Polish Psychological Bulletin

Original Papers 2023, vol. 54(3) 207–216

DOI: 10.24425/ppb.2023.148053

Teresa Chirkowska-Smolak1 , Tomasz Górecki1 , Mateusz Klakus1

Weronika Metzger1, Martyna Szargan1
1
Adama Mickiewicza University, Poznań, Poland
CORRESPONDING AUTHOR - Teresa Chirkowska-Smolak, chirko@amu.edu.pl

The factorial validity of the Maslach Burnout Inventory:

Student Survey (MBI-SS) in Poland

ABSTRACT: This paper explores the psychometric properties of the Polish adaptation of the MBI-SS questionnaire,
which measures academic burnout in students. The factorial structure of the tool was examined, and its validity was
evaluated using a sample from Polish state universities (N=935).
Findings suggest that the MBI-SS possesses strong psychometric parameters. Data supports the tool's 3-factor
structure and internal consistency. However, there are reservations regarding one item (MBI_6). Consequently, the 15-
item version of the MBI-SS is recommended as a suitable tool for assessing burnout in Polish students.

Keywords: students, burnout, validation, Maslach Burnout Inventory

INTRODUCTION Maslach, 2009). The most popular concept of burnout is

a multidimensional model developed by Christina Ma-
For many years, the phenomenon of professional slach, which has been widely supported by empirical
burnout has been recognized as a syndrome whose causes research, and it is on this approach that the WHO bases its
lie in the emotionally demanding contact of employees definition. Maslach initially defined burnout as a “psycho-
with other people, e.g., with patients, students, or clients. logical syndrome involving emotional exhaustion, deper-
Over time, it has been recognized that burnout not only sonalization, and a diminished sense of personal accom-
affects people working in social professions but is also plishment that occurred among various professionals who
a universal problem (Maslach, Schaufeli, Leiter, 2001) and work with other people in challenging situations”
can even affect people who are not employees (Schaufeli, (Maslach, 1993, p. 19) and emphasized that burnout
Martinez, Pinto, Salanova & Bakker, 2002). Professional involves a loss of concern for another person. Currently,
burnout syndrome has been included in the latest WHO a more universal concept of burnout is used, since it has
Classification of Diseases and Disorders, ICD-11, as been recognized that it is a phenomenon that can affect
a problem related to professional functioning, which representatives of various professions, not necessarily
highlights the growing recognition of burnout’s signifi- social ones. To describe the three basic dimensions of
cance and its potential impact on both individuals and the burnout, Maslach and colleagues used modified descrip-
broader community. tions of burnout symptoms — emotional exhaustion was
The issue of burnout is an important research topic in replaced with exhaustion in general, which refers to both
psychology and has been intensively studied since it was emotional functioning and the physical condition of
first presented in the psychological literature in the mid- an individual; depersonalization was replaced with the
1970s. In 2009 Schaufeli, Leiter, and Maslach estimated more general term “cynicism”, which was understood
that over 6000 books, chapters, dissertations, and journal as a distanced attitude toward work, while the sense
articles had been published on Burnout (Schaufeli, Leiter, of lack of personal achievements was narrowed down

© 2023. The Authors. Published by Committee for Psychological Science of Polish Academy of Sciences. This is an open
access article under the CC BY-NC-ND license (https://creativecommons.org/licenses/by-nc-nd/3.0/).
 The factorial validity of the Maslach Burnout Inventory: Student Survey (MBI-SS) in Poland 208
to the sense of reduced professional efficacy (Maslach, of the two-factor Oldenburg Burnout Inventory (Campos
Leiter, 2008). et al., 2012) or the Copenhagen Burnout Inventory
Exhaustion is not simply felt; rather, according to (Campos et al., 2013). However, they are not as widely
Maslach and Leiter (2008), it leads to actions that distance used as that developed by Schaufeli et al. (2002) MBI-SS.
the individual emotionally and cognitively from work, This article presents a questionnaire that may prove useful
arguably a way of coping with work overload. Attempts to in the analysis of academic burnout among Polish students.
distance oneself from various aspects of work (cynicism, The questionnaire is a modification of the world’s most
depersonalization) are a direct response to exhaustion. This popular test for measuring academic burnout, i.e., the
attitude manifests itself in negative, cynical, and overly Maslach Burnout Inventory (MBI-GSS).
defensive distancing and loss of idealism. The third
component of burnout reflects the employee’s self-esteem. METHOD
It refers to feelings of incompetence, lack of professional
achievement, and low productivity. Research problems and hypotheses
With the increasing knowledge of the determinants The main objective of the study was to prepare
and mechanisms of burnout, researchers have pointed out a Polish adaptation of the tool to measure academic
that people burn out not only in professional settings burnout in students, i.e., the MBI-SS, which is a special
(Stawiarska, 2016). Students were also identified as version of the world’s most popular questionnaire to
a particularly vulnerable social group (Hu, Schaufeli, measure professional burnout. Although the results of the
2009). Even though formally speaking, the academic validation of this tool in different countries have already
activity of students is not professional work, it can be said, been published, they show that further work is needed to
from a psychological point of view, that studying is a type confirm its value, as single items of this questionnaire raise
of work because it is an organized and preimposed, doubts among researchers and are eliminated from the
structured and goal-oriented activity that is subject to final version in a given language (Schaufeli et al., 2002;
external evaluation. Schaufeli and his team (Schaufeli Hu and Schaufeli, 2009; Portoghese, Leiter, Maslach,
et al., 2002) proposed that academic burnout could be Galletta, Porru, D’Aloja, Finco, Campagna,.., 2018).
defined analogously to occupational burnout, as suggested Above all, there is no Polish version that could be used
by the authors: “burnout refers to feeling exhausted in research in the area of mental health and well-being of
because of study demands, having a cynical and detached students. Therefore, the main research problem was to
attitude toward study, and feeling incompetent as a student” confirm the good validity of this tool in the Polish version
(p. 465). and to determine whether the three-factor structure is also
However, despite the domination of the Maslach applicable in Poland.
model of burnout in the field, a debate among scholars on The Maslach Burnout Inventory (MBI) questionnaire
the nature of burnout continues. This debate revolves for measuring burnout by Maslach, Jackson, Leiter,
around two interrelated issues: the dimensionality of Schaufeli, and Schwab was developed in 1981 (Maslach,
burnout and its scope (Schaufeli, Leiter, Maslach, 2009). Jackson, Leiter, 1996). There are several versions of this
There is no consensus on student burnout among tool, designed to survey people from different professions:
researchers: some researchers assume a three-factor for medical personnel (MBI-PM), for social professions in
conception of burnout (Salgado & Au-Yong-Oliveira, general (MBI-HSS), for teachers (MBI-ES) and a general
2021; Schaufeli et al., 2002), and others assume that version (MBI-GS) designed to study this phenomenon
burnout consists of only the two key factors of exhaustion among representatives of all professions. They differ not
and cynicism, excluding the third factor of reduced only in the indication of the context to which they refer but
professional efficacy (Alarcon et al., 2011). As for the also in their length (e.g., the PM version has 22 items,
scope of burnout, some scholars consider burnout as while the GS version has 16). All versions of the MBI
a generic, context-free phenomenon that may occur questionnaire are licensed by MindGarden.
outside work, not only work-related (Kristensen et al., However, the publisher does not have the Polish
2005). adaptation of the MBI-SS student burnout measurement
The concept of burnout may be applied to activities tool. The latest addition to the MBI group was developed
that are similar to work but are outside the occupational by Schaufeli, Martinez, Pinto, Salanova and Bakker
context, such as, for example, in studies. As Schaufeli and (2002). It is a modified version of the general ques-
Taris note, such activities may be exhausting and may also tionnaire for all occupations (MBI-GS), where for
allow people to withdraw from them (2005, p. 260). In the example, the term “work” has been replaced by “study”.
last decade, there has been a growing interest among The questionnaire is used in different countries in a 16-
psychologists in student academic burnout (Portoghese item version (as in the original) or a 15-item version,
et al., 2018; Rostami et al., 2014; Salgado & Au-Yong- omitting one test item from the cynicism scale. This is the
Oliveira, 2021; Schaufeli et al., 2002; Simancas-Pallares MBI_13 item, which was also questionable in the Polish
et al., 2017; Turhan et al., 2021). Many of these adaptation of the MBI-GS developed by Chirkowska-
researchers have struggled to find (or develop) a good Smolak and Kleka (2011).
measurement tool. To assess burnout in students, various General information on the student version is
instruments have been used, such as the students’ version presented in the current test manual (the 4th edition of
209 Teresa Chirkowska-Smolak, Tomasz Górecki, Mateusz Klakus, Weronika Metzger, Martyna Szargan

the Manual) of 2017, but the authors emphasize that its cacy is observed much less frequently (Lee & Ashforth,
psychometric properties has not yet been sufficiently 1996; Brennikmeijer, Van Yperen, 2003). This dimension
documented. However, since the development of the scale was, in fact, later included in the model (Maslach, 1993).
and the first validation studies in three countries, Spain, Some authors suggest that professional efficacy is related
Portugal, and the Netherlands (Schaufeli et al., 2002), to individual characteristics (Bandura’s sense of self-
increasingly more data have become available each year, efficacy) rather than being a component of burnout
along with successive language versions, providing (Shirom, 2003). The results of a study conducted by
researchers with the opportunity to make international a team of Spanish researchers allow us to suggest that it is
comparisons. the lack of professional efficacy that may contribute to the
development of burnout (Salanova, Peiró, & Schaufeli,
Tool description 2002) and that, in this case, it should be treated as an
The scale, similar to the general version, which was independent variable rather than a component of burnout
the basis for the construction of the student version, (Bresó, 2008). One may also wonder the extent to which
consists of 16 items that are assigned to three subscales: the wording of the test items contributes to this or whether
exhaustion, associated with the demands of studying (EX, the reverse wording of the items for the third subscale
5 items, e.g., “I feel emotionally drained by my studies does not account for the weaker correlation results
cynicism, showing cynical attitudes and lack of identifica- between the subscales (Chirkowska-Smolak, 2012).
tion with one’s studies (CY, 5 items, e.g., "I doubt the
significance of my studies") and professional/academic Research procedure
efficacy, i.e., the feeling that one is not competent in the Work on the Polish version of the questionnaire
role of student (EF, 6 items reverse scored, e.g., “In my proceeded in two stages: translation and validation of the
opinion, I am a good student”). tool.
Respondents provided answers on a 7-point Likert
scale ranging from 0 (never) to 6 (every day). High scores Development of the Polish language version.
on the Exhaustion and Cynicism scales suggest the The MBI-SS Burnout Questionnaire is a licensed tool
presence of burnout in the students surveyed, while in published by Mind Garden Inc. Before starting work on
the case of Academic Efficacy, on which all items are the Polish version, consent was obtained from the publish-
positively worded, burnout is indicated by a low score. It er, who accepted the conditions of the survey and allowed
should also be added that the authors make it clear that an us to administer the MBI as an online survey via non-Mind
overall score for burnout should not be calculated, e.g., as Garden survey system.
an average of all test items/items. The results of each First, translation was performed according to the
subscale are analyzed separately — they create individual procedure for translation of research tools from the area of
profiles for each respondent. quality of life and health, i.e., translation and retranslation
One of the important research goals related to performed by independent translators and comparison of
burnout and its measurement tools is identifying the the original version with the retranslation (WHO, 2016).
factor structure. Previous studies have explored the However, this procedure was extended, as we also took
dimensionality of the MBI. Still, they have yielded into account the suggestions of Beaton et al. (2000)
inconsistent and conflicting results because some re- regarding the cultural adaptation of tests.
searchers identified various factor solutions (Faraci, First, prior to translation, we made a preliminary
2018). Especially there are some doubts as far as the assessment of the individual test items, whether they were
inclusion of professional efficacy is concerned. Some appropriate for measuring burnout among students and
researchers have shown that lack of personal accomplish- whether there were any cultural differences, such as those
ment works differently from the two other MBI dimen- related to differences in the education system, which
sions, suggesting that lack of personal accomplishment needed to be taken into account.
might not be part of the burnout syndrome (Schaufeli and Then, the original version (in English) was translated
Taris, 2005). The analyses conducted by Pasikowski into Polish by the authors of the adaptation (translations
(1996) in the Polish language version of the MBI were conducted independently). After a discussion about
questionnaire for social professions and later by Chir- discrepancies and ambiguities between the different
kowska-Smolak and Kleka (2011) in relation to the language versions, the Polish version of the questionnaire
general scale indicated a satisfactory value of the Polish was established through consensus. The next step was to
versions of the tools and confirmed the three-factor perform retranslation with the help of two translators who
structure and reliability of the tool; however, there were did not know the original version (blind translation) and
doubts about the relationship between the individual translated the Polish version into English independently.
subscales. The main point is that between the first two Finally, the original version was compared with the two
subscales (exhaustion and cynicism), there are quite retranslated versions, and the differences were discussed
strong relationships, while they are weaker in the case with a psychologist working in the US. We wanted to
of a lack of professional efficacy, yet these are the ensure that the Polish-language version and the original
dimensions of the same phenomenon. Exhaustion and version would be equivalent, taking into account any
cynicism occur together, and reduced professional effi- differences between higher education systems. Once we
 The factorial validity of the Maslach Burnout Inventory: Student Survey (MBI-SS) in Poland 210
were satisfied that the original version and the retransla- ‘work’ has been replaced by ‘study’. Respondents are
tions were equivalent, we moved to the next stage related asked to indicate for each of 5 items on a 7-point scale
to the development of the Polish language version, which (1. Strongly disagree, 7. Strongly agree).
was the initial evaluation by the target group.
For this purpose, we used the help of 25 psychology Participants
students specializing in work and organizational psychol- We determined the required sample size a priori. The
ogy who formed an expert panel. Each completed the sample size was computed using recommendations in
questionnaire and evaluated the way in which individual literature (Kim, 2005) when conducting a confirmatory
test items were expressed, the comprehensibility of factor analysis (CFA). We took into account model
instructions and the way of answering on the proposed complexity: number of items per factor (5, 5, and 6),
scale. Additionally, they could make comments regarding relationships between factors (average factor correlation
the Polish translation. In this way, we checked how 0.4), and factor loading (average factor loading 0.6). For
participants might understand and interpret individual alpha, a value of 0.01 was selected as the target
items. Then, it was possible to move on to the actual study probability, and a value of 0.99 was selected for power.
to confirm the factorial structure of the questionnaire. We expected that CFA would confirm a good fit of the
model to the observed data (expected CFI 0.95). The
Organization of the validation study minimal sample size was calculated with a web calculator
To validate the tool, we checked its factorial retrieved from http://wnarifin.github.io (Arifin, 2023)
structure, verified the fit of the three-factor model, and as 574.
calculated the reliability of the tool (its internal consis- The Polish tertiary education system is based on
tency and test retest). a three-tier structure of studies. These are first degree
Students were informed about the ongoing study on studies (usually 3-year studies, bachelor’s degree studies),
well-being connected with their studies through social second degree studies (master’s degree studies, 2-year
media (e.g., departmental profiles on FB). Informed studies) or comprehensive master’s degree studies, which
consent was obtained from each respondent according to concern strictly defined majors, such as psychology, law,
APA guidelines. Participation in the study was voluntary or medicine (5-6 years). Studies can be conducted either as
and anonymous, as no personal data were gathered. full-time studies or extracurricular studies.
Respondents could withdraw from the study at any time A total of 935 students from state universities in
and close the form without saving the results. Respondents Poland participated in the actual study1. A total of 67%
were informed that they could contact the researchers if of the sample was female, and 28,6% was male. Most
they felt uncomfortable or had questions or concerns and of the respondents were studying full-time (70,9%).
necessary contact details were provided. Additionally, Approximately half of the respondents (53%) were
participants were informed of the possibility of seeing the students of social sciences (e.g., psychology, pedagogy,
results by contacting researchers. sociology), and the others studied science and natural
In the first step, only the MBI questionnaire was sciences (e.g., mathematics, chemistry, biology, geogra-
distributed to students. In the appropriate form’s section, phy, 23%), humanities (e.g., philology, cultural studies,
they were able to indicate their willingness to participate in 12%) and law and economics (e.g., law in business,
a further stage of the study (concerning construct validity). management, 12%).
A similar strategy was adopted after it was sent; They were students of all years (1st–5th year), 53%
participants could declare their willingness to participate were students at the bachelor’s degree level, 16.1% at the
in the planned retest (described in Section 3.2.). Retest was master’s degree level, and the rest were students at the
conducted after 3 months. comprehensive master’s degree level (majoring in psy-
chology or law).
Measure In the retest participated sample of 129 students
In the retest the following questionnaires were used: (48.4% were women). The average age of the participants
– Copenhagen Psychosocial Questionnaire COPSOQ was 21, most of them (62.1%) did not engage in additional
II (Pejtersen, Kristensen, Borg, Bjorner, 2010; Polish student activities, and 69.4% were not employed. 52% of
version: Baka, 2019): burnout (4 items) and self-efficacy them were social science students, 33% - studied science
beliefs (6 items) subscales. The COPSOQ II is designed to and natural sciences.
comprehensively measure psychosocial working condi-
tions, such as demands, organization, and content of work, RESULTS
social relations, and leadership or work values. Respon-
dents are asked to indicate their answers for all statements Description of statistical variables
on a 5-point scale (‘1. Always, 5. Never’ for the burnout The mean values for EX, CY, and EF were at
subscale and ‘1. To a very large degree, 5. To a very low a moderate level, with students reporting their efficacy at
degree’ for the self-efficacy beliefs subscale); the highest level and cynicism at the lowest level. The
– Brief Job Satisfaction Measure II (judge, Bono,
Locke, 2000; Polish version: Chirkowska-Smolak, Gro- 1
We share our dataset on OSF: https://osf.io/kuc7g/?view_on-
belny, 2016): in a modified version for students, the term ly=250757b377cc42afbf605cd4fcccfb9c
211 Teresa Chirkowska-Smolak, Tomasz Górecki, Mateusz Klakus, Weronika Metzger, Martyna Szargan

coefficients for variability (54% for exhaustion, 72% for

Table 1. Descriptive statistics and reliability coefficients
(Cronbach’s alphas if item deleted) for the three MBI-SS cynicism, and 48% for efficacy) indicate strong variability
subscales in the results, with respondents being the most diverse in
terms of their cynical approach to their studies.
Subscale and
M SD Reliability
items Reliability
1. Exhaustion 3.28 1.77 0.90 The reliability of the tool was measured using an
EX_1 2.62 1.40 0.88 internal consistency test. Cronbach’s alphas for EX, CY,
EX_2 2.95 1.46 0.87 and EF were 0.90, 0.89, and 0.83, respectively, indicating
EX_3 2.83 1.71 0.87 a high level of internal consistency for all three MBI-SS
EX_4 2.79 1.64 0.87 subscales.
EX_5 2.08 1.83 0.88 Figure 1 shows the correlations between the variables.
2. Cynicism 2.73 1.96 0.89 We can see that the variables are highly correlated within
CY_1 1.83 1.91 0.85 subscales. In addition, some correlation is observable
CY_2 2.17 2.01 0.84 between the variables forming the EX and CY subscales,
CY_3 3.45 1.84 0.92 especially between EX_5 and the variables that make up
CY_4 2.64 1.97 0.87 the CY subscale.
CY_5 2.22 2.01 0.86 Another way to test the reliability of the adapted tool
3. Efficacy 3.47 1.40 0.83 was to conduct a repeat measurement 3 months apart. In
EF_1 3.60 1.40 0.83 the end, 129 responses were obtained for the comparison
EF_2 2.37 1.64 0.80 of absolute stability. Table 2 shows the obtained Pearson's
EF_3 3.31 1.67 0.82 r correlation coefficients between individual subscales and
EF_4 3.46 1.60 0.81 for the entire questionnaire. The coefficients obtained
EF_5 1.87 1.63 0.81 range from r = 0.56 for the exhaustion subscale to r = 0.81
EF_6 3.04 1.55 0.80 for personal efficacy.

Figure 1. Correlation matrix

 The factorial validity of the Maslach Burnout Inventory: Student Survey (MBI-SS) in Poland 212
analyses were performed with maximum likelihood
Table 2. Test-retest estimated by Pearson’s r coefficient
estimation with robust standard errors and a Satorra-
Subscale M1 M2 r 95% CIs Bentler scaled test statistic.
First, we assessed the fit of the one-factor (M1)
Exhaustion 3.07 3.25 0.56** 0.46, 0.69
solution, which assumes that all three aspects of burnout
Cynicism 2.56 2.81 0.74** 0.65, 0.81 load are on one underlying dimension, as well as the fit of
Personal efficacy 4.09 3.27 0.81** 0.74, 0.86 two- (M2) and three-factor solutions (M3), which assumes
that two or three aspects of burnout are independent yet
MBI General
3.25 3.12 0.70** 0.60, 0.78 correlated factors. The two factor model (M2) result from
Score
combining two subscales: EX and CY, forming first factor
Note: **p < 0.01 and the PE subscale becoming second factor. Our results
(Table 3) show that the fit of the three-factor solution (M3)
Confirmatory factor analysis appears to be somewhat better than that of the one- and
To confirm the three-factor structure of the scale that two-factor solutions.
was corroborated for the original version, we performed In the next step, we decided to test the model by
confirmatory factor analysis using R project’s library leaving out the variable EX_5, which was highly
lavaan (Rosseel, 2012). Its results were also intended to correlated with CY scale (M4). This model obtained better
allow us to decide on a test item that raised our earlier results in terms of quality assessment. In this situation, we
concerns. decided to perform additional tuning of the model. We
To examine factorial validity, we assessed the fit of used a method called modification indices (Sörbom, 1989).
different models to identify which models best fit the data The modification index is an χ2 value, by which model fit
in our study. The goodness of fit was assessed using the χ2 would improve if a particular path was added or
test statistic, the comparative fit index (CFI), and the a constraint freed. Values larger than 3.84 mean that the
Tucker‒Lewis index (TLI). χ2 tests the null hypothesis that model could be ‘improved’, the p-value for the added
the predicted model and observed data are equal. Because parameter would be less than 5%, and values larger than
we want our predictions to match the actual data as closely 10.83 indicate that the parameter would have a p-value less
as possible, we do not want to reject it (a nonsignificant than 0.1%.
result indicates a good model fit). Based on the literature Modification indices (MIs) are useful diagnostic tools
(Hu & Bentler, 1999), the comparative fit index (CFI) and in Confirmatory Factor Analysis (CFA) that help identify
Tucker‒Lewis index (TLI) cutoff scores should be above model misspecification and suggest potential improve-
0.95 for well-fitted models. Furthermore, a root mean ments. They indicate how much the model's overall chi-
square error of approximation (RMSEA) less than 0.06 and square goodness-of-fit statistic would decrease if a parti-
a standardized root mean square residual (SRMR) less than cular parameter (e.g., a factor loading, covariance, or error
0.08 indicate an acceptable model fit (Hu & Bentler, covariance) were freely estimated instead of being fixed or
1999). The chi-square values for all models were constrained. In Table 4 we present MIs along with their
statistically significant (p<0.01) due to the large sample corresponding Expected Parameter Change (EPC) values.
size (Tanaka, 1987; Babyak & Green, 2010). MIs represent the expected decrease in the chi-square
By default, lavaan library (and Mplus) will always set statistic if a parameter is freely estimated. Larger MIs
the first loading of each factor to 1 (marker method). We suggest greater potential for model improvement. EPCs
checked multivariate normality. Mardia’s test revealed provide the expected change in the parameter value if it
multivariate non-normality. Base on this result our were freely estimated.

Table 3. Indices of overall fit for alternative factor structures

Model χ2(df) CFI TLI RMSEA (90% CIs) SRMR

M0: null model 8028.13 (120) <0.001 <0.001 0.275 (0.270, 0280) 0.375
M1: one-factor model 2432.42 (104) 0.674 0.623 0.169 (0.163, 0.175) 0.137
M2: two-factor model 1235.25 (103) 0.839 0.812 0.119 (0.113, 0.125) 0.077
M3: three-factor model 821.50 (101) 0.897 0.878 0.096 (0.090, 0.102) 0.064
M4: modified three-factor model 509.50 (87) 0.930 0.916 0.079 (0.073, 0.086) 0.061
M5: final three-factor model 141.15 (76) 0.989 0.985 0.033 (0.025, 0.042) 0.026
M6: MGCFA configural model 215.58 (152) 0.989 0.985 0.033 (0.022, 0.043) 0.028
M7: MGCFA weak model 234.62 (168) 0.989 0.986 0.032 (0.022, 0.042) 0.033
M8: MGCFA strong model 317.46 (180) 0.977 0.973 0.044 (0.036, 0.052) 0.038
213 Teresa Chirkowska-Smolak, Tomasz Górecki, Mateusz Klakus, Weronika Metzger, Martyna Szargan

We checked also MIs for M6, M7 and M8. However,

Table 4. Largest MIs for M4 (those MIs that have been added
no additional changes have been suggested for these
to the model are in bold).
models.
Expected Para- Using different observed variables in different models
Modification in-
Modification meter Change makes the chi-square difference test results uninterpreta-
dex (MI)
(EPC) ble. This test requires that the models be nested, one
CY_1~~CY_2 132.91 0.62 a special case of the next. Models with different observed
variables are not nested. Hence, in our situation is not
EX_2~~EX_4 91.30 0.56
possible to compare using this method models M3 (model
EF_4~~EF_5 84.54 0.52 with Ex5) and M4 (model without Ex5). Due to the
Cynicism=~EX_1 81.07 0.34 frequent challenge of determining if non-nested models
overlap, we generally employ a two-phase testing
Efficacy=~EX_2 48.36 0.37
approach outlined by Vuong (1989). In the initial phase,
CY_4~~CY_5 45.91 0.39 we assess if the two models can be differentiated from
Cynicism=~EX_2 38.69 -0.22 each other, which is feasible if they overlap. In the
subsequent phase, we evaluate if the fits of both models
EX_1~~EX_4 35.53 -0.34
are equivalent. From the first test (the variance test) we
CY_3~~CY_4 33.85 0.40 obtained p-value 3.33E-9. Hence, M3 and M4 are
CY_1~~CY_3 31.26 -0.31 distinguishable. Thus, we can move on to the second
(non-nested likelihood ratio test) test. For alternative “M4
CY_1~~CY_4 28.67 -0.28
fits better than M3” we obtained p-value <0.001. Exactly
Efficacy=~EX_3 24.43 -0.28 the same conclusion is reached when comparing the M5
CY_2~~CY_4 24.10 -0.26 and M3 models. M5 fits better than M3.
Finally, we evaluated the obtained model (M5) using
Efficacy=~EX_1 21.52 -0.27
the goodness-of-fit indices. We obtained a CFI value of
EF_3~~EF_6 21.34 0.24 0.983 and a TLI value of 0.976, which are good fit indices
Cynicism=~EF_2 21.18 -0.15 for internal validity. The value of RMSEA is 0.041, which
indicates a good fit. Similarly, the SRMR value of 0.038 is
EF_2~~CY_3 20.80 -0.31 below the cutoff score of 0.08 and indicates a good fit. The
CY_2~~CY_5 19.19 -0.24 final model is shown in Figure 2.
EF_4~~EF_6 17.41 -0.21 To provide evidence of MBI-SS measurement invar-
iance across different groups, we conducted multigroup
Efficacy=~CY_3 17.14 -0.32 CFA (MGCFA) across genders. Before we compare the
EX_4~~EF_4 15.48 0.19 values of latent means across multiple groups, we first
Cynicism=~EX_4 14.39 -0.14 need to establish measurement invariance. Testing for
measurement invariance involves a fixed sequence of
Exhaustion=~EF_2 13.97 -0.16 model comparison tests. A typical sequence involves three
EF_2~~EF_4 13.46 -0.22 models:
Exhaustion=~EF_6 13.42 -0.13 1. Configural invariance (M6). The same factor structure
is imposed on all groups.
EX_1~~CY_2 12.96 0.15 2. Weak (metric) invariance (M7). The factor loadings are
constrained to be equal across groups.
We made modifications one at a time, starting with 3. Strong (scalar) invariance (M8). The factor loadings
the largest MI. After each modification, re-ran the CFA, and intercepts are constrained to be equal across
assessed the model fit, and checked the MIs again. We groups.
continued this iterative process until we achieved an We prepared the three models and compared them
acceptable model fit (all MIs less than 10.83). with each other. It turned out that the first of the three
Final M5 specification: models (M6 in Table 7) is comparable to the next two
Exhaustion =~ EX_1 + EX_2 + EX_3 + EX_4 + EF_6 models. Therefore, there is no reason to think that gender
Efficacy =~ EF_1 + EF_2 + EF_3 + EF_4 + EF_5 + EF_6 + EX_2 affects the results. In addition, we can note that this model
Cynicism =~ CY_1 + CY_2 + CY_3 + CY_4 + CY_5 + EX_1 + EF_2 is inferior model to the proposed M5 model (all quality
CY_1 ~~ CY_2 measures are inferior), which did not account for gender.
EX_2 ~~ EX_4 We assumed that the factorial structure of the MBI-
EF_4 ~~ EF_5 SS is invariant across the samples of females and males.
EF_2 ~~ CY_3 However, compared to previous studies that used the MBI-
EX_1 ~~ CY_2 SS, we expected no such differences; many researchers
CY_1 ~~ CY_3 reported no significant effect of sex or that the factorial
EX_4 ~~ EF_4 structure of the MBI-SS also was invariant across females
 The factorial validity of the Maslach Burnout Inventory: Student Survey (MBI-SS) in Poland 214
Construct validity
The validity of the MBI-SS was measured in two
ways. First, MBI scores were correlated with satisfaction
from studies, the outcome that had been hypothesized to be
related to burnout. Second, to establish convergent validity
of the MBI, burnout scores were correlated with responses
on another measure of burnout — the subscale of the
COPSOQ II.
To determine convergent validity, the three dimen-
sions of the adapted questionnaire were subjected to
correlation analysis with the general level of burnout
[GLB], self-efficacy beliefs [SEB], and satisfaction with
Figure 1. Confirmatory factor analysis plot studies [SwS] (Table 8).
The obtained results indicate a statistically signifi-
and males (Portoghese et al., 2018; Simancas-Pallares cant correlation between the analyzed variables at the
et al., 2017; Turhan et al., 2021). p < 0.01 level. Dimensions of exhaustion and cynicism
The ANOVA tables below are for the models that can correlate strongly and negatively with self-efficacy (r =
be so compared. The Tables 5-7 show that the M0-M3 -0.29 and r = -0.28) and negatively and significantly with
models all differ significantly. We, therefore, choose the satisfaction with studies (r = -0.17 and r = -0.39). The
M3 model. Models M4 and M5 are also significantly opposite is found for the Personal Efficacy subscale,
different, so we choose model M5 from these. Models M6 which has positive, medium correlations (r = 0.51 and
and M7 do not differ significantly, but M7 and M8 do. We, r = 0.46). Moreover, the higher the overall burnout score
therefore, choose M6 from this group of models. (GLB), the higher the score on the exhaustion and

Table 5. ANOVA table comparing the M0-M3 models.

Model χ2(df) Difference in χ2 p-value

M3: three-factor model 821.50 (101)
M2: two-factor model 1235.25 (103) 413.75 0.00
M1: one-factor model 2432.42 (104) 1197.17 0.00
M0: null model 8028.13 (120) 5595.71 0.00

Table 6. ANOVA table comparing the M4-M5 models.

Model χ2(df) Difference in χ2 p-value

M5: final three-factor model 141.15 (76)
M4: modified three-factor model 509.50 (87) 368.35 0.00

Table 7. ANOVA table comparing the M6-M8 models.

Model χ2(df) Difference in χ2 p-value

M6: MGCFA configural model 215.58 (152)
M7: MGCFA weak model 234.62 (168) 19.04 0.29
M8: MGCFA strong model 317.46 (180) 82.84 0.001

Table 8. Pearson correlation coefficients for validity of MBI-GS(S)

GLB SEB SwS

r 95% CIs r 95% CIs r 95%CIs
EX 0.58** 0.5; 0.65 -0.29** -0.41; -0.21 -0,17** -0.28; -0.06
CY 0,41** 0.31; 0.5 -0,28** -0.38; -0.17 -0,39** -0.48; -0.29
PE -0,36** -0.45; -0.26 0.51** 0.42; 0.59 0.41** 0.31; 0.50
Note: ** p < 0.001, GLB – general level of burnout or general burnout, SEB – self-efficacy beliefs, SwS – satisfaction with studies
215 Teresa Chirkowska-Smolak, Tomasz Górecki, Mateusz Klakus, Weronika Metzger, Martyna Szargan

cynicism subscales. The findings allow us to accept the limiting our confidence in interpreting results. To mitigate
assumed hypotheses for the construct validity of the these concerns, we analyzed response patterns for straight-
adapted tool. lining and other indications of inattentiveness, even though
this is a more indirect measure than attention checks. We
DISCUSSION also considered the time participants took to complete the
survey and found that no responses were completed in an
The aim of the present study was to investigate the unrealistically short timeframe. While these steps help
value of a tool measuring burnout in Polish students. The alleviate some concerns, future iterations of this research
analysis of the full version (16 items) supported their should consider the incorporation of attention checks to
adequate three-factor structure, reliability, and validity – further ensure data quality.
consistent with our hypothesis. Notably, the Cronbach’s Further research should also consider criterion-
alpha for each scale exceeded 0.8, indicating strong relevance analysis and longitudinal studies for a more
internal consistency. A similar interpretation emerges holistic understanding of academic burnout in students.
from the correlations between variables, with the exception
of one item (EX_5/MBI_6; “I feel burned out from my CONCLUSIONS
studies”), which has a higher correlation with Cynicism
than Exhaustion. Removing this item enhanced the fit of The MBI-SS questionnaire is invaluable for research-
the three-factor model. The same item lowered the ers, university educators, managers, and psychologists
reliability of the EX scale in the Polish version of the aiming to gauge the level of academic burnout in students.
tool for measuring burnout in all professions (MBI-GS, Our findings suggest that the Polish adaptation has good
Chirkowska-Smolak, Kleka, 2011), leading the authors of validity. The three-factor structure of the questionnaire
the adaptation to suggest its removal from the Polish was supported, comprising distinct factors: exhaustion,
version of the questionnaire. cynicism, and academic efficacy. Due to higher factor
Considering the fit coefficients of the three-factor loadings for the MBI_6 item on another subscale (CY)
model, our findings suggest the validity of the tool. While rather than its original (EX), we recommend omitting this
the results for the standard three-factor model were item from the Polish version. As such, we advocate for the
satisfactory, testing the modified version with the excluded use of the 15-item version of the scale (excluding MBI_6).
item led to improved values. Comparable findings emerge
in studies of other adaptations of the MBI-SS, but they REFERENCES
involve the MBI_13 item (Hu, Schaufeli, 2009; Shin, Puig,
Lee, J., Lee, J. H. & Lee, S. M., 2011; Tsubakita, Alarcon, G. M., Edwards, J. M., Menke, L. E. (2011). Student Burnout
and Engagement: A Test of the Conservation of Resources Theory.
Shimazaki, 2016, Faye-Dumanget, Carré, Le Borgne,
The Journal of Psychology 145(3), 211-227. https://doi.org/10.1080/
Boudoukha, 2017). Additionally, we utilized MGCFA to 00223980.2011.555432
assess the invariance of the MBI-SS across genders. The Baka, Ł. (2019). Kopenhaski Kwestionariusz Psychospołeczny (COPSOQ
results indicate that the factorial structure of the MBI-SS II). Podręcznik do polskiej wersji narzędzia. Centralny Instytut
remains consistent across both student samples. Ochrony Pracy – Państwowy Instytut Badawczy.
Beaton, D.E., Bombardier, C., Guillemin, F., Ferraz, M.B. (2000).
Guidelines for the process of cross-cultural adaptation of self-report
LIMITATIONS OF THE STUDY measures. Spine, 25(24), 3186-3191. https://doi.org/10.1097/
00007632-200012150-00014
This study had its limitations. Primarily, our respon- Brenninkmeijer, V., Van Yperen, N. (2003). How to Conduct Research
on Burnout: Advantages and Disadvantages of an Unidimensional
dent sample was not entirely representative. The survey
Approach to Burnout, Occupational and Environmental Medicine,
was primarily among students at public universities 60(1), 16-20. https://doi.org/10.1136%2Foem.60.suppl_1.i16
studying major subjects; thus, findings may not apply to Campos, J. A. D. B., Carlotto, M. S., Marôco, J. (2012). Oldenburg
students at technical or medical universities. A significant Burnout Inventory-Student Version: Cultural Adaptation and Valida-
portion of the respondents were social science students, tion into Portuguese. Psicologia: Reflexao e Critica, 24(4), 709-718.
https://doi.org/10.1590/S0102-79722012000400010
leading to a higher female representation. While our test of
Campos, J. A. D. B., Carlotto, M. S., Marôco, J. (2013). Copenhagen
construct validity supported the adapted tool’s value, Burnout Inventory-Student Version: Adaptation and Transcultural
concerns about the sample's representativeness persist (as Validation for Portugal and Brazil. Psicologia: Reflexao e Critica,
noted above). Anticipated correlations between the surveys 26(1), 87-97. https://doi.org/10.1590/S0102-79722013000100010
used would benefit from validation in a broader, more Chirkowska-Smolak, T. (2012). Psychologiczny model zaangażowania
w pracę. Wydawnictwo Naukowe UAM.
representative study.
Chirkowska-Smolak, T., Grobelny, J. (2016). Konstrukcja i wstępna
A significant limitation is the lack of attention checks analiza psychometryczna Kwestionariusza Postrzeganego Stresu
in our survey. The inclusion of attention checks has w Pracy (PSwP). Czasopismo Psychologiczne, 22(1), 131-139.
become an increasingly common practice in survey http://dx.doi.org/10.14691/CPPJ.22.1.131
research to identify and potentially exclude respondents Chirkowska-Smolak, T., Kleka, P. (2011). The Maslach Burnout
Inventory–General Survey: validation across different occupational
who may not be reading or processing questions
groups in Poland. Polish Psychological Bulletin, 42(2), 86-94. http://
thoroughly. Their absence has several implications: dx.doi.org/10.2478/v10059-011-0014-x
inattentive responses could introduce noise into the data, Demerouti E., Bakker A.B., Nachreiner F., Schaufeli W.B. (2001). The
reducing the precision of our estimates and potentially job demands-resources model of burnout. Journal of Applied
 The factorial validity of the Maslach Burnout Inventory: Student Survey (MBI-SS) in Poland 216
Psychology, 86(3), 499–512. https://doi.org/10.1037/0021-9010. Rosseel, Y. (2012). lavaan: An R Package for Structural Equation
86.3.499 Modelling. Journal of Statistical Software, 48(2), 1-36. https://doi.
Esteve, E.B. (2008), Well-being and Performance in Academic Settings: org/10.18637/jss.v048.i02
The Predicting Role of Self-efficacy [unpublished doctoral disserta- Rostami, Z., Abedi, M. R., Schaufeli, W. B., Ahmadi, A., Sadeghi, A. H.
tion]. Castellón de la Plana: Universidad Jaume I. (2014). The Psychometric Characteristics of Maslach Burnout
Faraci, P. (2018). Testing for the dimensionality of the Maslach Burnout Inventory Student Survey: A Study Students of Isfahan University.
Inventory (MBI) with a sample of high school teachers. Clinical Zahdan Journal of Research in Medical Sciences, 16(9), 55-58.
Neuropsychiatry, 15(1), 50-59. http://dx.doi.org/10.18637/jss.v048.i02
Faye‐Dumanget, C., Carré, J., Le Borgne, M., Boudoukha, A. H. (2017). Salanova, M., Peiró, J. M., Schaufeli, W. B (2002). Self-efficacy
French validation of the Maslach Burnout Inventory-Students Survey Specificity and Burnout among Information Technology Workers:
(MBI-SS). Journal of Evaluation in Clinical Practice, 23(6), 1247– An Extension of the Job Demands-Control Model. European Journal
1251. https://doi.org/10.1111/jep.12771 on Work and Organizational Psychology, 11, 1-25. https://doi.org/
Hu, Q.,Schaufeli, W. B. (2009). The factorial validity of the Maslach 10.1080/13594320143000735
Burnout Inventory: Student survey in China. Psychological Reports, Salgado, S., Au-Yong-Oliveira, M. (2021). Student burnout: A case study
105(2), 394–408. https://doi.org/10.2466/PR0.105.2.394-408 about a Portuguese public university. Education Sciences, 11(1),
Judge, T.A., Bono, J.E., Locke, E.A. (2000). Personality and job 1–35. https://doi.org/10.3390/educsci11010031
satisfaction: The mediating role of job characteristics. Journal of Schaufeli, W. B., Leiter, M. P., & Maslach, C. (2009). Burnout: 35 years
Applied Psychology, 85(2), 237-49. http://dx.doi.org/10.1037/0021- of research and practice. Career Development International, 14(3),
9010.85.2.237 204–220. https://doi.org/10.1108/13620430910966406
Kaiser, H. F. (1974). An index of factorial simplicity. Psychometrika, Schaufeli, W. B., Martinez, I., Pinto, A. M., Salanova, M., Bakker, A. B.
39(1), 31–36. https://doi.org/10.1007/BF02291575 (2002). Burnout and engagement in university students: A cross-
Kristensen, T.S., Borritz, M., Villadsen, E. and Christensen, K.B. (2005) national study. Journal of Cross-Cultural Psychology, 33(5), 464–
The Copenhagen Burnout Inventory: A New Tool for the Assessment 481. https://doi.org/10.1177/0022022102033005003
of Burnout. Work and Stress, 19, 192-207. http://dx.doi.org/10.1080/
Schaufeli, W. B., Taris, T. W. (2005). The conceptualization and
02678370500297720
measurement of burnout: Common ground and worlds apart. Work &
Lee, R. T., Ashforth, B. E. (1996). A Meta-analytic Examination of the
Stress, 19, 256–262. https://doi.org/10.1080/02678370500385913
Correlates of the Three Dimensions of Job Burnout, Journal of
Simancas-Pallares, M. A., Fortich Mesa, N., González Martínez, F. D.
Applied Psychology, 81(2), 123-133. http://dx.doi.org/10.1037/0021-
(2017). Validity and internal consistency of the Maslach Burnout
9010.81.2.123
Inventory in dental students from Cartagena, Colombia. Revista
MacCallum, R. C., Browne, M. W., Sugawara, H. M. (1996). Power
Colombiana de Psiquiatría (English Ed.), 46(2), 103–109. https://
analysis and determination of sample size for covariance structure
doi.org/10.1016/j.rcpeng.2017.05.009
modelling. Psychological Methods, 1(2), 130–149. https://doi.org/
Shin, H., Puig, A., Lee, J., Lee, J. H., Lee, S. M. (2011). Cultural
10.1037/1082-989X.1.2.130
validation of the Maslach Burnout Inventory for Korean students.
Maslach, C. (1993). Burnout: A Multidimensional Perspective. In:
Asia Pacific Educational Review, 12(4), 633–639. https://doi.org/
Schaufeli, W., Maslach, C., Marek, T. (eds.), Professional Burnout:
10.1007/s12564-011-9164-y
Recent Developments and Research (19–32), Taylor & Francis:
Shirom, A. (2003). Job-related Burnout: A Review. In: Quick, J. C.,
London. http://dx.doi.org/10.4324/9781315227979-3
Tetrick, L. (eds.), Handbook of Occupational Health Psychology
Maslach, C., Jackson, S. E., Leiter, M. P. (1996). Maslach burnout
(245-264). American Psychological Association. https://doi.org/
inventory manual (3rd ed.). Consulting Psychologists Press.
10.1037/10474-012
Maslach, C., Jackson, S. E., Leiter, M. P. (2017). Maslach Burnout
Inventory Manual, 4th edition. Mind Garden. Sörbom, D. (1989). Model modification. Psychometrika, 54(3), 371-384.
Maslach, C., Leiter, M. P. (2008). Early predictors of job burnout and https://doi.org/10.1007/BF02294623
engagement, Journal of Applied Psychology, 93(3), 498-512. http:// Stawiarska, P. (2016). Wypalenie zawodowe w perspektywie współczes-
dx.doi.org/10.1037/0021-9010.93.3.498 nego świata. Difin.
Maslach, C., Schaufeli, W. B., Leiter, M. P. (2001). Job burnout. Annual Tsubakita, T., Shimazaki, K. (2016). Constructing the Japanese version of
Review of Psychology, 52(1), 397–422. https://doi.org/10.1146/ the Maslach Burnout Inventory – Student Survey: Confirmatory
annurev.psych.52.1.397 factor analysis. Japan Journal of Nursing Science, 13(1), 183-188.
Pasikowski, T. (1996). Polska adaptacja kwestionariusza Maslach https://doi.org/10.1111/jjns.12082
Burnout Inventory. In: Sęk, H. (eds.), Wypalenie zawodowe: Turhan, D., Scheunemann, A., Schnettler, T., Bäulke, L., Thies, D. O.,
Przyczyny, mechanizmy, zapobieganie (135-148). PWN. http://dx. Dresel, M., Fries, S., Leutner, D., Wirth, J., Grunschel, C. (2021).
doi.org/10.18559/SOEP.2018.3.2 Psychometric Properties of the German Short Version of the Maslach
Pejtersen, J. H., Kristensen, T. S., Borg, V., & Bjorner, J. B. (2009). The Burnout Inventory – Student Survey. European Journal of Health
second version of the Copenhagen Psychosocial Questionnaire. Psychology, 28(2), 45–58. https://doi.org/10.1027/2512-8442/
Scandinavian Journal of Public Health.https://doi.org/10.1177/ a000067
1403494809349858 Vuong, Q. H. (1989). Likelihood ratio tests for model selection and non-
Portoghese, I., Leiter, M. P., Maslach, C., Galletta, M., Porru, F., nested hypotheses. Econometrica, 57(2):307-333. https://doi.org/
D’Aloja, E., Finco, G., Campagna, M. (2018). Measuring Burnout 10.2307/1912557
Among University Students: Factorial Validity, Invariance, and Wickramasinghe N. D., Dissanayake, D. S., Abeywardena, G. S. (2018).
Latent Profiles of the Italian Version od the Maslach Burnout Validity and reliability of the Maslach Burnout Inventory-Student
Inventory Student Survey (MBI-SS). Frontiers in Psychology, 9, Survey in Sri Lanka. BMC Psychology, 6(1), article 52. https://doi.
article 2105. https://doi.org/10.3389/fpsyg.2018.02105 org/10.1186/s40359-018-0267-7
Revelle, W. (2021) psych: Procedures for Personality and Psychological World Health Organization (WHO) (2016). Process of translation and
Research, Northwestern University, Evanston, Illinois, USA, https:// adaptation of instruments. Available at: http://www.who.int/sub-
CRAN.R-project.org/package=psych Version = 2.1.6. stance_abuse/research_tools/translation/en/ (26 January 2022).