Journal of Psychosomatic Research 138 (2020) 110249

Contents lists available at ScienceDirect

Journal of Psychosomatic Research

journal homepage: www.elsevier.com/locate/jpsychores

The Occupational Depression Inventory: A new tool for clinicians and T

epidemiologists
Renzo Bianchia, , Irvin Sam Schonfeldb
⁎

a
Institute of Work and Organizational Psychology, University of Neuchâtel, Neuchâtel, NE, Switzerland
b
Department of Psychology, The City College of the City University of New York, New York City, NY, USA

ARTICLE INFO ABSTRACT

Keywords: Background: Depressive symptoms induced by insurmountable job stress and sick leave for mental health rea­
Bifactor analysis sons have become a focal concern among occupational health specialists. The present study introduces the
Burnout Occupational Depression Inventory (ODI), a measure designed to quantify the severity of work-attributed de­
Depression pressive symptoms and establish provisional diagnoses of job-ascribed depression. The ODI comprises nine
Job strain
symptom items and a subsidiary question assessing turnover intention.
Occupational health
Work stress
Methods: A total of 2254 employed individuals were recruited in the U.S., New Zealand, and France. We ex­
amined the psychometric and structural properties of the ODI as well as the nomological network of work-
attributed depressive symptoms. We adopted an approach centered on exploratory structural equation modeling
(ESEM) bifactor analysis. We developed a diagnostic algorithm for identifying likely cases of job-ascribed de­
pression (SPSS syntax provided).
Results: The ODI showed strong reliability and high factorial validity. ESEM bifactor analysis indicated that, as
intended, the ODI can be used as a unidimensional measure (Explained Common Variance = 0.891). Work-
attributed depressive symptoms correlated in the expected direction with our other variables of interest―e.g.,
job satisfaction, general health status―and were markedly associated with turnover intention. Of our 2254
participants, 7.6% (n = 172) met the criteria for a provisional diagnosis of job-ascribed depression.
Conclusions: This study suggests that the ODI constitutes a sound measure of work-attributed depressive
symptoms. The ODI may help occupational health researchers and practitioners identify, track, and treat job-
ascribed depression more effectively. ODI-based research may contribute to informing occupational health po­
licies and regulations in the future.

1. Introduction rewarding experiences on the one hand, and negative, punitive ex­
periences on the other hand [11–13]. Situations involving unresolvable
Depression is a major contributor to the burden of disease, with stress, in which individuals are sentenced to endure the harmful effects
more than 300 million individuals affected worldwide [1,2]. The life­ of stressors that cannot be neutralized, have long been identified as key
time prevalence of major depression exceeds 15% in countries such as depressogenic factors [14–17]. Depression is predictive of a constella­
the U.S. and appears to be on the rise for several decades [3–5]. De­ tion of health disturbances and morbidities fostered by unresolvable
pressive conditions are primarily characterized by dysphoric mood and stress, including immune and neurological alterations [18–20], cardi­
anhedonia (i.e., loss of pleasure and interest in activities previously ovascular disease [21,22], diabetes [23], osteoporosis [24], accelerated
experienced as enjoyable), with suicidal ideation an important severity aging [25], dementia [26], and cancer [27]. Depression is also a prime
marker [6–8]. While depression is nosologically defined and diag­ risk factor for suicide [28], consistent with the view that “suicide occurs
nosable [6], there is robust evidence that depression is best conceived when the perspective of dying has become definitely more rewarding
of as a dimensional phenomenon, on a continuum from euthymia to than the perspective of going on living” [29] (p. 192). In light of these
full-blown depressive disorders [9–11]. findings, preventing and treating depression is crucial for promoting
From an etiological standpoint, the development of depressive individuals' overall health and longevity.
symptoms has been linked to a discrepancy between positive, Over the last few decades, depressive symptoms induced by job

⁎
Corresponding author at: Institute of Work and Organizational Psychology, University of Neuchâtel, Émile-Argand 11, 2000 Neuchâtel, NE, Switzerland.
E-mail addresses: renzo.bianchi@unine.ch (R. Bianchi), ischonfeld@ccny.cuny.edu (I.S. Schonfeld).

https://doi.org/10.1016/j.jpsychores.2020.110249
Received 22 May 2020; Received in revised form 28 August 2020; Accepted 11 September 2020
0022-3999/ © 2020 The Authors. Published by Elsevier Inc. This is an open access article under the CC BY-NC-ND license
(http://creativecommons.org/licenses/BY-NC-ND/4.0/).
R. Bianchi and I.S. Schonfeld Journal of Psychosomatic Research 138 (2020) 110249

stress and sick leave for mental health reasons have become a focal to 500?”). Participants selecting any other option than “0” were ex­
concern among occupational health specialists [30,31]. In Switzerland, cluded. Second, we included an open-ended, qualitative question about
for instance, sick leave for mental health reasons has reportedly in­ life stress. Any out-of-scope or incomprehensible answer was elim­
creased by 50%–70% in less than 10 years [32]. The cost of depression inatory. Third, at the end of the survey, we asked respondents to in­
in the workplace is in billions of U.S. dollars in Western countries and is dicate whether they had responded randomly to any questions. Parti­
considered an individual-, an organizational-, and a society-level pro­ cipants who disclosed random responses were removed from the
blem. To date, however, no instrument has been developed to assess sample. Of the 350 respondents who initially took the survey, 10.9%
depressive symptoms that individuals specifically ascribe to their work (n = 38) were identified as careless and excluded. Fifty-seven percent
[12]. While numerous depression scales are available, such scales assess of the final respondents were females. Respondents' mean age was
depressive symptoms without etiological considerations. The absence of 41.28 (SD = 9.94). Sample 3 is further described in Supplementary
a measure of work-ascribed depressive symptoms is problematic for Material 3.
occupational health specialists, for example when it comes to deciding All participants completed Internet surveys administered with
whether work-centered interventions or new labor regulations are Qualtrics (https://www.qualtrics.com/). Internet surveys have proved
needed. Although worryingly high levels of depressive symptoms have as reliable and valid as paper-and-pencil surveys [41]. The study was
been documented in certain occupational groups [33], the extent to conducted in compliance with the ethical standards of the institutional
which affected individuals consider these symptoms job-related is, in review board of the University of Neuchâtel.
most cases, unclear [34].
The present study introduces the Occupational Depression
2.2. Measures of interest
Inventory (ODI), a measure designed to assess the severity of work-
attributed depressive symptoms and establish provisional diagnoses of
2.2.1. ODI
job-ascribed depression. The ODI thus approaches depression from both
We developed the ODI with reference to the nine diagnostic criteria
a dimensional (quantitative) and a categorical (qualitative) standpoint.
for major depression of the Diagnostic and statistical manual of mental
We report on the development of the ODI in two languages―English
disorders, fifth edition (DSM-5) [6]. The ODI thus includes symptom
and French―across three countries―the U.S., New Zealand, and
items aiming to assess anhedonia, depressed mood, sleep alterations,
France. We scrutinized the psychometric and structural properties of
fatigue/loss of energy, appetite alterations, feelings of worthlessness,
the ODI relying on exploratory structural equation modeling (ESEM)
cognitive impairment, psychomotor alterations, and suicidal ideation
[35]. We used ESEM for the purpose of both exploratory factor analysis
(Table 1). Consistent with DSM-5 diagnostic criteria for major depres­
(EFA) and bifactor analysis―bifactor analysis is particularly well-suited
sion, respondents are asked to report on symptoms experienced over the
for examining scale dimensionality [36]. We examined the ODI's no­
past two weeks. Items are rated on a 4-point scale, from 0 for “never or
mological network to assess the scale's criterion validity. We inspected
almost never” to 3 for “nearly every day.” Instead of assessing de­
the ODI's relationships with a variety of work-contextualized (e.g., job
pressive symptoms in a “cause-neutral” manner, each ODI item involves
satisfaction) and context-free (general health status) measures. By de­
causal attributions to respondents' work/job (e.g., “My experience at
veloping the ODI, our aim is to provide occupational health specialists
work made me feel like a failure”). The ODI also includes a subsidiary
with a tool that (a) allows for a better identification, monitoring, and
question related to turnover intention: “If you have encountered at least
treatment of job-ascribed depression and (b) helps to inform occupa­
some of the problems mentioned above, do these problems lead you to
tional health policies and regulations on a global scale. The develop­
consider leaving your current job or position?” Three response options
ment of the ODI responds to a long-expressed need for tailored as­
are provided: “yes,” “no,” and “I don't know.” This complementary item
sessment tools in occupational health science [34,37].
is intended to help investigators assess the concrete work implications
of the depressive symptoms reported. The instructions to respondents
2. Methods
stipulate that the questions asked concern the impact of the re­
spondents' work/job on themselves. In addition, the instructions to
2.1. Study samples
respondents emphasize that if respondents experienced the problems
presented for reasons they consider unconnected to their work/job or
A total of 2254 participants took part in this study. Participants
for reasons they cannot identify, they should select the “never or almost
came from three different samples recruited in three different countries.
never” option (reflected in a score of 0) when responding.
Most of the participants were employed as schoolteachers, an occupa­
The ODI was designed to (a) quantify the severity of work-attributed
tional group substantially affected by job stress [12].
depressive symptoms―dimensional approach―and (b) establish pro­
The first sample (Sample 1) comprised 1450 French schoolteachers
visional diagnoses of job-ascribed depression―categorical approach.1
(MAGE = 43.69, SDAGE = 9.56). Eighty-four percent were females.
The quantification of work-attributed depressive symptoms is
Respondents had been employed in the educational field for 18.56 years
straightforward. Work-attributed depressive symptoms are reflected in
on average (SD = 10.07). No compensation was offered. Sample 1 is
the ODI's sum (or mean) score, with higher scores signaling that an
further described in Supplementary Material 1.
individual is more severely affected. For establishing provisional diag­
The second sample (Sample 2) consisted of 492 schoolteachers
noses of job-ascribed depression, we created an algorithm inspired by
employed in New Zealand (MAGE = 47.09, SDAGE = 11.81). Eighty
the one developed for the PHQ-9 [42], a measure of reference in de­
percent were females. Respondents' mean length of employment was
pression research [3,12]. A provisional diagnosis is produced if an in­
18.54 years (SD = 12.59). Again, we offered no compensation. Sample
dividual exhibits a score of 3 on at least five of the nine ODI's symptom
2 is further described in Supplementary Material 2.
items and one of these symptom items is anhedonia (item 1) or de­
The third sample (Sample 3) was recruited through Amazon's
pressed mood (item 2). A score of 3 corresponds to symptoms experi­
Mechanical Turk (MTurk), an open online marketplace (https://www.
enced “nearly every day,” a frequency of symptoms that dovetails with
mturk.com/). Two qualification requirements were specified: (a) U.S.
DSM-5 diagnostic criteria for major depression [6]. The DSM-5 indeed
location and (b) full-time employment (i.e., 35+ hours per week). Each
indicates that “[t]he criterion symptoms for major depressive disorder
respondent was remunerated $0.50. MTurk can be used to obtain high-
must be present nearly every day to be considered present” (p. 162) [6].
quality data [38]; employing measures to detect careless respondents is
however recommended [39,40]. We relied on the following safeguards.
First, we included a bogus item (“On a scale from 0 to 10, and without 1
We talk of provisional diagnoses because the method of reference for diag­
speculating on possible advances in science, how likely are you to live nosing clinical forms of depression is the standardized clinical interview [3].

2
R. Bianchi and I.S. Schonfeld Journal of Psychosomatic Research 138 (2020) 110249

Table 1
Occupational Depression Inventory (ODI): Instructions to respondents and items.
Instructions to respondents

The following statements concern the impact your work could have had on you.
Please read each statement and indicate how often you experienced the problems mentioned over the PAST TWO WEEKS. Use the scale provided to respond:
0 = never or almost never
1 = a few days only
2 = more than half the days
3 = nearly every day
Here is an example: “I felt anxious because of my job.”
• If you did NOT feel anxious because of your job, select 0.
• If you felt anxious for reasons that you consider UNCONNECTED TO YOUR JOB (personal problems, marital problems, family problems, health problems, etc.), select 0 as
well.
• If you felt anxious but don't know why, again select 0.
• If it is clear for you that YOUR JOB caused you to feel anxious, select 1, 2 or 3 to indicate how often that happened.

Items

1. Anhedonia “My work was so stressful that I could not enjoy the things that I usually like doing.”
2. Depressed mood “I felt depressed because of my job.”
3. Sleep alterations “The stress of my job caused me to have sleep problems (I had difficulties falling asleep or staying asleep, or I slept much more than usual).”
4. Fatigue/loss of energy “I felt exhausted because of my work.”
5. Appetite alterations “I felt my appetite was disturbed because of the stress of my job (I lost my appetite, or the opposite, I ate too much).”
6. Feelings of worthlessness “My experience at work made me feel like a failure.”
7. Cognitive impairment “My job stressed me so much that I had trouble focusing on what I was doing (e.g., reading a newspaper article) or thinking clearly (e.g., to make
decisions).”
8. Psychomotor alterations “As a result of job stress, I felt restless, or the opposite, noticeably slowed down—for example, in the way I moved or spoke.”
9. Suicidal ideation “I thought that I'd rather be dead than continue in this job.”
SQ Turnover intention If you have encountered at least some of the problems mentioned above, do these problems lead you to consider leaving your current job or
position?

Notes. ODI forms are available in Supplementary Materials 4 (French version) and 5 (English version). An SPSS syntax implementing the provisional diagnosis
algorithm of the ODI is provided in Supplementary Material 6. SQ: subsidiary question.

Importantly, suicidal ideation (item 9) counts even with a score of 1 or 2.3. Data analyses
2 (symptoms experienced “a few days only” or “more than half the
days”). Suicidal ideation is given a special weight due to its intrinsic We examined the factor structure of the ODI based on ESEM EFA
gravity and alarm status [6,43].2 The state of the art indicates that there (using a geomin rotation) and ESEM bifactor analysis (using a bi-geomin
are no iatrogenic risks of assessing suicidality [45]. rotation). In the ESEM bifactor analysis, we ascertained whether the ODI
The full version of the ODI, which includes detailed instructions to can be viewed as essentially unidimensional by scrutinizing the loadings
respondents, is available in French in Supplementary Material 4, and in of ODI's items on the General factor and computing the Explained
English in Supplementary Material 5. An SPSS syntax implementing the Common Variance (ECV) index. The ECV index reflects the proportion of
provisional diagnosis algorithm of the ODI is provided in the common variance extracted that is accounted for by the General
Supplementary Material 6. factor; ECV values exceeding 0.80 are suggestive of essential uni­
dimensionality [47]. We treated the items as ordinal and employed the
weighted least squares—mean and variance adjusted—estimator [49].
2.2.2. Additional measures In addition, we employed ESEM bifactor analysis to examine the
In the interest of reducing response burden on participants, single- convergent and discriminant validity of the ODI vis-à-vis our cause-neutral
item measures were employed in all samples for assessing trait anxiety, measures of depression—the CES-D and the HADS-D. As our goal was
environmental quality, residential satisfaction, safety in daily life, confirmatory, we used a partially specified target rotation [35]. Because
general health status, social support in work life, social support outside all three scales are intended to assess depressive symptoms, we expected
of work, job satisfaction, and life satisfaction―the items are provided in the ODI to show convergent validity with the CES-D and the HADS-D. We
Supplementary Materials 1 to 3. thus anticipated that ODI, CES-D, and HADS-D items would all sub­
In Samples 1 and 2, we assessed (cause-neutral) depressive symp­ stantially load on the General factor. However, because the ODI assesses
toms with the 10-item version of the Center for Epidemiologic Studies work-attributed depressive symptoms whereas the CES-D and the HADS-D
Depression scale (CES-D; Cronbach's α = 0.831 and 0.850, respectively) assess depressive symptoms in a cause-neutral manner, we also expected
[46] and dedication to work with the dedication subscale of the Utrecht some degree of discriminant validity, as reflected in ECV indices markedly
Work Engagement Scale-Short Form (UWES-9; Cronbach's α = 0.855 below 0.80. All factor analyses were conducted with Mplus 8 [50].
and 0.845, respectively) [47]. In addition, willingness to stay in the job We estimated the reliability of the ODI based on Cronbach's α and
and active search for another job/position were assessed in Sample 1, McDonald's ω [51]. To investigate the ODI's nomological network, we
using single-item measures. In Sample 3, we assessed (cause-neutral) calculated Spearman's rank correlation coefficients. Finally, we relied
depressive symptoms with the depression subscale of the Hospital An­ on analysis of variance and Dunnett's T3 to examine the link between
xiety and Depression Scale (HADS-D; Cronbach's α = 0.869) [48]. The work-attributed depressive symptoms and turnover intention—as as­
HADS-D and the CES-D are widely used measures of depression [3]. sessed by the subsidiary question of the ODI.

3. Results

2
We did not include an independent “clinical significance” criterion in view Descriptive statistics pertaining to the measures employed are
of the uncertainty surrounding its use [44]. available in Supplementary Materials 1 to 3. In the three samples, the

3
R. Bianchi and I.S. Schonfeld Journal of Psychosomatic Research 138 (2020) 110249

Fig. 1. Exploratory structural equation modeling bifactor analysis of the Occupational Depression Inventory (ODI; N = 2254). The items load on average 0.819 on
the General factor. The items of the ODI are coded ODI1 to ODI9. GF: General factor; BF1: first bifactor; BF2: second bifactor.

most frequently endorsed ODI item was fatigue/loss of energy and the (no item loaded ≥0.30 on more than one factor), items 1, 2, 6, and 7
least frequently endorsed ODI item was suicidal ideation. In Sample 1, showed some degree of factorial complexity. The two factors correlated
7.7% of the participants (n = 111) met the criteria for a provisional 0.752 to 0.843 across the three samples.
diagnosis of job-ascribed depression; in Sample 2, 8.3% (n = 41); in
Sample 3, 6.4% (n = 20). The overall prevalence was 7.6% (n = 172). 3.2. ESEM bifactor analysis
Turnover intention was substantially linked to ODI scores in all sam­
ples, ps < 0.001 (see Supplementary Materials 1 to 3). Because the basic factor structure of the ODI was similar in the three
samples, we conducted our ESEM bifactor analysis merging all datasets
3.1. ESEM EFA (N = 2254). Results are summarized in Fig. 1 and Table 2. We extracted
two bifactors (one for anhedonic/somatic symptoms and one for dys­
ESEM EFAs were indicative of a similar two-factor structure in all phoric symptoms) in addition to the General factor. Our bifactor model
samples (Supplementary Material 7). The first factor was dominated by showed a good fit: RMSEA = 0.044; CFI = 0.999; TLI = 0.996. All ODI
anhedonic/somatic symptom items and the second factor, by dysphoric items loaded strongly on the General factor—from 0.758 to 0.897—and
symptom items. Although no substantial cross-loading was observed more strongly on the General factor than on the bifactors. The bifactors
did not collapse, however, with some bifactor loadings near or above
0.30 [52]. With a value of 0.891, the ECV index was indicative of es­
Table 2 sential unidimensionality [36]. Each item-level ECV index exceeded
Exploratory structural equation modeling bifactor analysis of the Occupational 0.80, suggesting that ODI items contributed homogeneously to the
Depression Inventory (ODI): Explained Common Variance. unidimensionality of the measure.
Item C I-ECV ECV
3.3. Reliability
ODI1 0.756 0.902 0.891
ODI2 0.925 0.818
Cronbach's α for the ODI was excellent, with values of 0.916 in
ODI3 0.700 0.823
ODI4 0.722 0.825 Sample 1, 0.915 in Sample 2, and 0.931 in Sample 3. McDonald's ω for
ODI5 0.657 0.875 the ODI was also highly satisfactory, with values of 0.924 in Sample 1,
ODI6 0.747 0.960 0.923 in Sample 2, and 0.938 in Sample 3.
ODI7 0.834 0.965
ODI8 0.787 0.944
ODI9 0.672 0.908
3.4. Convergent and discriminant validity

Notes. N = 2254. C: communality; ECV: Explained Common Variance; I-ECV: Results regarding the convergent and discriminant validity of the
item-level ECV. ODI and CES-D are summarized in Supplementary Material 8. We

4
R. Bianchi and I.S. Schonfeld Journal of Psychosomatic Research 138 (2020) 110249

extracted two bifactors in addition to the General factor because of our Pragmatically speaking, with only nine core items, the ODI is a brief
focus on two different scales. The targets were defined based on the measure that can be completed rapidly. Moreover, scale scoring is
items belonging to each scale. The model showed a satisfactory fit in straightforward and diagnostic information can be extracted in just a
both Sample 1 (RMSEA = 0.053; CFI = 0.987; TLI = 0.980) and few seconds once the diagnostic procedure is mastered. Such char­
Sample 2 (RMSEA = 0.057; CFI = 0.984; TLI = 0.977). In both acteristics can be helpful in clinical practice because they facilitate
samples, every CES-D and ODI item loaded substantially on the General occupational physicians' work while reducing patients' burden. Brevity
factor, signaling convergent validity of the two measures. As antici­ and coding simplicity are also advantageous in the research context. In
pated, however, the CES-D and the ODI also showed some degree of epidemiological studies, for instance, survey duration is a significant
discriminant validity. The ECV was 0.646 in Sample 1 (ODI scale-level concern for it bears on participant involvement and attrition, and
ECV = 0.570) and 0.691 in Sample 2 (ODI scale-level ECV = 0.596). coding simplicity can render data analysis and reporting less laborious.
Results regarding the convergent and discriminant validity of the The qualities of the ODI thus make it a handy and polyvalent tool.
ODI and HADS-D (Sample 3) can also be found in Supplementary While our study has noticeable strengths, such as the use of ad­
Material 8. As was previously the case, we extracted two bifactors in vanced statistical techniques, it also has limitations. First, the study
addition to the General factor. The model showed a satisfactory fit: samples were self-selected and their representativeness is unclear. Our
RMSEA = 0.045; CFI = 0.995; TLI = 0.992. All HADS-D and ODI items study may have, for instance, attracted a disproportionately high
loaded substantially on the General factor, signaling convergent va­ number of job-stressed individuals. Our prevalence estimates are thus
lidity of the two measures. As expected, some degree of discriminant sample-specific and offer no opportunity for generalization. Second, our
validity was concomitantly observed. Indeed, the ECV was 0.560 (ODI study involved a limited array of occupations—most of our participants
scale-level ECV = 0.633). were schoolteachers. Because some jobs may be more likely than others
to precipitate depression or to be perceived as depressogenic, we re­
3.5. Criterion validity commend that future studies focus on a much wider range of occupa­
tional groups. More broadly, it would be useful to estimate the pre­
Correlations among the study variables are displayed in valence of job-ascribed depression at multiple levels of observation,
Supplementary Materials 1 to 3. In all samples, work-attributed de­ e.g., across countries (based on samples representative of the general
pressive symptoms correlated in the expected direction with our other working population), occupational categories, and organizations within
variables of interest. Correlations were supportive of the ODI's criterion a given sector of activity. Such a mapping could, for instance, enable us
validity. to identify countries, occupational categories, and organizations in
Regarding work-contextualized variables, ODI-assessed symptoms which the prevalence of job-ascribed depression is abnormally high.
correlated substantially with job satisfaction (rhos [ρs] from −0.478 to Such information could then help us guide health-promoting inter­
−0.606), dedication to work (ρs of −0.464 and − 0.476), and will­ ventions. Third, our study had a cross-sectional design. Follow-up stu­
ingness to stay in the job (ρ = −0.457) and moderately with social dies are needed to examine test-retest reliability as well as sensitivity to
support in work life (ρs from −0.211 to −0.438) and active search for change—e.g., by comparing ODI scores before and after work-centered
another job/position (ρ = 0.331). Regarding context-free variables, interventions. Fourth, our study is based on self-reported measures.
ODI-assessed symptoms correlated substantially with (cause-neutral) Self-reported measures are subject to response biases (e.g., social de­
depressive symptoms (ρs from 0.432 to 0.722), trait anxiety (ρs from sirability bias). This being said, it is well-known that self-reported
0.451 to 0.478), general health status (ρs from −0.300 to −0.523), and measures are predictive of objective outcomes. For example, perceived
life satisfaction (ρs from −0.360 to −0.507). ODI-assessed symptoms occupational stress is prospectively associated with actual turnover
correlated to a (much) weaker extent with social support outside of [53], subjective assessments of health status and depressive symptoms
work (ρs from −0.173 to −0.261), environmental quality (ρs from predict mortality [54,55], and questionnaire-evaluated suicidal idea­
−0.125 to −0.168), residential satisfaction (ρs from −0.001 to tion is linked to attempted and completed suicides [56]. At a more
−0.232), and safety in daily life (ρs from −0.187 to −0.298). general level, patients' inputs constitute key sources of information for
researchers and practitioners in identifying symptoms experienced,
4. Discussion etiological pathways, and treatment efficacy and side-effects [57,58].
Patients can, in fact, provide information that would be otherwise un­
The aim of this study was to introduce the ODI, a measure devel­ available because neither technology nor any observer grants access to
oped to assess the severity of work-attributed depressive symptoms and it [59]. An examination of how the ODI behaves vis-à-vis objective
establish provisional diagnoses of job-ascribed depression. The ODI indicators of health and performance should, however, be put high on
showed strong reliability and high factorial validity. ESEM bifactor ODI users' agenda. Fifth, the ODI was examined only in its English and
analysis indicated that, as intended, the ODI can be used as a uni­ French versions. The ODI should be developed in other languages in the
dimensional measure. Incidentally, our results suggest that the DSM-5 future.
symptoms defining major depression show appreciable unity [6]. The Importantly, in the ODI, the link between depressive symptoms and
ODI exhibited both convergent and discriminant validity vis-à-vis work is approached through respondents' causal attributions. Causal
cause-neutral depression scales. These results are consistent with the attributions are an important aspect of how people make sense of their
notion that, at the population level, all individuals with a job-ascribed experiences and interpret the events they encounter, thereby con­
depression should be identified as depressed in a cause-neutral assess­ tributing to shaping subsequent emotion, motivation, cognition, and
ment of depressive disorders whereas only some of the individuals action [60]. An idea underlying the use of causal attributions is that
identified as depressed in a cause-neutral assessment of depressive individuals are often in a privileged position to synthesize information
disorders should meet the criteria for a job-ascribed depression. on what goes wrong in their lives, especially when “low-observability”
Work-attributed depressive symptoms correlated in the expected phenomena are at stake. In many cases, no one else has access to more
direction with both our work-contextualized (e.g., job satisfaction) and or better information [59]. However, causal attributions of course go
our context-free (e.g., general health status) variables of interest, with a risk of misattributions—a risk that exists in the context of both
speaking to the criterion validity of the ODI. ODI-assessed symptoms self-reports and clinician-supervised anamnesis and etiological in­
were clearly associated with turnover intention, a finding consistent vestigations. With respect to the reduction of that risk, we note that the
with the view that the ODI turnover intention item can help assess ODI's instructions to respondents have been designed to discourage
concrete work implications of the depressive symptoms reported. hasty attributions of depressive symptoms to work (see Table 1).

5
R. Bianchi and I.S. Schonfeld Journal of Psychosomatic Research 138 (2020) 110249

Through the guidance provided, respondents are primed to pay atten­ explicit regarding what the ODI does and does not assess. First, the ODI
tion to both nonwork and unidentified depressogenic factors when re­ intentionally assesses depressive symptoms in connection to an attributed
sponding. Respondents are invited to report symptoms only when they cause, namely, perceived job stress. Consequently, investigating whe­
feel able to establish a link between their symptoms and their work with ther perceived job stress predicts ODI scores would involve a tautolo­
clarity. We underline that the reliance on individuals' causal attribu­ gy—often referred to as the triviality or circularity trap [52]. Second,
tions is commonplace in clinical and health research. Major nationwide the ODI does not involve presuppositions as to the extent to which in­
surveys, such as the Stress in America™ survey commissioned by the ternal dispositions (e.g., personal incompetence) or external conditions
American Psychological Association [61], have relied on individuals' (e.g., management styles setting contradictory or unattainable job ob­
causal attributions to identify leading sources of stress among the jectives) should be held “responsible” for the emergence of the symp­
general public. Causal attributions are also key to the diagnosis of toms assessed. “Self-blame” and “self-excuse” issues are not in the scope
several disorders described in the DSM-5, such as posttraumatic stress of what the ODI assesses. In the ODI, causal attributions do not concern
disorder (PTSD), acute stress disorder (ASD), and adjustment disorders internal versus external explanatory factors; causal attributions concern
(ADs) [6]. The symptoms characterizing PTSD, ASD, and ADs derive a domain of life—work.3 On a related note, it is worth remembering
their diagnostic value from being imputable to specific traumatic/ that the etiology of depression is best understood through the dynamic
stressful events. Overall, causal attributions have been fruitfully used in interplay between internal dispositions and external conditions [11–17].
a variety of clinical and health research areas (e.g., common mental This study suggests that the ODI constitutes a sound measure of
disorders, specific forms of self-harm) in the context of etiological in­ work-attributed depressive symptoms. The ODI may help clinicians and
vestigations [57,62,63]. epidemiologists identify, track, and treat job-ascribed depression more
It might be argued that job-related suffering can already be in­ effectively. Ultimately, ODI-based research may contribute to informing
vestigated based on “burnout,” a work-contextualized construct that occupational health policies and regulations.
has gained popularity over the last decades [64]. Unfortunately, the
burnout construct is plagued by definitional and measurement pro­ Appendix A. Supplementary data
blems that undermine its usability in occupational health research and
practice [34,65,66]. First, the burnout syndrome is nosologically and Supplementary data to this article can be found online at https://
diagnostically uncharacterized [6,67]. Consequently, cases of burnout doi.org/10.1016/j.jpsychores.2020.110249.
cannot be identified and the prevalence of burnout cannot be estimated
[37,68,69]. This state of affairs renders the burnout construct virtually References
impossible to employ in medical decision-making [12,34]. Second, even
the most basic aspects of burnout's conceptualization and oper­ [1] S.L. James, D. Abate, K.H. Abate, et al., Global, regional, and national incidence,
ationalization remain controversial [34,60,65,66]. As an illustration, prevalence, and years lived with disability for 354 diseases and injuries for 195
countries and territories, 1990-2017: a systematic analysis for the global burden of
from one study to another, burnout has been equated with exhaustion, disease study 2017, Lancet 392 (10159) (2018) 1789–1858, https://doi.org/10.
psychological withdrawal, or professional inefficacy [65,69,70]. 1016/S0140-6736(18)32279-7.
Burnout has constituted a catch-all label [71], with no consensual de­ [2] World Health Organization, Depression and Other Common Mental Disorders:
Global Health Estimates. Author, (2017).
finition available despite nearly 50 years of sustained research. Third, [3] I.H. Gotlib, C.L. Hammen, Handbook of Depression, 3rd ed., The Guilford Press,
with their narrow symptomatic focus [60,64], assessments of burnout 2014.
overlook critical signs of suffering in the workplace such as suicidal [4] R.C. Kessler, P. Berglund, Demler, et al., The epidemiology of major depressive
disorder: Results from the National Comorbidity Survey Replication (NCS-R), JAMA
ideation. Such shortcomings are disquieting because symptoms such as 289 (23) (2003) 3095–3105, https://doi.org/10.1001/jama.289.23.3095.
suicidal ideation are crucial to identifying workers who urgently need [5] A.H. Weinberger, M. Gbedemah, A.M. Martinez, D. Nash, S. Galea, R.D. Goodwin,
help [12,52]. As a reminder, there are no identified iatrogenic risks of Trends in depression prevalence in the USA from 2005 to 2015: widening dis­
parities in vulnerable groups, Psychol. Med. 48 (8) (2017) 1308–1315, https://doi.
assessing suicidality [45]. Finally, the extent to which individuals as­
org/10.1017/S0033291717002781.
cribe their burnout symptoms to their occupational activity is unclear. [6] American Psychiatric Association, Diagnostic and Statistical Manual of Mental
As an illustration, a recent study found that fewer than half of the ex­ Disorders, 5th ed., American Psychiatric Publishing, 2013.
amined workers considered their job to be the driver of their burnout [7] E.T. Rolls, A non-reward attractor theory of depression, Neurosci. Biobehav. Rev. 68
(2016) 47–58, https://doi.org/10.1016/j.neubiorev.2016.05.007.
symptoms [72]. These findings are consistent with the fact that burnout [8] H. Wu, J. Mata, D.J. Furman, A.J. Whitmer, I.H. Gotlib, R.J. Thompson,
scales such as the Maslach Burnout Inventory include items that involve Anticipatory and consummatory pleasure and displeasure in major depressive dis­
explicit causal attributions to the job (e.g., “I feel frustrated by my job”; order: an experience sampling study, J. Abnorm. Psychol. 126 (2) (2017) 149–159,
https://doi.org/10.1037/abn0000244.
“I worry that this job is hardening me emotionally”; “I feel burned out [9] R. Bianchi, Do burnout and depressive symptoms form a single syndrome?
from my work”) but also items that do not (“I feel like I'm at the end of Confirmatory factor analysis and exploratory structural equation modeling bifactor
my rope”; “I feel very energetic”) [64]. All in all, such observations analysis, J. Psychosom. Res. 131 (2020) 109954, , https://doi.org/10.1016/j.
jpsychores.2020.109954.
render the need for a renewed approach to job-related suffering salient. [10] N. Haslam, E. Holland, P. Kuppens, Categories versus dimensions in personality and
As underlined by Schwenk and Gold (2018), “[c]linical depressive psychopathology: a quantitative review of taxometric research, Psychol. Med. 42
disorders have more solid grounding, methods of measurement, pa­ (5) (2012) 903–920, https://doi.org/10.1017/s0033291711001966.
[11] M. Wichers, The dynamic nature of depression: a new micro-level perspective of
thophysiological foundation, and empirically proven approaches to
mental disorder that meets current challenges, Psychol. Med. 44 (7) (2014)
treatment of varying levels of severity [than burnout]” (p. 1110) [34]. 1349–1360, https://doi.org/10.1017/S0033291713001979.
The ODI captures burnout researchers' initial intent to assess work-at­ [12] R. Bianchi, I.S. Schonfeld, J. Verkuilen, A five-sample confirmatory factor analytic
study of burnout-depression overlap, J. Clin. Psychol. 76 (4) (2020) 801–821,
tributed suffering [60,73] while (a) repatriating research on work-at­
https://doi.org/10.1002/jclp.22927.
tributed suffering to the long-established and well-defined framework [13] C.R. Pryce, D. Azzinnari, S. Spinelli, E. Seifritz, M. Tegethoff, G. Meinlschmidt,
of depression research, (b) covering depressive symptoms in their Helplessness: a systematic translational review of theory and evidence for its re­
variety (i.e., cognitive, affective, somatic aspects), (c) allowing for the levance to understanding and treating depression, Pharmacol. Ther. 132 (3) (2011)
242–267, https://doi.org/10.1016/j.pharmthera.2011.06.006.
assessment of critical signs of job-related suffering such as suicidal [14] A. Breier, M. Albus, D. Pickar, T.P. Zahn, O.M. Wolkowitz, S.M. Paul, Controllable
ideation, and (d) providing a diagnostic algorithm to identify likely
cases of job-ascribed depression and estimate their prevalence. In the
hands of occupational health specialists, the ODI has the potential to 3
Some people may think they are to blame for the depressogenic effect that
resolve many of the long-lamented difficulties linked to the use of the their work has on them; others may think their occupational environment is to
burnout construct [34,37,65]. blame for the depressogenic effect that their work has on them. The ODI does
Before concluding, two points may need to be rendered more not address such issues.

6
R. Bianchi and I.S. Schonfeld Journal of Psychosomatic Research 138 (2020) 110249

and uncontrollable stress in humans: alterations in mood and neuroendocrine and (2015) 877–902, https://doi.org/10.1146/annurev-psych-010814-015321.
psychophysiological function, Am. J. Psychiatr. 144 (11) (1987) 1419–1425, [42] K. Kroenke, R.L. Spitzer, The PHQ-9: a new depression diagnostic and severity
https://doi.org/10.1176/ajp.144.11.1419. measure, Psychiatr. Ann. 32 (9) (2002) 509–515, https://doi.org/10.3928/0048-
[15] B.P. Dohrenwend, The role of adversity and stress in psychopathology: some evi­ 5713-20020901-06.
dence and its implications for theory and research, J. Health Soc. Behav. 41 (1) [43] G.E. Simon, E. Johnson, J.M. Lawrence, et al., Predicting suicide attempts and
(2000) 1–19, https://doi.org/10.2307/2676357. suicide deaths following outpatient visits using electronic health records, Am. J.
[16] M.E.P. Seligman, Helplessness: On Depression, Development, and Death, W. H. Psychiatr. 175 (10) (2018) 951–960, https://doi.org/10.1176/appi.ajp.2018.
Freeman, 1975. 17101167.
[17] P. Willner, J. Scheel-Krüger, C. Belzung, The neurobiology of depression and anti­ [44] R.L. Spitzer, J.C. Wakefield, DSM-IV diagnostic criterion for clinical significance:
depressant action, Neurosci. Biobehav. Rev. 37 (10) (2013) 2331–2371, https://doi. does it help solve the false positives problem? Am. J. Psychiatr. 156 (12) (1999)
org/10.1016/j.neubiorev.2012.12.007. 1856–1864, https://doi.org/10.1176/ajp.156.12.1856.
[18] J. Blume, S.D. Douglas, D.L. Evans, Immune suppression and immune activation in [45] C.R. DeCou, M.E. Schumann, On the iatrogenic risk of assessing suicidality: a meta-
depression, Brain Behav. Immun. 25 (2) (2011) 221–229, https://doi.org/10.1016/ analysis, Suicide Life Threat. Behav. 48 (5) (2018) 531–543, https://doi.org/10.
j.bbi.2010.10.008. 1111/sltb.12368.
[19] M.C. McKinnon, K. Yucel, A. Nazarov, G.M. MacQueen, A meta-analysis examining [46] J.C. Cole, A.S. Rabin, T.L. Smith, A.S. Kaufman, Development and validation of a
clinical predictors of hippocampal volume in patients with major depressive dis­ Rasch-derived CES-D short form, Psychol. Assess. 16 (4) (2004) 360–372, https://
order, J. Psychiatry Neurosci. 34 (1) (2009) 41–54 https://jpn.ca/vol34-issue1/34- doi.org/10.1037/1040-3590.16.4.360.
1-41/. [47] W.B. Schaufeli, A.B. Bakker, M. Salanova, The measurement of work engagement
[20] G.M. Slavich, M.R. Irwin, From stress to inflammation and major depressive dis­ with a short questionnaire: a cross-national study, Educ. Psychol. Meas. 66 (4)
order: a social signal transduction theory of depression, Psychol. Bull. 140 (3) (2006) 701–716, https://doi.org/10.1177/0013164405282471.
(2014) 774–815, https://doi.org/10.1037/a0035302. [48] A.S. Zigmond, R.P. Snaith, The hospital anxiety and depression scale, Acta
[21] R.M. Carney, K.E. Freedland, Depression and coronary heart disease, Nat. Rev. Psychiatr. Scand. 67 (6) (1983) 361–370, https://doi.org/10.1111/j.1600-0447.
Cardiol. 14 (2017) 145–155, https://doi.org/10.1038/nrcardio.2016.181. 1983.tb09716.x.
[22] D.L. Hare, S.R. Toukhsati, P. Johansson, T. Jaarsma, Depression and cardiovascular [49] C. DiStefano, G.B. Morgan, A comparison of diagonal weighted least squares robust
disease: a clinical review, Eur. Heart J. 35 (21) (2014) 1365–1372, https://doi.org/ estimation techniques for ordinal data, Struct. Equ. Model. Multidiscip. J. 21 (3)
10.1093/eurheartj/eht462. (2014) 425–438, https://doi.org/10.1080/10705511.2014.915373.
[23] B. Mezuk, W.W. Eaton, S. Albrecht, S.H. Golden, Depression and type 2 diabetes [50] L.K. Muthén, B.O. Muthén, Mplus User’s Guide, 8th ed., Muthén & Muthén,
over the lifespan: a meta-analysis, Diabetes Care 31 (12) (2008) 2383–2390, 1998–2017.
https://doi.org/10.2337/dc08-0985. [51] A.F. Hayes, J.J. Coutts, Use omega rather than Cronbach’s alpha for estimating
[24] G. Cizza, S. Primma, M. Coyle, L. Gourgiotis, G. Csako, Depression and osteoporosis: reliability. But…, Commun. Methods Meas. 14 (1) (2020) 1–24, https://doi.org/10.
a research synthesis with meta-analysis, Horm. Metab. Res. 42 (7) (2010) 467–482, 1080/19312458.2020.1718629.
https://doi.org/10.1055/s-0030-1252020. [52] J. Verkuilen, R. Bianchi, I.S. Schonfeld, E. Laurent, Burnout-depression overlap:
[25] J.E. Verhoeven, D. Révész, E.S. Epel, J. Lin, O.M. Wolkowitz, B.W.J.H. Penninx, exploratory structural equation modeling bifactor analysis and network analysis,
Major depressive disorder and accelerated cellular aging: results from a large psy­ Assessment. (2020), https://doi.org/10.1177/1073191120911095 Advance online
chiatric cohort study, Mol. Psychiatry 19 (8) (2014) 895–901, https://doi.org/10. publication.
1038/mp.2013.151. [53] Y. Kachi, A. Inoue, H. Eguchi, N. Kawakami, A. Shimazu, A. Tsutsumi, Occupational
[26] S. Holmquist, A. Nordstrom, P. Nordstrom, The association of depression with stress and the risk of turnover: a large prospective cohort study of employees in
subsequent dementia diagnosis: a Swedish nationwide cohort study from 1964 to Japan, BMC Public Health 20 (1) (2020) 174, https://doi.org/10.1186/s12889-
2016, PLoS Med. 17 (1) (2020) e1003016, , https://doi.org/10.1371/journal.pmed. 020-8289-5.
1003016. [54] S.D. Barger, M.R. Cribbet, M.F. Muldoon, Participant-reported health status predicts
[27] D. Spiegel, J. Giese-Davis, Depression and cancer: mechanisms and disease pro­ cardiovascular and all-cause mortality independent of established and nontradi­
gression, Biol. Psychiatry 54 (3) (2003) 269–282, https://doi.org/10.1016/S0006- tional biomarkers: evidence from a representative US sample, J. Am. Heart Assoc. 5
3223(03)00566-3. (9) (2016) e003741, https://doi.org/10.1161/jaha.116.003741.
[28] E. Chesney, G.M. Goodwin, S. Fazel, Risks of all-cause and suicide mortality in [55] P. Cuijpers, N. Vogelzangs, J. Twisk, A. Kleiboer, J. Li, B.W. Penninx,
mental disorders: a meta-review, World Psychiatry 13 (2) (2014) 153–160, https:// Comprehensive meta-analysis of excess mortality in depression in the general
doi.org/10.1002/wps.20128. community versus patients with specific illnesses, Am. J. Psychiatr. 171 (4) (2014)
[29] R. Bianchi, I.S. Schonfeld, E. Laurent, Burnout syndrome and depression, in: Y.- 453–462, https://doi.org/10.1176/appi.ajp.2013.13030325.
K. Kim (Ed.), Understanding Depression: Volume 2. Clinical Manifestations, [56] G.E. Simon, C.M. Rutter, D. Peterson, M. Oliver, U. Whiteside, B. Operskalski,
Diagnosis and Treatment, Springer Nature, 2018, pp. 187–202. E.J. Ludman, Does response on the PHQ-9 depression questionnaire predict sub­
[30] Health Promotion Switzerland, Workplace Health Management: Job Stress Index. sequent suicide attempt or suicide death? Psychiatr. Serv. 64 (12) (2013)
Author, https://healthpromotion.ch/workplace-health-management/studien- 1195–1202, https://doi.org/10.1176/appi.ps.201200587.
wirkung-bgm/job-stress.html, (2020, May 10). [57] D. Ashoorian, R. Davidson, D. Rock, S. Gudka, R. Clifford, A review of self-report
[31] U.S. Bureau of Labor Statistics, Workplace Injuries. Author, https://www.bls.gov/ medication side effect questionnaires for mental health patients, Psychiatry Res.
data/#injuries, (2020, May 5). 219 (3) (2014) 664–673, https://doi.org/10.1016/j.psychres.2014.05.060.
[32] A. Steck, Burnout: Arbeitsausfälle steigen auf Rekordhoch [burnout: Absences from [58] I.B. Olsen, S. Øverland, S.E. Reme, C. Løvvik, Exploring work-related causal attri­
work rise to record high]. NZZ am Sonntag, https://nzzas.nzz.ch/wirtschaft/ butions of common mental disorders, J. Occup. Rehabil. 25 (3) (2015) 493–505,
burnout-arbeitsausfaelle-steigen-auf-rekordhoch-ld.1533349?reduced=true, https://doi.org/10.1007/s10926-014-9556-z.
(2020, January 11). [59] W. Baldwin, Information no one else knows: The value of self-report, in: A.A. Stone,
[33] D.A. Mata, M.A. Ramos, N. Bansal, R. Khan, C. Guille, E. Di Angelantonio, S. Sen, J.S. Turkkan, C.A. Bachrach, J.B. Jobe, H.S. Kurtzman, V.S. Cain (Eds.), The Science
Prevalence of depression and depressive symptoms among resident physicians: a of Self-Report: Implications for Research and Practice, Lawrence Erlbaum
systematic review and meta-analysis, JAMA 314 (22) (2015) 2373–2383, https:// Associates Publishers, 2000, pp. 3–7.
doi.org/10.1001/jama.2015.15845. [60] T.S. Kristensen, M. Borritz, E. Villadsen, K.B. Christensen, The Copenhagen burnout
[34] T.L. Schwenk, K.J. Gold, Physician burnout—a serious symptom, but of what? inventory: a new tool for the assessment of burnout, Work Stress. 19 (3) (2005)
JAMA 320 (11) (2018) 1109–1110, https://doi.org/10.1001/jama.2018.11703. 192–207, https://doi.org/10.1080/02678370500297720.
[35] H.W. Marsh, A.J. Morin, P.D. Parker, G. Kaur, Exploratory structural equation [61] American Psychological Association, Stress in AmericaTM: Paying with our Health,
modeling: an integration of the best features of exploratory and confirmatory factor Author, 2015.
analysis, Annu. Rev. Clin. Psychol. 10 (2014) 85–110, https://doi.org/10.1146/ [62] M.Z. Brown, K.A. Comtois, M.M. Linehan, Reasons for suicide attempts and non­
annurev-clinpsy-032813-153700. suicidal self-injury in women with borderline personality disorder, J. Abnorm.
[36] A. Rodriguez, S.P. Reise, M.G. Haviland, Evaluating bifactor models: calculating Psychol. 111 (1) (2002) 198–202, https://doi.org/10.1037/0021-843X.111.1.198.
and interpreting statistical indices, Psychol. Methods 21 (2) (2016) 137–150, [63] A.J. Edmondson, C.A. Brennan, A.O. House, Non-suicidal reasons for self-harm: a
https://doi.org/10.1037/met0000045. systematic review of self-reported accounts, J. Affect. Disord. 191 (2016) 109–117,
[37] L.S. Rotenstein, M. Torre, M.A. Ramos, R.C. Rosales, C. Guille, S. Sen, D.A. Mata, https://doi.org/10.1016/j.jad.2015.11.043.
Prevalence of burnout among physicians: a systematic review, JAMA 320 (11) [64] C. Maslach, W.B. Schaufeli, M.P. Leiter, Job burnout, Annu. Rev. Psychol. 52 (1)
(2018) 1131–1150, https://doi.org/10.1001/jama.2018.12777. (2001) 397–422, https://doi.org/10.1146/annurev.psych.52.1.397.
[38] M. Buhrmester, T. Kwang, S.D. Gosling, Amazon’s mechanical Turk: a new source of [65] R. Bianchi, I.S. Schonfeld, E. Laurent, Burnout: moving beyond the status quo, Int. J.
inexpensive, yet high-quality, data? Perspect. Psychol. Sci. 6 (1) (2011) 3–5, Stress. Manag. 26 (1) (2019) 36–45, https://doi.org/10.1037/str0000088.
https://doi.org/10.1177/1745691610393980. [66] T. Cox, M. Tisserand, T. Taris, The conceptualization and measurement of burnout:
[39] W. Mason, S. Suri, Conducting behavioral research on Amazon’s mechanical Turk, questions and directions, Work Stress. 19 (3) (2005) 187–191, https://doi.org/10.
Behav. Res. Methods 44 (1) (2012) 1–23, https://doi.org/10.3758/s13428-011- 1080/02678370500387109.
0124-6. [67] World Health Organization, Burn-out an “occupational phenomenon”: International
[40] A.S.M. Niessen, R.R. Meijer, J.N. Tendeiro, Detecting careless respondents in web- Classification of Diseases. Author, https://www.who.int/mental_health/evidence/
based questionnaires: which method to use? J. Res. Pers. 63 (2016) 1–11, https:// burn-out/en/, (2019) (May 28, 2019).
doi.org/10.1016/j.jrp.2016.04.010. [68] R. Bianchi, I.S. Schonfeld, E. Laurent, The dead end of current research on burnout
[41] S.D. Gosling, W. Mason, Internet research in psychology, Annu. Rev. Psychol. 66 (1) prevalence, J. Am. Coll. Surg. 223 (2) (2016) 424–425, https://doi.org/10.1016/j.

7
R. Bianchi and I.S. Schonfeld Journal of Psychosomatic Research 138 (2020) 110249

jamcollsurg.2016.05.012. (Eds.), Mental Health and Productivity in the Workplace: A Handbook for
[69] R. Bianchi, I.S. Schonfeld, E. Laurent, Can we trust burnout research? Ann. Oncol. Organizations and Clinicians, Jossey-Bass, 2003, pp. 433–457.
28 (9) (2017) 2320–2321, https://doi.org/10.1093/annonc/mdx267. [72] R. Bianchi, R. Brisson, Burnout and depression: causal attributions and construct
[70] R. Brisson, R. Bianchi, On the inconsistency of burnout conceptualization and overlap, J. Health Psychol. 24 (11) (2019) 1574–1580, https://doi.org/10.1177/
measurement, J. Am. Coll. Surg. 224 (1) (2017) 87, https://doi.org/10.1016/j. 1359105317740415.
jamcollsurg.2016.09.006. [73] H.J. Freudenberger, Staff burnout, J. Soc. Issues 30 (1) (1974) 159–165, https://
[71] P.M. Liu, D.A. Van Liew, Depression and burnout, in: J.P. Kahn, A.M. Langlieb doi.org/10.1111/j.1540-4560.1974.tb00706.x.