Dimensionality of burnout – Is the Mini Oldenburg Burnout Inventory suitable for measuring separate burnout dimensions?

©

Theoretical background: Research on burnout has widespread interest in mental health sciences.The Demands-Resources Model of Burnout represents a new direction in research.Similarly to the most popular model of burnout, the Multidimensional Model, it embeds burnout in an organizational framework.Goals: The purpose of this study is to analyze the dimensionality and divergent validity of the Mini Oldenburg Burnout Questionnaire (MOLBI), the measurement tool of the Demands-Resources Model.Our work also tests the multidimensional theory, which posits that the burnout dimensions are independent of one another and do not form a global burnout factor.Method: Participants' (n = 406 people, 59.1% female) mean age was 39.4 (SD = 11.06) years and most of them had a graduate degree.In addition to the MOLBI questionnaire, participants completed the Work-Family Balance and Mental Health Test.We conducted a parallel analysis and exploratory factor analysis to analyze the structure of MOLBI; a bifactor analysis and model-based reliability test to analyze the validity of global and specific factors.We assessed the relationships between MOLBI and other parameters with Kendall's tau-b correlation.Results: Our results showed that the original two-factor structure of the questionnaire fit our data best (χ 2 = 78.489,DF = 26, p < 0.001; CFI = 0.977; NNFI = 0.960; RMSEA = 0.068; RMSEA CI 90 :0.066-0.070).The reliabilities of factors and global scores were adequate.Besides, the bifactor analysis showed that the global burnout dimension and disengagement subscale had enough explanatory power.The exhaustion factor of

Introduction
Burnout is a widely researched construct in the social and health sciences.The phenomenon, which became a common research topic in the 1970s, has taken several approaches in the last forty years.Multidimensional- (Maslach & Jackson, 1981), existential-(Pines, 1993), and conservation of resources theories (Shirom, 2003) are well known in the international literature.These approaches highlight three important characteristics of the phenomenon.Trends agree that burnout is (1) characteristic of people who have no other mental illness.(2) It usually occurs in a workplace situation and ( 3) it has a broad spectrum of negative emotions.However, there is no consensus on whether it has motivational, physiological, and performance-related aspects too (Maslach & Jackson, 1981;Shirom, 2003).Also, controversial theories have arose if burnout is related only to work situations (Maslach & Jackson, 1981) or it appears in other areas of life (eg.marriage or parenting; Pines 1993).
Recently, a new theory has appeared in burnout literature, which attempts to unify the common points of previous theories.It highlights and integrates parts that can be useful elements for further detailed burnout testing and provides a basis for diagnostic practice.The Burnout Demands-Resources Model (BDR Model) was developed by Demerouti and her coworkers (2001).According to the model, burnout is due to relatively high work requirements and the limited resources available to manage them.The limited ability to cope with the resources leads to negative emotional states.Thus, the discrepancy between resources and challenges creates a significant negative emotional state.Therefore, the BDR Model introduces the exhaustion dimension that characterizes work-related emotional and physical fatigue.In addition, the model is complemented by disengagement, which maps the motivational background of how open an individual is to deal with a discrepancy.Thus, the disengagement dimension measures the lack of work-related motivation (Bakker, Demerouti, & Verbeke, 2004;Demerouti et al., 2001).The BDR Model narrows the definition of burnout and addresses only these two dimensions of burnout, omitting the reduction in performance that appears in the multidimensional model.Some researchers state that performance loss is not necessarily part of the burnout phenomenon.Some authors view it as a consequence of burnout (Shirom, 2003), while others view it as a personality trait (as part of self-efficacy) that does not characterize the transient state of burnout (Cordes & Dougherty, 1993).Demerouti and Bakker (2008) introduce refinements to the multidimensional model not only at the theory level but also at the measurement model level.The dimension of exhaustion and disengagement is characterized by simple and clearly-worded items, which can also be very helpful in translating the questionnaire into any other language.As a result of conceptual and semantic clarification, the Oldenburg Burnout Questionnaire (OLBI; Demerouti & Bakker, 2008) has been created, which contains 16 items.Eight items to describe exhaustion, eight to describe disengagement.The authors also took care to include both straight and reversely worded items in both dimensions.As opposed to Maslach's approach, which only describes the dimensions of emotional exhaustion and cynicism with negative terms, while using only positive items to characterize personal performance.The downside of this solution is that when analyzing the factor-structure of a measurement tool, one-way items, by their very wording, are condensed into a single factor, so they do not necessarily show the properties of the characterized dimension (Demerouti & Bakker, 2008).Therefore, in the Oldenburg Burnout Questionnaire, each scale has both straight and reverse items.
The factor structure of OLBI has been studied in several studies (Campos, Carlotto, & Marôco, 2012;Halbesleben & Demerouti, 2005;Mahadi et al., 2018;Schuster & Dias, 2018;Sedlar, Šprah, Tement, & Sočan, 2015;Subburaj & Vijayadura, 2016).Some confirmed the original version of the questionnaire (Halbesleben & Demerouti, 2005;Subburaj & Vijayadura, 2016), but many had to remove items from the measurement tool or keep only negative items because positive items generated response bias (Campos, et al., 2012;Mahadi et al., 2018;Schuster & Dias, 2018;Sedlar et al., 2015).To use the questionnaire more effectively, Thun, Fridner, Minucci and Løvseth have developed an abbreviated version of OLBI, the Mini Oldenburg Questionnaire (MOLBI; Thun et al., 2014).This version of the questionnaire turned out to be more applicable in intercultural and longitudinal studies.MOLBI also measures exhaustion and disengagement.Each factor contains 3-3 straight and 2-2 inverted items.During the Hungarian adaptation of OLBI and MOLBI, comparative studies showed that the factor structure of MOLBI shows a better fit in the domestic environment (Ádám et al., 2020).Therefore, the present study aims to analyze MOLBI on a heterogeneous Unauthenticated | Downloaded 09/14/23 03:23 PM UTC sample, which is important both because it analyzes the applicability of the questionnaire outside the health sector and on the other hand provides further evidence for the two-factor structure.Based on the above, we assume that: H1: MOLBI has two dimensions, named exhaustion and disengagement.
We aimed to confirm the two-factor structure of MOLBI on our sample.We also aimed to examine if exhaustion and disengagement constitute independent or correlated factors.In our previous study, we found that the two factors of MOLBI show a relatively high correlation (r = 0.810, p < 0.001) (Ádám et al., 2020).Also, some burnout studies question the independence of the subscales in the multidimensional approach.For example, regarding the Maslach Burnout Inventory (MBI; Maslach & Jackson, 1981), researchers often note that exhaustion and the other major burnout dimension (called: depersonalization, cynicism, or disengagement) are correlated at a medium or high level (Gil-Monte, 2005;Mäkikangas, Hätinen, Kinnunen, & Pekkonen, 2011;Vanheule, Rosseel, & Vlerick, 2007).Moreover, certain studies question the independence of the emotional exhaustion dimension, and show that this dimension is more likely to merge into "non-specific psychological distress" (Schonfeldt, Verkuilen, & Bianchi, 2019) factor and correlate more intensively with depression than with other burnout subscales (Bakker et al., 2000).
The other aim was to measure how subfactors and how global factor could explain the total variance of variables and how much specific information is provided by subfactors.In our previous study, we found that global burnout scores explain significantly more of the variance than the specific factors (such as emotional exhaustion and depersonalization dimensions; Mészáros, Ádám, Szabó, Szigeti, & Urbán, 2014).We have also found that burnout can be positioned within the multidimensional space of mental problems, as certain burnout characteristics overlap with characteristics of depression and anxiety.However, research needs to unfold the syndrome-specific characteristics of burnout.
Testing other forms of validity can also guide the meaning of the subscales.Therefore we would like to test MOLBI in the relationship to other constructs.As previously described, the emotional component of burnout is most commonly researched.Some studies highlight the pervasive nature of negative feelings related to fatigue.Thus, they experience negative emotions during work that can affect other areas of life, such as family life or leisure activities (Westman, 2001, Westman & Vinkour, 1998).Bakker, Demerouti & Schaufeli (2005) narrow his analysis to the emotional component of burnout.It describes how emotional exhaustion affects positive feelings about privacy, such as happiness or satisfaction.Therefore, it can be assumed that: H2: Burnout and its subscales are in correlation with work-family balance and mental health.

Sample
In our previous study about the factor structure of MOLBI, we used a relatively homogenous sample of health care workers (Ádám et al., 2020).Therefore, we aimed here to test the questionnaire in a relatively heterogeneous sample.We have analyzed only complete questionnaires, there were no missing values.Table 1 shows the demographic profile of the sample.Our sample consisted of 406 people, 240 female (59.1%) and 166 male (40.9%).The mean age was 39.39 (SD=11,06) years.They mainly have master's degree (n = 132, 32.5%) and are typically married (n = 192, 47.3%).The inclusion criteria of the study were: 1) age 18 or above, and 2) having no known mental disorder within the past year.It is a 20-item questionnaire.The items are rated on a 6-point Likert scale, ranging from 1 ('not typical at all') to 6 ('typical at all'), so the higher scores indicate better mental health.The meaning of the subscales are as follows: 1) Well-being: the feeling of well-being in the emotional, psychological, social, and spiritual aspects of life.2) Savoring: positive cognition processes of life in service of well-being.3) Creative, executive individual, and social efficiency: collection of the strategies and actions, which help the individual to change the circumstances in a difficult adjustment situation, to achieve the goals and plans.4) Self-regulation: the control of emotions, attention, and cognition.5) Resilience: psychological flexibility and resistance (Vargha et al., 2020;Zábó & Vargha, 2019).The reliability of the scales in the Hungarian sample are WB = 0.784; S = 0.747; CE = 0.719; SR = 0.815; R = 0.777 (Zábó & Vargha, 2019).
Work-Family Balance Questionnaire (WFB; Carlson et al., 2006;Ádám et al., 2019) measures the global balance of work and family.This questionnaire gives a system-wide perspective that goes beyond the individual.In Hungarian, it is a 5-item scale, which measures the balance in a 5-point Likert-scale, ranging from 1 ('totally disagree') to 5 ('fully agree)', with a Cronbach's α = 0.919.
Table 2 represent the descriptive statistics, normality test results, and internal reliabilities.The latter is in the acceptable to good range.As it is seen, their reliability is in the optimal range -the Cronbach's α between 0.7 and 0.8 is acceptable and between 0.8 and 0.9 is good -, but their normality is violated in all cases (Cohen, 1996).

Procedure and statistical analysis
We chose an online method to reach a relatively heterogeneous sample for the study.We collected our sample with the help of students of the KRE Institute of Psychology, and used the snowball sampling method.The demographical questions and the questionnaires were administered in a Google form.The answers we collected were voluntary and anonymous.
The plan of the study was accepted by the ethical committee of the Reformed Church of Károli Gáspár University (ethical numbers are 406/2018P, 407/2018P), the study started in January, 2019 and ended in April, 2019.Participants agreed to participate in the study in an online form.
For the statistical analysis, we used FACTOR (Lorenzo-Seva, 2019), Mplus 8.4.(Muthén & Muthén, 2019), Omega Calculator (Watkins, 2013) and ROPstat (Vargha, 2019).To test the first hypothesis, we calculated parallel analysis and exploratory factor analysis with FACTOR.Additionally, we tested the bifactor model with Mplus 8.4., and calculated omega and omega hierarchical with Omega calculator.Lastly for testing the second hypothesis, we used Kendall's tau-b correlation with ROPstat.
In the study, we marked the Pearson correlation with "r", and the fit indices of the factor analysis with χ², RMSEA, CFI, TLI, NNFI.The optimal range of the fit indices is: CFI, NNFI and TLI above 0,95.The RMSEA under 0,05 is excellent; under 0,08 is good and above 0,10 is not acceptable.Besides these parameters, not significant χ² is really good, but in bigger sample it's rather preferably, than rule (Bentler, 1990).
We calculated model-based reliability for the bifactor model.The omega coefficient reflects the proportion of variance of the scale score that is attributable to the global and specific factors.For example, in the disengagement scale, score coefficient omega indicates the proportion of variance that is attributable to the global and the specific disengagement factor.In contrast, the omega hierarchical indicates how much variance is attributable to only the specific factor (e.g.disengagement factor).Reise, Bonifay and Haviland (2013) suggest that the value of the omega hierarchical is preferable around 0.50, but Szigeti and her coworkers (2016) use lenient criteria.They define the coefficient around 0.20, because in this case, the specific factor explains at least 20% of the variance in specific factor scores.
We use p = 0.05 as the upper limit of statistical significance and report its exact value unless it is less than 0.001, in which case we present it as p < 0.001.

Results
At first, we would like to test our first hypothesis, if MOLBI has two factors.We tested our sample with an exploratory factor analysis and a parallel analysis.We used the Maximum Likelihood method.This is a Pearson correlation-based method.According to Baglin (2014) and Vargha (2019), this solution is advisable if the skewness and Kurtosis of the variables are under 1 in every case.In our sample, all items of MOLBI fulfilled this criterion (see Table 3).We used oblique rotation with a direct oblimin method, because this rotation properly evaluates the factor loadings if fewer complex variables (variables, which load onto more than one factor) are present and it is advisable if there is a medium or high correlation between the factors (Lorenzo-Seva, 1999;Vargha, 2019).
The parallel analysis advised that one-factor solution fits the best for the questionnaire.But the fit indices of this solution were worse (χ 2 = 285.488,DF = 35, p < 0.001; CFI = 0.927; NNFI = 0.906; RMSEA = 0.105; RMSEA CI 90 :0.100-0.110)than the two factor solution (χ 2 = 78.489,DF = 26, p < 0.001; CFI = 0.977; NNFI = 0.960; RMSEA = 0.068; RMSEA CI 90 :0.066-0.070).This usually can happen, when the eigenvalue of the second factor is around one, therefore the one-factor solution is statistically more economical than the two-factor solution.In our sample, the eigenvalue of the first factor was 3.932, and the second was 1.507.Additionally, the two-factor solution nearly matched the original model (refer to Table 4).Only item 3 ('talk about my work in derogatory way') loaded equally in both factors.Exhaustion factor can be described by the following items: 'after work I feel worn out', 'I feel tired before work'.The items with the highest factor loading are: 'lost internal relationship to work' and 'I think less and execute tasks mechanically'.A medium correlation were found between the two factors (r = 0.487).In the next step, to test the legitimacy of the two specific factors, we ran a bifactor model for the two-factor solutions.We examined how much variance was attributable to the global factor and how much variance was attributable to the specific factors.The fit indices of the bifactor model were in the proper range (Satorra-Bentler χ 2 = 67.852,DF = 23, p = 0.001; CFI = 0.962; TLI = 0.926; RMSEA = 0.068; RMSEA CI 90 :0.050-0.087).The model-based reliability (omega and omega hierarchical) indices showed that the disengagement factor explained 47% of the variability alone, the exhaustion factor explained only 16.1% (see Table 5).The reliability of the two scales in the sample was exhaustion α = 0,870; disengagement α = 0,740, which also underlines the adequacy of the structure.
Finally, we wanted to test the MOLBI's divergent validity from mental health and work-family balance constructs.The Kendall's tau-b correlations are shown in Table 6.Exhaustion had more medium correlations with particular aspects of mental health and work-family balance.The global burnout, and also its subscales showed a relatively large difference from mental, physical, and psychological aspects of well-being.In the case of disengagement well-being and creative, executive individual and social efficiency showed a medium correlation, however, exhaustion had a medium negative relationship with resilience, well-being, and work-family balance.Thus, being more exhausted is linked to less work-family balance and more stressful life situations.

Discussion
Burnout, similar to other psychological problems, contains specific elements, which can be described with emotional exhaustion, depersonalization, and reduced personal accomplishment.Besides the specific elements, there are some general features, which could overlap with other mental health problems (e.g. in the case of burnout with depression, or anxiety) according to the dimensional approach of mental health problems.
One of our previous study, we have studied the dimensionality of MBI-HSS, and we have found that only the overall burnout factor and the specific reduced personal accomplishment factor had high enough explanatory power (Mészáros et al., 2014).Although Maslach and Jackson (1981) emphasize from the 1980s that burnout has distinct, non-additive dimensions, Hungarian researches either with MBI-HSS (Mészáros et al., 2014) or MBI-ES (Szigeti et al., 2017) have not confirmed this.Moreover, in our other previous research on studying the structure of MOLBI, we concluded that exhaustion and disengagement dimensions were highly correlated (Ádám et al., 2020).These leave the question open on whether the dimensions have individual added value or whether the questionnaire measures a global burnout dimension just like the various forms of MBI.
This study aimed to unfold the underlying factor structure of the MOLBI questionnaire and to test its validity.Our results showed that the correlation between the factors is moderate and the bifactor model confirmed the separate explanatory power of the specific disengagement factor.Disengagement has significantly higher self-explanatory variance than the exhaustion factor.This is close to the criteria suggested by Reise, Bonifay and Haviland (2013) and high above the criteria suggested by Szigeti and coworkers (2017).In conclusion, MOLBI is a suitable instrument for measuring global burnout and specific disengagement subscale, so the questionnaire contains elements specific to disengagement and contains global burnout elements as well.However, exhaustion factor does not provide much specific information.
We also explored the associations of MOLBI with mental health and work-family conflict.As described by Bakker et al. (2005), burnout has an impact on positive feelings in other areas of life.Therefore, we hypothesized that burnout and also its subscales correlate with these constructs.Our results showed that burnout has a negative relationship with different aspects of mental health and work-family balance.The dimension of exhaustion -characterized as an emotional and physical aspect -shows a similar pattern to global burnout.This confirms that (1) the exhaustion dimension is not separated from the whole construct, and (2) its relationship with the work-family balance and mental health dimensions has a complex (emotional and physical) nature.The disengagement's connection with creative, executive individual, and social efficiency confirm the motivational and behavioral aspects of the dimension.
Our study has some shortcuts.Our sample consists of only 406 people.We have administered our questionnaire in an online way with snowball sampling, which can affect the quality of the responses and select the sample in a specific way (to computer users, which significantly limit the generalizability of the results).Only Hungarian people filled in our questionnaires so our sample is culturally homogeneous, results can be conditionally generalized to other cultures.But we believe this analysis can be a first step to introduce a methodically and semantically well designed burnout questionnaire.

Conclusions
The focus of our study was to examine whether the MOLBI questionnaire can be used as a whole and whether each dimension -disengagement and exhaustion -has separate meaning as well.We also tested the divergent validity because we wanted to get closer to their meaning.Our analysis showed that the burnout dimension can be used as a whole, and the disengagement dimension also has a significant amount of separate meaning.When testing the divergent validity, it was seen that the relationships of the exhaustion dimension was very similar to the relationships of the global burnout scale, which underpin the similarity to the global phenomenon, and it also became apparent that this dimension covered not only the emotional aspects of the phenomenon but more than that.The relationships of the disengagement dimension suggest that this is primarily a motivational and behavioral aspect of burnout.

Table 1 .
Vargha et al., 2020;ám et al., 2020)denburg Questionnaire (MOLBI;Thun et al., 2014;Ádám et al., 2020)is a 10-item scale, which measures the two aspects of the burnout.Its first aspect is exhaustion, which is physical and emotional tiredness connected to work.The second aspect is disengagement, which contains a lack of motivation in work and intensive depersonalization.Five items belong to each scale, and two of them are reversed.Responses are rated on a 4-point Likert-scale, ranging from 1 ('totally disagree') to 4 ('totally agree').The higher score indicates more burnout symptoms.The reliability (i.e.Croncabch's α) of the scales in the health care workers' sample is: Exhaustion = 0.790; Disengagement = 0.762(Ádám et al., 2020).Mental Health Test (MHT;Vargha et al., 2020; Zábó & Vargha, 2019) measures the emotional, psychological, social, and spiritual characterizing of the person.The four aspects of the questionnaire are measured by 5 scales, named: Well-being (SWB), Savoring (S), Creative, executive individual and social efficiency (CE), Self-regulation (SR), and Resilience (R).

Table 2 .
Unauthenticated | Downloaded 09/14/23 03:23 PM UTC The Mini Oldenburg Burnout Inventory Parameters of the used scales in our sample

Table 4 .
Item labels and standardized factor loadings of the MOLBI items

Table 6 .
Correlations between burnout, mental health and work-family balance (Kendall's tau-b) Only self-regulation and disengagement has p = 0.029 correlation, the other correlations' significance are p < 0.001.