Psychological Medicine, 1979, 9, 139-145

Printed in Great Britain

A scaled version of the General Health Questionnaire

D. P. GOLDBERG1 AND V. F. HILLIER2
From the Department of Psychiatry, University of Manchester

SYNPOSIS This study reports the factor structure of the symptoms comprising the General Health
Questionnaire when it is completed in a primary care setting. A shorter, 28-item GHQ is proposed
consisting of 4 subscales: somatic symptoms, anxiety and insomnia, social dysfunction and severe
depression. Preliminary data concerning the validity of these scales are presented, and the perform-
ance of the whole 28-item questionnaire as a screening test is evaluated. The factor structure of the
symptomatology is found to be very similar for 3 independent sets of data.

INTRODUCTION from a set of 60 symptoms in order to identify a

probable case indicates that there are various
The General Health Questionnaire (GHQ) is a
dimensions of symptomatology even at this
self-administered screening questionnaire de-
relatively undifferentiated diagnostic level.
signed for use in consulting settings aimed at
The original principal components analysis of
detecting those with a diagnosable psychiatric
the GHQ used subjects who were either identified
disorder (Goldberg, 1972). It concerns itself with
psychiatric patients or who were psychologically
2 major classes of phenomena: inability to carry
healthy: patients with intermediate degrees of
out one's normal 'healthy' functions, and the
disturbance, such as those encountered in a
appearance of new phenomena of a distressing
primary care setting, were not included. In that
nature. It has already been shown that, at certain
analysis a general factor was found to account
critical symptom levels, a psychiatrist using a
for 45 % of the variance, while factors concerned
standardized psychiatric interview is likely to
with depression and anxiety accounted for a mere
make psychiatric diagnoses: below this level, a
3-3% and 2-1 % of the variance respectively.
respondent is likely to be perceived as 'sub-
clinically disturbed' (Goldberg & Blackwell, The present study uses GHQs completed by
1970; Wing et al. 1977). consecutive attenders in primary care settings in
The GHQ may, therefore, be thought of as an attempt to investigate the feasibility of pro-
comprising a set of questions which form a ducing a scaled version of the GHQ, and presents
'lowest common multiple' of symptoms which some data concerned with the validity of these
will be encountered in the various differentiated scales.
syndromes of mental disorder, consisting as it
does of symptoms which best differentiate psych-
PREPARATION OF THE 28-ITEM GHQ
iatric patients as a general class from those who
consider themselves to be well. No theoretical The first step was to carry out an unrotated
assumptions are made about the nature of the principal axes analysis on 523 questionnaires
diagnostic hierarchy withi n the class of psychiatric completed by consecutive attenders to a group
illness, as the questionnaire is focused on the practice in South Manchester for a survey previ-
hinterland between psychological sickness and ously described by Goldberg et al. (1976a). For
health. However, the fact that the full-length this analysis, all 60 questions were scored using
version of the GHQ requires any 12 symptoms Likert scoring (0—1—2—3). Eleven significant fac-
1
tors accounted for 63-4% of the total variance,
Address for correspondence: Professor David Goldberg,
Department of Psychiatry, University Hospital of South but only the first 6 could be readily conceptual-
Manchester, West Didsbury, Manchester M20 8LR. ized. These factors account for 53-5 % of the total
* Present address: Department of Community Medicine, variance, and are shown as Table 1.
University of Manchester, Oxford Road, Manchester M13
9PL. The first factor is a general unipolar factor,
139
140 D. P. Goldberg and V. F. Hillier

Table 1. Unrotated principal axes analysis on 60-item GHQ

Variance
accounted for
Factor number Eigenvalue Name of factor

1 20-8 34-6 Severity of illness

2 35 5-8 Psychic v. somatic depression
3 2-5 41 Social dysfunction
4 20 3-4 Suicidal depression
5 1-9 3-2 Sleep disturbance
6 1-4 2-4 Head pains

Table 2. Varimax rotation of first 6 factors possible to identify each factor with larger
accounting for 53-5% of variance clusters of items. The Appendix shows the 7 items
with the highest loads on each scale which were
Factor Questions chosen for the 'Scaled G H Q ' ; on this analysis
number Name of scale comprising scale no item had a load of less than 0-51. No item
1 General illness 1. 2, 3, 4, 16 loaded significantly on any other scale. It was
2 Somatic symptoms 5, 6. 8, 9, 10 found that one question - ' Have you recently
3 Sleep disturbance 11. 12, 17, 18,20
4 Social dysfunction 21, 28, 30, 36, 42 felt unhappy and depressed ?' - loaded equally on
5 Anxiety and dysphoria 45, 47, 49, 50, 55 scale B and scale D, and was therefore omitted.
6 Suicidal depression 52, 56, 57, 59, 60 This omission had the effect of making the items
remaining on scale B more homogeneous (see
Table 3).
Table 3. Varimax rotation of first 4 factors ac-
At this stage the analysis was repeated, using
counting for 48% of variance of the 60-item
only the 28 chosen items, and now the first 4
questionnaire
factors were found to account for 59 % of the
total variance. All questions loaded heavily on
Factor Questions comprising
number Name of scale scale their assigned scales except no. 18 and no. 58.
Question 18,' Have you recently had difficulty in
1 A. Somatic symptoms 1, 7 3 4 5 6 9 staying asleep once you are off?' had previously
2 H Anxiety and 14, 18 19 44 45 47 S5
insomnia loaded only on Scale B, 'anxiety and insomnia',
3 C. Social dysfunction 21, 22, 28, 30, 35, 36, 42 but now also loaded on Scale A, 'somatic symp-
4 D. Severe depression 51, 52, 56, 57, 58, 59, 60 toms'. Question 58, 'Have you recently found at
times you couldn't do anything because your
nerves were too bad ?' had previously loaded only
while subsequent factors are generally bipolar, on Scale D, 'severe depression', but it now
with fewer items loading on each successive loaded equally on Scale D and Scale B, 'anxiety
factor. When these factors were subjected to a and insomnia'.
varimax rotation, it became even easier to con- Since a general factor accounts for 35 % of the
ceptualize the factors since the more equal dis- variance in the unrotated analysis, it is inevitable
tribution of variance caused the factors to have that the various scales will not be pure measures
approximately equal numbers of items loading of the 4 factors. The extent to which they overlap
on each one. It was possible to construct 6 five- is shown in Table 4, which shows the mean factor
question scales by taking the 5 questions with loads for each scale on each factor (see Table 4).
the highest loads on each factor (see Table 2).
It was thought to be of interest to examine the
solution in which the scale concerning severe de-
REPLICATION WITH OTHER DATA
pression can be kept separate from the scale
mainly concerned with anxiety on the least number Principal components analyses on the chosen 28
of dimensions: this turns out to be the 4-factor items were then repeated on 2 additional sets of
solution and it accounts for 48 % of the total GHQs that were available to the authors: 552
variance. Since there are fewer factors, it is consecutive attenders in Blackwell's practice
 Scaled version of the GHQ 141

Table 4. Mean factor loads of the 4 scales on the 4 factors (59 % of variance on these factors:
varimax rotation of the 2%-item questionnaire)

Factor A Factor B Factor C Factor D

A scale questions 0-55 019 0-26 010
B scale questions 0-25 0-59 0-26 0-21
C scale questions Oil 019 0-59 013
D scale questions 013 0-28 017 0-67

Table 5. Correlation coefficients between GHQ scale scores and independent clinical measures

A scale B scale C scale D scale

'somatic 'anxiety and 'social 'severe Total score
symptoms' insomnia' dysfunction' depression' 28-itemGHQ

Somatic symptoms 0-32 0-28 0-23 0-21 0-32

Anxiety and worry + 0-47 0-70 0-43 0-51 0-67
'anxious'
Despondency + ' depressed' 0-49 0-71 0-54 0-56 0-73
Psychiatrist's severity 0-55 075 0-56 051 0-76

in Croydon in 1968 (Goldberg & Blackwell,
1970), and 4247 attenders in 92 general practices
in Greater Manchester in 1976 (Marks et al.
1979). For the Croydon data, 6 significant factors
accounted for 62% of the total variance, while
for the Manchester data 5 significant factors
accounted for 62% of the total variance. The
first 4 factors accounted for 53 % and 58 % of
the total variance respectively. The loads of the
varimax rotation of these 4 factors were very
similar to the original data described above: the
same 2 items loaded on 2 factors, while the
remaining 26 items loaded heavily on the pre-
dicted factors.
CONCURRENT VALIDITY OF THE SCALES
In the original validity study of the GHQ 200
patients completed the 60-item questionnaire and
were also independently interviewed by one of
us (D.G.) using the Clinical Interview Schedule
(CIS) (Goldberg et al. 1970). It was possible to
extract clinical ratings from the Clinical Interview
Schedule which might best validate the A, B and
D scales. Four measures were chosen: 'somatic
symptoms' for the A scale; the sum of the
reported symptom 'anxiety and worry' and
'observed anxiety' for the B scale; the sum of
reported ' despondency' and' observed depressed
mood' for the D scale; and the sum of all the
morbid ratings - called the 'severity rating' - for
the total score (see Table 5). It was not possible
to examine the validity of the C scale, since
ratings of social dysfunction were not made in
this study.
It can be seen that the rating for 'somatic
symptoms' correlates more highly with the A
scale than with the others, but it is of interest
that the A scale itself correlates even more highly
with affective disturbance rated at interview.
The psychiatrist's ratings of morbid anxiety
correlate very highly with the B scale, but the B
scale itself also correlates equally well with the
psychiatrist's rating of depression (the latter
finding is hardly surprising, since the psychia-
trist's anxiety and depression ratings are them-
selves correlated +0-71). Finally, the D scale
correlates more highly with the psychiatrist's
depression rating than with his anxiety rating. It
would therefore seem that the D scale does
provide some additional information concerning
depression rated at interview. (The correlation
between total GHQ score and the depression
rating is +0-73; if the D scale is removed this
falls to +0-70.)
It should perhaps be noted that the CIS rating
'somatic symptoms' can only be made if the
research psychiatrist has reason to believe that
the symptoms complained of have been precipi-
tated, exacerbated or maintained by psychologi-
cal factors, whereas the A scale does not have
this restriction. This may account for the rather
lower validity coefficient for the A scale.
142 D. P. Goldberg and V. F. Hillier

Table 6. Relationship between 28-item GHQ score and clinical status for 200 patients

Clinical status

GHQ score Normal Subclinical Mild case Moderate Severe case

High scores
12-28 5 21 14
6-11 12 34 8
5 4 4 0
Low scores
0-4 42 36 5 1 0
Total (50) (57) (64) (23) (6)

Table 7. Cutting score 4/5

Raw frequencies Adjusted frequencies

Non-cases Cases Non-cases Cases

High scores 29 87 20-55 61-65
Low scores 78 6 109-39 8-41

Table 8. Screening characteristics ofli-item GHQ for different cutting scores

Threshold Sensitivity Specificity Overall misclassi-

score (%) (%) fication rate(%)

4/5 880 84-2 14-5

5/6 800 88 8 14-2

THE 28-ITEM GHQ AS A SCREENING TEST
If the scaled GHQ is to be used as a screening
test, the simpler 'GHQ scoring method' (0-0-
1-1) gives results which are if anything better
than the Likert method. Table 6 shows the GHQ
scores on the scaled GHQ for the 200 patients
by clinical group based on the clinical assessment
using the clinical interview schedule. It can be
seen that the threshold score 4/5 gives slightly
better overall results than 5/6.
In order to calculate sensitivity and specificity
of this version of the GHQ, it is necessary to
correct for the stratified sampling strategy that
was used in order to select the 200 patients for
interview. It can be seen that, of the patients
interviewed, 116 (58%) had scores of 5 or above.
However, these patients were drawn from 553
consecutive attenders, of whom 41-1 % had such
scores. Had the 200 patients been drawn at
random, there would therefore have been 82-2
high scorers and 117-8 low scorers. The raw fre-
quencies can therefore be adjusted by assuming
that the high scorers would have been in a ratio
of 29:87 and low scorers in a ratio of 78:6 of
non-cases to cases, and these results are shown
in Table 7.
Since specificity is the number of true negatives
as a proportion of non-cases, and sensitivity the
number of true positives as a proportion of the
number of cases, these statistics can be computed
from the adjusted frequencies, and they are
shown in Table 8. (The figures for the threshold
score of 5/6 were calculated taking into account
that fact that 34-8 % of consecutive attenders had
a score of 6 or above.)
DISCUSSION
The great similarity between the rotated solutions
of multivariate analyses of data drawn from 3
quite different populations, collected over a
period of 8 years, suggests that the proposed
scales are fairly stable. Results described in this
paper contrast with those already published for
a 30-item GHQ in Philadelphia by Goldberg et al.
(1976 b) because the American data started with
 Scaled version of the GHQ 143

Table 9. Intercorrelations between the 4 scales and total GHQ score

28-item total
B scale C scale D scale GHQ score

A scale 0-58 0-52 0-33 0-79

B scale 0-57 0-61 0-90
C scale 0-44 0-75
D scale 0-69

only 30 items, while the present study derived
the 4 scales by selecting items from the larger
pool of 60 items.
The 30-item GHQ used in the United States
consisted of the best discriminators between the
original calibration groups, which in practice
meant that they are items which load highly on
the first, general factor in the data. The method
of selecting items for the 'scaled G H Q ' has
meant that other dimensions have been able to
emerge: it is of interest that 14 items in the scaled
GHQ do not appear in the 30-item GHQ.
The 4 subscales are by no means independent
of one another, although correlations between
the scales are lower than those between the sub-
scales and the total score (see Table 9). This ten-
dency for the subscales to intercorrelate reflects
the presence of a general factor in the unrotated
data. For the 28-item GHQ, this general factor
accounts for 32% of the total variance in the
London data, and 37 % for the 2 sets of Man-
chester data.
The correlation coefficient of +0-90 between
the B scale and the total score supports the view
that anxiety is a core phenomenon which under-
lies the common syndromes of psychiatric dis-
order, but the existence of the 3 other scales
allows investigators with special interests to
measure other dimensions of symptomatology.
The complete questionnaire is shown as the
Appendix.
The scaled version of the GHQ is intended for
studies in which an investigator requires more
information than is provided by a single severity
score. For example, an investigator wishing to
select depressed patients might select from the
population of high scorers only those respondents
with high scores on the D scale. In studies where
the GHQ is administered before and after some
event such as a treatment, or a social change, the
ratio of the 2 sets of scaled scores might reveal
that the external event had its greatest effect in
one particular symptom area.

APPENDIX
The scaled GHQ
GENERAL HEALTH QUESTIONNAIRE
Please read this carefully:
We should like to know if you have had any medical complaints, and how your health has been in general,
over the past few weeks. Please answer ALL the questions on the following pages simply by underlining the
answer which you think most nearly applies to you. Remember that we want to know about present and
recent complaints, not those that you had in the past.
It is important that you try to answer ALL the questions.
Thank you very much for your cooperation.
HAVE YOU RECENTLY:
Al. Been feeling perfectly well and in Better Same Worse Much worse
good health ? than usual as usual than usual than usual
A2. Been feeling in need of a good Not No more Rather more Much more
tonic? at all than usual than usual than usual
A3. Been feeling run down and Not No more Rather more Much more
out of sorts? at all than usual than usual than usual
A4. Felt that you are ill? Not No more Rather more Much more
at all than usual than usual than usual
144 D. P. Goldberg and V. F. Wilier

A5. Been getting any pains in your head? Not No more Rather more Much more
at all than usual than usual than usual
A6. Been getting a feeling of tightness Not No more Rather more Much more
or pressure in your head? at all than usual than usual than usual
A7. Been having hot or cold spells? Not No more Rather more Much more
at all than usual than usual than usual
Bl. Lost much sleep over worry ? Not No more Rather more Much more
at all than usual than usual than usual
B2. Had difficulty in staying asleep Not No more Rather more Much more
once you are off? at all than usual than usual than usual
B3. Felt constantly under strain? Not No more Rather more Much more
at ail than usual than usual than usual
B4. Been getting edgy and bad-tempered? Not No more Rather more Much more
at all than usual than usual than usual
B5. Been getting scared or panicky Not No more Rather more Much more
for no good reason ? at all than usual than usual than usual
B6. Found everything getting on Not No more Rather more Much more
top of you? at all than usual than usual than usual
B7. Been feeling nervous and strung-up Not No more Rather more Much more
all the time? at all than usual than usual than usual
Cl. Been managing to keep yourself More so Same Rather less Much less
busy and occupied? than usual as usual than usual than usual
C2. Been taking longer over the Quicker Same Longer Much longer
things you do? than usual as usual than usual than usual
C3. Felt on the whole you were Better About Less well Much
doing things well? than usual the same than usual less well
C4. Been satisfied with the way More About same Less satisfied Much less
you've carried out your task ? satisfied as usual than usual satisfied
C5. Felt that you are playing a More so Same Less useful Much less
useful part in things? than usual as usual than usual useful
C6. Felt capable of making decisions More so Same Less so Much less
about things? than usual as usual than usual capable
C7. Been able to enjoy your normal More so Same Less so Much less
day-to-day activities? than usual as usual than usual than usual
Dl. Been thinking of yourself as a Not No more Rather more Much more
worthless person ? at all than usual than usual than usual
D2. Felt that life is entirely hopeless? Not No more Rather more Much more
at all than usual than usual than usual
D3. Felt that life isn't worth living? Not No more Rather more Much more
at all than usual than usual than usual
D4. Thought of the possibility that you Definitely I don't Has crossed Definitely
might make away with yourself? not think so my mind have
D5. Found at times you couldn't do Not No more Rather more Much more
anything because your nerves were at all than usual than usual than usual
too bad ?
D6. Found yourself wishing you were Not No more Rather more Much more
dead and away from it all ? at all than usual than usual than usual
D7. Found that the idea of taking your own Definitely I don't Has crossed Definitely
life kept coming into your mind? not think so my mind has

c
A B
D D 1 TOTAL

Copyright © General Practice Research Unit 1978 (De Crespigny Park, London SE5 8AF)
 Scaled version of the GHQ 145

Note
The version of the GHQ-28 used in the United States has 4 questions reworded:
Al. Been feeling in need of some medicine to pick you up?
B2. Had difficulty staying asleep?
B7. Been feeling nervous and uptight all the time?
D4. Thought of the possibility that you might do away with yourself.

Goldberg, D. P., Kay, C. & Thompson, L. (1976a). Psych-

REFERENCES
Goldberg, D. P. (1972). The Detection of Psychiatric Illness
by Questionnaire. Oxford University Press: London.
Goldberg, D. P. & Blackwell, B. (1970). Psychiatric illness in
a suburban general practice. A detailed study using a new
method of case identification. British Medical Journal ii,
439-443.
Goldberg, D. P., Cooper, B-, Eastwood, M. R., Kedward,
H. B. & Shepherd, M. (1970). A standardised psychiatric
interview suitable for use in community surveys. British
Journal of Preventive and Social Medicine 24, 18-23.
iatric morbidity in general practice and the community.
Psychological Medicine 6, 565-569.
Goldberg, D. P., Rickels, K., Downing, R. & Hesbacher, P.
(19766). A comparison of two psychiatric screening tests.
British Journal of Psychiatry 129, 61-67.
Marks, J., Goldberg, D. P. & Hillier, V. F. (1979). Deter-
minants of the ability of general practitioners to detect
psychiatric illness. Psychological Medicine 9 (in the press).
Wing, J. K., Nixon, J., Mann, A. & Left", J. P. (1977). Relia-
bility of the PSE (ninth edition) used in a population study.
Psychological Medicine!, 505-516.

PSM 9