Journal of Affective Disorders Reports 10 (2022) 100437

Contents lists available at ScienceDirect

Journal of Affective Disorders Reports

journal homepage: www.sciencedirect.com/journal/journal-of-affective-disorders-reports

Research Paper

The overlap between burnout and depression through a different lens: A

multi-method study
Makilim Nunes Baptista a, Nelson Hauck-Filho a, *, Hugo Ferrari Cardoso b
a
Universidade São Francisco, Graduate School of Psychology, Campinas, Brazil
b
Universidade Estadual Paulista, Graduate School of Developmental and Learning Psychology, Bauru, Brazil

A R T I C L E I N F O A B S T R A C T

Keywords: Introduction: Depression and burnout are highly overlapping constructs, according to many correlational, factor,
Distress network, and mixture analysis studies. However, the latent correlation between the unique factors of burnout
Affective disorders and depression might be confounded by a general factor of distress, and acquiescence (i.e., the tendency to agree
Occupational health
more than disagree when responding to self-report items). In the current study, we performed an in-depth
Job stress
investigation on the latent structure of burnout and depression controlling for a general factor and acquies­
Cluster analysis
Construct overlap cence, and using a combination of random-intercept bifactor, network, and latent profile analysis.
Method: Participants were 584 nurse professionals (79% nurse technicians and assistants, 21% nurses), with ages
ranging from 20 to 65 years (M = 35.41; SD = 9.54), who responded to measures of depression symptoms,
burnout, and work stress.
Results: Results revealed a latent overlap between depression and burnout, even once acquiescence (r = .74), and
a general factor was accounted for (r = .57). Burnout and depression indicators formed a coherent network of
associated symptoms, with fatigue and lack of energy acting as bridge symptoms. The latent profile analysis
yielded five classes that once again suggested a high dependence between burnout and depression.
Discussion: We conclude that controlling for acquiescence and a general factor does not eliminate the high
overlap between burnout and depression, that fatigue symptoms bear clinical importance as a trigger to more
severe mental suffering in occupational settings, and that burnout involves a continuum of work stress.

1. Introduction
Depression and burnout are overlapping constructs. Evidence sug­
gests these concepts refer to syndromes with blurred boundaries, that
tend to display high comorbidity, and simultaneous onset (Aloha et al.,
2005; Bianchi et al., 2015). Burnout symptoms correlate more strongly
with depression than with each other, they share a general dimension
with depression (Bianchi, 2020), and they appear to represent the same
continuum of distress (Schonfeld and Bianchi, 2021). All this evidence
puts serious questions as to whether burnout holds as a standalone
syndrome, independent from depression. In the current study, we
combine bifactor, network, and latent profile analysis to further explore
the latent association between burnout and depression. We put emphasis
on the need to simultaneasouly control for response bias in self-report
questionnaires, and to account for a general factor of depression (and
burnout).
The main source of confusion in the distinction between depression
and burnout is because they are likely part of the same latent phenom­
enon. Many studies have suggested that these two constructs are man­
ifestations of a shared latent condition, even if specific combinations of
symptoms might exist in each case (Tavella and Parker, 2020). Bianchi
et al. (2015) conducted a systematic review of the literature and found
lack of conceptual and empirical distinction between depression and
burnout. Althouhg some have claimed that the connection of burnout
and depression are primarily driven by fatigue indicators (Maslach and
Leiter, 2016), controlling for fatigue have yielded no decrease in the
overlap between these constructs (Bianchi, 2020). While the
meta-analyses of Koutsimani et al. (2019) and (Chen and Meier, 2021)
reported correlations of r = .52 and .40, respectively, an updated
meta-analysis of Bianchi et al. (2021) using data from 14 independent
studies found a much higher positive association of r = .80. Depression
and burnout can be both be elicited by workplace stressors (Bianchi and
Brisson, 2019), and workers in occupational environments that are
highly demaning and stressful, such as health settings, tend to have

* Corresponding author.
E-mail address: hauck.nf@gmail.com (N. Hauck-Filho).

https://doi.org/10.1016/j.jadr.2022.100437
Received 4 April 2022; Received in revised form 13 October 2022; Accepted 15 November 2022
Available online 17 November 2022
2666-9153/© 2022 The Author(s). Published by Elsevier B.V. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-
nc-nd/4.0/).
M.N. Baptista et al.
comorbid depression and burnout (Wurm et al., 2016). This has led
Schonfeld and Bianchi (2021) to reframe burnout as a form of “occu­
pational depression.” In sum, a fair amount of evidence poses serious
questions to the status of depression and burnout as standalone noso­
graphic entities.
One possible form of approaching the separability of burnout and
depression is bifactor modeling. Bifactor models control for a general
common factor across a set of indicators, and they allow researchers to
test if cluster (“specific”) factors exist in the data that explain why a
subset of indicators display residual and meaningful correlations to each
other (Reise, 2012). This strategy has already been implemented in the
field. A series of exploratory and confirmatory structural equation
modeling studies have repeatedly reported that a general factor ac­
counts for the major part of the shared variance across indicators of
burnout and depression (Bianchi, 2020; Schonfeld et al., 2019; Verkui­
len et al., 2020). In a similar fashion, this general dimension, denomi­
nated “nonspecific psychological distress” (Schonfeld et al., 2019), has
also been found in second-order confirmatory factor analysis in­
vestigations (Bianchi et al., 2020). Moreover, in the meta-analysis of
Bianchi et al. (2021), bifactor models achieved good fit to the data of 14
independent samples from many countries and occupational domains.
The traditional specification of a bifactor model involves orthogo­
nalizing all cluster/specific factors, i.e., constraining them to be non-
correlated. This makes sense when one wants to detect the existence
of independent partial/unique common variance across a subset of in­
dicators. However, an alternative specification is to orthogonalize the
general factor, but leave the cluster/specific factors to freely correlate
(Mansolf and Reise, 2016). For instance, in the context of exploratory
factor analysis, Jennrich and Bentler (2012) developed a bifactor rota­
tion that produces correlated cluster/specific factors. This is an inter­
esting approach in the case of burnout and depression. One can theorize
that burnout and depression indicators share a common dimension, but
that they could also comprise residual factors that are not entirely in­
dependent. In other words, the overlap between the two constructs
might not be completely attributable to a “non specific psychological
distress” (Schonfeld et al., 2019), as other latent independent sources of
variance might exist. We tested such hypothesis in the current study.
Another question pertains to how much of the variance that is shared
among burnout and depression indicators is true construct variance and
not response bias. As repeatedly demonstrated, self-report inventories
might capture variability that is due to the idiossincratic styles with
which some individuals use the response scales (van Vaerenbergh and
Thomas, 2013). One of this biases is acquiescent responding, the ten­
dency to agree regardless of item content or at the expense of semantic
consistency (Paulhus, 1991). Accordingly, as self-report is the method of
choice in most empirical studies in the area, a remaining question is
whether acquiescence might confound estimates of the true overlap
between burnout and depression. For instance, would the estimated
large and positive association between the two constructs (r = .80 in the
study of Bianchi et al., 2021) remain high if acquiescence is taken into
account? Yet, given that acquiescence will contaminate all items from a
questionnaire, it can contaminate (inflate) the loadings of burnout and
depression indicators on a general factor. In this study, we explored the
correlation between burnout and depression before and after imple­
menting statistical controls of acquiescence, as a further strategy to
investigate their potential status as independent constructs. We also
inspected the magnitude of loadings of indicators of such constructs on a
general factor after accounting for acquiescent responding. To do so, we
controlled acquiescence using two distinct strategies: structural equa­
tion modeling, and exploratory graph (network) analysis.
Moreover, we supplemented our analytical strategy with latent
profile analysis. If depression and burnout are independent phenomena,
then we would expect to find in the population discrete clusters of in­
dividuals with high levels in only one of the constructs, but not the other.
Previous latent profile/class analysis investigations have almost dis­
charded this possibility, as they reported parallel profiles of increasing
2
Journal of Affective Disorders Reports 10 (2022) 100437
levels in both depression and burnout symptoms across classes (Ahola
et al., 2014; Bianchi et al., 2016; Bianchi and Brisson, 2019). Never­
theless, depression and burnout exhibit a low prevalence in the popu­
lation (depression is estimated to occur around 5% in the population;
Auerbach et al., 2018). Hence, complex score patterns might be rela­
tively rare among the general population and difficult to find in latent
profile/class studies. An interesting approach would be to have data
collected from populations where the relative prevalence of burnout and
depression is expected to be higher than in the general population. For
instance, nurses who suffer from depression represent almost twice the
rate of other professionals (Brandford and Reed, 2016). In this case, the
number of individuals who score high in only one of the constructs can
be sufficiently large, so that they can appear as members of an inde­
pendent latent profile or class in the analysis.
In the current study, we investigated the status of depression and
burnout as distinct psychological syndromes by using a combination of
bifactor, network, and latent profile analysis in a sample of nurses. If
depression and burnout are overlapping constructs, we should find that:
(1) they remain correlated after partialling a general factor in a bifactor
model with correlated cluster/specific factors; (2) their correlation is
still positive and moderate to large after we control for acquiescent
responding; (3) their symptom communities will overlap in an explor­
atory graph (network) analysis even after controlling acquiescence; and
(4) no salient latent profiles will be found, among nurses, that charac­
terize a high level in one construct, but low in the other. We also used
work stress to externally validate our latent profiles, as both depression
and burnout associate with stressors at the workplace (Plieger et al.,
2016).
2. Method
2.1. Participants
Sample was composed of 584 nurse professionals (79% nurse tech­
nicians and assistants, 21% nurses), with ages ranging from 20 to 65
years (M = 35.41; SD = 9.54). Most participants we females (82.7%),
who acted in three hospitals in the state of São Paulo, Brazil. As to
relational status, 46.9% of participants reported being married, and
41.1% single. Approximately 60% of the professionals that took part in
the study worked in their respective institutions and were in their work
positions for up to five years. Among these health care professionals,
36% declared their mean workday lasted 12 hours, while 33.6% worked
6 hours on mornings, and 24.8% six hours on afternoons. More than half
(64%) indicated having no second job.
2.2. Instruments
2.2.1. Brazilian burnout scale (Cardoso et al., 2022)
BBS is a 28-item self-report inventory devised to the assessment of
burnout symptoms and dimensions as conceptualized by (Maslach and
Jackson, 1981). Items are rated on a four-point Likert scale (1 = never, 2
= rarely, 3 = often, 4 = always), and they divide into two correlated
scales, with higher scores implying higher burnout: Emotional Exhaus­
tion/Frustration (22 items), and Depersonalization/Distancing (6
items). In the current study, reliability estimates were .92 for Exhaus­
tion, and .63 for Depersonalization, according to Omega Total coeffi­
cient (ωT).
2.2.2. Baptista’s depression scale-brief version (Gonçalves et al., 2021)
EBADEP-B is a brief, 15-item inventory of broad perceived symptoms
of depression. Items are scored on a four-point Likert scale (1 = disagree;
4 = agree), which provide a broad score of depression. Gonçalves et al.
(2021) reported sensitivity of .96, and specificity of .82 when the scale’s
total score is employed for screening purposes. Reliability in the current
study was .90 according to Alpha coefficient, and .92 according to
Omega Total (ωT).
M.N. Baptista et al.
2.2.3. Vulnerability to work stress scale (Sisto et al., 2007)
EVENT is a self-report inventory that lists 40 situations capable of
eliciting stress at the workplace, including features of the organizational
environment, infrastructure, and work routine. Items are rated on a
three-point Likert scale (0 = never, 2 = very often). In the present study,
reliability was estimated as .93 according to Omega coefficient (ωT).
2.3. Procedures
The research project was approved by a Research Ethics Committee.
Data were collected in three hospitals located in the countryside of the
state of São Paulo, Brazil. Upon authorization of the Board of Review of
each hospital, the nurse professionals, nurse technicians, and assistants
were approached and invited to take part in the study. Before
responding to the questionnaires, each participant signed an Informed
Consent containing the full description of the study aims, procedures,
and risks. The mean duration for taking the full survey was approxi­
mately 20 minutes.
2.4. Data analysis
The aim of our study was to investigate the extent to which
depression and burnout symptoms do overlap and can be regarded as
distinct components of the same syndrome. To do so, we implemented
the following strategies. First, we tested the fit to the data of our uni­
dimensional measures of burnout and depression. We then proceded to
their combined modeling, by testing a two- correlated factors model, an
oblique bifactor model, and an oblique bifactor model with a random
intercept. Random intercept factor models include a method factor hy­
pothesized to exert equal influence over all trait factor indicators, be
them semantically positive or negative, exactly what occurs when
acquiescent responding takes place (Maydeu-Olivares and Coffman,
2006). We compared the estimated latent correlation between burnout
and depression that was produced in each case. These analyses were
carried out using the robust Weighted Least Squares Mean- and
Variance-adjusted (WLSMV) estimator from the Mplus software.
To compare the variance proportion of each component—general
factor versus residual/specific factors—, we relied on the omega coef­
ficient. Omega hierarchical ωH captures the amount of reliable variance
attributable to a general factor, while omega total ωT also considers the
reliable variance captured by the residual/specific factors. If the residual
factors of depression and burnout add reliable variance beyond the
general factor, this would be reflected in an expressive difference from
omega hierarchical ωH to omega total ωH. This analysis was conducted
using the psych package, which computes the aforementioned co­
efficients from a Schmid-Leiman bifactor solution (Revelle and Revelle,
2015).
Next, we performed an exploratory graph analysis (EGA) of the
combined set of depression and burnout indicators. EGA is a network
analysis technique that relies on the graphical Lasso (glasso) estimation
and community detection algorithms to identify clusters of partially
correlated items (Golino and Epskamp, 2017). We implemented our
glasso analysis combined to the triangulated maximally filtered graph
and walktrap algorithms. Whereas the further limits the number of
partial correlations in the network model, the later identifies item
communities based on “random walks” between items (i.e., nodes). The
final network admits a visual representation of item communities, which
is useful for gathering a comprehension about bow connected are
depression and burnout symptoms. If the two concepts refer to distinct
syndromes, then they should appear as having their items clustering
apart. Comparatively, we also estimated a random intercept EGA
network, which implements an algorithm that estimates the strength of
association between nodes after accounting for acquiescent responding.
These analyses were conducted on R using the EGAnet package (Golino
and Epskamp, 2017).
Lastly, we also conducted an exploratory latent profile analysis. We
3
Journal of Affective Disorders Reports 10 (2022) 100437
hypothesized that, if depression and burnout are the same construct, no
crossed profiles should exist in our sample of respondents that display
high scorers in one variable but not the other. We extracted an
increasing number of latent profiles using the z-standardized scores of
depression, and burnout as indicators. To aid in the class identification,
we used: the loglikelihood (the smaller, the better), the Bayesian In­
formation Criterion (the smaller, the better), the entropy coefficient (the
larger, the better), and the Vuon-Lo-Mendell Rubin test (a non-
significant p-value indicates that the model does not fit better than a
model with a one less class) (Nylund et al., 2007). We inspected the
relative size of the retained classes as well in each solution, as we did not
want to keep classes with potentially spurious or unreplicable response
patterns. Accordingly, we added as a criterion that the best solution
should not contain any classes comprising less than 5% of the sample.
Lastly, we compared the latent classes as to their means on work stress.
The data that support the findings of this study are public available and
can be downloaded from https://github.com/hauck-nf/Depression-a
nd-burnout-symptoms.
3. Results
First, we conducted a series of factor analyses to test the structural
validity of the scales employed in this study. Whereas the one-factor
model of the BBS (burnout scale) yielded a poor fit to the data, χ2
(170) = 1962.81, RMSEA = .135, CFI = .839, TLI = .82, an exploratory
model with two correlated factors exhibited a much-improved fit, χ2
(151) = 543.29, RMSEA = .067, CFI = .965, TLI = .956. However, a
closer inspection of the resulting pattern matrix of factor loadings
clearly indicated that the two factors comprised items with positive
versus negative semantics. As mathematically proved and empirically
demonstrated many times, the separation of items from the same
construct into distinct negatively correlated factors might be a symptom
of acquiescent responding (Ferrando and Lorenzo-Seva, 2010). For that
reason, we suspected that a unidimensional model would fit the data
once we control acquiescence using a random intercept factor model
(Aichholzer, 2014). This was confirmed, χ2(169) = 550.44, RMSEA =
.062, CFI = .966, TLI = .962. Therefore, although this provided evidence
for the unidimensionality of the BBS, it indicated as well that any
combined analysis of the BBS and the EBADEP-brief should also be
conducted controlling for acquiescence. The EBADEP-brief one-factor
model yielded a reasonable fit to the data, χ2(90) = 631.39, RMSEA =
.102, CFI = .931, TLI = .920. Including a residual correlation between
two items about poor sleeping increased fit, χ2(89) = 513.79, RMSEA =
.091, CFI = .946, TLI = .936.
Next, we wanted to test the latent correlation between depression
and burnout unique factors after controlling for a general factor, and
acquiescence. As our intention was to estimate the correlation between
the unique components of both depression and burnout, we specified our
bifactor modeling without constraining these specific factors to be
orthogonal to each other. An initial simpler model with no control of a
general factor or acquiescence, but only the two factors of depression
and burnout connected to their corresponding items fitted poorly to the
data, χ2(559) = 2,541.09, RMSEA = .078, CFI = .773, TLI = .759.
However, it also revealed a large latent correlation of r = .75 between
depression and burnout, consistent with the findings from Bianchi et al.
(2021), in which the correlation was estimated at .80. Sequentialy, we
tested our target models controlling for acquiescence, and for acquies­
cence plus a general factor. The parameter estimates for each model are
presented in Table 1. While the random intercept model yielded a good
fit to the data, χ2(558) = 1,394.25, RMSEA = .051, CFI = .955, TLI =
.952, the bifactor random intercept yielded an even better fit, χ2(524) =
1,150,12, RMSEA = .045, CFI = .966, TLI = .962. Acquiescence proved
to be a non-negligble influence on item responses, as the factor loadings
on the random intercept factor were .32. However, even after control­
ling for acquiescence, the large depression and burnout latent factors
remained almost unchanged, r = .74; including the control of a general
M.N. Baptista et al. Journal of Affective Disorders Reports 10 (2022) 100437

Table 1 Table 1 (continued )

Factor loadings of the random intercept models. Item Random intercept Bifactor, random intercept
Item Random intercept Bifactor, random intercept model model
model model DEP BURN ACQ FG DEP BURN ACQ
DEP BURN ACQ FG DEP BURN ACQ
18. I feel important − .70 .32 − .78 − .17 .32
1. I have felt like .63 .32 .26 .62 .32 in the performance
crying of my tasks
2. I have felt more .70 .32 .29 .70 .32 19. I feel willing − .67 .32 − .62 − .30 .32
anguish or when I wake up
distressed and think about
3. I felt less able to .66 .32 .56 .37 .32 the activities I will
face my problems perform
4. I no longer feel like .66 .32 .51 .43 .32 20. I feel − .69 .32 − .57 − .39 .32
doing things I used professionally
to like fulfilled
5. I don’t believe .58 .32 .38 .44 .32 Correlation .74 .57
things will get depression ×
better burnout
6. I was happier .73 .32 .33 .70 .32
Note. GF = General factor, DEP = specific factor of depression, BURN = specific
before
7. I can no longer .73 .32 .68 .32 .32 factor of burnout, ACQ = random intercept factor (acquiescent responding).
concentrate
8. I am slower to .76 .32 .69 .37 .32 factor decreased the correlation to r = .57. We used the Omega co­
perform my tasks
efficients to weight the relative size of these variance components.
9. My life is getting .76 .32 .35 .72 .32
worse Omega hierarchical ωH was .75, which confirms the salience of the
10. I have been .66 .32 .41 .53 .32 general factor over the residual specific factors from the bifactor model.
thinking that it The explained common variance due to this general factor was 59%.
would be better to However, omega total was ωT was .95, which also indicates the residual
be dead
11. I can no longer .56 .32 .31 .49 .32
factors did add some reliable variance in the model.
sleep through the Next, we also addressed the separability of the burnout and depres­
night sion constructs using a graphic procedure. Results can be visualized in
12. I am more tired .81 .32 .59 .56 .32 Fig. 1. The first network was performed with no control of acquiescent
13. I have been .63 .32 .39 .49 .32
responding, while the second included a random intercept procedure
sleeping a lot
14. I have been losing .11 .32 .17 − .02 .32 that computes the interconections of variables after taking this response
weight without bias into account. One can see that the left-side graph reveals a
dieting distinction between burnout and depression, but also the same spurious
15. I have been .52 .32 .36 .37 .32 separation between positively- and negatively-framed burnout items
feeling guilt about
problems
seen in the factor analysis. The right-side graph offers a more inter­
1. I can do my − .79 .32 − .57 − .56 .32 pretable network. Bearing in mind that red and green edges represent
activities in a good negative and positive associations, respectively, the five clusters consist
mood of: (1) the core burnout features of emotional exhaustion, (2) lack of self-
2. I feel I have − .75 .32 − .52 − .54 .32
confidence, (3) sleep problems, (4) depressive mood, and (5) lack of
enough energy to
complete my tasks energy and lack of prestige/fullfilment. One the one hand, one can see
3. My job demands .46 .32 .33 .31 .32 the network analysis somehow separated depression and burnout
more than I can symptoms, as clusters 1 and 5 represent features of burnout, while
handle groupings 2, 3, and 3 represent facets of depression. On the other hand,
4. I get tired after .51 .32 .29 .43 .32
completing my day
interconnections between both constructs were numerous, as they
5. I feel frustrated .69 .32 .38 .59 .32 indeed formed a network of associated indicators. Worth mentioning is
6. I feel exhausted .75 .32 .37 .67 .32 that central nodes in the network were three burnout symtoms: BURN15
7. I feel energized by − .57 .32 − .51 − .29 .32 (“My job is exciting”), BURN02 (“I feel I have enough energy to com­
the tasks of my job
plete my tasks”), and BURN07 (“I feel energized by the tasks of my job”).
8. I am dissatisfied .59 .32 .28 .55 .32
with my job In other words, bridge symptoms were mainly features of lack of energy
9. I am uninterested .68 .32 .37 .59 .32 or excitement with job-related tasks— indicators of fatigue.
in my job Next, we were interested in investigating the relationship between
10. I am drained of .72 .32 .41 .60 .32 depression and burnout from a latent profile analysis perspective. To do
energy
11. My dedication − .63 .32 − .48 − .41 .32
so, we extracted an increasing number of latent profiles, avoiding classes
continues as before that comprised less than 5% of the sample. Results can be found in
12. I am worn out .77 .32 .36 .72 .32 Table 2. As can be seen in Table 2, by following the VLMRT and the BIC
13. I feel disqualified .40 .32 .30 .26 .32 criteria, we would need to increase the number of classes to more than
14. I am .66 .32 .20 .71 .32
six. Nevertheless, the six-class solution resulted in a class with only
disappointed with
my job 1.23% of the sample (6 individuals), the reason why we decided to keep
15. My job is exciting − .48 .32 − .46 − .21 .32 only five classes.
16. The thought of .71 .32 .47 .53 .32 The resulting latent profiles are described in Fig. 2. Whereas Classes
having to work 1 (33.25 %) and 5 (6.14 %) represented individuals with low and high
makes me tired
17. I have been − .60 .32 − .65 − .16 .32
scores in both depression and burnout, respectively, the remaining
progressing classes consisted of profiles with distinct level combinations. We found
professionally classes in which elevations in depression where not followed by eleva­
tions in burnout, and vice-versa. For instance, Class 2 (9,50 %) included

4
M.N. Baptista et al. Journal of Affective Disorders Reports 10 (2022) 100437

Fig. 1. EGA of depression and burnout symptoms.

Note. Left side: with no control of acquiescence. Right side: with control of acquiescence.

linear relationship between class membership and level of work stress.

Table 2
The Tukey post hoc test indicated significantly higher means on work
Results of the latent profile analyses.
stress for classes 4 and 5 in the comparison with the remaining classes.
Model LL BIC Entropy VLMRT (p % smallest Class 3 was located in-between, with significantly higher scores on work
value) class
stress relative to classes 1 and 2, which did not differ to each other. As
2− class − 1207.34 2458.90 .61 p < .001 45.89 the classes resulted primarily ordered relative to burnout scores, this
3− class − 1158.75 2380.68 .74 p < .001 7.87 ordering was predominantly related to burnout.
4− class − 1140.48 2363.09 .68 .030 5.21
5− class − 1118.11 2337.29 .72 .049 6.14
6-class − 1101.90 2323.84 .76 p < .001 1.23 4. Discussion
Note. LL = Loglikelihood, BIC = Bayesian Information Criterion, BLR = Boot­
strap Likelihood Ratio test. In this study, we hypothesized that if depression and burnout
symptoms form overlapping syndromes, they should: (1) remain corre­
lated after partialling a general factor in a bifactor model with correlated
individuals with slightly above the mean scores in depression, but
cluster/specific factors; (2) still positively correlate after we control for
slightly below the mean scores in burnout, while Class 3 (25.88%)
acquiescent responding; (3) still have their symptom communities
included adults with higher scores in burnout than in depression. Latent
overlapping in an exploratory graph (network) analysis even after
classes were ordered as to their means on burnout, but this ordering did
controlling acquiescence; and (4) produce no salient latent profiles that
not occur regarding depression.
would characterize a high level in one construct, but low in other. Our
Next, we were interested in testing the relationship between being a
findings largely supports the notion that depression and burnout are
member in each of these classes and displaying work stress, by
elements connected to the same latent phenomenon.
comparing means between groups, which can be seen in Fig. 3. The
The evidence we found confirms that the linear dependence of
general linear model indicated significant differences between classes in
depression and burnout is not attributable either exclusively to a gen­
their work stress levels, F(4, 391) = 25.06, p < .001, η2 = .26. Post hoc
eral, non-specific distress factor [hypothesis 1] or to response bias
tests (Tukey) were also performed to investigate more specific mean
[hypothesis 2]. The estimated latent correlation between burnout and
differences between classes. As one can see in Fig. 3, there was a clear
depression was .75, and virtually the same (.74) after acquiescence was

Fig. 2. Latent profile analysis of depression and burnout symptoms.

5
M.N. Baptista et al.
Fig. 3. Latent classes and z-standardized means in work stress.
controlled in the model. This is in line with the r = .80 obtained in the
meta-analysis of Bianchi et al. (2021), which further supports the con­
clusions reached by those authors, that burnout problematically over­
laps with depression. We also went further and partialled out a general
domain in an oblique bifactor model with a random intercept factor. As
reported in a large twin-study, a general factor common to major
depression, generalized anxiety disorder, and burnout is largely influ­
enced by genetics (58%) (Mather et al., 2016). The estimate was now
.57, which is smaller, but still high for two putatively independent
constructs. Just for the sake of comparison, a meta-analysis of person­
ality scales indicated that the typical convergent validity coefficient for
scales measuring the same construct falls in the range of .30—.60.
Hence, our findings add to the empirical literature questionnaing the
status of burnout and depression as distinct constructs. Apparently, their
overlap is (1) not an artifact of response biases, and (2) not solely
because they share variance due to a general mental disorder liability.
Our investigation also considered the overlap issue from the
perspective of exploratory graph analysis [hypothesis 3]. The separation
between burnout and depression was clear before controlling acquies­
cence, but not after controlling this response bias. As found in a previous
network analysis study (Verkuilen et al., 2020), small clusters of in­
dicators with similar wording might cause items to cluster together. We
reached a similar result: More specifically, item groups emerged that
contained wording related to 1) the core burnout features of emotional
exhaustion, 2) lack of self-confidence, 3) sleep problems, 4) depressive
mood, and 5) lack of energy and lack of prestige/fullfilment. Never­
theless, the overall closeness of the network nodes was observably larger
once acquiescence was taken into account, and the network resembled a
coeherent single syndrome of symptoms and signs. Once again, this not
only confirmed the lack of empirical distinction between these con­
structs, but also indicated the importance of considering response biases
as potentially confounding variables. We should stress that acquiescence
is a sample-dependent phenomenon, i.e., this is a feature of persons, not
of items or instruments. The bias could represent much of a problem in
some samples more than others.
Another interesting finding from our network analysis was the bridge
nodes capturing fatigue. Indeed, some have claimed before that the
connection between burnout and depression are primarily driven by
fatigue indicators (Maslach and Leiter, 2016). Fatigue is one of the main
somatic symptoms in depressive disorder, which tends to cause a slow
response to antidepressant treatment and psychotherapy (Demytte­
naere et al., 2005). Moreover, “fatigue and lack of energy nearly
everyday” is one of the DSM-5-TR diagnostic criteria for Major
Depression that has explicit implications for occupational impairment
(APA, 2022). The importance of this indicator have also been recognized
in the Bianchi and Schonfeld (2020) 9-item recently published inventory
6
Journal of Affective Disorders Reports 10 (2022) 100437
of “occupational depression” (“4. Fatigue/loss of energy “I felt exhaus­
ted because of my work””). Hence, the findings from our network
analysis not only confirm the salience of fatigue to burnout/occupa­
tional depression, but also reveal that this is exactly what might trigger
all the other symptom components of the common burnout/depression
network. Even if depressed mood and anhedonia are core to depression
conditions (APA, 2022), fatigue may be the main component that is
responsible for activating those symptoms in patients. In other words,
the emergence of fatigue can trigger the many other symptoms in the
network, the reason why fatigue has a clinical value, as it may signal the
initial course of more severe mental suffering at work.
Lastly, we found only weak evidence of specific latent profiles in
which depression and burnout symptoms were not comorbid [hypoth­
esis 4]. Crossed profiles indeed occurred in our sample of nursing pro­
fessionals, namely, classes 3 and 4 (approximately 35%). However, even
if these two classes represented more than 35% of our entire sample, in
none of them have we observed extremely contrasting scores of burnout
and depression. Rather, one can say that classes 2, 3, and 4 merely
represent moderate levels in both constructs. As expected to variables
that are pathological, the most prevalent class was class 1, the one
comprising low scorers in both burnout and depression. Conversely,
class 5 included individuals with the highest scores in both variables.
Similarly to previous latent profile/class studies, our findinds support
the view of burnout and depression as intrinsically comorbid entities
(Ahola et al., 2014; Bianchi et al., 2016; Bianchi and Brisson, 2019).
One more result is worth a mention. The classes exhibited mean­
ingful differences in their work stress scores. Issues such as low work
satisfaction, unprepared staff, and lack of organizational support can
elicit or intensify burnout symptoms (Costello et al., 2019), especially
emotional exhaustion, one of the core burnout dimensions (Seidler et al.,
2014). We detected a linear trend, so that the same ordering across the
latent profiles regarding burnout replicated when the classes were
compared for their work stress scores. One possible explanation is that
the latent classes reflected a continuum of occupational distress. This
might be taken as evidence of dimensionality of burnout, then rejecting
the conception of the construct as a categorical latent entity, confirming
the modern conceptions about the construct (Bianchi, 2020; Bianchi
et al., 2021; Schonfeld and Bianchi, 2021). We should also recognize
that the specific occupational content both in the items of burnout and
work stress inventories can increase the correlation between them
(Bianchi et al., 2021). We do not state that work stress is more connected
to burnout than depression, but only that the ordering in the latent
classes we found involves a continuum of occupational distress.
This study is not exempt of limitations. It is worth mentioning that
our findings resulted entirely from analysis conducted on data collected
using self-report inventories, the reason why replication with data
collected with other sources of information is encouraged. Using gold
standard measures (e.g., clinical diagnostic according to DSM-5 criteria
for Major Depressive Disorder) would provide more reliable assessments
of depression. Moreover, despite our data intentionally included only
nursing professionals, they were mainly females. Gender invariance
should not be taken as granted, but rather tested, the reason why maybe
other replication studies are still necessary. By contrast, we consider our
findings to be in line with results from many other sites and using much
more diverse data (Bianchi et al., 2021). Even if having limitations, the
current investigation add further support to the growing body of evi­
dence that depression and burnout are part of the same latent entity.
Author statement
Contributors
Prof. Dr. Makilim Nunes Baptista
Participated in the design of the study, the data analysis, and the
writing of the manuscript.
Prof. Dr. Nelson Hauck-Filho (corresponding author)
Participated in the data analysis, and the writing of the manuscript.
M.N. Baptista et al.
Prof. Dr. Hugo Ferrari Cardoso
Participated in the design of the study, the data collection, and the
writing of the manuscript.
Role of the funding source
This work received no financial assistance from any funding in­
stitutions. The research was conducted in accordance with the Helsinki
Declaration as revised 1989.
Declaration of Competing Interest
This work received no financial assistance from any funding in­
stitutions. All the authors declare that they have no financial interest in
the research. The data that support the findings of this study are public
available and can be downloaded from https://github.com/hauck-nf/De
pression-and-burnout-symptoms. Prof. Dr. Makilim Nunes Baptista,
Universidade São Francisco, Graduate School of Psychology, Campinas,
Brazil. Prof. Dr. Nelson Hauck-Filho (corresponding author), Uni­
versidade São Francisco, Graduate School of Psychology, Campinas,
Brazil. Prof. Dr. Hugo Ferrari Cardoso, Universidade Estadual Paulista,
Graduate School of Developmental and Learning Psychology, Bauru,
Brazil.
Acknowledgments
None.
References
Ahola, K., Hakanen, J., Perhoniemi, R., Mutanen, P., 2014. Relationship between
burnout and depressive symptoms: a study using the person-centred approach. Burn.
Res. 1 (1), 29–37. https://doi.org/10.1016/j.burn.2014.03.003.
Ahola, K., Honkonen, T., Isometsä, E., Kalimo, R., Nykyri, E., Aromaa, A., Lönnqvist, J.,
2005. The relationship between job-related burnout and depressive
disorders—results from the finnish health 2000 study. J. Affect. Disord. 88 (1),
55–62. https://doi.org/10.1016/j.jad.2005.06.004.
Aichholzer, J., 2014. Random intercept EFA of personality scales. J. Res. Pers. 53, 1–4.
https://doi.org/10.1016/j.jrp.2014.07.001.
APA, A.P.A., 2022. Diagnostic and Statistical Manual of Mental Disorders: Text Revised -
DSM-5-TR (Fifth Edition). American Psychiatric Association Publishing.
Auerbach, R.P., Mortier, P., Bruffaerts, R., Alonso, J., Benjet, C., Cuijpers, P.,
Demyttenaere, K., Ebert, D.D., Green, J.G., Hasking, P., Murray, E., Nock, M.K.,
Pinder-Amaker, S., Sampson, N.A., Stein, D.J., Vilagut, G., Zaslavsky, A.M.,
Kessler, R.C., 2018. WHO world mental health surveys international college student
project: prevalence and distribution of mental disorders. J. Abnorm. Psychol. 127
(7), 623–638. https://doi.org/10.1037/abn0000362.
Bianchi, R., 2020. Do burnout and depressive symptoms form a single syndrome?
confirmatory factor analysis and exploratory structural equation modeling bifactor
analysis. J. Psychosom. Res. 131, 109954 https://doi.org/10.1016/j.
jpsychores.2020.109954.
Bianchi, R., Brisson, R., 2019. Burnout and depression: causal attributions and construct
overlap. J. Health Psychol. 24 (11), 1574–1580. https://doi.org/10.1177/
1359105317740415.
Bianchi, R., Schofield, I.S., Mayor, E., Laurent, E., 2016. Burnout-depression overlap: a
study of New Zealand schoolteachers. N. Z. J. Psychol. 45 (3), 4–11.
Bianchi, R., Schonfeld, I.S., 2020. The occupational depression inventory: a new tool for
clinicians and epidemiologists. J. Psychosom. Res. 138, 110249 https://doi.org/
10.1016/j.jpsychores.2020.110249.
Bianchi, R., Schonfeld, I.S., Laurent, E., 2015. Burnout–depression overlap: a review.
Clin. Psychol. Rev. 36, 28–41. https://doi.org/10.1016/j.cpr.2015.01.004.
Bianchi, R., Schonfeld, I.S., Verkuilen, J., 2020. A five-sample confirmatory factor
analytic study of burnout-depression overlap. J. Clin. Psychol. 76 (4), 801–821.
https://doi.org/10.1002/jclp.22927.
Bianchi, R., Verkuilen, J., Schonfeld, I.S., Hakanen, J.J., Jansson-Fröjmark, M., Manzano-
García, G., Laurent, E., Meier, L.L., 2021. Is burnout a depressive condition? A 14-
sample meta-analytic and bifactor analytic study. Clin. Psychol. Sci. 9 (4), 579–597.
https://doi.org/10.1177/2167702620979597.
Brandford, A.A., Reed, D.B., 2016. Depression in registered nurses. Work. Health Saf. 64
(10), 488–511. https://doi.org/10.1177/2165079916653415.
Journal of Affective Disorders Reports 10 (2022) 100437
Cardoso, H.F., Filho, N.H., Baptista, M.N., Valentini, F., 2022. Brazilian burnout scale
(BBS): internal structure and acquiescence control. Psicologia: Teoria e Pesquisa.
Chen, C., Meier, S.T., 2021. Burnout and depression in nurses: a systematic review and
meta-analysis. Int. J. Nurs. Stud. 124, 104099 https://doi.org/10.1016/j.
ijnurstu.2021.104099.
Costello, H., Walsh, S., Cooper, C., Livingston, G., 2019. A systematic review and meta-
analysis of the prevalence and associations of stress and burnout among staff in long-
term care facilities for people with dementia. Int. Psychogeriatr. 31 (08),
1203–1216. https://doi.org/10.1017/S1041610218001606.
Demyttenaere, K., de Fruyt, J., Stahl, S.M., 2005. The many faces of fatigue in major
depressive disorder. Int. J. Neuropsychopharmacol. 8 (1), 93–105. https://doi.org/
10.1017/S1461145704004729.
Ferrando, P.J., Lorenzo-Seva, U., 2010. Acquiescence as a source of bias and model and
person misfit: a theoretical and empirical analysis. Br. J. Math. Stat. Psychol. 63 (2),
427–448. https://doi.org/10.1348/000711009X470740.
Golino, H.F., Epskamp, S., 2017. Exploratory graph analysis: a new approach for
estimating the number of dimensions in psychological research. PLoS One 12 (6),
e0174035. https://doi.org/10.1371/journal.pone.0174035.
Gonçalves, A.P., Baptista, M.N., Villemor-Amaral, A.E., Carvalho, L., de, F., 2021.
Diagnosis accuracy of baptista depression scale: adult and screening versions.
Psicologia: Teoria e Pesquisa 37. https://doi.org/10.1590/0102.3772e37449.
Jennrich, R.I., Bentler, P.M., 2012. Exploratory bi-factor analysis: the oblique case.
Psychometrika 77 (3), 442–454.
Koutsimani, P., Montgomery, A., Georganta, K., 2019. The relationship between burnout,
depression, and anxiety: a systematic review and meta-analysis. Front. Psychol. 10
https://doi.org/10.3389/fpsyg.2019.00284.
Mansolf, M., Reise, S.P., 2016. Exploratory bifactor analysis: the schmid-leiman
orthogonalization and jennrich-bentler analytic rotations. Multivar. Behav. Res. 51
(5), 698–717. https://doi.org/10.1080/00273171.2016.1215898.
Maslach, C., Jackson, S.E., 1981. The measurement of experienced burnout. J. Organ.
Behav. 2 (2), 99–113. https://doi.org/10.1002/job.4030020205.
Maslach, C., Leiter, M.P., 2016. Understanding the burnout experience: recent research
and its implications for psychiatry. World Psychiatry 15 (2), 103–111. https://doi.
org/10.1002/wps.20311.
Mather, L., Blom, V., Bergström, G., Svedberg, P., 2016. An underlying common factor,
influenced by genetics and unique environment, explains the covariation between
major depressive disorder, generalized anxiety disorder, and burnout: a Swedish
twin study. Twin Res. Hum. Genet. 19 (6), 619–627. https://doi.org/10.1017/
thg.2016.73.
Maydeu-Olivares, A., Coffman, D.L., 2006. Random intercept item factor analysis.
Psychol. Methods 11 (4), 344–362. https://doi.org/10.1037/1082-989X.11.4.344.
Nylund, K., Asparouhov, T., Muthén, B., 2007. Deciding on the number of classes in
latent class analysis and growth mixture modeling: a Monte Carlo simulation study.
Struct. Equ. Model. 14 (4), 535–569.
Paulhus, D.L., 1991. Measurement and control of response styles. J. P. Robinson, P. R.
Shaver, & L. S. Wrightsman Measures of Personality and Social Psychological
Attitudes. Academic Press, pp. 17–59.
Plieger, T., Melchers, M., Montag, C., Meermann, R., Reuter, M., 2016. Life stress as
potential risk factor for depression and burnout. Pers. Individ. Differ. 101, 505.
https://doi.org/10.1016/j.paid.2016.05.255.
Reise, S.P., 2012. The rediscovery of bifactor measurement models. Multivar. Behav. Res.
47 (5), 667–696. https://doi.org/10.1080/00273171.2012.715555.
Revelle, W., & Revelle, M. (2015). Package ‘psych.’ In The Comprehensive R Archive
Network.
Schonfeld, I.S., Bianchi, R., 2021. From burnout to occupational depression: recent
developments in research on job-related distress and occupational health. Front.
Public Health 9. https://doi.org/10.3389/fpubh.2021.796401.
Schonfeld, I.S., Verkuilen, J., Bianchi, R., 2019. An exploratory structural equation
modeling bi-factor analytic approach to uncovering what burnout, depression, and
anxiety scales measure. Psychol. Assess. 31 (8), 1073–1079. https://doi.org/
10.1037/pas0000721.
Seidler, A., Thinschmidt, M., Deckert, S., Then, F., Hegewald, J., Nieuwenhuijsen, K.,
Riedel-Heller, S.G., 2014. The role of psychosocial working conditions on burnout
and its core component emotional exhaustion – a systematic review. J. Occup. Med.
Toxicol. 9 (1), 10. https://doi.org/10.1186/1745-6673-9-10.
Sisto, F.F., Baptista, M.N., Noronha, A.P.P., Santos, A.A.A., 2007. Escala de
vulnerabilidade ao estresse no trabalho (EVENT). Vetor.
Tavella, G., Parker, G., 2020. Distinguishing burnout from depression: an exploratory
qualitative study. Psychiatry Res. 291, 113212 https://doi.org/10.1016/j.
psychres.2020.113212.
van Vaerenbergh, Y., Thomas, T.D., 2013. Response styles in survey research: a literature
review of antecedents, consequences, and remedies. Int. J. Public Opin. Res. 25 (2),
195–217. https://doi.org/10.1093/ijpor/eds021.
Verkuilen, J., Bianchi, R., Schonfeld, I.S., Laurent, E., 2020. Burnout–depression overlap:
exploratory structural equation modeling bifactor analysis and network analysis.
Assessment. https://doi.org/10.1177/1073191120911095, 107319112091109.
Wurm, W., Vogel, K., Holl, A., Ebner, C., Bayer, D., Mörkl, S., Szilagyi, I.S., Hotter, E.,
Kapfhammer, H.-P., Hofmann, P., 2016. Depression-burnout overlap in physicians.
PLoS One 11 (3), e0149913. https://doi.org/10.1371/journal.pone.0149913.