Workaholism,work engagement,and burnout among academics in Montenegro: A psychometric network approach

Abstract

BACKGROUND:

The academic environment is known for its high demands in research, teaching, and administration, that along with increasing publish or perish culture can lead to reduced psychological well-being and mental health issues.

OBJECTIVE:

This study aimed to investigate the associations between workaholism, work engagement, and burnout among academics in Montenegro.

METHODS:

A cross-sectional design was used to develop anonymous online survey. Data was collected from 131 participants employed as teaching and research staff at public and private universities. To measure the variables of interest we used: ultra-short Utrecht Work Engagement Scale (UWES-3), the work-related burnout subscale from the Copenhagen Burnout Inventory (CBI-7) and the Dutch Work Addiction Scale (DUWAS-10). Psychometric network analysis was employed to examine the relationships among variables.

RESULTS:

The findings revealed two distinct clusters: the first containing the dimensions of work engagement and the second containing burnout and the dimensions of workaholism. The two clusters were connected with the dimensions of dedication - burnout having the strongest edge (–0.25 and –0.40). In the cross-sample network the strongest connection was burnout –working excessively (.35). No significant differences in network density (0.80 (12/15 edges)) and global strength (p = 0.159) in the networks of public and private universities were found.

CONCLUSION:

Results of the network centrality and the edge strength analyses suggest that the interventions focused at increasing dedication while not fostering a work environment that encourages working excessively might be the key to preventing and reducing burnout in academia across contexts of public and private universities.

Keywords

Workaholism work engagement burnout academia psychometric network analysis Montenegro

1 Introduction

Academia is a high-pressure environment characterised by elevated research, teaching, and administrative demands, increasing bureaucracy, long working hours, and job insecurity. Unfortunately, such circumstances frequently take a toll on the mental well-being of academics, resulting in heightened levels of psychological distress, anxiety, depression, and burnout when compared to the general population [1]. Recent reports on mental health in the UK have shed light on the extent of these challenges. Among various professions, academics exhibit the highest prevalence of common mental disorders, with rates hovering around 37% [2]. Also, a systematic review conducted by Urbina-Garcia [3] underscores the vulnerability of young academic staff, particularly those in private universities, to poorer well-being, marked by heightened stress and burnout. Furthermore, in the last 20 years the academic landscape has been significantly influenced by neoliberalism as western universities followed the model of ‘competition-based logics’ [4]. This ideology emphasises performance management and the use of performance indicators [5]. Within such a context, securing research funding and publishing in high-impact journals have become crucial for both universities’ competitiveness and researchers’ career progression, fostering a culture of competition and often leading to a “publish or perish” mindset [1]. As Ryan-Flood and Gill [6] point out, neoliberalism’s impact on academics’ mental health is palpable, given their inherent drive to “work hard” and “excel” [6]. Consequently, the academic environment may also inadvertently expose researchers to the risk of workaholism.

More specifically, workaholism refers to a strong inner drive to work excessively hard and to allocate an exceptional amount of time to work [7]. It is characterised by the tendency to compulsively, persistently, and frequently think about work or be obsessed with work, even when not working [8]. Empirical evidence shows that academic workers often engage in working practices indicative of workaholism, such as working on weekends, bringing work home, and working in the evenings [9, 10]. Workaholism prevalence among academic staff is between 50% and 66% [9, 11]. Moreover, academic work is open-ended and absorbing [9], with each academic largely responsible for deciding the scope of their workload. Finally, the university context is claimed to have ‘overtime culture’ [12] which is one of the main contributors to the onset of workaholism [13] and presenteeism [14]. But, some authors argued that, although especially in contexts like academia, workaholism manifests itself through the same dysfunctional characteristics as other addictive behaviours, it comes with a little social stigma attached to it, or might even be actively encouraged by the reward system that stimulate quantity over quality of output [15]. Because the nature and intensity of job demands as well as the climate in academia may be both challenging for mental health and conducive to workaholism, it is important to assess relationships between workaholism and other indicators of wellbeing. Therefore, in the current study, we build on the circumplex model of work wellbeing [16] to analyse the relationships between workaholism, work engagement, and burnout as the emotional states central to the work domain.

The circumplex model [16] represents a broad theoretical framework that is useful to represent wellbeing at work because it distinguishes and encompasses both, positive and negative types of work wellbeing. It further proposes that there are two primary dimensions of wellbeing in the workplace, namely, activation (the level of energy and arousal that an individual experiences at work) and pleasure (the level of positive emotions and satisfaction an individual experiences). On one hand, work engagement refers to a positive, fulfilling, work-related state of mind that is characterised by vigor, dedication, and absorption [17] and is reflective of high levels of pleasure and activation related to one’s work. Based on the circumplex model, workaholism and engagement share the dimension of activation because they are both characterised by high energy and arousal related to work but are different in pleasure that work entails in that engaged workers feel fulfilled by their jobs, whereas those who are work-addicted feel frustrated and continually dissatisfied. Empirical studies provide some support for this theorising as they showed that the two might be both negatively [19] and positively [20] related. Additionally, work engagement and workaholism might both be reflected in identical behaviours and, in the short-term, lead to similar positive work outcomes such as higher productivity, organisational commitment, and long working hours [21]. On the other hand, in contrast to work engagement, burnout reflects a negative emotional state that is most dominantly characterised by general feeling and experience of extreme chronic exhaustion or draining of physical and psychological resources due to continuous exposure to demanding working conditions, which is reflective of low levels of pleasure and activation related to one’s work [18]. Within the circumplex model, workaholism and burnout share the dimension of pleasure because they are both characterised by low satisfaction and negative affect related to one’s work but are different in the level of activation at work. In sum, despite the evidenced distinctions between work engagement, burnout, and workaholism [7, 17], the relationships among the three might be more complex than expected and some conceptual similarities persist. Hence, the current study employs the psychometric network approach to provide deeper insight into the relationships between work engagement, workaholism, and burnout among academic employees in Montenegro.

The network approach [23] has become popular in the psychological sciences for studying mental disorders [24 –26] and, recently, it has been used in occupational behaviours for investigating the complex interplay of different occupational health dimensions such as job satisfaction, job crafting, and meaningfulness at work [28]. In brief, a network refers to a graph that consists of nodes, the observed variables, and edges that represent the relationships among the nodes [29]. In the network perspective, constructs are theorised as a network system of pairwise associations among variables where a change in one variable is associated to a change in the remaining variables or in the whole network [30, 31]. In this sense, work engagement, burnout, and work addiction are the nodes that are connected by edges [29]. Absent edges indicate zero partial correlations, whereas non-absent edges indicate the association between each two variables after controlling for all other variables [29, 32]. The network methodology may offer a more comprehensive representation of how work engagement, burnout, and work addiction are organised.

1.1 Montenegrin academic context

Since 2003, Montenegro’s higher education system has been transformed in accordance with European documents, policies, and legal agreements, common structural reforms, and shared tools as a member of the Bologna Process/European Higher Education Area (EHEA) and European Research Area (ERA) [33]. All of these transformations, in combination with other factors such as overall social and economic development, the pandemic, change in all spheres of life, digitalization, modernization of higher education, awareness of the need for quality improvement in teaching, introduction of state accreditation and quality control, and a stronger requirement of higher education to be more responsive to the needs of employers and employability of students, created a working context with multiple demands and sources of pressure for academic staff at the universities in Montenegro [33]. In addition, social pressure for universities to place highly on the international ranking lists is strikingly evident.

As a part of the current Law on Higher Education, the Council for Higher Education of Montenegro issued the Criteria on the Conditions and Requirements for Promotion of Teaching Staff Working at Higher Education Institutions (Official Gazette of Montenegro, No. 44/14 and 47/15). These criteria are related to selection, promotion to a higher position or re-election to the same position i.e. academic title.

Privatization of higher education is one of the most significant trends in education over the past few decades around the world [34]. As a mixture of national and international systems, the environment in which public and private higher education institutions operate and want to establish themselves clearly affects them in different ways [35]. Although accredited public and private universities operate under the same law and perform the same core teaching and research functions, they differ in how they go about doing them. Given the fact that there seem to be no differences in terms of core job functions and work being done by academic staff at public and private universities, we wanted to explore whether there are structural differences in experiencing burnout at work, workaholism, and work engagement.

The main objective of our study was to investigate relationships between work addiction and burnout with work engagement among academic workers in Montenegro. From a practical perspective, the prevalence of workaholism among academic workers is a growing concern, given the negative consequences of this condition for individuals and organizations. Academic institutions should recognize the signs of workaholism and provide support and resources to prevent and manage this condition.

2 Method

2.1 Participants and procedure

The population of the current study consisted of academic staff at public and private universities in Montenegro. According to the Statistical Office of Montenegro [36] the total number of academic employees in higher education institutions for the academic year 2021/2022 was 1,289, with men making up 678 and women 611. There is no official data on the precise number of teaching staff members working in public and private institutions. The current study is based on a convenience sample of 131 academic researchers from two universities in Montenegro (55% were from the public university, and 45% were from one private university, Table 1). In calculating sample size for cross-sectional network model, we used the Powerly package in R developed by Constantin and colleagues [37]. Specifically, for a network model of 6 nodes with a sensitivity of 0.6, a probability of 0.8, and a density of 0.3, a sample size of 153 was recommended.

Table 1
Sample characteristics for public and private universities

University

Sample characteristics Public Private Total

Gender % Men 27.5 36.2 31.5

Women 72.5 63.8 68.5

Relationship % Yes 80.3 71.2 76.3

Children % No 36.8 57.9 46.4

One or more 63.2 42.1 53.6

Education % Undergrad 1.4 13.6 6.9

Postgrad 33.3 40.6 36.6

PhD 65.3 45.8 56.5

Position % Associate^* 40.6 48.1 43.9

Assistant professor 20.3 18.5 19.5

Associate professor 18.8 13 16.3

Full professor 17.4 13 15.4

Researcher (all ranks) 2.8 7.4 4.9

Tenure (in years) M (SD) Academia 15.03 (7.53) 8.95 (8.03) 12.23 (8.31)

University 14.08 (8.22) 5.05 (3.75) 9.86 (7.91)

Current position^** 6.78 (7.15) 4.20 (3.39) 5.58 (5.84)

			University
Gender %	Men	27.5	36.2	31.5
	Women	72.5	63.8	68.5
Relationship %	Yes	80.3	71.2	76.3
Children %	No	36.8	57.9	46.4
	One or more	63.2	42.1	53.6
Education %	Undergrad	1.4	13.6	6.9
	Postgrad	33.3	40.6	36.6
	PhD	65.3	45.8	56.5
Position %	Associate^*	40.6	48.1	43.9
	Assistant professor	20.3	18.5	19.5
	Associate professor	18.8	13	16.3
	Full professor	17.4	13	15.4
	Researcher (all ranks)	2.8	7.4	4.9
Tenure (in years) M (SD)	Academia	15.03 (7.53)	8.95 (8.03)	12.23 (8.31)
	University	14.08 (8.22)	5.05 (3.75)	9.86 (7.91)
	Current position^**	6.78 (7.15)	4.20 (3.39)	5.58 (5.84)

Note. n = 131 ^*Associates together –associate with master, associate on PhD studies, associate with PhD. ^**Length of experience on the present position –calculated without full professors.

Participants took part in an anonymous online survey in June 2022. There were 66.40% women, 30.50% men, and 3.10% did not declare their gender. The majority had a PhD degree (56.50%). The majority were in the position of teaching associate (43.90%), whereas the smallest number held positions in research (4.90%). Three quarters of participants were in a relationship, and slightly over half of the sample had one child or more.

All measures were translated into Montenegrin using a translation–back translation procedure [38]. To ensure the quality of the questionnaire and the translation, the ReMO Pilot Survey research team went through several iterations of survey adjustment and fine-tuning. Through discussions, the team reached consensus regarding the final layout and wording of the survey.

2.2 Measures

2.2.1 Work engagement

We used the ultra-short version of the Utrecht Work Engagement Scale (UWES-3) [39], asking participants to rate their vigor, dedication, and absorption with one item each. The responses are provided on a 5-point scale ranging from 1 (never) to 5 (everyday). An example item is: “At my work, I feel bursting with energy” (α= .84).

2.2.2 Burnout

The work-related burnout subscale (7 items) from the Copenhagen Burnout Inventory (CBI) has been used [18]. It measures the degree of physical and psychological fatigue related to work (an example item: “Do you feel worn out at the end of the working day?”). The responses are provided on a 5-point scale, ranging from 1 (very low degree/never) to 5 (very high degree/always) (α= .90).

2.2.3 Workaholism

We used the 10-item Dutch Work Addiction Scale (DUWAS-10) [40]. The scale comprised two 5-item subscales measuring working excessively (e.g., “I seem to be in a hurry and racing against the clock”; α= .84) and working compulsively (e.g., “I feel obliged to work hard, even when it’s not enjoyable”; α= .84). The responses are provided on a 5-point scale ranging from 1 (almost never) to 5 (almost always).

2.2.4 Analytical strategy

We performed a confirmatory factor analysis (CFA) to test the psychometric properties of the measures using R package lavaan 0.6.14 [41] with the weighted least square mean and variance (WLSMV) adjusted estimator. We considered the following model fit indices and reference values: root mean square error of approximation (RMSEA) <0.06, Comparative Fit Index (CFI) >0.95 and Tucker-Lewis Index (TLI) >0.95 [42].

Network analyses were performed with R version 4.2.2 [43] and qgraph 1.9.3 package [44]. We followed steps described by Fried and colleagues [29] in estimating networks from multiple samples: (a) network estimation, (b) network stability, (c) network inference, and (d) network comparison. Furthermore, in reporting results in cross-sectional dataset, we followed guidelines developed by Burger and colleagues [45].

2.2.5 Network estimation

The fused graphic lasso (FGL) method and the EstimateGroupNetwork 0.3.1 package were used for jointly estimating the two networks (private and public universities) [46]. We averaged the layouts for the two individually estimated networks for the visualization of both networks. Finally, we used a spin-glass algorithm implemented in the igraph 1.3.5 package [47] for searching for clusters of nodes within the two networks.

2.2.6 Network stability

Stability of each network was investigated using the bootnet 1.5.0 package [29], with nonparametric bootstrapping and case bootstrapping based on 1000 bootstrap samples. Network stability was measured considering the correlation stability coefficient. Specifically, a correlation stability coefficient > 0.50 suggests good stability, and a correlation stability coefficient > 0.25 suggests acceptable stability [29].

2.2.7 Network inference

The bridge strength developed by Bereznowski and colleagues [27] was performed for assessing node centrality. In comparing node centrality of networks, we calculated Spearman correlation coefficients between both versions of the node strength for the networks. In estimating nodes’ predictability, we used the mgm 1.2.13 package [48]. In our study, node predictability represents the percentage of variance explained by all its neighbors (R²) [48].

2.2.8 Network comparison

In comparing networks, we calculated Spearman correlation coefficients of edge weights using the NetworkComparisonTest 2.2.1 package [49] with seed set to 1. In investigating whether all edges of the two networks were equal, the omnibus test was performed. If networks were significantly different, the post hoc test with Holm-Bonferroni method was performed for inspecting which edge weights were different between the two networks. Then, the networks’ global strengths were estimated for both networks and tested whether they differed. Finally, we estimated a cross-sample network (combining both samples into one general sample) to investigate the similarities between the networks. Furthermore, to investigate the differences between both networks, we assessed a cross-sample variability network where each network’s edge correspond to the standard deviation of this edge between the networks [29].

3 Results

3.1 Preliminary analyses

In the first step, we fit and compared a series of factor models. Specifically, we tested five different factor structures (1) a one factor CFA (all items form one general factor), (2) a three-factor CFA (burnout, items from working excessively and working compulsively dimensions were combined to form a single factor of workaholism, and work engagement), and (3) a four-factor CFA (burnout, working compulsively, working excessively, and work engagement). As shown in Table 2, results showed that the theoretical four-factor CFA showed the better fit to the data (CFI = .966, TLI = .962, RMSEA = .132, 95% CI = .120 –.144).

Table 2
Goodness of fit statistics of the measurement model

Model χ² Df CFI _(>0.95) TLI _(>0.95) RMSEA (90% CI)

One-factor CFA 1561.86 171 .882 .869 .244 (.233 –.255)

3-factor CFA 666.26 170 .958 .953 .146 (.134 –.158)

4-factor CFA 567.88 168 .966 .962 .132 (.120 –.144)

Model	χ²	Df	CFI _(>0.95)	TLI _(>0.95)	RMSEA (90% CI)
One-factor CFA	1561.86	171	.882	.869	.244 (.233 –.255)
3-factor CFA	666.26	170	.958	.953	.146 (.134 –.158)
4-factor CFA	567.88	168	.966	.962	.132 (.120 –.144)

Note: n = 131; one-factor CFA = all items load in one general factor; 3-factor CFA = (1) burnout, (2) items from working excessively and working compulsively dimensions were combined to form a single factor of workaholism, and (3) work engagement; 4-factor CFA = (1) burnout, (2) working compulsively, (3) working excessively, and (4) work engagement). χ² = chi-square, CFA = Confirmatory Factor Analysis; df = Degrees of freedom; CFI = Comparative Fit Index; TLI = Tucker-Lewis Index; RMSEA = Root Mean Square Error of Approximation, 90% CI = 90% confidence interval for RMSEA.

3.2 Network estimation

The jointly estimated networks for the two samples are presented in Fig. 1.

Fig. 1

The two regularized partial correlation networks estimated jointly for the two samples. Note. Line thickness indicates the strength of a relationship. The lighter gray area in the ring around a node represents predictability based on the proportion of explained variance (R2) by all of its neighbors, and the darker gray area in the ring around a node represents predictability based on the marginal distribution of a node. ABS = Absorption; BUR = Burnout, COM = Working Compulsively; DED = Dedication; EXC = Working Excessively; VIG = Vigor.

The network density was 0.80 (12/15 edges) for both networks. The mean absolute edge weights was 0.15, and 0.16 for network 1 (Private University), and network 2 (Public University), respectively. The spin-glass algorithm found the same two clusters in both networks. Specifically, the first cluster included the three dimensions of work engagement (vigor, absorption, and dedication), the second cluster included burnout, and both dimensions of workaholism (working excessively and working compulsively). Mainly, the cluster of work engagement was linked to the cluster of burnout and work addiction by a significant edge (see Fig. 1). The strongest edge was dedication —burnout.

3.3 Network stability

Stability analyses suggested that both networks were accurately estimated, with small to moderate confidence intervals around the edge weights. The correlation stability coefficients suggested acceptable stability [29] for network 1 = .51, and.44 for network 2.

3.4 Network inference

Concerning network 1, dedication was the most central node (unstandardized value = 1.25) and work compulsively was the least central node (unstandardized value = 0.67). Concerning network 2, dedication was the most central node (unstandardized value = 1.73) and work compulsively was the least central node (unstandardized value = 0.80). Spearman correlation coefficients of the standard version of the node strength were 0.83. For more details, see on Fig. 2.

Fig. 2

The unstandardized values of the standard version of the node strength (centrality) in the private and public networks. Notes: 1 = Vigor; 2 = Dedication; 3 = Absorption; 4 = Burnout, 5 = Working Compulsively; 6 = Working Excessively.

Predictability analysis showed that dedication was the most predictable variable (average predictability equaled 68.1%) and working excessively was the least predictable node (average predictability equaled 38.3%; see Fig. 1). Average predictability equaled 55.1% in network 1, and 51.3% in network 2.

3.5 Network comparison

In the omnibus test of the comparison, networks did not differ significantly (p = 0.169). We inspected edge differences, finding these two edges (13.3%) differed significantly: dedication-working excessively (p = 0.017), and absorption-working excessively (p = 0.017). Global strength did not differ significantly (p = 0.159), and its values were 2.48 and 3.35.

Figure 3a shows the cross-sample network with averaged edge weights, Fig. 3b the cross-sample variability network, the unstandardized values of the standard version of the node strength in the cross-sample network are presented in Fig. 3c, and the unstandardized values of the bridge strength in the cross-sample network are in Fig. 3d.

Fig. 3a

The cross-sample network. Note. The cross-sample network was obtained by pooling all observations into one sample. Solid lines represent positive edges, and dashed lines represent negative edges. Line thickness and darkness indicate the strength of a relationship. Note: ABS = Absorption; BUR = Burnout, COM = Working Compulsively; DED = Dedication; EXC = Working Excessively; VIG = Vigor.

Fig. 3b

The cross-sample variability network. Note. Edge weights represent the standard deviation of edge weights between the jointly estimated networks. Note: ABS = Absorption; BUR = Burnout, COM = Working Compulsively; DED = Dedication; EXC = Working Excessively; VIG = Vigor.

Fig. 3c

The unstandardized values of the standard version of the node strength in the cross-sample network. Note: ABS = Absorption; BUR = Burnout, COM = Working Compulsively; DED = Dedication; EXC = Working Excessively; VIG = Vigor.

Fig. 3d

The unstandardized values of the bridge strength in the cross-sample network. Note: ABS = Absorption; BUR = Burnout, COM = Working Compulsively; DED = Dedication; EXC = Working Excessively; VIG = Vigor.

The strongest edges connecting the variables were working compulsively—working excessively (edge_w = .48), vigor -dedication (edge_w = .38), and burnout-working excessively (edge_w = .35).

The correlation stability coefficient of the cross-sample network was 0.67 and beyond the recommended threshold (= 0.50) for stable estimation of centrality indices [29]. Node strength showed that dedication was the most central node (unstandardized value = 1.16), working compulsively was the least central node (unstandardized value = 0.48).

4 Discussion

Our study aimed to explore the relationships between workaholism, burnout, and work engagement among academics. Specifically, adopting a psychometric network perspective, we jointly estimated two networks and combined two samples of academics (private and public universities) into one to estimate the cross-sample network. Concerning private and public universities, both networks did not significantly differ, and there were only two differences in edge weights between the networks (edges between work engagement dimensions and working excessively). In this sense, the relationship between workaholism, burnout, and work engagement was not significantly different among public and private universities. Mainly, our results suggest that the relationship between workaholism, burnout and work engagement is context independent and is common in Academia. Furthermore, those results might confirm that, in Montenegro, as public and private universities are similar in core teaching and research functions [33], they did not differ in terms of academic staff’s mental health. Those results are in line with previous research [50, 51] that suggested that Academic culture (‘overtime culture’[52]) and deep changes in the academic working environment may have facilitated conditions prone to the development of workaholic behaviours and poor mental health.

Considering the jointly estimated networks, we observed two distinct clusters of nodes, (1) work engagement and (2) workaholism-burnout clusters. In the first cluster, only vigor and dedication showed a strong connection, whereas absorption was weakly linked to the other two dimensions of workaholism. In the second cluster, working excessively was strongly connected to burnout and working compulsively. No direct connection between working compulsively and burnout was found. Despite the bivariate correlation between workaholism and burnout is well known [7, 40], very little research has investigated this relationship adopting different measure, such as global burnout instead of measuring exhaustion, cynicism, and personal accomplishment. Moreover, it is possible that the relationship between working compulsively and job burnout would exhibit a delayed effect that manifests itself after some time has elapsed, then requesting for a longitudinal perspective.

Furthermore, work engagement cluster was connected to the other cluster through the negative edges between dedication and vigor with job burnout and working excessively, and the positive (weak) relationship between absorption and both dimensions of workaholism. Concerning the private university network, vigor and dedication nodes are not connected to any of the workaholism nodes, whereas absorption is positively linked to both dimensions of workaholism. It might indicate that “being fully concentrated and deeply engrossed in one’s work, whereby time passes quickly and one has difficulties with detaching oneself from work” (p.74) [17] share overlapping mechanisms of workaholism, especially for working excessively component [21]. Furthermore, vigor and dedication nodes are connected to job burnout, suggesting that having “high levels of energy and mental resilience while working, the willingness to invest effort in one’s work, and persistence even in the face of difficulties” and “experiencing a sense of significance, enthusiasm, inspiration, pride, and challenge“ (p.74) [17] are directly connected to job burnout. Concerning the public university network, vigor and dedication are directly negatively and positively linked to working excessively, respectively. Those results are in line with previous research [27, 53], suggesting that work engagement had direct relationship with workaholism, especially for working excessively component. Furthermore, absorption was (weakly) negatively linked to working excessively and positively linked to working compulsively. Those results are in contrast with previous research [21 , 53], suggesting that the relationship between absorption and workaholism need more investigation.

When we considered the cross-sample network, we observed the same two distinct clusters of nodes, where work engagement dimensions clustered in the first one, and workaholism components and burnout clustered in the second one. Both clusters were connected by the negative edges between dedication and vigor with job burnout. Interestingly, considering workaholism, working excessively was connected to burnout, and no significant edges were identified with work engagement dimensions. In general, those results are partially in line with previous research that showed how work engagement and workaholism are (weakly) correlated. In fact, in their systematic review and meta-analysis, Di Stefano and Gaudiino [21] showed that absorption has a medium-size association with working excessively (g = .34, 95% CI [.25,.43]), whereas only dedication showed a significant (g = .14 [.08,.21]) weak association with working excessively. In this sense, our results confirmed Di Stefano and Gaudiino [21] results about the not-overlapping concepts hypothesis, showing that workaholism and work engagement are distinct constructs. That is in line with Taris, Schaufeli, and Shimazu [49], who suggested that workaholism and work engagement are intrinsically different, and “engaged workers lack the typical compulsive drive that is characteristic of any addiction, including an addiction to work” (p. 51). Concerning studies that adopted a network perspective, our results were partially in line with Bereznowski and colleagues [27, 53], who found that work addiction and work engagement form separate clusters, and job burnout form another distinct cluster. In fact, we found that job burnout and workaholism form a cluster, suggesting that both variables share some unique variance. Furthermore, our results are in line with Bereznowski and colleagues [27, 53], who found a dense network characterized by weak edges between work engagement and workaholism (edge_w = .10 [.10,.10]).

Additionally, our results confirm the positive association of workaholism with occupational health risks, such as job burnout [54 –56]. In particular, people with workaholism patterns are characterized by a personal tendency to invest more energies into their jobs, taking on heavy workloads and spending more time at work, which can drain their personal resources such as physical and psychological energies, and personal/social life, and exposing workers to higher burnout risk [40 , 58].

The correlation stability coefficients indicated that node strength was stable for both jointly estimated networks. There are two strong bridge nodes in our study: dedication and working excessively. In this sense, we found that dedication and working excessively had the strongest connections with burnout. This suggests that promoting interventions aimed at increasing dedication might reduce job burnout, as dedication “refers to being strongly involved in one’s work, and experiencing a sense of significance, enthusiasm, inspiration, pride, and challenge” (p. 3) [17]. However, dedication shares with working excessively the characteristics of time and energy investment (for example, in terms of significance and enthusiasm) into the work, thus it might represent a potential additional risk for developing burnout. Moreover, both networks showed the same predictability. Mainly, dedication and working excessively were the most predictable nodes, whereas working compulsively was the least predictable node. This suggests that work engagement, workaholism, and burnout are a complex network of interrelated nodes where both types of employees who work excessively and are dedicated to their work are at risk of developing burnout [7].

Based on the main results of our study, potential interventions aimed at reducing workaholism can be proposed. Firstly, interventions aimed at promoting work engagement should focus on enhance sense of significance, enthusiasm, inspiration, pride, and challenge might directly reduce job burnout. At the same time, fostering a work environment that does not encourage work excessively, condemning a culture that emphasizes high investment in work, may directly reduce job burnout. It is important to remark that work excessively refers to the “behavioural component of hard working, which is spending too much time on work-related activities” [21]. In this sense, it is linked to the workaholic person rather than to organizational demands [21]. Then, interventions aimed at reducing job demands would not impact workaholics.

4.1 Limitations and future study directions

Although the current study provides valuable insights into the relationships between workaholism, work engagement, and burnout among academic workers in Montenegro, some limitations are to be considered when interpreting the findings. First, it is important to acknowledge that the data were collected in a single country context, and as such, the generalizability of the results to other academic settings and cultural contexts may be limited. However being more inclusive of the samples from middle and lower-income countries is good because these samples are generally underrepresented in wellbeing research. That may represent an important limitation in our analysis, as we did not control for these differences. Future research should aim to replicate the findings using data from multiple universities in different cultural contexts to offer a broader understanding of the relationships between workaholism, work engagement, and burnout among academic workers.

Second, despite the fact that the study was endorsed by all the public and private universities in Montenegro we obtained responses from the public university and one larger private university, but no participants were obtained from any other private university. In addition, the estimated response rate (18%) in the current study was relatively low, resulting in a small sample. In fact, our sample was smaller than the recommended sample size for cross-sectional network models. According to Constantin and colleagues [37], “for psychological networks, the necessity for making informed decisions about the sample size is further emphasized by the fact that the number of parameters that needs to be estimated increases rapidly with the number of variables included in the network” (p. 2). In this sense, future studies will have to consider a larger sample when replicating our results. Relatedly, due to limited sample size, we were not able to perform further comparisons among other sub-groups possibly existing within the broader population of academics in Montenegro. For example, in addition to comparison between the academics from public and private universities, it would have been interesting to assess the networks of the study variables across academic majors and disciplines. Future research, on larger populations and samples, should perform more fine-grained comparisons among different majors, disciplines, and other relevant groupings.

Third, the collected data is cross-sectional which does not enable us to test causal relationships between workaholism, work engagement, and burnout. Longitudinal data would be beneficial in assessing the temporal relationships among these constructs and determining whether workaholism and burnout predict changes in work engagement or vice versa. Moreover, in the context of the network approach, longitudinal data would have been used to assess the dynamic interplay between workaholism, work engagement, and burnout, allowing for a more comprehensive understanding of the relationships among these constructs over time.

5 Conclusion

In the present study, we investigated the network structure of workaholism, work engagement, and burnout within a sample of academics in Montenegro. Mainly, our results suggest that interventions should consider workaholism as an important burnout risk factor and dedication as a protective factor for academics. Furthermore, our contribution highlights that work engagement and workaholism are distinct and not directly connected, suggesting that interventions should be developed considering this empirical distinction.

Author contributions

All authors contributed to the study conception and design. Data collection: Sabina Osmanovic; Analysis: Igor Portoghese; Writing –original draft preparation: Igor Portoghese; All authors revised the draft critically and contributed to the writing of the manuscript. The final version of the manuscript was read and approved by all authors.

Footnotes

Acknowledgments

The authors acknowledge COST Action ReMO Survey SIG, in particular Stefan T. Mol, Jana Lasser and Carla da Silva Matos for their contribution. The authors are also grateful to Veselin Micanovic, Sandra Tinaj, Ana Maksimovic, and Darko Lacmanovic for their support in disseminating the survey.

Conflict of interest

The authors declare no conflict of interest.

Ethical approval

The study was approved by the Ethics Committee of the University of Amsterdam number 20220523100547.

Informed consent

Informed consent was obtained from all participants for being included in the study.

Funding

No funding was received to assist with the preparation of this manuscript.

References

Nicholls

, Nicholls

, Tekin

, Lamb

, Billings

The impact of working in academia on researchers’ mental health and well-being: A systematic review and qualitative meta-synthesis. PLOS ONE. 2022;17(5):1–29.

Guthrie

, Lichten

, Belle van

, Ball

, Knack

, Hofman

Understanding mental health in the research environment: A Rapid Evidence Assessment. Santa Monica, CA: RAND Corporation; 2017.

Urbina-Garcia

What do we know about university academics’ mental health? A systematic literature review. Stress and Health. 2020;36(5):563–85.

Ball

Neoliberal education? Confronting the slouching beast. Vol. 14, Policy Futures in Education. 2016. p. 1046–59.

Deem

The Early Stage Academic and the contemporary university: communities of practice versus new managerialism. In: Research Handbook on Academic Careers and Managing Academics. Edward Elgar Publishing; 2022. p. 405–19.

Ryan-Flood

, Gill

Breaking the silence: The hidden injuries of neo-liberal academia. In: Ryan-Flood R, Gill R, curatori. Secrecy and silence in the research process: feminist reflections. Abingdon, Oxon; NewYork,NY: Routledge; 2010. p. 228–44.

Schaufeli

, Taris

, Bakker

It takes two to tango: Workaholism is working excessively and working compulsively. The long work hours culture: Causes, consequences and choices. 2008;203–26.

Oates

Confessions of aworkaholic: The facts aboutwork addiction. New York: World. 1971.

Hogan

, Hogan

, Hodgins

A study of workaholism in Irish academics. Occup Med (Lond).. 2016;66(6):460–5.

10.

Leung wing

, Siu ling

, Spector

Faculty stressors, job satisfaction, and psychological distress among university teachers in Hong Kong: The role of locus of control. International Journal of Stress Management. 2000;7: 121–38.

11.

Bartczak

, Ogińska-Bulik

Workaholism and Mental Health Among Polish Academic Workers. International Journal of Occupational Safety and Ergonomics. 2012;18(1):3–13.

12.

Mazzetti

, Schaufeli

, Guglielmi

, Depolo

Overwork climate scale: psychometric properties and relationships with working hard. Journal of Managerial Psychology. 2016;31(4):880–96.

13.

Kinman

Effort-reward imbalance in academic employees: Examining different reward systems. International Journal of Stress Management. 2019;26(2):184.

14.

Kinman

, Wray

Presenteeism in academic employees—occupational and individual factors. Occupational Medicine. 2018;68(1):46–50.

15.

Clark

, Michel

, Zhdanova

, Pui

, Baltes

All work and no play? A meta-analytic examination of the correlates and outcomes of workaholism. Journal of Management. 2016;42(7):1836–73.

16.

Russell

Core affect and the psychological construction of emotion. Psychological Review. 2003;110(1):145.

17.

Schaufeli

, Salanova

, González-Romá

, Bakker

The measurement of engagement and burnout: A two sample confirmatory factor analytic approach. Journal of Happiness Studies. 2002;3: 71–92.

18.

Kristensen

, Borritz

, Villadsen

, Christensen

The Copenhagen Burnout Inventory: A new tool for the assessment of burnout. Work & Stress. 2005;19(3):192–207.

19.

Hakanen

, Bakker

, Schaufeli

Burnout and work engagement among teachers. Journal of School Psychology. 2006;43(6):495–513.

20.

Shimazu

, Schaufeli

, Kamiyama

, Kawakami

Workaholism vs.Work Engagement: the Two Different Predictors of Future Well-being and Performance. International Journal of Behavioral Medicine. 2015;22(1):18–23.

21.

Di Stefano

, Gaudiino

Workaholism and work engagement: How are they similar? How are they different? Asystematic review and meta-analysis. European Journal of Work and Organizational Psychology.. 2019;28: 329–47.

22.

Halbesleben

, Harvey

, Bolino

Too engaged? A conservation of resources view of the relationship between work engagement and work interference with family. Journal of Applied Psychology. 2009;94(6):1452.

23.

Borsboom

A network theory of mental disorders. World Psychiatry. 2017;16(1):5–13.

24.

Contreras

, Nieto

, Valiente

, Espinosa

, Vazquez

The Study of Psychopathology from the Network Analysis Perspective: A Systematic Review. Psychother Psychosom. 2019;88(2):71–83.

25.

Fried

Moving forward: how depression heterogeneity hinders progress in treatment and research. Expert Review of Neurotherapeutics. 2017;17(5):423–5.

26.

Robinaugh

, Hoekstra

RHA

, Toner

, Borsboom

The network approach to psychopathology: A review of the literature –and an agenda for future research. Psychological Medicine. 2020;50(3):353–66.

27.

Bereznowski

, Bereznowska

, Atroszko

, Konarski

Work Addiction and Work Engagement: a Network Approach to Cross-Cultural Data. International Journal of Mental Health and Addiction. 2021;1–25.

28.

Pimenta de Devotto

, de Freitas

CPP

, Wechsler

Basic Psychological Need Satisfaction, Job Crafting, and Meaningful Work: Network Analysis. Trends in Psychology. 2022.

29.

Epskamp

, Fried

A tutorial on regularized partial correlation networks. Psychological Methods. 2018;23(4):617–34.

30.

Fried

, Cramer

AOJ

Moving Forward: Challenges and Directions for Psychopathological Network Theory and Methodology. Perspectives on Psychological Science. 2017;12(6):999–1020.

31.

Schmittmann

, Cramer

AOJ

, Waldorp

, Epskamp

, Kievit

, Borsboom

Deconstructing the construct: A network perspective on psychological phenomena. New Ideas in Psychology. 2013;31(1):43–53.

32.

Epskamp

, Rhemtulla

, Borsboom

Generalized Network Psychometrics: Combining Network and Latent VariableModels. Psychometrika. 2017;82(4):904–27.

33.

NACRT STRATEGIJE VISOKOG OBRAZOVANJA CRNE GORE ZA PERIOD 2021–2025. (s.d.). [Internet]. Available at:https://www.gov.me/cyr/clanak/nacrtstrategije-visokog-obrazovanja-crne-gore-za-period-2021-2025.

34.

Brankovic

Positioning of Private Higher Education Institutions in the Western Balkans: Emulation, Differentiation and Legitimacy Building. In 2014. p. 121–44.

35.

Klemencic

The effects of Europeanisation on institutional diversification in the Western Balkans. In 2013. p. 117–38.

36.

Draft of Higher Education Strategy in Montenegro 2021–2025. Ministry of Education 2021.

37.

Constantin

, Schuurman

, Vermunt

A general Monte Carlo method for sample size analysis in the context of network models. Psychological Methods. 2023;

38.

Brislin

Back-Translation for Cross-Cultural Research. Journal of Cross-Cultural Psychology. 1970;1(3):185–216.

39.

Schaufeli

, Shimazu

, Hakanen

, Salanova

, De Witte

An Ultra-Short Measure for Work Engagement. European Journal of Psychological Assessment. 2019;35(4):577–91.

40.

Schaufeli

, Bakker

, van der Heijden

FMMA

, Prins

Workaholism Among Medical Residents: It Is the Combination of Working Excessively and Compulsively That Counts. International Journal of Stress Management. 2009;16(4):249–72.

41.

Rosseel

, Oberski

, Byrnes

, Vanbrabant

, Savalei

, Merkle

, et al. Package ‘lavaan’. Retrieved June. 2023;17(1).

42.

Schermelleh-Engel

, Moosbrugger

, Müller

Evaluating the Fit of Structural Equation Models: Tests of Significance and Descriptive Goodness-of-Fit Measures. 2003;8(2).

43.

R Core Team R. R: A language and environment for statistical computing. R Foundation for Statistical Computing [Internet]. 2023; Available at:https://www.R-project.org/.

44.

Epskamp

, Cramer

AOJ

, Waldorp

, Schmittmann

, Borsboom

qgraph: Network Visualizations of Relationships in Psychometric Data. J Stat Soft. 2012;48(4):1–18.

45.

Burger

, Isvoranu

, Lunansky

, Haslbeck

JMB

, Epskamp

, Hoekstra

RHA

, et al. Reporting Standards for Psychological Network Analyses in Cross-Sectional Data. Psychological Methods 2022.

46.

Costantini

, Epskamp

, Costantini

Package ‘EstimateGroupNetwork’. 2017.

47.

Csardi

, Nepusz

The igraph software package for complex network research. InterJournal, Complex Systems. 2006;1695(5):1–9.

48.

Haslbeck

mgm: Estimating Time-Varying k-Order Mixed Graphical Models [Internet]. 2023. Available at:https://CRAN.R-project.org/package=mgm.

49.

van Borkulo

, van Bork

, Boschloo

, Kossakowski

, Tio

, Schoevers

, et al. Comparing Network Structures on Three Aspects: A Permutation Test. Psychological Methods, 2022;

50.

Torp

, Lysfjord

, Midje

Workaholism and work–family conflict among university academics. Higher Education. 2018;76(6):1071–90.

51.

Guidetti

, Viotti

, Converso

The interplay between work engagement, workaholism, emotional exhaustion and job satisfaction in academics: A person-centred approach to the study of occupational well-being and its relations with job hindrances and job challenges in an Italian university. Higher Education Quarterly. 2020;74(3):224–39.

52.

Mazzetti

, Schaufeli

, Guglielmi

Are workaholics born or made? Relations of workaholism with person characteristics and overwork climate. International Journal of Stress Management. 2014;21(3):227–54.

53.

Bereznowski

, Atroszko

, Konarski

Work addiction, work engagement, job burnout, and perceived stress: A network analysis. Frontiers in Psychology. 2023;14: 1130069.

54.

Hartmann

, Mathieu

The relationship between workaholism, burnout and personality: a literature review. Sante Mentale Au Quebec. 2017;42(2):197–218.

55.

Cheung

, Tang

CSK

, Lim

MSM

, Koh

Workaholism on Job Burnout: A Comparison Between American and Chinese Employees. Frontiers in Psychology. 2018;9.

56.

Engelbrecht

, de Beer

, Schaufeli

The relationships between work intensity, workaholism, burnout, and self-reported musculoskeletal complaints. Human Factors and Ergonomics In Manufacturing. 2020;30(1):59–70.

57.

Demirel

, Erdirençelebi

The relationship of burnout with workaholism mediated by work-family life conflict: a study on female academicians. Journal of Language and Linguistic Studies. 2019;15(4):1300–16.

58.

Moyer

, Aziz

, Wuensch

From workaholism to burnout: psychological capital as a mediator. International Journal of Workplace Health Management. 2017;10(3):213–27.

			University
Sample characteristics		Public	Private	Total
Gender %	Men	27.5	36.2	31.5
	Women	72.5	63.8	68.5
Relationship %	Yes	80.3	71.2	76.3
Children %	No	36.8	57.9	46.4
	One or more	63.2	42.1	53.6
Education %	Undergrad	1.4	13.6	6.9
	Postgrad	33.3	40.6	36.6
	PhD	65.3	45.8	56.5
Position %	Associate^*	40.6	48.1	43.9
	Assistant professor	20.3	18.5	19.5
	Associate professor	18.8	13	16.3
	Full professor	17.4	13	15.4
	Researcher (all ranks)	2.8	7.4	4.9
Tenure (in years) M (SD)	Academia	15.03 (7.53)	8.95 (8.03)	12.23 (8.31)
	University	14.08 (8.22)	5.05 (3.75)	9.86 (7.91)
	Current position^**	6.78 (7.15)	4.20 (3.39)	5.58 (5.84)

Workaholism,work engagement,and burnout among academics in Montenegro: A psychometric network approach

Abstract

BACKGROUND:

OBJECTIVE:

METHODS:

RESULTS:

CONCLUSION:

Keywords

1 Introduction

1.1 Montenegrin academic context

2 Method

2.1 Participants and procedure

2.2.1 Work engagement

2.2.2 Burnout

2.2.3 Workaholism

2.2.4 Analytical strategy

2.2.5 Network estimation

2.2.6 Network stability

2.2.7 Network inference

2.2.8 Network comparison

3 Results

3.1 Preliminary analyses

Table 2 Goodness of fit statistics of the measurement model Model χ2 Df CFI (>0.95) TLI (>0.95) RMSEA (90% CI) One-factor CFA 1561.86 171 .882 .869 .244 (.233 –.255) 3-factor CFA 666.26 170 .958 .953 .146 (.134 –.158) 4-factor CFA 567.88 168 .966 .962 .132 (.120 –.144)

3.4 Network inference

4.1 Limitations and future study directions

5 Conclusion

Author contributions

Footnotes

Acknowledgments

Conflict of interest

Ethical approval

Informed consent

Funding

References

Table 2
Goodness of fit statistics of the measurement model

Model χ² Df CFI _(>0.95) TLI _(>0.95) RMSEA (90% CI)

One-factor CFA 1561.86 171 .882 .869 .244 (.233 –.255)

3-factor CFA 666.26 170 .958 .953 .146 (.134 –.158)

4-factor CFA 567.88 168 .966 .962 .132 (.120 –.144)