British

Journalof Psychiatry
(1989),
155,178—185

The Factor Structure of the General Health Questionnaire (GHQ-30)

A Reliability Study on 6317 Community Residents
FELICIA A. HUPPERT, DAVID EUROF WALTERS, NICHOLAS E. DAY and B. JANE ELLIOTT

An individual's responses to Goldberg's 30-item General Health Questionnaire are usually

representedas a single score which provides a measureof the number of psychiatric symptoms
reported. No account is taken of the nature of the symptoms. Factor analyses of the GHQ-30
were undertaken in ten randomly selected samplesof 600 adults each, and also on 12 age—sex
groupings covering the age range 18—98.The results indicate an impressive degree of
consistency of the factor structure, and the identification of five distinct factors corresponding
to anxiety, feelings of incompetence, depression, difficulty in coping, and social dysfunction.
The instrument of choice for screening psychiatric (Cox et al, 1987). The survey collected a great deal
disorder in patient and community samples is the of information on physical health and health-related
General Health Questionnaire (GHQ; Goldberg, behaviour. In selecting an instrument to measure
1972). The widespread use of this instrument means mentalhealth we wantedtofocuson psychological
that methods for increasing the amount of informa and psychosocial symptoms rather than somatic
tion yielded by the questionnaire would be of great symptoms. For this reason we selected the 30-item
value in advancing research. The traditional way of version of the GHQ (GHQ-30), which was derived
scoring the GHQ is to rate each response in terms from the original 60-item version by excluding
of whether the symptom is present or absent, and symptoms that were commonly present in subjects
to sum the number of symptoms. Although this with entirely physical illness. The other version which
simple, additive approach is useful as a measure of we considered was the GHQ-28, which has been
the number of psychiatric symptoms declared by an derived from the GHQ—60on the basis of factor
individual, it neglects qualitative aspects of the data, analysis. A quarter of the items in the GHQ—28 are
i.e. the nature of the symptoms. The list of questions concerned with somatic symptoms and a further
in the GHQ covers many different types of behaviour, quarter with severe depression, which we felt to be
including sleep patterns, worries, relationships, unsuitable for a normal population sample.
depressed mood, etc. Therefore, it may be possible The number of respondents who completed the
to derive subscales, each concerned with a particular questionnaire was 6317, which is by far the largest
type of symptom, and subsequently to derive a score sample ever studied with this instrument, and is a
for each individual on each of these subscales. good representation of the British population (see
An objective approach to deriving subscales is below). Basic descriptions of the GHQ scores and
factor analysis. This is a statistical method for their relationship to socio-demographic and other
determining which items cluster together to form variables are available elsewhere (Huppert et al,
‘¿factors'.
Sometimes factors are easy to interpret, 1987). A comparative study of the conventional
because the items in the cluster bear a meaningful GHQ score and a score taking account of chronic
relationship to each other. At other times it is symptoms (CGHQ) has also been reported on this
difficult to understand why a group of items cluster population sample (Huppert et a!, 1988).
together to form a factor. One reason why
interpretative difficulties often arise following factor Method
analysis is limited sample size. When a sample is
small, random fluctuations can have a marked The 30-item GHQ was introduced to respondents in their
effect. Moreover, with smallsamples,itisrarely own home by a nurse at the conclusion of a series of
possible to examine the reliability of the factor physiological measurements and simple tests of cognitive
function (see Cox et a!, 1987; Huppert et a!, 1987). The
structure.
nurse's visit was the second phase of the Health and
The exceptionally large database which we have Lifestyle Surveyandtookplace 1-2weeksafter anextensive
established provides a unique opportunity to investi initialinterview.The GHQ was acceptedby 7304participants,
gate the factor structure of the GHQ. We administered who were instructed to complete it on their own and return
the GHQ to an adult sample of the British population it in a stamped, addressed envelope. The questionnaire was
participating in the Health and Lifestyle Survey returned by mail by 6572 participants (88.6%). Only

178
 RELIABILITY OF GHQ FACTOR STRUCTURES 179

completequestionnaireswereanalysed,the fmalfigurebeing analyses were carried out in order to explore any variation
6317(86.5%). To determine if the sample was representative in factor structure due to sex and/or age.
of the British population, the age and sex distribution of Orthogonalvarimaxrotation facilitatesthe identification
our sample was compared with figures from the 1981census of factors that are uncorrelated. Although this can simplify
(Blaxter, 1987).Overall, our sample was found to be a good interpretation of what the factors mean, it is likely that in
representative sample, with the exception of men aged the real world factors are correlated. For this reason the
18—29,
who wereunder-represented(21.3% v. 24.8%), as factor analysis on the whole sample was repeated using an
were women aged 70 plus (9.2°lo
v. 16.2%). ‘¿oblimin'(oblique) rotation which does not retain the
The first factor analysis was carried out on a randomly orthogonality of the factors. All analyses were carried out
selectedsampleof 6000individualswhichwas taken from on the raw data (Likert scores).
the complete database of 6317 individuals. The size of this
sample was adjusted for the convenience of the following Results
subsampling. These 6000 individuals were then randomly
allocatedto one of 10subsamplesof 600adults, and factor It should be appreciated at the outset that there will always
analyses were carried out on each of the 10 subsamples. be a measure of subjectivity in the number of factors
Factor analyses were also carried out on the 12 groups of selected, and in the items considered to be making an
individuals obtained when the main sample of 6000 was important contribution to a factor. This is an inevitable
divided by sex and by age into the following groupings: result of the data set containing no a priori classification
18—24, 25—34, 35—44, 45—54, 55—64, 65+. These last of either cases or items. The criterion for identifying

TABLE I
Varimax rotated factor structure of the GHQ—30(n = 6000)1@2
ItemABFactor
CDE1Could
concentrate39450109152Lostsleep66100909083Restless
not

nights11—090138174Not
occupied—04592201285Not
busy or
house15390203546Not
out of the
well09551808007Not
managing
well24750706058Not
doing things
task20740213059No
satisfied with
affection131209031510Could
warmth and
others140712066111Not
not get on with
others090806097912Not
chatting with
part114826074513Could
playing a useful
decisions041606500714Felt
not make
strain820600090615Could
under
difficulties682022130316Found
not overcome
struggle210204690417Not
life a
activities0512—10660418Taking
enjoying
hard720511140719Scared
things
panicky27032255—0320Could
or
problems253436090721Felt
not face
top801317120722Unhappy
everything on
depressed751528091023Lost
and
confidence622341080824Felt
worthless442061040825Felt
hopeless21005942—0326Not
life
future152353111727Not
hopeful about
happy01042761—0128Nervous
feeling
up751230110529Felt
and strung
living361270041030Nerves
life not worth
bad4917520506@1o
too
variance accounted for28.97.66.24.13.8
1. Loadingsgreaterthan0.5 arein italics.
2. All decimalpoints on loadingsare omitted.
180 HUPPERT ET AL
TABLE II
Items making a sign@ficantcontribution to GHQ factors

Factor Total variance accounted for inOrder

Factorsubsamples ofitems2'3 GHQitem
(%)Whole
sub for whole
sample10 samplesloadings sample
A 26.6—31.2 1 14 Felt constantly under strain? 0.82
2 2 21 Foundeverythinggettingon top of you? 0.80
3 3 28 Beenfeelingnervousandstrung-upall the
time? 0.75
4 4 22 Been feeling unhappy and depressed? 0.75
5 5 18 Been taking things hard? 0.72
6 6 15 Felt you couldn't overcome your difficulties? 0.68
7 7 2 Lost much sleep over worry? 0.67
8 8 23 Been losing confidence in yourself? 0.63
(9) (9) (30) Found at timesyou couldn't do anything
becauseyour nerveswere too bad? (0.49)
(10) (10) (24) Been thinking of yourself as a worthless
person? (0.44)
B 7.2—8.3 1 1 7 Felt on the whole you were doing things
well? 0.75
2 2 8 Been satisfied with the way you've carried
out your task? 0.74
3 4 4 Been managing to keep yourself busy and
occupied? 0.59
4 3 6 Been managing as well as most people
would in your shoes? 0.55
(5) (5) (12) Felt that you are playing a useful part in
things? (0.48)
(6) (7) (1) Been able to concentrate on whatever
you're doing? (0.45)
(7) (6) (5) Been getting out of the house as much as
usual? (0.39)

C 5.2—6.8 1 29 Felt that life isn't worth living? 0.70

2 2 24 Been thinking of yourself as a worthless
person? 0.61
3 3 25 Felt that life is entirely hopeless? 0.59
4 4 26 Been feeling hopeful about your own future? 0.53
5 5 30 Found at times you couldn't do anything
becauseyour nerves were too bad? 0.52
(6) (6) (23) Been losing confidence in yourself? (0.41)
D4 3.9—5.0 1 16 Been finding life a struggle all the time? 0.69
2 2 17 Been able to enjoy your normal day
to-day activities? 0.66
3 3 27 Been feeling reasonably happy, all things
considered? 0.61
4 4 19 Been getting scared or panicky for no good
reason? 0.55
5 5 13 Felt capable of making decisions about
things? 0.50
(6) (7) (25) Been feeling that life is entirely hopeless? (0.43)
(7) (6) (3) Been having restless, disturbed nights? (0.38)

E 3.5-4.0 1 Il Spent much time chatting with people? 0.80

2 2 10 Been finding it easy to get on with other
people? 0.61
3 3 5 Been getting out of the house as much as
usual? 0.54
(4) (4) (12) Felt that you are playing a useful part in
things? (0.45)
1. The order in which the factors (A to E) are presentedis the order obtained for the whole sampleand for most subsamples.
2. The orderof itemswithin a factor isin descendingorderof importancefor the wholesample.The most frequentitem order for the subsamples
is also indicated. Items were includeduntil a natural break occurredin the factor loadingsaround 0.50.
3. Numbersin parenthesisindicate itemswhich appearedin only a few (2—3) subsamples.The position and factor loadingsof theseitems for
the whole sampleare also given in parenthesis.
4. For one subsample,D was split into two: Dl correspondsto questions27, 17, 13; D2 correspondsto questions16, 3, 19.
 RELIABILITY OF GHQ FACTOR STRUCTURES 181

plausible factors was the standard SPSS-X property that strongest factor in every subsample and included the same
the corresponding eigenvalues of the correlation matrix eight items as in the total sample, although in three of the
should be greater than unity. The number of items making subsamples one or two additional items made a significant
an important contribution to the factor was determined by contribution. Factors B, C and D were also identifiable in
a visual inspection of the plots of the coefficient values all ten subsamples although in one sample, D was split into
corresponding to the item. Items were progressively included two. Factor E was present in nine subsamples. For Factors
until there appeared to be a natural break in the coefficient B, C, D and E there were slight variations in which items
values, which often occurred at approximately 0.50. All contributed significantly to each factor. The order in which
items with values of 0.50 or greater were included. the factors emerged (i.e. their strength) differed to some
extent from sample to sample. This is depicted in Table
IV. For all subsamples, there were only one or two items
Large sample
which made a significant contribution to the sixth factor.
The factor analysisusingvarimaxrotation on the sampleof The question numbers corresponding to these items are
6000 cases identified six factors with eigenvalues greater than shown in parentheses.
1. The six factors combined accounted for 50.6% of the It can be seen that the extent of overall agreement
variance. We refer to these factors alphabetically in order between the 10subsamples and between the subsamples and
of decreasing importance. Becausethe sixth factor contained the large sample (Table II) inspires confidence that the
only a single item (Question 9) which reached our inclusion factors are robust and that the itemswhichconstituteeach
criterion, details of the factor structure of only the five factor are consistent to an acceptable degree.
major factors are presented in Table I. The items contri
buting significantly to each factor, and the percentage of TABLE IV
the variance they account for are presented in Table II. Order of factors for each subsample
The oblique rotation resulted in the same five main
factors being identified in the same order of importance factors'1ABCEDlD22ACBDE(9)3ABCDE(3)4ABDCE(9)5ACBED(4,5)6ABDCE(9,10
SubsampleOrderof
as in the varimaxrotation, exceptthat factorD (asidentified
by the varimaxrotation)accountedfor moreof the variance
than factor C. Table III shows the close similarity in the
items which contribute significantly to oblique and varimax
rotated factors.
TABLE III
Comparisonof items making a signjficant contribution to
varimax rotated factors and oblique rotated factors
FactorGHQitemsmaking a significant
contributionVarimaxA
1. Numbers in parenthesesare question numbers.
18
23VarimaxB
Oblique 114 1421 2128 1822 28 2215 22 1523 Analysis by age—sexgroupings
Principal-components factor analysis followed by a varimax
Oblique27
6VarimaxC 78 84 46 rotation was repeated for each of the 12 age—sexgroups.
30 All groups had at least six significant factors; additional
Oblique
26VarimaxD
429 2924 2525 2426 factors usually contained only one or two significant items.
The data are presented in Table V. Factor A was the
19 strongestfactor in each of the 12groups and containedthe
Oblique 316
19VarimaxE 1717 1627 2713 13 same eight items as in the total sample. Although the order
in which the other main factors emerged varied to some
Oblique 511 1110 105 extent from group to group, all of the factors identified
in the large sample and the ten subsamples also appeared
in each of the 12 age—sexgroups. This suggests that the
This demonstrates that the factors identified by the factors describedin Tablesland II are not only robust for
varimax rotation in this data set are a good summary of the population as a whole but also for individual sections
the structure of the GHQ, and that imposing orthogonality of the population designated by age and sex.
does not distort the results.
Interpretation of the factors
Analysis of random subsamples
Deciding what a particular factor means and applying a
Varimax rotation in each of the 10 random subsamples of suitable label is a rather subjective affair. We have
600 cases, revealed six significant factors (eigenvalues attempted to do this in a conunonsense fashion, rather than
greater than 1) in each subsample. Factor A was the being influenced by theory or diagnostic conventions, since
182 HUPPERT ET AL
TABLE V
Order of factors for each age—sex
grouping
offactorsMen18—24343ACD2BE(4,
SexAgenOrder

(1)25—34495ACBDE(9)(6)(5)35-44520ACBDE(3, 5)Dl(20)

5)45—54403ABCDE(3)55—64451AD2(4,5)CBEDl65+437ABDCE(3)(13,1,9)Women18—24406ACBDE(3)(4,

5)(9)25—34688ACEDB(3)35-44710ACBEDD245—54552ABDCE(3)55-64490ABCE(26)D(3)65+505ABCDE(9,10)(3)

our respondents would have used a lay or commonsense be seen that scores on Factor A decrease with age, while
interpretation of the questions. The five factors have scores on the other four factors increase with age. Factor A
accordingly been labelled as follows: also shows the most pronounced sex difference, particularly
for the youngeragegroups,whileFactorF showsa reversal
Factor A anxiety, worry and tension of the usual sex-difference,with women obtaining lower
Factor B feelings of incompetence, low self-esteem scores than men except in the oldest age group.
Factor C depression, hopelessness
Factor D difficulty in coping, dispirited Discussion
Factor E social dysfunction
There have been a number of factor analyses carried
The first three items in Factor A (seeTable II) were clearly out on the GeneralHealthQuestionnaire,
mostly
related to feelings of anxiety, worry and tension, and the
last four items could also be interpreted as reflecting these using a varimax rotation. Burvill & Knuiman (1983)
feelings. The only item which presented difficulty was employed the full 60-item GHQ and examined a
the fourth item (Q. 22: “¿Been feeling unhappy and community sample of 2044 in Perth, Australia. They
depressed?―).However, we did not feeljustified in labelling identified five factors which were, in order, physical
this factor anxiety/unhappiness on the basis of a singleitem. illness, sleep disturbance, social dysfunction, anxiety!
It is noteworthy that this item does not have a significant dysphoria, and suicidal ideation. Goldberg & Hillier
loading on the depression factor in our study. (1978) chose a four-factor solution for the GHQ—60
The only other factor which presented interpretive on a sample of 523 general-practice patients in
difficulties was Factor D, where the items were rather Manchester. The factors were, in order, somatic
inhomogeneous. It was particularly difficult to accommo
symptoms, anxiety/insomnia, social dysfunction,
date Q. 19: “¿Been
getting scared or panicky for no good
reason?― On the whole, however, it was concluded that and severe depression.
in a general population sample, this item probably bears The shorter 30-item GHQ was derived from the
little relationship to diagnosable phobias or panic attacks GHQ-60 by excluding symptoms that were commonly
and could be subsumed under difficulty in coping. present in subjects with entirely physical illness.
Thus the GHQ-30 could be regarded as a measure
Factor scores of more purely psychological or psychosocial
Although we have found a great deal of consistency in the symptoms. There have been several factor analyses
underlyingfactor structure of the GHQ for different age of the GHQ—30 in relatively large community
sex groups, the scores of these groups on individualfactors samples. Goldberg et a! (1976) report data from a
may show different patterns. Figure 1 presents the mean sample of 1310 whites and 1310 blacks in the USA.
valuesof the standardisedfactor scores(overallmean= 0, For both groups, the main factor was labelled
s.d. =1) for each age-sex group, calculated on the basis depression and anxiety, and it accounted for about
of the orthogonalfactor structureof the sampleas a whole. 21°loof the variance. The authors comment that no
It canbe seenthat for four factorsand the majorityof age matter how many factor solutions they examined (up
groups women have higher factor scoresthan men, reflecting
the factthat womentendto obtainhigherscoreson theGHQ to seven factors) it was not possible to separate
as a whole(e.g.Goldberg,1972;Huppertet al, 1988).How anxiety and depression on the GHQ-30. It is
ever, the advantage of producing individual factor scores is interestingtocomparetheir findingswithourown
that we can provide a more detailed characterisation of age results. An examination of the nine items which
and sex differences on the GHQ. For example, it can make up their anxiety/depression factor shows that
 RELIABILITYOF GHQ FACTOR STRUCTURES 183

0.4 Factor A 0.4 Factor B

V. 0I
a
@ 0
U
0.2
(‘I U)
0
@ U
a
0.0

-0.2 -0.2

1-0.4 -0.4

18-24 25-34 35-44 45-54 55-64 65+ 18-24 25-34 35-44 45-54 55-64 65+

Ag Ag.
0.4 Factor C
U,

0.2 0
U
U)
0 0
V
a 0.0 - U
‘¿5

Cl)
-0.2

-0.4

-0.6
18-24 25-34 35-44 45-54 55-64 65+
@1 65+
Ag. Ag.

If

I -0.2'

-0.4

18-24 25-34 35-44 45-54 55-64 65+

Age
Fio. 1 Standardised factor scores as a function of age and sex (9 male, •¿
female).

four items correspond to our anxiety factor, four Another factor analysis of the GHQ—30 was
correspond to our depression factor and one comes undertaken in a community sample of 2000 + in
from our difficulty-in-coping factor. It is not clear Saskatchewan by D'Arcy (1982). The questionnaire
why anxiety and depression were reliably separated was mailed to respondents and there was a 53010
in our analyses (and others mentioned below) but not response rate. This raises the possibility that the
in theirs. This difference is unlikely simply to reflect sample was biased. Indeed, 47010of the sample
random fluctuation due to differences in sample size, reported no psychiatric symptoms, compared with
since our subsamples of 600 are smaller than their only 29°lo in the Health and Lifestyle Survey, where
samples and yet in each of our 10 subsamples, the response rate was 86.5°lo.The four factors which
anxiety and depression emerged as separate factors D'Arcy identified were anxietylmsomnia, depression!
(Table IV). The possibility of cultural differences in anhedonia, anergia and social dysfunction. The first
the factor structure or in the interpretation of the factor is a combination of our anxiety and difficulty
items cannot be excluded. in-coping factors. Despite the large number of items
184 HUPPERT ET AL

with significant loadings on D'Arcy's first factor (13 comparison is between anxiety and feelings of
items), it accounts for only 16°lo
of the variance. This incompetence. Levels of anxiety are very high at age
contrasts with the eight significant items in our 55—64
in both sexes, and show a marked drop in the
anxiety factor, which accounts for 28.901o of the age group 65—74. At the same time, feelings of
variance. Five of the nine items in D'Arcy's incompetence show a marked rise, particularly in
depression/anhedonia factor clearly correspond to men. A likely explanation for the finding is that
our depression factor, the remaining four not leaving paid employment is associated with reduced
forming part of any coherent factor in our analyses. anxiety, but also with feelings of incompetence or
The anergia factor has no counterpart in our sample, low self-esteem.
largely because two of the items exist only in the The GHQ was developed as a screening instrument
North American version of the GHQ—30. The items for minor psychiatric morbidity: when symptoms are
which they replace in the original British version have added to yield a score ranging from 0—30(the GHQ
the highest loading on our social dysfunction factor score) the GHQ has been found to have acceptable
(item 11) and the second highest loading on our sensitivity and specificity when validated against
feelings-of-incompetence factor (item 8). clinical diagnosis (e.g. Tarnopolsky eta!, 1979). We
Some factor-analytic studies of the GHQ—30have have previously shown that a different scoring
examined data from highly selected groups. These method (the CGHQ, which takes account of the
include the study by D'Arcy & Siddique (1984) on chronicity of symptoms) identified different indivi
1038 Canadian adolescents and the study by Chan duals as cases (Huppert et a!, 1988).The identification
& Chan (1983) on 255 first-year undergraduates at of a reliable factor structure for the GHQ—30 raises
the Chinese University of Hong Kong. The factor the possibility that the detection of cases might be
structures which they report are unlikely to be typical improved by examining an individual's profile of
of the general adult population, which is the primary scores on the different factors. Further validation
focus of this paper. However, it is worth noting that studies are required to determine whether the
both studies report separate anxiety and depression effectiveness of the GHQ as a general psychiatric
factors. screening instrument can be improved in this way,
None of the previous studies of the GHQ-30 has and whether the factors relate to specific diagnostic
had a large enough sample size to examine the categories.
reliability of the factor structure in a representative The GHQ can also be used as a descriptive
population sample. Our study has demonstrated the measure of psychiatric symptomsapart from its use
reliability of a qualitative approach to analysing as a screening instrument. The description can be
GHQ data, which yields information not only about sharpened by the development of empirically derived
the number of symptoms, but also about the nature subscales corresponding to different types of psychia
of these symptoms. We have clearly identified five tric symptoms. We have recently derived five
distinct and robust factors in the GHQ-30. The same subscales based on our factor analysis. We have
factors have emerged in each of the following found that the profile of performance on these
analyses: subscales varies between groups selected as having
a high risk of psychiatric disorder (e.g. the elderly
(a) comparing 10 independent random samples of
living alone, middle-aged unemployed men). Findings
600 cases each such as these indicate the potential value of GHQ
(b) comparing men and women in each of six age
subscales for refining the descriptions of psychiatric
groups from 18 to 98
symptoms.
(c) comparing orthogonal and non-orthogonal factor
Our study, along with most other factor-analytic
analyses.
studies of the GHQ, shows a separation between
The factors can be labelled anxiety, feelings of symptoms of anxiety and symptoms of depression.
incompetence, depression, difficulty in coping, and These studies have been based on community
social dysfunction. samples, but similar findings have been reported in
Although age differences are minimal on conven studies of patients with diagnosed affective disorders.
tional GHQ scores (e.g. Huppert eta!, 1987), there Mullaney (1984) reviewed 40 factor-analytic studies
are pronounced age differences on most of the of patients with affective disorders. Although a wide
factors. Anxiety shows a marked decrease with age, variety of interview schedules was used, all revealed
while the other four factors tend to increase with age. a clear separation between anxiety and depression
Middle-aged adults obtain high anxiety and low symptoms. Mullaney concluded that anxiety and
depression scores, while older adults obtain high depression are separate syndromes. A similar view
depression and low anxiety scores. Another interesting is implicit in DSM—III—R(American Psychiatric
 RELIABILITYOF GHQ FACTOR STRUCTURES 185

Association, 1987). There seems little doubt, therefore, structure of the General Health Questionnaire in a Chinese
that individuals are capable of recognising anxiety context. Psychological Medicine, 13, 363—371.
symptoms and differentiating them from symptoms Cox, B. D., BLAXTER,
M., BUCKLE,
A. C. J. et al (1987)The
Health and Lifestyle Survey: Preliminary Report of a Nationwide
of depression or hopelessness. Whether the processes Survey of the Physical and Mental Health, Attitudes and
underlyinganxietyand depressionare different Lifestyle of a Random Sample of 9(W.)3 British Adults. London:
cannot be determined by data of the type we have Health Promotion Research Trust.
D'ARCY, C. (1982) Prevalence and correlates of nonpsychotic
presented. psychiatric symptoms in the general population. Canadian
Journal of Psychiatry, 27, 316—323.
—¿ & SIDDIQUE, C. M. (1984) Psychological distress among

Canadian adolescents.Psychological Medicine, 14, 615-628.

Acknowledgement GOLDBERG,D. P. (1972) The Detection of Psychiatric Illness by
Questionnaire. London: Oxford University Press.
We wish to thank Alison Jonesand Mauvis Gore for their assistance
—¿, RICKELS, K., DOwNING, R. & HESBACHER, P. (1976) A
in collating the data, and the two referees for their helpful
comparison of two psychiatric screening tests. British Journal
comments.This researchwassupportedby a grant from The Health
of Psychiatry, 129, 61—67.
Promotion Research Trust.
—¿ & HILLER, V. (1978) A scaled version of the General Health

Questionnaire. Psychological Medicine, 9, 139—145.

References HIJPPERT, F. A., GoRE, M. & Rom, M. (1987) Psychological
factors. In The Health and Lifestyle Survey: Preliminary Report
AMERICAN P5YcHIAi-iuc ASSOCIATION(1987) Diagnostic and of a Nationwide Survey of the Physical and Mental Health,
Statistical Manual of Mental Disorders (3rd edn, revised). Attitudes and Lifestyle of a Random Sample of 9003 British
Washington DC: APA. Adults (eds B. D. Cox, M. Blaxter, A. C. J. Buckle et at),
BLAXTER, M. (1987) Sample and data collection. In The Health and pp. 51-58. London: Health Promotion Research Trust.
Ljfestyle Survey: Preliminary Report of a Nationwide Survey of —¿, —¿ & Ewoi-r, B. J. (1988) The value of an improved

the Physical and Mental Health, Attitudes and Lifestyle of a scoring system (CGHQ) for the General Health Questionnaire
Random Sample of 9003 British Adults (eds B. D. Cox, M. in a representative British sample. Psychological Medicine, 18,
Blaxter, A. C. J. Buckle et at), pp. 1-3. London: Health 1001-1006.
Promotion Research Trust. MULLANEY,J. A. (1984) The relationship between anxiety and
BuRvILL, P. W. & KNUIMAN,M. W. (1983) Which version of the depression:a reviewof someprincipalcomponentanalytic
General Health Questionnaire should be used in community studies.Journalof AffectiveDisorder,7, 139-148.
studies? Australian and New Zealand Journal of Psychiatry, 17, TARNOPOLSKY, A., HAND, D. J., McL@, E. K. et al (1979)
237—242. Validity and useof a screeningquestionnaire
(GHQ)in the
Ciwi, D. W. & CHAi4,
T. S. C. (1983)Reliability,validityand the community. British Journal of Psychiatry, 134, 508-515.

*Felicia A. Huppert, PhD, Department of Psychiatry, University of Cambridge; D. E. Walters, PhD,

Churchill College, Cambridge; N. Day, PhD, MRC Biostatistics Unit; B. Jane Elliott, BA(Cantab),Child
Care and Development, University of Cambridge

Correspondence: Level 4, Addenbrooke's Hospital, Hills Road, Cambridge CB2 2QQ