British Journal of Health Psychology (1999), 4, 257–275 Printed in Great Britain 257

q 1999 The British Psychological Society

Measures of perceived work characteristics

for health services research: Test of a
measurement model and normative data
Clare E. Haynes, Toby D. Wall*, Richard I. Bolden
and Chris Stride
Institute of Work Psychology, University of ShefŽeld, UK

Jo E. Rick
Institute for Employment Studies, University of Sussex, UK

Objectives. The aim of this paper is to describe the development and reŽnement of nine
measures of perceived work characteristics for use in studies of the psychological well-
being of health services employees. The constructs measured are: autonomy/control,
feedback, inuence, leader support, professional compromise, role clarity, role conict,
peer support and work demands.

Design. The study was a large scale cross-sectional survey.

Methods. National Health Service Trust staff were selected from seven major occupa-
tional groups (nurses, doctors, administrative staff, managers, professions allied to
medicine, professional and technical staff, and ancillary staff ) within a sample of Trusts
chosen to be representative of those in England. Questionnaires incorporating the scale
items were completed by over 9,000 staff. The Žt with the a priori nine-dimensional
measurement model was tested using conŽrmatory factor analysis. Scale internal
reliabilities and norms were calculated, and aspects of construct validity were examined.

Results. The Žndings show a good Žt to the measurement model for the sample as a
whole and across occupational groups. All the scales exhibit acceptable internal
reliabilities; and normative data is provided for all seven occupations (with norms for
more speciŽc job categories available on request). Analyses suggest good construct
validity, showing the scales discriminate as anticipated across occupational groups
and job categories, and they have the expected relationships with job satisfaction and
psychological strain.

Conclusions. The measures developed are usable for research purposes across the major
occupational groups, with the possible exception of ancillary staff. Further work is
desirable to conŽrm the measurement properties and extend the normative database.

It is widely recognized that job-related factors can affect people’s attitudes and mental
health at work. The substantial body of empirical research on job satisfaction over the last
*Requests for reprints should be addressed to Toby Wall, Institute of Work Psychology, University of ShefŽeld, ShefŽeld
S10 2TN, UK (e-mail: T.D.Wall@shefŽeld.ac.uk).
258 Clare E. Haynes et al.
20 years, and that focused more particularly on work stress, exempliŽes the enduring
interest of occupational, organizational and health psychologists in this issue, as do the
many reviews of the area and associated theoretical developments (e.g. Arsenault, Dolan
& Van Ameringen, 1991; Caplan, Cobb, French, Van Harrison & Pinneau, 1975; Karasek
& Theorell, 1990; Landeweerd & Boumans, 1994; Russell, Altmaier & Van Velzen, 1987;
Schabracq, Winnubst & Cooper, 1996; Warr, 1990). Recent initiatives in the Želd of
health and safety in Europe and elsewhere (e.g. EEC, 1989; Jenkins & Warman, 1993;
Sauter, Murphy & Hurrell, 1990), coupled with prominent recent legal cases in which
employees have received compensation for work-related psychological strain (e.g.
Howard, 1995), indicate that concern about stress at work will remain high on the
research and social agenda.
Whilst research on work stress to date has covered a wide range of occupations, from
blue collar workers, through teachers to police ofŽcers (e.g. Kauffmann & Beehr, 1989;
Russell et al., 1987; Wallace, Levens & Singer, 1988), investigation of those working in
health care has long been of especial interest. This is for three reasons. First, the number
of people employed in the health services in Britain, the rest of Europe and the USA is
large relative to many other sectors. Second, there are indications of high levels of strain
among health care employees, at least for those in ‘front-line’ occupations (e.g. Bond,
1984; Charlton, Kelly, Dunnell, Evans & Jenkins, 1993; Guralnick, 1963; Hingley,
1984). For instance, a large-scale investigation covering all the major occupational
groups within National Health Service (NHS) provider units (i.e. Trusts) reveals that
doctors, nurses and managers report higher levels of minor psychiatric disorder than do
their counterparts in the general working population (Wall et al., 1997).
The third reason for the interest in health care employees is perhaps the most
important. Health care employees are likely to encounter the full range of exposure to
many work factors theoretically implicated as determinants or moderators of strain. Due
to the nature of their jobs it might be expected, for example, that some nurses and doctors
will experience levels of role conict, role ambiguity or work demands, which, in other
occupations, may not be so extreme. Thus the study of work stress among health care
employees is important for theoretical as well as practical reasons.
From a methodological point of view, research on work stress in the health service to
date has one important weakness. Whereas considerable effort has been put into
developing measures of outcome variables, much less attention has been given to the
measurement of work characteristics implicated as causal factors. Thus there are many
good, and widely applicable measures of strain, which have been carefully constructed,
validated and supported by normative data, such as the Maslach Burnout Inventory
(Maslach & Jackson, 1986); the General Health Questionnaire (Goldberg, 1972;
Goldberg & Williams, 1991) and Warr’s measures of job-related anxiety and depression
(1990). Similarly there are established measures of job satisfaction (e.g. Warr, Cook &
Wall, 1979; Smith, Kendall & Hulin, 1969). With regard to work-related stressors,
however, few if any equivalent measures exist. Researchers have typically developed their
own scales. This is not too surprising since few research papers fully describe the scales
used, and normative data is generally not available. Nevertheless, the result is that it is
almost impossible to make direct comparisons across studies. As a consequence, there are
many, potentially good and varied measures of such constructs as autonomy or control,
social support, role conict and work demands, but few which have been used in the same
 Measures of work characteristics 259
form in more than a few studies. The only attempt to rectify this situation of which we are
aware is the recent study by Petterson & Arnetz (1997) which describes the properties of
several ‘work environment’, ‘modifying’ and ‘health’ (outcome) measures in three large
samples of Swedish health service employees.
This weakness in the measurement of predictor variables potentially undermines
individual studies. As Parkes (1982) noted some while ago, when discussing methodo-
logical problems in stress research, more adequate measures of the work environment are
needed in the form of ‘published test materials, rather than ad hoc measures designed
with reference to a particular work setting and rarely adequately validated’ (p. 786).
Relatedly, the weakness in the measurement of such independent variables also hinders
the development of cumulative knowledge. One way of systematizing knowledge across
studies is through meta-analysis. This method, however, is also limited by such
measurement deŽcits. Tett and colleagues (Tett, Meyer & Roese, 1994) reviewed Žve
meta-analyses examining the relationship between work features and strain. They
concluded that ‘Overall, the meta-analyses reviewed . . . provide support for expected
relations involving work-related stress’ (p. 100), but note that: ‘There seems little
question that meta-analysis is an imperfect research tool. In addition to its sensitivity to
researcher judgements, it cannot correct for widespread problems at the study-level’ (p.
106), and they point to the inconsistent measurement of independent variables as one
such type of problem.
The ability to compare the levels of stressors across studies is also important for
theoretical reasons. This is because of the emergence of propositions that relate
psychological effects either to the existence of relatively extreme levels of salient job
features, or to particular combinations of those features. Warr’s ‘Vitamin Model’ (1987) is
an example of the Žrst type. He proposes that job properties inuence strain in a way
analogous to the effects of vitamins on health: ‘that their absence tends towards an
impairment of mental health, but their presence above a required level does not yield
further beneŽt’ (p. 10). The ‘Demand–Control Model’ proposed by Karasek & Theorell
(1990) illustrates the second type of proposition. This predicts that psychological strain
and physical illness occur where jobs entail high ‘psychological demands’ (e.g. work
demands) and low ‘decision-latitude’ (e.g. autonomy or control). Proper tests of these
kinds of hypotheses manifestly require measures which allow assessment of the levels of
job properties in particular studies. Sample-speciŽc measures do not meet this need
because there is no way to determine whether scores deŽned as high (or low) in one study
are equivalent to those in another. In the absence of any absolute metric, such
comparability can only be achieved by measures supported by normative data, which
in turn require that they be widely applicable. In addition, there is a need for such
measures to be as short as possible (while retaining acceptable levels of reliability), so
that multivariate investigations are viable.
For the several reasons given above, the aim in this paper is to present psychometric
data on nine scales measuring work-related factors relevant to research into well-being
and job satisfaction in health service settings. The choice of constructs to be measured
was guided by both theoretical and practical considerations. The project from which
this paper comes was an investigation into the mental health of the NHS workforce
(Borrill et al., 1996) funded by the Department of Health. To establish the set of
constructs to be measured, an analysis was made of the variables identiŽed within leading
260 Clare E. Haynes et al.
theoretical frameworks, including the work of Katz & Kahn (1978), Caplan (1971,
1979), Warr (1987) and Karasek (e.g. Karasek & Theorell, 1990), and the integrative
accounts of this domain such as those provided by Arnold, Robertson & Cooper (1991)
and Cooper, Cooper & Eaker (1988). This eclectic approach led to the identiŽcation of
eight constructs common to two or typically more approaches, namely: autonomy/
control, feedback on work performance, inuence over decisions, leader support, role
clarity, role conict, peer support and work demands. Qualitative pilot work conŽrmed
the salience of these constructs to practitioners and also suggested the importance of a
ninth, which was named professional compromise. This is deŽned in the Method section,
along with the other constructs.
The scales were developed on a pilot sample of 825 NHS Trust staff and a main sample
of over 9,000 NHS Trust staff. The latter sample was collected so as to be representative
of the population of Trust employees as a whole and covered seven major occupational
groups (i.e. nurses, doctors, administrative staff, managers, professions allied to medicine
(PAMs), professional and technical staff (P&Ts) and ancillary staff ). Three questions were
addressed: (1) Can short scales be developed to Žt the assumed measurement model, and
to show good internal reliability? (2) Do the scales show evidence of construct validity?
(3) What are the norms for different occupational groups and job grades or types within
those occupational groups?

Method

Pilot work
The initial stage of the pilot work involved discussions with staff from three NHS Trusts to determine and
develop the applicability of a large number of potential items for the measures. Then a pilot questionnaire,
incorporating a large number of items to measure each construct, was devised and administered to employees
in those Trusts, to which responses were obtained from 825 people. The aim was to develop short scales of
six or fewer items which showed good face validity, a clear factor structure and high internal reliabilities.
To accomplish this reduction, exploratory factor analysis was Žrst carried out on a subsample (N = 389). On
the basis of this analysis, items that duplicated others, were not required to obtain high scale reliabilities,
or which did not load unequivocally on their proper factor, were identiŽed and omitted from further analysis.
The analyses were then rerun with the reduced item set on the remainder of the pilot sample (N = 436). This
showed all nine work-related scales had good internal reliabilities and gave a clear nine-factor structure. The
pilot work thus established the item set to be used in the main study.

Main sample and data collection

The main sample was obtained in two stages. First, using data from ofŽcial records and the expertise of
members of the NHS Executive, 20 Trusts were selected. Trusts were chosen to provide an overall proŽle
consistent with that of the 350 Trusts operating in England at the outset of data collection in 1993
(e.g. geographical location, size, type, Žnancial and market position). All 20 Trusts agreed to collaborate;
but for logistic reasons it was necessary to drop two, for one of which there was time to substitute another.
The second stage was to select individuals from the major occupational groups within the Trusts. These
were nurses, doctors, administrative staff, managers, professions allied to medicine (PAMs), professional and
technical staff (P&Ts) and ancillary staff. For the smaller occupational groups, where the number of possible
respondents was 100 or fewer, all staff were included. For larger occupational groups individuals were
randomly selected from staff lists, with the minimum proportion being 20%. Thus, under this method of
stratiŽed random sampling, greater proportions were chosen from smaller occupational groups and from
smaller Trusts. A total of 25 352 names were selected, and questionnaires were distributed to each named
 Measures of work characteristics 261
person’s place of work. Completed questionnaires were returned by 11 637 people (46%). The return rate
by occupational group was as follows: nurses, 50%; doctors, 42%; administrativestaff, 48%; managers, 81%;
PAMs, 58%; P&Ts, 44%; ancillary staff, 30%.
The above return rate for the whole sample (i.e. 46%) is not a proper index of response rate because the
staff lists were not entirely accurate. They included, for example, some people no longer working for a Trust.
To estimate the actual response rate, two tracer studies were carried out. These studies of named individuals
(N = 123 and 147) revealed that 25% and 29%, respectively, of those selected were not in a position to
respond at the time of the survey, due to a variety of factors, including labour turnover, sickness absence,
maternity leave, wrong address etc. Applying these attrition rates to the sample as a whole (reducing the
denominator accordingly) leads to an estimated overall response rate of 61% to 65%. For a fuller description
of this analysis see Wall et al. (1997).
Two questionnaires were used in the study. The main questionnaire, completed by two thirds of
respondents (randomly selected), included the work-related measures which are reported on in this paper.
A second questionnaire, completed by one third of respondents, focused on organizational climate. Thus, the
number of respondents in the main sample who completed the work-related measures is 7698.
In the course of the above planned Želdwork, three further Trusts asked to be included in the study. Staff
were selected from these volunteer sites at random as for the original sample. A total of 3945 names were
selected and questionnaires distributed.Questionnaires were returned by 2064 people (52%), of whom 1629
completed the work-related measures. To determine whether it was appropriate to include data from these
additional respondentsin our analyses, a comparison was made between the scale means obtained in volunteer
sites and those in the original sample, both for the samples as a whole and by occupational group. For seven of
the nine measures there was no statisticallysigniŽcant difference between the planned and volunteer samples.
For the autonomy/control and professional compromise measures, however, differences were observed for the
full sample ( p < .001 and p < .007, respectively). Nevertheless, the size of the effect was extremely small,
accounting for less than 0.1% of the variance in both cases. Moreover, examination by occupation showed few
differences between the volunteer and planned sites, and none which were substantial.The difference between
the means was less than 0.10 in all cases, except for ancillary staff, where it was around 0.50. Thus it was
decided to include this additional data in the study because it held no material threat to the Žndings relating
to norms (with the possible exception of those for ancillary staff ), yet had the beneŽt of providing larger
numbers of especial value for the Žner grained breakdown. The size of the Žnal sample, therefore, was 9327,
collected from 22 Trusts. The mean ages overall and for men and women separately in each occupational
group is shown in Table 1.

Table 1. Age proŽles for the sample as a whole, and for men and women by occupational group
Mean Number of Mean Number Mean
Total age SD females age SD of males age SD
Full sample 9327 39.61 10.53 7045 39.20 10.51 2282 40.87 10.63
Nurses 3441 38.51 10.15 3116 38.63 10.17 325 37.41 9.98
Doctors 1075 39.87 10.30 353 37.24 9.87 722 41.15 10.26
Admin. 1467 41.42 11.34 1318 41.53 11.15 149 40.48 12.91
Managers 750 40.50 8.89 374 39.82 8.78 376 41.18 8.97
PAMs 1409 38.18 10.32 1221 37.81 10.25 188 40.67 10.45
P&Ts 591 38.39 10.49 353 37.14 10.33 238 40.27 10.47
Ancillary 594 44.55 11.31 310 44.53 11.00 284 44.58 11.65

The work characteristics measures

The measures are based predominantly on previously developed instruments, adapted where necessary for
use in a health service context. A full list of the items and response scales is presented in Appendix 1. Space
constraints prevent inclusion of the main source(s) from which items were drawn, but these are available
on request from the authors. A brief description of each measure follows.
262 Clare E. Haynes et al.
The measure of autonomy/control (A/C) focuses on the extent to which individuals can choose how they
carry out their work. The measure of feedback (F) on work performance was designed to tap the extent to
which individuals receive feedback about their work performance. The inuence (I) scale measures the extent
to which individuals perceive they are able to contribute to decision making at work. The leader support (LS)
scale measures the extent to which individuals receive support and encouragement from their immediate
superior. A measure of professional compromise (PC) was specially devised for the current study to tap the
construct which pilot work suggested was potentially detrimental to well-being, but which was not well
covered by other scales. It is designed to measure the extent to which individuals believe that staff within
the Trust have to compromise professional standards in carrying out their work in order to meet conicting
objectives, such as reducing Žnancial costs and staff levels. The role clarity (RCL) scale is designed to measure
the extent to which individuals are clear about their work role; whilst that of role conict (RCO) focuses
on the extent to which individuals receive conicting instructions from others about their own work
requirements. The peer support (PS) measure was designed to capture the extent to which individuals receive
practical and emotional support from their peers (as distinct from leader support, see above). Finally, the work
demands (WD) scale was designed to measure the extent to which individuals feel they have the time and
resources to carry out their job properly.
The items were grouped according to their respective measures in the questionnaire,as pilot work revealed
this was the method preferred by respondents and scales had different response formats. For eight of the nine
scales all items were positively worded. The question of whether or not to include negatively worded items
is a difŽcult one to resolve. Certainly the inclusion of negative items can produce methodological artifacts, as
Schmitt & Stults (1985) have clearly demonstrated. Using items all worded in the same direction, however,
carries the risk of inated inter-item correlations (and hence internal reliabilities) as a result of such bias as
generalized positive (or negative) affect. However, in a study such as this, where focus is on the distinctiveness
across separate scales, such biases would show up as a common factor and mitigate against Žnding a clear
factor structure.On balance, therefore, it was decided to opt for wording all items in the same direction. The
one exception was for the feedback scale, where there was no previous evidence of problems associated with
negatively worded items, and all original sources included both positively and negatively worded items.

Outcome measures
Also included in the study were three established outcome measures for use to help establish construct
validity. The Žrst is the 15-item Job Satisfaction Scale developed by Warr et al. (1979). An additional item
concerning how satisŽed people are with in-service training was added, resulting in a 16-item measure. This
has a 7-point response scale running from ‘extremely dissatisŽed’ to ‘extremely satisŽed’. Scale scores are
derived by averaging item scores, with higher values representing greater satisfaction. This measure had an
internal reliability for the present sample of 0.88. Job-related anxiety and depression were tapped using the
relevant parts (i.e. three items measuring anxiety and three items measuring depression) from the set of
measures of well-being and mental health developed by Warr (1990). A 5-point response scale was used and
scored so that higher values (average item score) represent greater anxiety and depression. The internal
reliabilities for the present sample were 0.85 and 0.88 respectively.

Statistical analyses
ConŽrmatory factor analysis (CFA) was carried out using the EQS programme (Bentler, 1988), and involved
testing a sequence of measurement models of increasing complexity. The baseline was a null model with zero
covariances between items (Bentler & Bonnett, 1980). Stage 2 tested for a one-factor model, in case the items
could be subsumed under a single construct as a result, for example, of common method variance or
generalized positive or negative affect. This was followed, in stages 3 to 5, by tests for a two-factor model,
a four-factor model and the predicted nine-factor model. For these latter stages, which involve two or more
factors, fully oblique models were speciŽed, because the dimensions of interest are expected to be correlated
(e.g. role conict and role clarity, or role conict and professional compromise).
Having set out to measure nine different constructs, a large number of alternative intermediate
measurement models of two or more dimensions could be speciŽed. To reduce the analysis to manageable
proportions, four alternatives were formulated. The two-factor model distinguished between an overall
 Measures of work characteristics 263
demands factor (work demands, role conict, role clarity and professional compromise items combined) and
an overall support factor (peer support, leader support, autonomy/control, feedback and inuence items
combined). The four-factor model, reecting the distinctions made by Karasek & Theorell’s demands–
control model (1990), speciŽed a demands factor (work demands and professional compromise items
combined), a control/inuence factor (autonomy/control, inuence and feedback items combined), a support
factor (peer support and leader support items combined) and a role characteristics factors (role conict and
role clarity items combined). Finally, the nine-factor model speciŽed each of the above scales as separate
factors.
The appropriateness of each model was examined using a number of approaches. The overall Žt of a
measurement model to the data has most commonly been tested using the chi-square test statistic, and the
ratio of the chi-square to its degrees of freedom. However, this approach is not entirely satisfactory on its own
because the magnitude of chi-square, and hence its statistical signiŽcance, varies according to sample size.
For this reason, Bentler & Bonett’s (1980) non-normed Žt index (NNFI: rho) was also used. This compares
the Žt of the speciŽed model with that of a null model, taking account of the degrees of freedom used
in moving from one model to another. Acceptance of a more complex model over a simpler one requires
both a substantial decrease in chi-square and a large increase in rho; and only where these criteria are met is
preference for the more complex model over the simpler one justiŽed.
In addition, the comparative Žx index (CFI) (Bentler, 1990) was used as this shares the desirable properties
of the rho index but has a lower standard error. Bentler (1992) suggested that a value of 0.90 or more is
required to indicate an acceptable Žt for the data. Both rho and CFI are independent of sample size (Bentler,
1990; Marsh, Balla & MacDonald, 1988); and both are incremental Žt indices which depend on the choice
of a particular null model. In the case of CFA, they indicate the extent to which item covariances may be
reproduced, relative to a model of zero common factors. The sensitivity of the chi-square statistic has resulted
in numerous alternative indices being proposed and evaluated (for reviews see Marsh et al., 1988). One such
index is Akaike’s (1987) information criterion (AIC) and we report an evaluation of all models based on this.
Considering these several indices together means that a good Žt for the predicted nine-factor model
requires: (a) successive signiŽcant decreases in chi-square as one moves from the simpler to the more complex
full model; (b) equivalent decreases in the AIC index; and (c) progressive increases in both rho and CFI, with
Žnal values of the order of 0.90.

Results

Applicability of the nine-factor model

The Žrst aim of the study was to determine whether the factor structure established in the
original development of the scales would be replicable in a separate larger sample, and
would represent a good Žt to the data. This was addressed initially through the use of
CFA on the data from all employees. Of the 9327 respondents, 1978 were dropped from
the analysis because of listwise deletion of missing data.
Fit indices for the competing models are shown in Table 2. The one-factor model
shows a clear improvement over the null model (D v 2 = 58 376.59) but nevertheless a
poor Žt to the data ( v 2 (860) = 135 766.58, p < .001; q = 0.27; CFI = 0.30). This
indicates that all items cannot be subsumed under a single construct of, say, generalized
positive or negative affect. The distinctiveness of demands and support is shown by the
improved Žt given by the two-factor oblique solution, indicated by the further decrease
in chi-square in the sample (D v 2 = 30 271.86). Similarly the substantial decrease in chi-
square associated with the move from a two-factor to a four-factor model indicates that
demands, autonomy, support and role characteristics, are all distinguishable from one
another (D v 2 = 30 033.16). Following the same pattern, there is a large decrease in chi-
square (D v 2 = 58 346.57) with the move to the predicted nine-factor model. In this case
the Žnal chi-square ( v 2 = 17 042.99) is relatively small for a sample of this size, and the
264 Clare E. Haynes et al.
Žt indices reach Bentler’s (1992) recommended levels for accepting the measurement
model (i.e. q = 0.91; CFI = 0.92). The AIC Žgures complement the emerging picture
with a statistic for the predicted model substantially smaller than that for all the
alternatives considered. Moreover, the Žt for the rho and the CFI indices is achieved
without making any adjustments for correlated errors.

Table 2. CFA Žt indices for competing models

2 2
Competing model v d.f. q CFI AIC D v
Sample (N = 7349)
1. Null model 194 229.17 903 NA NA 192 423.17 NA
2. One-factor model 135 766.58 860 0.27 0.30 134 046.58 58 376.59
3. Two-factor model 105 492.72 859 0.43 0.46 103 774.72 30 271.86
4. Four-factor model 75 449.56 854 0.59 0.61 73 741.56 30 033.16
5. Nine-factor model 17 042.99 824 0.91 0.92 15 394.99 58 346.57
Note. NA—not applicable.

Table 3 shows the item loadings for the nine-factor model. The pattern is very clear.
All items load heavily on their ‘own’ factor (0.58 to 0.89). Factor intercorrelations are
presented in Table 4. This shows that, while the factors are typically correlated with
one another, even the most strongly interrelated share no more than 30% of variance.

Table 3. Factor loadings for the nine-factor oblique model

Itemsa A/C F I LS PC RCL RCO PS WD
1 .74 .71 .64 .81 .73 .61 .78 .80 .84
2 .78 .72 .78 .76 .61 .58 .85 .87 .89
3 .68 .74 .83 .83 .85 .73 .79 .87 .81
4 .78 .76 .88 .84 .85 .87 .75 .84 .75
5 .83 .81 .82 .80
6 .67 .82 .68
a
Items are shown in Appendix 1.
Note. All factor loadings are signiŽcant at p < .001. A/C = autonomy/control; F = feedback; I = inuence; LS = leader
support; PC = professional compromise; RCL = role clarity; RCO = role conict; PS = peer support; WD = work
demands.

Supporting analyses
Three additional sets of analyses were conducted to further explore the Žndings. First,
because of the substantial loss of respondents resulting from listwise deletion of data
(22%), a CFA was run on the total sample using the covariance matrix derived by pairwise
deletion for missing values. The range of sample sizes for the covariances was from 8080
to 9260, with a mean of 8995. Therefore, a much greater proportion of the full sample
(9327) was included (i.e. 87%–99%, averge 96%). An equally acceptable level of Žt was
achieved for the full nine-factor model, with the rho and CFI Žt indices being 0.91 in
both cases.
Although the full nine-factor measurement model provides an acceptable Žt to the data
as a whole, this does not necessarily mean it applies equally to all the occupational groups
 Measures of work characteristics 265
Table 4. Factor correlations for the nine-factor oblique model
Factor A/C F I LS PC RCL RCO PS WD
Autonomy/control
Feedback .09*
Inuence .55* .28*
Leader support .12* .39* .53*
Prof. compromise .05* - .24* - .06* - .18*
Role clarity .25* .54* .30* .36* - .20*
Role conict - .06* - .37* - .20* - .34* .46* - .38*
Peer support .17* .32* .38* .42* - .10* .33* - .26*
Work demands .24* - .28* .10* - .18* .52* - .26* .52* - .08*
*p < .001.
Note. N = 7349. A/C = autonomy/control; F = feedback; I = inuence; LS = leader support; PC = professional compro-
mise; RCL = role clarity; RCO = role conict; PS = peer support; WD = work demands.

in the sample. Thus the second set of analyses involved Žtting a multigroup model. The
sample was partitioned into the seven major occupational groups, and, initially, the three
main measurement parameters (i.e. factor loadings, factor intercorrelations and item
variances) were allowed to vary freely. This is equivalent to Žtting the model separately
to each group in turn, but is more parsimonious in that it yields Žt indices for the sample
as a whole. The Žt indices obtained (i.e. rho 0.90, CFI 0.91) were virtually the same as
those for the original analysis, indicating a reasonable Žt to the nine-factor model across
all occupational groups. Subsequent analyses constraining various combinations of the
three measurement parameters (i.e. item variances, factor intercorrelations and factor
loadings) to be equal across occupational groups reduced the Žt of the model (e.g. rho and
CFI both 0.89, when all three parameters are constrained). This indicates that these
parameters within the core nine-factor model differ across the occupational groups;
however, the model itself appears robust across them.
The Žnal set of analyses compared the Žt of two alternative partial oblique models with
the fully oblique model as tested. The partial models also had nine factors, but allowed
correlations only among speciŽed sets of factors rather than amongst them all. The factors
allowed to intercorrelate were those as speciŽed earlier when describing the two and the
four-factor models forming part of the original CFA. Neither yielded as good a Žt as
the fully oblique model.

Scale reliabilities
The internal reliabilities (Cronbach’s alpha) for each of the seven occupational groups are
presented in Table 5 (also included in this table are the scale means and standard
deviations for each occupational group, and one-way ANOVAs on occupational differ-
ences). It is evident that all nine scales consistently show good levels of reliability. The
range across occupational groups is from 0.70 to 0.92. Only in the case of the feedback
scale for ancillary staff (0.70) is the internal reliability inadequate (which may be due to
misresponding to reverse-worded items which only this scale contains). A point worth
noting is that the use of positively worded items in all but one of the scales (i.e. feedback),
and the fact that the items were blocked into sets representing scales, may have slightly
inated some of the alpha ratings.
266 Clare E. Haynes et al.
Table 5. Scale internal reliabilities (Cronbach’s alpha, a ), means and standard deviations for the
seven occupational groups
Autonomy/ Leader Professional
control Feedback Inuence support compromise
Job category a M SD a M SD a M SD a M SD a M SD
Nurses .88 3.54 0.93 .80 3.55 0.77 .86 2.95 1.00 .92 3.32 1.03 .84 3.22 0.99
Doctors .86 3.47 0.90 .87 3.60 0.83 .85 3.20 1.06 .90 3.12 1.01 .80 3.31 0.90
Admin. .87 3.56 0.99 .82 3.66 0.77 .83 2.73 1.04 .92 3.29 1.10 .88 3.12 1.08
Managers .83 4.34 0.57 .86 3.53 0.80 .86 3.90 0.89 .90 3.40 0.95 .83 3.19 0.87
PAMs .89 3.73 0.97 .85 3.57 0.80 .84 3.31 0.98 .92 3.47 1.00 .83 3.40 0.93
P&Ts .88 3.28 1.00 .82 3.68 0.77 .87 2.99 1.08 .92 3.13 1.08 .84 3.20 0.97
Ancillary .88 2.90 1.09 .70 3.56 0.76 .82 2.29 1.02 .93 3.02 1.15 .87 2.86 1.20
One-way
ANOVA F 151.25 5.67 193.99 21.03 21.99
d.f. 6,9217 6,9087 6,9224 6,8949 6,8699
p <.001 <.001 <.001 <.001 <.001
Var.
explained % 8.9% <1% 11.1% 1.3% 1.4%
Note. The number in each occupational group varies as mean substitution was used in computing mean scale scores: nurses
(3420–3278); doctors (1068–864); admin. (1458–1143); managers (750–653); PAMs (1404–1247); P&Ts (590–511);
ancillary (578–391).

Validity
Clearly, the face validity of the scales is high because items, largely drawn from
established measures, were selected speciŽcally to represent the constructs of interest.
Moreover, the Žndings from the CFA, by demonstrating that the constructs are factorially
distinct, provide support for one important aspect of construct validity. Consideration
of the validity of the scale is now extended by examining: (a) their sensitivity to
occupational and job differences; and (b) their relationships with psychological outcomes.

Discrimination across occupational and job categories. At the most general level it is to be
expected that the scales will show differences in mean scores across the seven occupational
groups in the sample, and across different jobs within each of those occupations. Scale
scores were derived by averaging item scores. The results from one-way ANOVAs on
occupational differences are shown in Table 5. Also included in this table are the means
and standard deviations for each of the occupational groups, which provide part of the
normative data commented upon later.
The Žndings show that for all nine scales there are statistically signiŽcant ( p < .001)
occupational differences, with the amount of variance accounted for ranging from almost
9% (autonomy/control) to less than 1% (feedback). (The equivalent analyses for
differences across jobs within occupations are available on request from the authors.)
These also reveal consistent, and often large, differences in means for virtually all the
scales. Among hospital nurses, for example, the mean scores for autonomy/control
systematically increase in line with job seniority, with the latter job factor accounting
for almost 31% of the variance in the scale scores. Similarly, among doctors, inuence
scores increase in line with job level, and show 25% shared variance.
 Measures of work characteristics 267
Table 5. (continued)
Role Role Peer Work
clarity conict support demands
Job category a M SD a M SD a M SD a M SD
Nurses .83 3.91 0.71 .87 2.42 1.02 .90 3.77 0.87 .91 2.83 1.05
Doctors .82 3.78 0.77 .78 2.27 0.90 .90 3.64 0.86 .89 3.16 1.03
Admin. .84 3.94 0.79 .88 2.13 1.04 .92 3.49 1.03 .92 2.58 1.12
Managers .85 3.72 0.75 .85 2.47 0.97 .90 3.51 0.85 .89 3.21 0.95
PAMs .83 3.87 0.72 .86 2.11 0.93 .90 3.83 0.81 .91 2.95 1.07
P&Ts .84 3.83 0.80 .89 2.28 1.07 .91 3.60 0.92 .91 2.43 1.05
Ancillary .84 3.87 0.89 .90 2.26 1.20 .92 3.35 1.14 .90 2.28 1.05
One-way
ANOVA F 11.29 27.26 40.33 90.02
d.f. 6,9227 6,9236 6,9231 6,920
p <.001 <.001 <.001 <.001
Var.
explained % <1% 1.7% 2.5% 5.5%

For construct validity, of course, it is not simply the case that scales should be sensitive
to differences across occupations and jobs, but also that they should reect expected
differences. Given the number of occupations, jobs and scales involved in the present
study, literally hundreds of such predicted relationships could be considered; but this is
outside the scope of this paper. However, a few examples serve to illustrate how the
measures show expected effects. One example concerns the inuence over decisions
reported by nurses compared with doctors and managers. Despite many attempts to
rectify the situation, it has long been recognized that nurses’ views in the NHS have not
carried the same weight as their medical and managerial colleagues, and recent changes
have favoured increasing managerial inuence. This being the case, one would expect a
clear order for the means on the inuence scale, which is conŽrmed by the Žndings (see
Table 5) where managers report the highest scores (3.90), doctors intermediate levels
(3.20) and nurses the least (2.95). This pattern is also evident amongst the most senior
members of each of these three occupational groups, where the Žgure for senior managers
is 4.21, whilst the mean inuence scores for consultant doctors and senior nurses are 3.66
and 3.68 respectively.
A second example concerns the autonomy/control scale. A key feature of job level is
that, as it increases, it confers on people greater discretion over how to carry out their
work. It follows that the A/C measure should produce scores positively related to job
level. For those occupations in which jobs are classiŽed mainly in terms of job level,
namely hospital nurses, doctors, managers and ancillary staff, the Žndings are entirely
consistent with the expected pattern; senior nurses (Grades H and I) yield a mean of 4.25
on the A/C scale, which progressively decreases down the grades to 2.55 for the junior
nurses (Grade A). Similarly, the mean scores on the autonomy/control scale for
consultants is 3.83, whereas for registrars it is 3.22 and for junior doctors (i.e. house
ofŽcers and senior house ofŽcers) it is 2.85.
A Žnal example relates to the scale for professional compromise, which focuses on
trade-offs which may have to be made between cost responsibilities and providing quality
268 Clare E. Haynes et al.
patient care. Whereas one might expect professional compromise to affect all occupa-
tional groups in the NHS to some extent, it would be likely to affect most strongly
those in direct clinical contact with patients. This is conŽrmed by the data. As shown in
Table 5, the scores for nurses, doctors, PAMs and P&Ts (3.22, 3.31, 3.40 and 3.30,
respectively) are higher than those for managers, administrative staff and ancillary staff
(3.19, 3.12 and 2.86, respectively).
The above Žndings are not surprising, nor should they be. They, along with other
relationships evident within the data, serve to conŽrm that the scales are detecting the
kind of associations which they should, and thus contribute towards the overall
assessment of construct validity.

Relationships with outcomes. The Žnal aspect of construct validity to be addressed concerns
the relationships of the scales with psychological outcome measures. Correlation
coefŽcients among the nine scales and three outcome variables are shown in Table 6.
These are consistent with expectations from previous research. It would be predicted, for
example, that those with higher role clarity would report greater psychological well-
being (e.g. Fang & Baba, 1993; Revicki, Whitley & Gallery, 1993). Our Žndings show
this to be the case, with role clarity positively associated with job satisfaction (r = .42)
and negatively related to anxiety and depression (r = - .030 in both cases). Similarly,
work demands have been found to be negatively associated with well-being (e.g. Agius,
Blenkin, Deary, Zealley & Wood, 1996; Richardson & Burke, 1991). Again, the results
reect this, as scores on the work demands scale are negatively correlated with job
satisfaction (r = - .29) but positively associated with anxiety and depression (r = .44
and r = .31, respectively). More generally, as discussed earlier, the rationale for the choice
of all the work characteristics measures developed in this study was that they were
potential predictors of well-being at work, so all should show such relationships. The
only exception among the 27 relevant correlations (those between each of the nine job
characteristics measures and the three outcomes) is the lack of a statistically signiŽcant
relationship between autonomy/control and job-related anxiety. Lack of an association
between these two variables is repeated within occupational groups (not reported in table
format), with the exception of managers and doctors. With these two groups, autonomy/
control is related negatively to job-related anxiety (managers; r = - .13, p < .001;
doctors; r = - .16, p < .001). The fact that there is a relationship with these two
occupations is not surprising given the nature of their jobs. For example a manager who is
responsible for budgets, but has little control over them, may feel some anxiety. The
inconsistent relationship between autonomy/control and anxiety across occupational
groups more generally might also be explained by the fact that the effects of this variable
on outcomes is often predicted to be contingent on other variables (e.g. work demands,
see Karasek & Theorell, 1990).

Normative data
The Žnal aim is to provide normative data for the nine scales. The means and standard
deviations of each scale for the seven occupational groups are presented in Table 5, which
thus provides relevant information. These seven occupational groups were further
Table 6. Correlation coefŽcients between work factors and outcome measures
Factor 1 2 3 4 5 6 7 8 9 10 11 12
1. Autonomy/control
2. Feedback .08*
3. Inuence .53* .23*
4. Leader support .12* .32* .46*
5. Prof. compromise .05* - .21* - .04* - .16*
6. Role clarity .24* .45* .27* .32* - .18*
7. Role conict - .05* - .32* - .16* - .32* .41* - .33*
8. Work demands .22* - .25* .11* - .19* .48* - .24* .48*
9. Peer support .17* .27* .34* .39* - .09* - .29* - .23* - .08*
10. Job-related anxiety .00 - .28* - .06* - .22* .29* - .30* .39* .44* - .18*
11. Job-related depression - .09* - .27* - .19* - .26* .26* - .30* .37* .32* - .22* .70*
Measures of work characteristics

12. Job satisfaction .29* .39* .51* .56* - .36* .42* - .47* - .29* .42* - .36* - .47*
*p < .001.
Note. N = 8449- 9243 depending on missing values.
269
270 Clare E. Haynes et al.
subdivided into more speciŽc job categories. (These norms are available from the
authors.) For example, doctors were split into three groupings; house ofŽcers/senior
house ofŽcers, registrars/senior registrars and consultants.

Discussion
The aims of this paper were to answer three questions: First, could nine scales designed to
measure work-related factors relevant to research into the well-being of health service
staff be developed to Žt an assumed measurement model? Second, do the scales show
evidence of construct validity? Third, what are the norms for the different occupations
and job groupings within those occupations?
In response to the Žrst question, Žndings from a conŽrmatory factor analysis showed
empirical support for the conceptual distinction between the constructs as the predicted
nine-factor model exceeded Bentler’s (1992) recommended minimum CFI of 0.90;
whereas none of the alternative models tested provided as good a Žt. Furthermore, the
model was shown to apply across all major occupational groups, as well as to the sample as
a whole. Correspondingly, scale reliabilities were consistently good or acceptable across
all nine scales, both for the sample as a whole and within the seven occupational groups.
Examination of construct validity was approached initially by determining the
sensitivity of the scales to differences across occupational groups and to differences
across jobs within these occupational groups. Findings showed that all nine scales
discriminated as expected across occupations and jobs. In a further effort to address the
construct validity of the scales, the relationships of the nine scales with psychological
outcome variables, namely job satisfaction, job-related anxiety and job-related depression
were examined. The pattern of these results provides further evidence of construct
validity, as they were consistent with predictions from the literature.
The Žnal aim was to provide normative data. This was presented for seven major
occupational groups within the NHS, ranging from nurses and managers to ancillary
staff; and data has been made available on request for more speciŽc job categories within
each occupational group. This information is a useful comparative base for future studies.
A point to note, however, is that this data is based on a response rate of about 60%.
Whilst this compares favourably with the response rates achieved by many studies of this
size, it leaves open the possibility that the data may be affected by non-response bias.
Overall, the evidence supporting the nine scales is convincing; however, a number
of points are worthy of comment. First, the general applicability of the professional
compromise scale, devised speciŽcally for the current study, may be questioned. Though
in principle it can be answered by all health service employees, it is instructive that
administrative and ancillary staff were more likely to omit responding to items in this
scale than the other groups. The implication is that they found it less relevant to their
work, which is understandable given two of the four items are concerned with clinical
contact with patients. Regarding ancillary staff more generally, the low return rate (30%)
from this group may have been a function of the nature of their contracts (i.e.
predominantly short-term, therefore, high turnover); however, it could also indicate
that the scales are less appropriate for these workers, and the normative data presented
should, therefore, be interpreted with caution. The leadership support scale may have a
similar limitation. In particular, it should be noted that a substantial minority of doctors
 Measures of work characteristics 271
failed to complete all the items in this scale (i.e. 198 out of 1075 had missing values).
Of these ‘non-responders’, 85% were consultants. It is not unreasonable to assume this
was because they did not see themselves as having a manager to whom they report. Thus
the scale may not be appropriate for this particular job group.
Two additional points are worthy of comment. First, it is desirable to extend further
the empirical base with a view to improving the status of the normative data. More data
on these kinds of jobs, as well as others not included here, is ultimately required. The
recent paper by Petterson & Arnetz (1997) is an encouraging step in this direction.
Second, it will be recognized that we chose to operationalize the constructs through
self-report scales. It is important to have such perceptual measures since it is to be
expected that effect of job properties on psychological and behavioural outcomes depends,
at least in part, on employees being aware of them. Nevertheless, for wider research
and practical purposes, it remains important that measurement be extended to include
independent job ratings as well as the development of parallel measures based on
objective criteria. This paper provides a springboard for such developments.

Acknowledgements
The support of the NHS Executive is gratefully acknowledged. The work was funded as part of the
programme on the Mental Health of the NHS Workforce. Thanks are also due to the Trusts and
the employees who participated; Fiona Hill; David Woods for his statistical support and assistance;
and to our other colleagues, Carol Borrill, Angie Carter, David Golya, Gillian Hardy, David Shapiro and
Michael West who have contributed to the larger project from which the data for this paper were drawn.

References
Agius, R. M., Blenkin, H., Deary, I. J., Zealley, H. E. & Wood, R. A. (1996). Survey of perceived stress and
work demands of consultant doctors. Occupational and Environmental Medicine, 53(4), 217–224.
Akaike, H. (1987). Factor analysis and AIC. Psychometrika, 52, 317–332.
Arnold, J., Robertson, I. T. & Cooper, C. L. (1991). Work psychology. Understanding human behaviour in the
workplace. London: Pitman.
Arsenault, A., Dolan, S. L. & Van Ameringen, M. R. (1991). Stress and mental strain in hospital work:
Exploring the relationship beyond personality. Journal of Organizational Behaviour, 12, 483–493.
Bentler, P. M. (1988). Causal modelling via structural equation systems. In J. R. Nesselroade & R. B. Cattell
(Eds), Handbook of multivariate experimental psychology, 2nd ed., pp. 317–335. New York: Plenum.
Bentler, P. M. (1990). Comparative Žt indexes in structural models. Psychological Bulletin, 107, 238–246.
Bentler, P. M. (1992). On the Žt of models to covariances and methodology to the bulletin. Psychological
Bulletin, 112, 400–440.
Bentler, P. M. & Bonett, D. G. (1980). SigniŽcance tests and goodness of Žt in the analysis of covariance
structures. Psychological Bulletin, 88, 588–606.
Bond, M. (1984). Stress and self-awareness: A guide for nurses. London: Heinemann.
Borrill, C. S., Wall, T. D., West, M. A., Hardy, G. E., Shapiro, D. A., Carter, A. J., Golya, D. A. &
Haynes, C. E. (1996). Mental health of the workforce in NHS trusts. Phase I Report. Institute of Work
Psychology, University of ShefŽeld.
Caplan, R. D. (1971). Organizational stress and individual strain: A social-psychological study of risk factors in
coronary heart disease among administrators, engineers and scientists. Ann Arbor, MI: Institute for Social
Research, University of Michigan and University MicroŽlms No. 72–14822.
Caplan, R. D. (1979). Social support, person-environment Žt and coping. In L. A. Ferman & J. P. Gordus
(Eds), Mental health and the economy. Kalamazoo, MI: Institute for Employment Research.
Caplan, R. D., Cobb, S., French, J. R. P., Van Harrison, R. & Pinneau, S. R. (1975). Job demands and worker
health. Ann Arbor, MI: UMI.
272 Clare E. Haynes et al.
Charlton, J., Kelly, S., Dunnell, K., Evans, B. & Jenkins, R. (1993). Suicide deaths in England and Wales:
Trends in factors associated with suicide deaths. Population Trends, 69, 34–42.
Cooper, C. L., Cooper, R. D. & Eaker, L. H. (1988). Living with stress. Harmondsworth: Penguin.
EEC (1989). Council directive of 12 June 1989 on the introduction of measures to encourage improvements
in the safety and health of workers at work. OfŽcial Journal of the European Communities, 32, 1–8.
Fang, Y. Q. & Baba, V. V. (1993). Stress and turnover intention: A comparative study among nurses.
International Journal of Comparative Sociology, 34, 24–38.
Goldberg, D. P. (1972). The detection of minor psychiatric illness by questionnaire. Oxford: Oxford University
Press.
Goldberg, D. P. & Williams, P. (1991). A user’s guide to the General Health Questionnaire. Windsor, UK: NFER-
Nelson.
Guralnick, L. (1963). Mortality by occupation and cause of death (No. 3). Mortality by industry and cause of
death (No. 4). Mortality by occupation level and cause of death (No. 5) among men 20–64 years of age vs.
1950. Vital Statistics: Special Reports, 53. Washington DC: U.S. Government Printing OfŽce.
Hingley, P. (1984). The human face of nursing. Nursing Mirror, 59, 19.
Howard, G. (1995). Stress and the law. Occupational Health, ‘Watching Brief’, February.
Jenkins, R. & Warman, D. (1993). Promoting mental health in the workplace. London: HMSO.
Karasek, R. & Theorell, T. (1990). Healthy work: Stress, productivity and the reconstruction of working life. New
York: Basic Books.
Katz, D. & Kahn, R. L. (1978). The social psychology of organizations, 2nd ed. New York: Wiley.
Kaufmann, G. M. & Beehr, T. A. (1989). Occupational stressors,individual strains and social support among
police ofŽcers. Human Relations, 42, 185–197.
Landeweerd, J. A. & Boumans, N. P. G. (1994). The effect of work dimensions and need for autonomy on
nurses’ work satisfaction and health. Journal of Occupational and Organizational Psychology, 67, 207–217.
Marsh, H. W., Balla, J. R. & MacDonald, R. P. (1988). Goodness-of-Žt indexes in conŽrmatory factor
analysis: The effect of sample size. Psychological Bulletin, 103, 391–410.
Maslach, C. & Jackson, S. E. (1986). Maslach Burnout Inventory manual, 2nd ed. Palo Alto, CA: Consulting
Psychologists Press.
Parkes, K. R. (1982). Occupational stress among student nurses: A natural experiment. Journal of Applied
Psychology, 67, 784–796.
Petterson, I. & Arnetz, B. B. (1997). Measuring psychosocial work quality and health: Development of health
care measures of measurement. Journal of Occupational Health Psychology, 2(3), 229–241.
Revicki, D. A., Whitley, T. W. & Gallery, M. E. (1993). Organizational characteristics,perceived work stress
and depression in emergency-medicing residents. Behavioral Medicine, 19(2), 74–81.
Richardson, A. M. & Burke, R. J. (1991). Occupational stress and job satisfaction among Canadian
physicians. Work and Stress, 5(4), 301–313.
Russell, D. W., Altmaier, E. & Van Velzen, D. (1987). Job-related stress, social support, and burnout among
classroom teachers. Journal of Applied Psychology, 72, 269–274.
Sauter, S. L., Murphy, L. R. & Hurrell, J. J., Jr. (1990). Prevention of work-related psychological disorders:
A national strategy proposed by the National Institute for Occupational Safety and Health (NIOSH).
American Psychologist, 45, 1146–1158.
Schabracq, M. J., Winnubst, J. A. M. & Cooper, C. L. (Eds). (1996). Handbook of work and health psychology.
Chichester, UK: Wiley.
Schmitt, N. & Stults, D. M. (1985). Factors deŽned by negatively keyed items: The result of careless
respondents? Applied Psychological Measurement, 9(4), 367–373.
Smith, P. C., Kendall, L. M. & Hulin, C. L. (1969). The measurement of satisfaction in work and retirement.
Chicago: Rand-McNally.
Tett, R. P., Meyer, J. P. & Roese, N. J. (1994). Applications of meta-analysis: 1987–1992. In C. L. Cooper &
I. T. Robertson (Eds), International review of industrial and organizational psychology, vol. 9. Chichester, UK:
Wiley.
Wall, T. D., Bolden, R. I., Borrill, C. S., Carter, A. J., Golya, D. A., Hardy, G. E., Haynes, C. E., Rick, R. E.,
Shapiro, D. A. & West, M. A. (1997). Minor psychiatric disorder in NHS Trust staff: Occupational and
gender differences. British Journal of Psychiatry, 171, 519–523.
Wallace, M., Levens, M. & Singer, G. (1988). Blue collar stress. In C. L. Cooper & R. Payne (Eds), Causes,
coping and consequences of stress at work. Chichester, UK: Wiley.
 Measures of work characteristics 273
Warr, P. B. (1987). Work, unemployment and mental health. Oxford: Oxford University Press.
Warr, P. B. (1990). The measurement of well-being and other aspects of mental health. Journal of Occupational
Psychology, 63, 193–210.
Warr, P. B., Cook, J. D. & Wall, T. D. (1979). Scales for the measurement of some work attitudes and aspects
of psychological well-being. Journal of Occupational Psychology, 52, 129–148.

Received 10 June 1997; revised version received 27 August 1998

Appendix 1: Items for each scale

Autonomy and control

The following questions concern the amount of choice you have in your job.
To what extent do you:
1. Determine the methods and procedures you use in your work?
2. Choose what work you will carry out?
3. Decide when to take a break?
4. Vary how you do your work?
5. Plan your own work?
6. Carry out your work in the way you think best?

Response scale: 1 = not at all, 2 = just a little, 3 = moderate amount, 4 = quite a lot, 5 = a great deal.

Feedback
The following statements concern the information you get about your work performance.

1. I usually know whether or not my work is satisfactory in this job.

2. I often have trouble Žguring out whether I’m doing well or poorly on this job.
3. Most people on this job have a pretty good idea of how well they are performing their work.
4. Most people on this job have trouble Žguring out whether they are doing a good or bad job.

Response scale: 1 = strongly agree, 2 = agree, 3 = neither agree nor disagree, 4 = disagree, 5 = strongly
disagree.

Inuence over decisions

The questions below concern the inuence you have over decisions at work.
To what extent:
1. Can you inuence what goes on in your work area as a whole?
2. Does your immediate superior ask for your opinion before making decisions affecting your work?
3. Do you have the opportunity to contribute to meetings on new work developments?
4. Are you allowed to participate in decisions which affect you?

Response scale: 1 = not at all, 2 = just a little, 3 = moderate amount, 4 = quite a lot, 5 = a great deal.

Leader support
The following questions deal with your working relationship with your immediate supervisor, that is, the
person who most immediately supervises you and to whom you are responsible for your work.
274 Clare E. Haynes et al.
How much does your immediate superior:
1. Encourage you to give your best effort?
2. Set an example by working hard him/herself?
3. Offer new ideas for solving job-related problems?
4. Encourage those who work for him/her to work as a team?

Response scale: 1 = to a very little extent, 2 = to a little extent, 3 = to some extent, 4 = to a great extent,
5 = to a very great extent.

To what extent can you count on your immediate superior:

5. To listen to you when you need to talk about problems at work?
6. To help you with a difŽcult task at work?

Response scale: 1 = not at all, 2 = to a small extent, 3 = neither great nor small extent, 4 = to a great
extent, 5 = completely.

Professional compromise
In your opinion, how often do staff in the Trust meet the following problems in carrying out their work?
1. Having to make trade-offs between quality of patient care and cost savings.
2. Lack of clarity/agreement about the different responsibilities of doctors and nurses.
3. Being unable to achieve quality in their work because there are staff shortages.
4. Having to do an acceptable minimum of work rather than doing the best quality work possible.

Responsible scale: 1 = not at all, 2 = just a little, 3 = moderate amount, 4 = quite a lot, 5 = a great deal.

Role clarity
How true are the following of your job?
1. I have clear planned goals and objectives for my job.
2. I know that I have divided my time properly.
3. I know what my responsibilities are.
4. Explanation is clear of what has to be done.
5. I know exactly what is expected of me.

Response scale: 1 = not at all, 2 = just a little, 3 = moderate amount, 4 = quite a lot, 5 = a great deal.

Role conict
How often do you Žnd these issues arising in carrying out your job?
1. I receive conicting instructions from two or more people.
2. Professionals make conicting demands of me.
3. Managers make conicting demands of me.
4. I do things which are accepted by one person, but not by another.

Response scale: 1 = not at all, 2 = just a little, 3 = moderate amount, 4 = quite a lot, 5 = a great
deal.

Peer support
The following questions ask about the extent to which other people provide you with help or support.
 Measures of work characteristics 275
To what extent can you:
1. Count on your colleagues to listen to you when you need to talk about problems at work?
2. Count on your colleagues to back you up at work?
3. Count on your colleagues to help you with a difŽcult task at work?
4. Really count on your colleagues to help you in a crisis situation at work, even though they would have to
go out of their way to do so?

Response scale: 1 = not at all, 2 = to a small extent, 3 = neither great nor small extent, 4 = to a great
extent, 5 = completely.

Work demands
How often do you Žnd yourself meeting the following problems in carrying out your job?
1. I do not have enough time to carry out my work.
2. I can not meet all the conicting demands made on my time at work.
3. I never Žnish work feeling I have completed everything I should.
4. I am asked to do work without adequate resources to complete it.
5. I can not follow best practice in the time available.
6. I am required to do basic tasks which prevent me completing more important ones.

Response scale: 1 = not at all, 2 = just a little, 3 = moderate amount, 4 = quite a lot, 5 = a great deal.