You are on page 1of 21

Journal of Occupational and Organisational Psychology (2000), 73, 221-240 Printed in Great Britain 2 2 1

© 2000 The British Psychological Society

Revisions and further developments of the


Occupational Stress Indicator: LISREL
results from four Dutch studies

Ame Evers* and Michael Frese


Work <& Organisational Psychology, University of Amsterdam, The Netherlands

Cary L. Cooper
University of Manchester Institute of Science and Technology, UK

The Occupational Stress Indicator (OSI) is a popular instrument for the diagnosis
of stress and stress-related personality and outcome variables. However, one
weakness of the OSI was the low reliability of some of its scales. This paper
describes in a series of four studies improvements of the reliabilities of most of
the scales of the Dutch version of the OSI. All new scales were tested for
unidimensionality. Compared to the original OSI, the personality scales (type A,
locus of control, and coping styles) have been changed completely and nearly all
original items were replaced. The other scales are revisions of the original OSI,
with some items being deleted, rewritten, or added. A new scale for satisfaction
with pay was included. All but two of the revised scales now show sufficient
reliabilities and unidimensionality. This revision led us to suggest two versions of
the OSI: an elaborate version with 28 scales and 188 items, and an abridged
version with 15 scales and 94 items.

The Occupational Stress Indicator (OSI) by Cooper, Sloan, and Williams (1988) is
a frequently used set of scales to understand stress at work. For example between
1990 and 1997 at least 38 articles using the OSI have been published (PsycLIT
Database) and many translations of the indicator exist. It is used as a tool to inform
organizadonal decisions regarding the management of stress or to suggest changes
at work, and as an evaluation instrument for stress management programmes or in
scientific research. Thus, it is an important inventory and therefore we need to
know its psychometric properties. The objective of this article is to describe four
studies with the Dutch version of the indicator (Spanningsmeter) which attempt to
optimize the unidimensionality and reliability of the component scales. This led to
a substantial revision of the scales. Given this work, it may pay off to start using the
revised scales which are (back)transiated into English.

*Rcqucsts for reprints should be addressed lo Dr Arnc Evers, Work & Organizadonal Psychology, University of
Amstetdam, Roctcrsstraat 15, 1018 WB Amsteidam, The Netherlands (e-mail: ao_evers@macmail.psy.uvs.al).
222 A. Evers et al
The Occupational Stress Inventory
The OSI is founded on the notion that stressors do not influence everyone in the
same way. Stress is seen as the 'lack of fit' between the individual and his/her
environment (BagUoni, Cooper, & Hingley, 1990). Therefore, the importance of
subjective perceptions and interpretations of the environment, the process of
cognitive appraisal and the effect of personality characteristics and demographic
factors is emphasized. The model states that work pressures lead to negative
outcomes (lowered job satisfaction and mental and physical health) and that this
relationship may be moderated by individual difference variables. The individual
difference variables may be linked to occupational stress as direct predictors of
stress-related outcomes as well (Cooper & Baglioni, 1988).
The OSI consists of 25 subscales with a total of 167 items. The scales look at
sources of pressure (61 items, six subscales: Intrinsic to the job. Organizational
role. Relationships with others. Career and achievement. Organizational structure
and climate. Home/work interface), type A behaviour (14 items, three subscales:
Attitude to living, Style of behaviour. Ambition), locus of control (12 items, three
subscales: Organizational forces. Management process. Individual influence), cop-
ing styles (28 items, six subscales: Social support. Task strategies. Logic, Home/
work relations. Time management. Involvement), job satisfaction (22 items, five
subscales: Achievement, value and growth. Job itself. Organizational design and
structure. Organizational processes. Personal relationships) and health (30 items,
two subscales: Mental ill-health, Physical ill-health). The names of the subscales
give an indication of the content of the scales. A more comprehensive description
can be found in the OSI Management Guide (Cooper et a/., 1988). The original
Dutch version (Broers, Evers, & Cooper, 1995) was a literal translation of the OSI.
One problem of the OSI was that the reliabilities of many scales have been on
the low side (in some cases as low as .10—.20) and some were clearly suboptimal
from a psychometric point of view (Broers et al, 1995; Davis, 1996; Ingledew,
Hardy, & Cooper, 1992; Robertson, Cooper, & Williams, 1990). Therefore most
studies only used composite scores (for instance the total score for coping styles or
locus of control) to circumvent the low reliabilities of the subscales. However, this
reduces the richness of knowledge potentially gained from the subscales. Thus a
revision to improve the reliabilities of the subscales is called for. Additionally, we
wanted to check the unidimensionality for all scales with a confirmatory factor
analysis which to our knowledge has not yet been done. Finally, we will examine the
independence of the new subscales.
I

Scale revisions
The reliability problem was particularly pronounced in the original type A, locus of
control and coping styles scales which included many items with very low item-total
correlations. For this purpose we constructed new scales for these three sections of
the inventory.
Theoretically, type A falls into three factors: impatience, achievement orientation,
and hostility (Friedman & Powell, 1984). In contrast to Jenkins Activity Scale QAS;
Revisions and further developments of the OSI 223
Jenkins, Zyzanski, & Rosenman, 1979), the best-known questionnaire for measur-
ing type A behaviour, we developed items for these three factors that were only
work related. The three global theoretical factors were subdivided into components
to get a good spread of the items and to investigate the possibility of developing
subscales. The components were: impatience, irritation and time pressure for the
impatience factor; work orientation, goal orientation and drive for the achievement
orientation factor; anger, expressive hostility and competitiveness for the hostility
factor. For each of the nine components six items were written, leading to a total
of 54 items. There was an overlap of two items with the original OSI scales, and
none with the JAS.
Locus of control implies three concepts which are customarily differentiated
(Levenson, 1981): helplessness with regard to powerful others, luck and internality.
For a clearer understanding of the locus of control construct and to get a better fit
with concepts within social-learning theory Palenzuela (1988) redefined these
concepts as non-contingency, chance and contingency. According to Palenzuela's
conceptualization six items for each of the three scales were constructed. However,
in contrast to the content of the items of the Palenzuela questionnaire (MASLOC),
which was academic-life oriented, the new OSI items were work-life oriented.
There was an overlap of two items with the original OSI scales, and none with the
MASLOC questionnaire.
Consultation of the literature on coping and coping questionnaires such as Lazarus
and Folkman's (1984) Ways of Coping, Billings and Moos (1984), Cohen, Reese,
Kaplan, and Riggio (1986) and a Dutch coping list (Schreurs, van de Willige,
Brosschot, Tellegen, & Graus, 1993), led to the definition of eight aspects of coping
that were considered to be most relevant in the field of work: direct action, seeking
social support (emotional as well as instrumental support), acceptance, denial,
planning, positive reinterpretation, and compensatory behaviour. In all, 80 items
were constructed (10 for each scale). None of the original OSI items were retained,
mainly because they did not match with the eight constructs. There was no overlap
in items with any of the other questionnaires. Again, tbe items are work-life
oriented whereas the other questionnaires are more general-life oriented.
For the sources of pressure scales item analysis on the original OSI items in their
Dutch translation had revealed some weaknesses. Some items were too abstract,
others were placed in the wrong scale, some were not appropriate for most jobs and
still others had too low item-total correlations. Some new items were constructed,
however the structure of the original OSI with six scales was maintained.
The original OSI items for job satisfaction and health were used and complemented
with a few new items. One new scale on satisfaction with pay was developed,
because this aspect was thought to be an important issue of job satisfaction
(Thierry, 1997). These items were based on the five subdimensions of satisfaction
with pay by Heneman and Schwab (1985). Some of the items in the health scales
were rephrased. No new items were constructed.
The objective was to construct scales that each consisted of four to eight items.
Scales of that length are generally considered to be a good compromise between the
practical demands for short scales and the theoretical demands for adequate
reHability and content validity. An exception with respect to scale length was made
224 A. Evers et al.
for the two health scales, because they contain more items in the original OSI (18
and 12 respectively), as well.

Method
Overview
Four studies were done that build upon each other to improve the scales of the OSI. Study 1 dealt
with type A behaviour and locus of control scales. Study 2 with coping styles, Study 3 with sources
of pressure and Study 4 with satisfaction and health scales. Because in some of these studies the new
scales were constructed, they are referred to as calibration studies. In addition. Study 4 was used as a
replication study of the scales constructed in Studies 1 to 3. However, because Studies 3 and 4 were
done concurrendy, the scales on job satisfaction and health developed in Study 4 were replicated in
Study 3. I 1 '

Participants
Participants in Study 1 were professional soldiers from the Dutch army. The questionnaire was sent
to 20 randomly selected soldiers from each of 40 barracks spread over Holland. A return envelope was
enclosed. In all, 310 usable questionnaires were received, which means a response rate of 39%. The
mean age of the group was 39 years, the group consisted of two women and 308 men and the ranks
ranged from corporal to major. Soldiers with a relatively low rank made up the majority of the group.
Pardcipants in Study 2 were Dutch street car drivers. The questionnaires were distributed by the
managers. To secure anonymity the questionnaires had to be sent back in a return envelope addressed
to the researchers. From the 765 distributed questionnaires 184 were returned, which translates into
a response rate of 24%. The mean age of the respondents was 40 years and 87% of them were males.
These figures closely resemble those of tbe population of street car drivers (mean age 39 years, 83V(i
males).
Participants in Study 3 were police officers in two districts in a big Dutch city. The method of
administration was the same as in Study 2: distribution by managers, returning the questionnaires by
mail to the researchers. Out of 509 questionnaires 126 were returned (25%). Tbe mean age of the
respondents was 35 years and 83% were males. No differences were found in age and gender between
respondents and non-respondents.
Participants in Study 4 were nurses from a hospital in the south of The Netherlands. The method
of administration was the same as in Study 1. Out of 553 questionnaires 400 were returned (72%). The
mean age of the respondents was 34 years, 80% of them were females. Tbese figures corresponded to
those from tbe entire population of 553 nurses in this hospital.

Analyses •
Especially for the new scales of type A, locus of control and coping, more items were included than
used in the final scales, because we expected that not all new items would fit well. In each calibration
study the item selection started with computing the corrected item-total correlations and deleting the
items with the lowest values. The minimal acceptable level of alpha was set to be .60.
Since a high alpha is no guarantee for unidimensionality (Drenth 5c Sijtsma, 1990), LISREL analyses
were used to check for unidimensionality. When the fit indices were too low, those items that were
responsible for the mulddimensionality were removed from the scale—using the modification indices
for this optimization. MacCallum, Roznowski, and Necowitz (1992) argue against the use of
modification indices to improve model fit because of capitalization on chance effects (in particular

'For tests that are primarily meant to study groups, the Dutch Comniittee on testing qualifies a reliability
coefficient between .60 and ,70 is 'sufficient' and a coefficient of .70 or higher as 'good' (Evers, Van Vllet-MuJder,
& Groot, 2000).
Revisions and further developments of the OSI 225
with relatively small sample sizes of 100 to 400 participants). Our study was arranged in such a way
that the susceptibility for this effect was minimized following the su^estions by MacCallum et al.
(1992). The first precaution has to do with the number of modifications. Only few modifications were
done as a consequence of the LISREL analyses, whereby we carefully checked the interpretability of
the modification indexes. As we were looking for unwanted high correlations between pairs of items
that were too similar in content or formulation, that is items that cause an 'unfair' increase in alpha,
we checked the content of the items before deleting them. The second precaution was to replicate the
LISREL analyses in the replication samples. In each case, the scales were optimized in one study and
then the optimized version was cross-checked in another study. Of course, there was no more attempt
to optimize fit parameters in the second study. In each case we used the fit indices of y^ (divided by
degrees of freedom), the Comparative Fit Index (CFI) and Root Mean-square Residual (RMR), which
arc important for scale development (Bentler, 1990; Schumacker & Lomax, 1996). A low %^ (when
judged against degrees of freedom) is suggestive of good fit. A CFI of above .90 is indicative of good
fit and of above .95 of excellent fit. The RMR should be about .05 or smaller for a good fit.

Results
Two types of results are presented here. First, the results on the scale development
and on their Cronbach's a and unidimensionality in the confirmatory factor
analyses are given. Secondly, we present the intercorrelations for each area of the
revised OSI.
After optimizing reliability and unidimensionality Study 1 resulted in three type
A scales with eight items each. The scales were labelled Irritation, Achievement
orientation and Hostility. Within these three domains no subscales were con-
structed, because within a domain most of the items were highly intercorrelated.
Locus of control was finally made up of three scales with five items each. In Study
2 the eight intended coping scales could be constructed. Each scale had six items.
In Study 3 the number of items of the sources of pressure scales were reduced from
12-14 items to eight items for each of the scales. Of the remaining 48 items 16
were new. All of the above-mentioned scales were replicated in Study 4.
The revision of the satisfaction and health scales in Study 4 resulted in six
satisfaction scales consisting of four to six items and in a reduction of the number
of items of the two health scales to 13 items each. Of the total 27 satisfaction items
nine items were new (five items for the new satisfaction with pay scale included).
These scales were replicated in Study 3.
Table 1 presents alphas and the three fit indices xV^-f-, CFI and RMR which
resulted from the LISREL analyses. Residuals were not allowed to correlate in the
LISREL analyses. We used the item-total correlations and the modification indices
to select the best items in each calibration study. In many cases, these two criteria
were not necessarily congruent. As a matter of fact, LISREL analyses often
suggested deleting items that ultimately led to a decrease in the alpha. This was
done because we wanted a joint optimization of internal consistency and
unidimensionality as tested by confirmatory factor analyses. For this reason,
some of the alphas are still not quite satisfactory, especially the coping scales
Compensating behaviour and Acceptance.
On the other hand, the results were encouraging because in all other cases, the
scales not only maintain an adequate alpha, but also show good fit indices in the
replication study. Thus, the three scales of type A behaviour kept their level of
226 A. Evers ct al.

ooo ooo oooooooo oooooo

1
a
o
•n

ro

i n " ^ ^ " ^ C M ^ T — i c o i n o o i — ' T — ' r o O r— CN to o r-


r^^ so r^ so t^-* r^ NO r^ r^ iri in so t^ 00 oo CO CO oo 00

"^ ^ r-l iri CO ND in rj NO NO NO


o ooo s o o o o o s o

oo r- es
oo O\ Ov
a
o
•a

O O O to r- fo
00 00 oo 00
oo t

CO oo oc 00 00
Revisions and further developments of the OSI 121

Or vO »-' CO SO I^
o o ooo o o

m CO ••-• 00 m
s o r ^ NO NO 00

rO rO <M ro (N (M NO
OOOOOO o

CO O N oo oo Ov Ov
Ov Os O N O N O V O S

# * *
Os * CM rg CN in
in in
JM VO VO
to NO o O CN to oo
o C N NO CN CN o CN
190.
154,

CM CO NO 00 vd 00
CM

o ooo S O ND
o r-- 00
00
o
00 00 00 00

Si
H
228 A. Evers et al.
Table 2. Intercorrelations between type A and locus of control scales (below diagonal:
calibration Study 1, above diagonal: replication Study 4)

1 2 3 4 5 6

1. Achievement orientation -.01 .36 -.27 -.17 .37


2. Irritadon .08 — .39 .29 .26 -.16
3. Competiaveness .31 .40 — .14 .16 .03
4. Helplessness -.15 .27 .09 — .53 -.32
5. Chance -.11 .15 .05 .51 — -.40
6. Internal] ty -.08 -.12 -.09 -.44 -.63 —

internal consistency and the fit indices. The sani,e was also true of the three locus
of control scales. Six of the eight coping style scales showed adequate or good
reliabilities in the first round and also in their replication in Study 4. Sources of
pressure showed good alphas and fit indices in both studies. Thus, there was little
reason to change these scales from the perspective of unidimensionality. Job
satisfaction and state of health also showed adequate and good reliabilities (note:
Study 3 is here used to replicate Study 4). However, in the replication study, there
were three scales that showed a low fit index Qob itself. Organizational processes
and Physical health). This suggests the need to develop better measures in this area
in the future.
Tables 2, 3, 4 and 5 give the intercorrelations within each area for both the
calibration and the replication studies. Table 2 shows the intercorrelations of the
type A and locus of control scales. The intercorrelations in the two studies were
quite similar, showing medium-sized correlations between the type A scales and
between the locus of control scales. Most clearly differentiated was Irritation from
the other two type A scales. The intercorrelations of the locus of control scales
were a bit higher in Study 1. But the three scales were reasonably distinct.
Table 3 shows the intercorrelations of the coping styles. Again, the correlations
were very similar in the two studies. A few correlations were very high, suggesting
that one can combine scales.
Table 4 shows the intercorrelations between the different sources of pressure.
These intercorrelations were very high and again quite regular across the two
studies. As a matter of fact, all scales together produced an alpha of .95 with the
scales as items. This suggests that one can make one scale out of the different
sources, which has been done by some of the originators of the OSI (Kirkcaldy &c
Cooper, 1992; Robertson, Cooper, & Williams, 1990). If one is interested to reduce
the length of the inventory, there is room for reduction in this area.
Table 5 shows the intercorrelations of satisfaction and health. It shows that it
was useful to develop a scale on pay because it was not highly correlated with the
other scales at least in one study (thus, it captures non-redundant information).
Study 3 shows somewhat higher intercorrelations between the scales than Study 4.
However, in general the intercorrelations were rather high (again, with the
Revisions and further developments of the OSI 229
Table 3, Intercorrelations between coping styles scales (below diagonal: calibration
Study 2, above diagonal: replication Study 4)
1 2 3 4 5 6 7 8

1. Active .45 .51 .54 -.01 .20 -.33 .51


2. Planning .52 — .50 .39 .07 .30 .11 .40
3. Social support (instrumental) .56 .42 — .62 .01 .35 -.02 .41
4. Social support (emotional) .51 .31 .72 — .16 .28 -.16 .26
5. Compensating behaviour -.07 -.12 .18 .27 — .13 .35 .02
6. Acceptance .09 .13 -.02 -.06 -.26 — .25 .34
7. Denial -.31 -.07 -.34 -.34 .12 .12 — -.03
8. Positive reinterpretation .50 .50 .30 .18 -.17 .23 -.03 —

Table 4. Intercorrelations between sources of pressure scales (below diagonal:


calibration Study 3, above diagonal: replication Study 4)
1 2 3 4 5 6

1. Intrinsic to the job .83 .72 .69 .76 .75


2. Organizational role .82 — .76 .76 .76 .76
3. Relationships with others .68 .73 — .79 .75 .80
4. Organizational structure and climate .62 .70 .67 — .64 .74
5. Home/work interface .69 .69 .73 .50 — .74
6. Career and achievement .68 ,68 .75 .70 .68 —

exception of satisfaction with pay), which may speak for a further reduction of
these scales or of the items of these scales.
The health scales show a characteristic pattern of being related on a medium to
high level which has been shown in other studies as well (Mohr, 1986).
The results above showed that in a few domains, such as coping, sources of
pressure, satisfaction and health, the intercorrelations of the scales were rather
high. For these domains we tried to reduce the number of scales and items to
get a more parsimonious instrument. First, an exploratory factor analysis was
performed using principal axis factoring and varimax rotation on all items for each
domain. Items with high factor loadings were retained. Just as done originally, we
then selected items by using item-total correlations. We used LISREL to check and
improve the unidimensionality and cross-validated the results on the replication
sample. The results are summarized in Table 6.
In the coping domain the number of items was reduced by dropping the
Acceptance items, due to the low reliabilities of the scale in both studies. The
four factor solution appeared to be the most adequate for interpretation. This
solution revealed a Social support factor (including both the Emotional and the
230 A. Evers et al.
Table 5. Intercorrelations between sadsfaction and health scales (below diagonal:
calibration Study 4, above diagonal: replication Study 3)

1. Achievement, value and growth .59 .58 .75 .38 .61 -.13 -.12
2. The job itself .42 .34 .54 .31 .50 -.20 -.18
3. Organizational design and structure .46 .49 — .64 .49 .44 .07 -.01
4. Oi^nizational processes .59 .49 .66 — .51 .56 -,23 - . 1 6
5. Personal relationships .45 .31 .45 .55 — .21 -.21 - . 1 2
6. Pay .38 .13 .17 .10 ,04 .00 - . 0 6
1. Mental ill-health ' , -.24 -.32 -.18 -.28 -.25 -.03 —• ,57
8. Physical ill-health -.14 -.24 -.14 -.20 -.12 -.17 .57 —

Instrumental Social Support items), a Denial factor, a Compensating behaviour


factor, and a fourth factor with loadings on most of the Active, Planning and
Positive reinterpretation items (although some of them showed double loadings
on the Denial—negative—or Social support factors). The four factors accounted
for 40% of the variance. The new scales—Social support and Active positive
attitude—showed good alphas and fit indices. The scales for Denial and
Compensating behaviour were left unchanged. The intercorrelations within the new
set of coping scales were much lower than those among the former eight scales.
In the domain sources of pressure nearly all items showed high loadings on the
first unrotated factor that accounted for 40% of the variance (this dropped to 5%
for the second factor). Therefore, one new sources of pressure scale was
constructed. For each of the six original sources of pressure scales the one item that
showed the highest factor loading was selected for this scale. The new scale showed
good reliability and fit.
Factor analysis of the satisfaction items resulted in a good interpretable
two-factor solution (the two factors explaining respectively 28 and 13% of the
variance). All of the Pay items showed high loadings on the second factor, all other
items except one loaded high on the first factor. From each of the five original
satisfaction scales (except Pay) the one item with the highest loading was selected
to construct a scale for General job satisfaction. This new scale showed good
reliability and fit. The Pay scale was left unchanged.
A factor analysis of the health items showed clearly distinct loadings for the
Mentai and Physical items on the first two factors (accounting for respectively 28
and 8% of the variance). Consequently, no attempt was made to construct one scale
despite the rather high correlation between the two scales. Both scales could be
reduced to seven items by omitting the five items with the lowest item-total
correlations and one item with the highest modification indices. Although this
inevitably resulted in the loss of some important symptoms (e.g. *low irritation
threshold', "worrying about mistakes in the past', 'muscles trembling', 'tendency to
sweat"), the new scales still have an adequate distribution of symptoms (see the
Appendix for the text of items of this shortened version). The correlations between
Revisions and further developments of the OSI 231

ooo ooo o o o o

O O 00
C C C
u
C
o
•a
«
*ooo* * *
*
CN (N CN in in in
a. 00 in
CN in in O CM
in »—' O '-^'
q 00
CD
o 8 CM
CN
1—•
I—'
CO m CN f o Csj i/i r^
CO CM 1 -

c
0 00 in CM r~ r - 00 -F-
Nn NO
[^ r^ so in ^

•*-^^ rjinm CNOONoin CO CM CM

•c c5c5c5 o o o o o o o o o o
C4

c/: •-Nl fM CN X in <N oo so


0 ^^ ^^ CN CN CTv CN H CN oc: C N CN CTv
CN C S
u
C

2
u
SO Cv
CS) r-
.s in 00 r- CN
CM CN

00 rO so so in
r^ t*^ so
c/5

0
g
0
u
11 I
•a
c
u 0 -a .
C c o 2. ^ 5 = 3
S-jj a, o
o u "^ !75 V <
o c O
U g do O •B
U
232 A. Evers et al.
Table 7. Intercorrelations between coping styles scales, abridged version (below
diagonal: calibration Study 2, above diagonal: replication Study 4)

1. Social support .44 .15 -.04


2. Active positive attitude 31 — .04 -.03
3. Compensating behaviour 22 -.21 — .35
4. Denial 29 -.10 .12

the satisfaction and health scales are shown in Table 7. The correlation between
satisfaction with pay and general job satisfaction was low in Study 4 and somewhat
higher in Study 3.
Tables 8 and 9 show the intercorrelations of all the scales in Study 4 for the
elaborate and the abridged versions of the Dutch OSI. It shows the familiar
pattern. In general the correlations among scales within the areas of the OSI were
higher than those between scales from different areas; this is true especially for the
elaborate version.
Correlations between sources of pressure scales and personality and coping styles
scales were low, though there were medium-sized correladons between the sources
of pressure scales and the scales for external locus of control (note: a correlation
coefficient of .15 is statistically significant at the \%, level). Moreover respondents
with higher external control reported higher dissatisfaction and higher ili-health.
These findings parallel the findings reported by Rees and Cooper (1992b) for the
original OSI. Respondents with high scores on dysfunctional coping styles
(compensating behaviour and denial) reported more health problems as well.
Concerning the relations between sources of pressure and outcome variables there
were medium-sized negative correiations between sources of pressure and satisfac-
tion scales (except for personal relationships and pay) and positive correlations
between sources of pressure scales and ill-health scales. This means that a high level
of pressure went along with a low level of satisaction and bad health scores.

Discussion
We attempted to improve the scales of the OSI with respect to reliability and
unidimensionality in a two-step process. In the first step we kept the original
structure of the OSI, For each of the domains we constructed specific subscales for
concepts that are well known in the literature on stress research. For most of the
scales our efforts were successful. However, there remained some problems with
some of the scales. In the replication study, two coping scales showed low
reliabilities (though higher than for the coping styles scales in the original OSI)
despite good fit indices, and for two satisfaction scales and one health scale the fit
indices were not good enough despite satisfactory reliabilities. In particular, the
sharp decline in the fit indices in the replication study (Study 3) for the three
Revisions and further developments of the OSI 233

S ?-•

Ss

m M —

EC O M!l r^ Q ^ ,_
•-; ^ cs (N) ^ q f^
t' r I

07

a?1

04 -
1 1
- -
1 1 1 1
S i!i S 2 8 Pi o ft <^ <^
1 1 1 1 1 )
•o m

15
01
sq s a

01
s *— o

u
u

on
O

o
•a

O
u

00
u
Si
H
234 A. Evers et al.

00
in

^— oo
o—

CM
oo m
I I
O O
O
I I

H
SS
1 1
3
in 00
•-j •—•
CM tN O — '- O
11 11
ON •^ CM '-
O CN T-" (N CM

-a
pi O NO
n tn r- o CN o
CN T— o CM *o
CM
1 1 1 1 1
00 (N in o 00 r- o
1—< CM O 1—' 1—• CN CN
o o
o 1 1
in
<^ (N O m CN
U ^- o
I I
-a
CN
O
CM o in
CM
CN
O o s o •4-
g
i 1 I 1 1 1 1
r- o — o
.14

.06
.08

.08

.07
.17

.17

in
.01

r-a T-« y—

(A
C
o
•a
I
c
u
-I
II
« •c
o
•o

I 3 JJ
oo
1 2 £
S i>
73

<
3 CM i n NO t~- "3 00
o
n u
H
Revisions and further developments of the OSI 235
outcome scales was somewhat surprising, because the fit indices for the other scales
in all the replication studies were remarkably stable. Maybe the participants in this
study (police officers who were rather dissatisfied with some aspects of their work)
were responsible for this effect.
Another finding was the high correlations among some scales within some areas
of the OSI. Therefore, a second round of psychometric analyses was carried out to
reduce the number of items and to construct composite scales. This effort resulted
in a reduction of the number of scales from 28 to 15 and of the number of items
from 188 to 94.
The number of coping styles scales was reduced from eight to four. Three of the
constructed coping scales correspond to the three coping meta-strategies that were
distinguished by Ben-Zur (1999): problem/accommodation (active, planning,
reinterpretation), support/emotion (instrumental and emotional support) and
avoidance/disengagement (denial). The fourth scale for compensating behaviour is
clearly a separate factor, although it showed the weakest reliability and fit indices.
However, one might argue that internal consistency is irrelevant for this scale,
because the items are causal indicators and that there is no underlying common
construct (for the difference between causal and effect indicators, see e.g. Spector
& Jex, 1998). The sum of the scores on items like 'watching TV', 'going out',
'drinking alcohol' and 'getting nasty' will definitely give an impression of someone's
behaviour, but the scores on these items need not necessarily correlate (see the
Appendix for the full text of the items).
Given the high correlations among the sources of pressure scales we were not
surprised that the factor analysis revealed only one strong factor in this area.
Apparently pressure can be measured by means of only six items, but this reduction
may make the OSI in total less of an indicator of stressful working conditions and
more of an indicator of personality characteristics and personal well-being. As a
consequence the abridged version of the OSI may be more valuable in stress
management programmes directed at individual rather than organizational change.
The abridged version of the revised OSI may also be used as a quick screening
instrument in practice. In the Appendix the English translations of all the items of
the abridged version, as well as an abbreviated form of the instructions, are given.
Practitioners or researchers who are interested in differential information can use
the elaborate version although they have to take into account a few weaker scales.^
Finally, we would like to point to some methodological issues. We did not
address issues in stress research that are potential threats for research results. These
are well-known in the literature (see e.g. Frese & Zapf, 1988; Spector, 1992):
common method variance, unwanted influence of 'third' variables such as negative
affectivity, uncertainty about the causality of the relations found in a cross-sectional
design, etc. Further research on these issues is still needed for the revised OSI, but
they were beyond the scope of the present study. One might argue that the
response rates in some of our samples were rather low. This may have influenced
the results in an unknown way. However, we think that the effect, if any, is minimal

^Because of copyright regulations (copyright held by Swcts &. Zeitiingcr and NFER-Nelson) the text of the
elaborate edibon cannot be published in this paper, but can be obtained from the iirst authoi.
236 A. Evers et al.
because of the great similarity of the results of the calibration and the replication
studies.
In all, we think the revised OSI provides a valuable tool for the diagnosis of
stress in organizations. The comprehensiveness of the instrument allows for the
assessment of many different aspects of the stress process. Parts of the OSI may be
useful in other than stress research as well. However, we realize that further
research on the construct validity and criterion validity is needed.

Acknowledgementa
The authors want to thank students Eiso Bos and Erwin Tabak (Study 1), Thomas dc Vries (Study 2),
Casper Bcntinck (Study 3) and Maud van Aalderen and Jacqueline van Heeswijk (Study 4). Their
enthusiasm and creativity was essential for the making of this revised OSI.

References
Baglionj, A. J. Jr., Cooper, C. L., St Hingley, P. (1990). Job stress, mental health and job sadsfaction
among UK senior nurses. Stress Medicine, 6, 9—20.
Bender, P. M. (1990). Comparative fit indexes in structural models. Psycholoffcal Buiielin, 107, 238—246.
Ben-Zur, H. (1999). The effecdveness of coping meta-strategies: Perceived efficiency, emodonal
correlates and cognidve performance. Personality and Individual Differences, 26, 923—939.
Billings, A. G., & Moos, R. H. (1981), The role of coping responses and social resources in attenuadng
the stress of life events. American Journal of Family Therapy, 10, 26—38.
Broers, P., Evers, A., & Cooper, C. L. (1995). Differences in occupadonal stress in three European
countries. International Journal of Stress Management, 2, 171-180,
Cohen, F., Reese, L. B., Kaplan, G. A., &c Riggio, R. E. (1986). Coping with the stresses of arthritis.
In R. W. Moskowitz & M. R. Haug (Eds.), Arthritis and tbe elderfy (pp. 47-56). New York: Springer.
Cooper, C. L., & Baglioni, A. J. Jr. (1988). A structural model approach toward the development of
a theory of the link between stress and mental health. British Journal of Medical Psycholog}^, 61, 87-102.
Cooper, C. L,, Sloan, S. J., & Williams, S. (1988). Occupational Stress Indicator management guide. Windson
NFER-Nelson.
Davis, A. J. (1996). A re-analysis of the Occupadonal Stress Indicator. V^ork and Stress, 10, 174-182.
Drenth, P. J. D., &c Sijtsma, K, (1990). Testtheorie. Houten/Antwerpen: Bohn Staflen Van Loghum.
Evers, A., Van Vliet-MtUder, J. C , & Groot, C. J. (2000). Documentatie van Tests en Testresearch in
Nederland. Amsterdam/Assen: NIP/Van Gorcum.
Frese, M,, & Zapf, D, (1988). Methodological issues in the study of work stress: Objecdve versus
subjecdve measurement of work stress and the quesdon of longitudinal studies. In C. L. Cooper &
R. Payne (Eds.), Causes, coping, and consequences of stress at work (pp. 375—411). Chichester: Wiley,
Friedman, M., & Powell, L. (1984). The diagnosis and qualitadve assessment of type A behavior
Introducdon and descripdon of the videotaped structured interview. Integrative Psychiatry, 2,
123-136.
Hcneman, H. G., & Schwab, D. P. (1985). Pay satisfacdon: Its mulddimensional nature and
measurement. International Journal of Psychology, 20, 129-141.
Ingledew, D. K,, Hardy, L., & Cooper, C. L. (1992). On the reliability and validity of the locus of
control scale of the Occupadonal Stress Indicator. Personality and Individual Differences, 13,
1183-1191.
Jenkins, C. D,, Zyzanski, S,, & Roscnman, R. H. (1979), Tbe Jenkins Actipity Survry manual. New York:
The Psychological Corporadon,
Kirkcaldy, B. D., & Cooper, C. L. (1992). Managing the stress of change: Occupational stress among
senior police officers in Berlin. Stress Medicine, 8, 219-231.
La2arus, R. S., & Folkman, S. (1984). Stress, appraisal and coping. New York: Springer.
Revisions and further developments of the OSI 237
Levensoo, H . (1981). Diffcrentiatiog among intemality, powerful others and chance. In H. M.
Lefcourt (Ed.), Research with the iocus cf control construct: Vol. 1. Assessment methods (pp. 15-63). N e w
York: Academic Press.
MacCallum, R. C , Roztiowski, M., & Necowitz, L- B. (1992). Model modifications in covariance
structure analysis: T h e problem of capitaJi2ation o n chance. Psycholoffcat Bulletin, 111, 490-504.
M o h r , G. (1986). Die Erfassung psychischer Befindensbeeintrachtigungen bei Industriearbeiten. Fratikfurt am
Main: Peter Lang.
Palenzuela, D . L. (1988). Refining the theory and measurement of expectancy of internal vs external
control of reinforcement. Personality and Individual Differences, 9, 607-629.
Rees, D., & Cooper, C. L. (1992d). Occupational stress in health service workers in the U K . Stress
Medicine, 8, 79-90.
Rees, D . W., & Cooper, C. L. (1992i). T h e Occupational Stress Indicator locus of control scale:
Should this be regarded as a state rather than trait measure? Work and Stress, 6, 4 5 - 4 8 .
Robertson, I. T., Cooper, C. L., & Williams, j . (1990). The validity of the Occupational Stress
Indicator. Work and Stress, 4, 29-39.
Schreurs, P. J. G., Willige, G. van de, Brosschot, J. F., Tellegen, B., Sc Graus, G . M. H. (1993). De
Utrechtse Coping Lijst. Omgaan metproblemen en gebeurtenissen. Her^ene handteidung (The Utrecht Coping
Inventory. Dealing with problems and events. Revised manual). LJsse: Swets & Zeitlinger.
Schumacker, R. E., & Lomax, R. G . (1996). -^ beginners guide to structural equation modelling. Mahwah, NJ:
Eribaum.
Spector, P. E. (1992). A consideration of the validity and meaning of self-report measures of job
conditions. I n C. L. Cooper & I. T. Robertson (Eds.), International review of industrial and organisational
psychology (Vol. 7, p p . 123-151). Chichester: WUey.
Spector, P. E., & Jex, S. M. (1998). Development of four self-report measures of job stressors and
strain: Interpersonal conflict at work scale, organizational constraints scale, quantitative workload
inventory, and physical symptoms mvzntory. foumal of Occupational Health Psychology, 3, 356-367.
Thierry, Hk. (1997). Motivation and satisfaction. In P. J, D . Drenth, Hk. Thierry, & Ch.J. D e Wolff
(Eds.), Handbook of work and organisational psychology: Vol. 4. Organisational psychology (2nd cd,
p p . 89-119). Hove: Psychology Press.

Receieed 8 April 1998; revistd version received 25 May 1999

Appendix
Items of the short version of the Dutch OSI (Spanningsmetet)
Type A behaviour
(You are required to indicate the extent to which you agree or disagree with statements about yourself
and your behaviour.)

Achievement orientation
I work quicker than other people
I demand a lot of the quality of my work
I take my job seriously
I like a challenging job
I feel responsible for my work
I keep trying to improve my work
I want to perform well in my work
I am a conscientious worker
<>
Irritation
I can not stand it when people do not immediately understand what I a m explaining
W h e n someone annoys me I will speak curtly to that person
People sometimes say that I easily get a temper
A. Evers et al.
• I am very impatient when people arc late foe an appointment
• When I get angry at someone, I do not want to work with him or Tier anymore
• I cannot work with people who have difficulty making decisions
• When my collegues worked as hard as I do, we would be able to do a lot more
• When someone makes an error, I will be very explicit about it to him/her
Competitiveness
• I always compare my achievements at work with those of others
• I strive to do my work better than others
• I am an achievement-oriented person who has the need to win
• When I do my very best I do not have to compete with colleagues
• I am annoyed when a colleague of mine is promoted
• It irritates me to see someone else being lazy, while I have to work hard
• I do my best to please the management
• I do not go home hefore I have finished what I have planned

ZJJCUS of control •,
(You are asked to indicate your agreement with the following statements.)

Helplessness
• I have no say in the assignments I get at my work
• I fed like I am only a number at my work
• My work is just a cog in the machine over which I have little control
• It is useless to try hard, since there is no relation between my efforts and the appreciation for it
• The management does not appreciate good ideas and creativity, since they do the thinking and
make the decisions
Chance
• Finding another job is like taking part in a lottery
• Success in your work is mainly determined by favourable conditions
• When there is a reorganization in your company, you are just lucky not being fired
• The way my career will develop is something I have little influence on
• Being promoted depends on whether there happens to be a vacancy
Internality ,
• I am in control of my own career
• When I am promoted 1 owe that to my own efforts
• I owe a promotion to my own abilities
• I am fully responsible for the errors I make at work
• Enjoying your work depends on what you make out of it yourself

Coping styles
(Rate these potential coping strategies in terms of the eitent to which you actually use them as ways
of coping with stress,)

Social support
• I go to colleagues for advice when my superior and I disagree
• I ask people who have had the same problem as I what they have done to solve it
• I talk to colleagues who can actually help me out with a problem
• When 1 have problems I discuss them with my partner or my friends
• I talk about how I feel about my work
• When there are problems I look for understanding and sympathy
Revisions and further developments of the OSI 239
Denial
• When there are difficulties I pretend that everything is OK
• When problems arise I avoid thinking about them
• Whenever possible I try to stay away from problems
• I avoid annoying confrontations with my supervisors and/or colleagues
• In problematic situations I just wait and see what will happen
• When there are problems I just pretend nothing has happened
Active positive attitude
• I think I can learn from certain unpleasant expedences as well
• In times of stress I try to think of the nice elements of my work
• When there are problems I stay optimistic about the future
• I try to see problems in a different perspective so as to make them look more positive
• In problematic situations I try to convince the person who is in charge
• In times of stress I can plan my work well
• I arrange my work in such a way that 1 do not have to hurry
Compensating behaviour
• I notice that I drink more alcohol when I have problems
• When there are problems I tend to watch more TV
• I go out more often when there are problems
• Under pressure I lose myself in daydreaming or fantasies
• When there are problems I tend to take more risks
• Under pressure I get nasty

Sources of pressure
(Rate the items below in terms of the degree of pressure you perceive each may place on you.)
• Insufficient control over your job
• Ambiguity about your responsibilities
• Feeling isolated
• Covert discrimination and favouritism
• Pursuing a career at the expense of home life
• Being undervalued

Job satisfaction
(Indicate the extent to which you feel satisfied ot dissatisfied with your joh.)

General job satisfaction


• The degree to which you can personally develop or grow in your job
• Your working conditions
• The way in which conflicts are resolved in your company
• The psychological 'feel' or climate that dominates your organizadon
• The style of supervision
Satisfaction with pay
• The prospect of a raise
• Your current salary
• Changes in your salary in the course of time
• Your benefit package (e.g, car of the company, bonuses, pension plan)
• The company's pay policy
240 A. Epers et al.
Health
(Assess your current state of health. Consider the questions with reference to how you have felt over
the last three months,)

Mental ill-health
• During an ordinary working day are there times when you feel unsettled though the reasons for this
might not always be clear?
• Are there times at work when you feel so exasperated that you think to yourself that 'life is all really
too much effort'?
• As you do your job have you noticed yourself questioning your own ability and judgment?
• Do you usually feel relaxed and at ease or do you tend to fee! restless and tense?
• If the jobs you are doing start to go wrong do you sometimes feel a lack of confidence and panicky?
• Concerning work and life in genral, would you describe yourself as a 'worrier'?
• Do you find yourself experiencing fairly long periods in which you feel melancholy for no apparent
reason?
Physical ill-health
• Inability to get to sleep
• Headaches
• Feeling unaccountably tired
• Decrease in sexual interest
• Pricking sensations or twinges in parts of your body
• Feeling as though you do not want to get up in the mormng
• Feeling dizzy

You might also like