment and decision making as well as memory. She is also interested in aging and development with a focus on the retirement process.  Her research also spans topics of well-being and health.