development and validation of a questionnaire for anxiety and depression in pakistan
TRANSCRIPT
www.elsevier.com/locate/jad
Journal of Affective Disord
Brief report
Development and validation of a questionnaire for anxiety
and depression in Pakistan
David B. Mumford a,*, Muhammad Ayub b, Raheel Karim b, Nasir Izhar b,
Aftab Asif b, John T. Bavington c
aCentre for Medical Education, University of Bristol, 41 St Michael’s Hill, Bristol BS2 8DZ, UKbDepartment of Psychiatry, King Edward Medical College, Lahore, Pakistan
cTranscultural Psychiatry Unit, Lynfield Mount Hospital, Bradford, UK
Received 9 August 2004; received in revised form 23 May 2005; accepted 26 May 2005
Available online 2 August 2005
Abstract
Background: Currently clinicians and researchers in Pakistan have to use translated western instruments to screen for anxiety
and depressive disorders. This study investigated the local idioms of emotional distress in Pakistan to develop a culturally valid
and easy-to-use instrument to screen for common mental disorders in general clinical settings.
Methods: A systematic survey was conducted of psychiatric case notes of patients attending clinics in Peshawar and Lahore,
diagnosed with anxiety or depressive disorders, to identify the range of common idioms of psychological distress. A pilot
version of the questionnaire was refined and validated among a composite sample of 330 patients in inpatient, outpatient and
rural community settings. ICD-10 Diagnostic Criteria for Research were used to define cases and patients’ relatives acted as
normal controls.
Results: The pilot version of the questionnaire was reduced to 42 items based on odds ratios between cases and controls.
Anxiety symptoms were generally reported by depressed patients, but not vice versa. Finally 30 items were selected, in two sub-
scales. This final version achieved sensitivity, specificity, and positive and negative predicted values of over 90% when
comparing cases and controls.
Limitations: This questionnaire was based on what patients tell doctors and may not capture the entire repertoire of local idioms
of distress. The validation study was conducted only in an Urdu/Punjabi speaking population, in Lahore and surrounding areas.
Conclusions: The Pakistan Anxiety and Depression Questionnaire consists of an anxiety/depression scale and a depression
scale, each of 15 items. It demonstrates excellent validity as screening instrument for anxiety and depressive disorders in clinical
settings in Pakistan.
D 2005 Published by Elsevier B.V.
Keywords: Psychiatric status rating scale; Anxiety disorders; Depression disorder; Pakistan; Cultural diversity; Cross-cultural comparison
0165-0327/$ - s
doi:10.1016/j.jad
* Correspondi
E-mail addre
ers 88 (2005) 175–182
ee front matter D 2005 Published by Elsevier B.V.
.2005.05.015
ng author. Tel.: +44 117 954 6522; fax: +44 117 954 6525.
ss: [email protected] (D.B. Mumford).
D.B. Mumford et al. / Journal of Affective Disorders 88 (2005) 175–182176
1. Introduction
Questionnaires for detecting and monitoring
common mental disorders in Pakistan need to
reflect the way that people experience and express
these disorders in this culture. Currently there is a
very restricted choice of Pakistan-derived instru-
ments. Some western questionnaires, translated
and adapted for Pakistan, demonstrate good validity
with literate and well-educated subjects (Hospital
Anxiety and Depression Scale: Mumford et al., 1991a;
General Health Questionnaire: Minhas and Mubba-
shar, 1996; Self-Reporting Questionnaire: Minhas et
al., 1995; Saeed et al., 2001), but their content and
their format are less appropriate for general use in
Pakistan.
Since many people with anxiety or depressive
disorders in Pakistan present with somatic symp-
toms rather than psychological complaints, one
option is to use a somatic questionnaire to screen
for common mental disorders. The Bradford So-
matic Inventory (BSI: Mumford et al., 1991b) was
developed for this purpose, and has demonstrated
good sensitivity and specificity when screening for
common mental disorders both in medical clinics
(Mumford et al., 1991c) and community-based epi-
demiological studies (Mumford et al., 1996, 1997,
2000). However the BSI score does not allow a
differentiation between anxiety and depressive dis-
orders, and its efficiency as a screening instrument is
sometimes affected by variations in physical health
status.
In the development of any new questionnaire,
careful attention needs to be given to the language
and idiom of emotional distress in Pakistan. One
data source is case notes of people who have been
diagnosed with depressive illness or anxiety disor-
ders. This method has already been used successfully
in the development of the BSI (Mumford et al.,
1991b): the first source of items for the present
questionnaire was the same set of case notes from
the Mental Health Centre in Peshawar, but now
identifying the non-somatic idioms and phrases. To
broaden the ethno-linguistic base to include Urdu
and Punjabi speakers as well as Pushto, items were
also derived from case notes from the Department of
Psychiatry at Mayo Hospital/King Edward’s Medical
College, Lahore.
2. Methods
2.1. Item derivation
A systematic search was made of a consecutive
series of 75 psychiatric case notes from the Mental
Health Centre, Peshawar, in which the clinical diag-
nosis was anxiety disorder or depressive illness.
These case notes often included verbatim accounts
of presenting complaints and symptoms. Lists were
complied of all psychological or behavioural expres-
sions of distress, excluding somatic symptoms. Items
occurring twice or more were used to draft 37 ques-
tions, taking into account the research team’s clinical
experience of symptoms presented by Pakistani
patients.
Experienced clinicians at Mayo Hospital, Lahore,
made a systematic search of a further 200 sets of case
notes using the initial 37-item questionnaire as a
template. Items not represented in the Peshawar case
notes that occurred twice or more in the Lahore case
notes were identified. The eventual pilot version of
the questionnaire consisted of 74 items, redrafted in
Urdu.
2.2. Item validation
The validation study was conducted using equal
numbers of (a) Mayo Hospital inpatients, (b) Mayo
Hospital outpatients and (c) patients attending rural
community clinics 200 km from Lahore. The pilot
version of the questionnaire was administered in Urdu
to 150 patients (75 men and 75 women) with common
mental disorders and 150 controls (75 men and 75
women) who were mostly these patients’ relatives in
attendance. The questionnaire was either completed
by the subject unaided or administered by a doctor
who was blind to the diagnosis. Psychiatric diagnoses
were made systematically according to the ICD-10
Diagnostic Criteria for Research (World Health Orga-
nisation, 1993) by doctors with at least 2 years’
experience in psychiatry.
The total cohort of 300 patients and controls was
randomly allocated into two sub-samples, 200 for the
first stage of analysis (item validation) and 100 for the
second stage (scale evaluation and refinement), stra-
tified to maintain balanced of male/female and case/
control ratios.
D.B. Mumford et al. / Journal of Affective Disorders 88 (2005) 175–182 177
2.3. Item selection
In the first stage, items were selected on the basis of
(i) a high odds ratio between cases and controls, (ii)
item specificity, and (iii) eliminating the less powerful
item when there was a pair of similar questions. In the
second stage, items were selected on the basis of (i) the
frequency with which they were reported in patients
with anxiety and depressive disorders respectively, and
(ii) eliminating any remaining redundancy among clo-
sely related items that were highly inter-correlated.
2.4. Scale evaluation
Scores of patients and controls were compared on
each version of the questionnaire. Sensitivity and
specificity in identifying common mental disorders
were calculated using an optimal threshold.
3. Results
The mean age of patients was 33.2 years and of
controls 33.9 years (t =0.55, NS), with no significant
difference between men (34.0) and women (33.1).
The ICD-10 diagnoses of the 150 patients were:
depressive episode 92%, generalized anxiety disorder
31%, panic disorder 5%, dysthymia 3%, and somati-
zation disorder 2%. Some patients attracted more than
one psychiatric diagnosis because the dexclusionclausesT in the ICD-10 DCR were ignored.
3.1. Initial item validation and selection
The 100 patients and 100 controls were compared in
their responses to each of the 74 items of the pilot ques-
tionnaire. The odds ratios ranged from 1.90 to 88.78, all
but one being statistically significant. 42 items were
selected for inclusion in the next working version of the
questionnaire based on (i) an odds ratio over 7.0 and (ii)
specificity of 75% or more. Some items were eliminated
when there was a pair of very similar questions.
3.2. Initial scale evaluation
The second sub-sample of 50 patients and 50 con-
trols was used to evaluate the 42-item version of the
questionnaire. Scores were calculated from the sum of
responses positive for the symptom. Patients’ mean
score (28.0) was very much higher than controls’ (5.5)
(t =15.6, df =98, p b0.001). Scores were not signifi-
cantly related to either age or gender.
3.3. Specificity for depressive and anxiety disorders
To refine the scale further, we wished to identify
which items were associated with depressive and
anxiety disorders respectively. To increase the number
of anxiety disorders unaccompanied by a diagnosis of
depression, we recruited 30 additional patients with
simple anxiety disorders (15 males and 15 females).
This yielded three principal groups for comparison:
pure depressive disorders (N =30), pure anxiety dis-
orders (N =33), and normal controls (N =50). A fourth
group consisted of patients receiving diagnoses of
both anxiety and depressive disorders (N =14).
About half of the items were reported much more
frequently by depressed patients than by patients with
anxiety disorders. The other half of the items showed
similar response rates between these two diagnostic
groups, with one item occurring significantly more
frequently in anxiety disorders (dDo you have repeti-
tive thoughts?T). The response rates to all items were
much lower in the control group. Patients receiving a
diagnosis of both an anxiety and a depressive disorder
had the highest response rate to most items.
3.4. Final item selection
In the light of these findings, we constructed two
subscales: one consisting of symptoms frequently
reported by patients with either anxiety or depressive
disorders, and the other of symptoms largely reported
by depressed patients. A further 12 items were elimi-
nated at this stage, either because of poor sensitivity
(less than 50% in either the anxiety or the depressive
group), or because a closely similar item performed
better. The resulting dAD scaleT and dD scaleT each
consisted of 15 items (Appendixes A and B). Mean
scores for each diagnostic group and normal controls
are given in Table 1.
3.5. Thresholds and validity/reliability coefficients
A histogram of AD scores showed a clear bimodal
distribution between controls and cases, with a point
Table 1
Mean questionnaire scores by diagnostic group
Scale Depressive disorders
(N =30)
Anxiety disorders
(N =33)
Diagnosis of depression
and anxiety disorder (N =14)
Normal controls
(N =50)
ANOVA
AD scale 11.1 (3.1) 9.9 (2.8) 12.7 (1.8) 2.5 (3.0) F =89.3
df =126
p b0.001
D scale 10.0 (3.3) 4.6 (3.3) 11.4 (2.0) 1.7 (2.2) F =80.2
df =126
p b0.001
Standard deviation in brackets.
D.B. Mumford et al. / Journal of Affective Disorders 88 (2005) 175–182178
of attenuation around a score of 6. The histogram of D
scores was weakly bimodal with a point of attenuation
also around 6.
The sensitivity, specificity, and positive and nega-
tive predictive values are given in Table 2. The mis-
classification rate for the AD scale was 7%, and for
the D scale 7% (depressed versus controls) and 25%
(depressed versus anxiety disorders). A discriminant
function analysis using the AD and D scores to predict
group membership achieved overall 82% of the ori-
ginal group correctly classified between control
group, anxiety and depression groups. 92% of the
controls were correctly classified.
The a coefficient of reliability for the AD scale was
0.92 (95% CI: 0.90–0.94) and for the D scale was
0.91 (95% CI: 0.89–0.93).
4. Discussion
The use of psychiatric case notes to derive items for
a culturally-sensitive questionnaire is a method which
Table 2
Validity coefficients
Scale Cases/controls Threshold Sensitivi
(%)
AD scale 77 all cases 5/6 96
50 controls
D scale 44 depressive disorders 5/6 91
50 controls
D scale 44 depressive disorders 5/6 91
33 anxiety disorders
AD+D scales 100 all cases 11/12 95
100 controls
has successful precedents. Anthropologists might
argue that conducting open-ended interviews in non-
medical settings would have elicited a fuller range of
cultural expressions of distress. However the case note
method used here does focus on dwhat patients tell
doctorsT, which replicates the clinical setting in which
the questionnaire is designed to be used.
The two psychiatric centres, in Peshawar and
Lahore, served both urban and rural patients. Most
people in Peshawar are Pushto-speakers, whereas
Urdu and Punjabi are spoken in Lahore. The idiom
of expression varied slightly between these languages,
although there was a very large overlap in symptoms
derived from the two centres.
How do the symptoms elicited in this study com-
pare with typical symptoms of anxiety and depression
reported by western patients? Reviewing the 74 items
of the pilot version, the most striking difference is the
emphasis on the social impact of mental disorder,
rather than on subjective feelings experienced by
the individual. A typical western patient might
describe their personal feelings of depression,
ty Specificity
(%)
Positive predictive
value (%)
Negative predictive
value (%)
88 93 94
94 93 92
64 69 84
91 91 95
Table 3
Validity coefficients of screening questionnaires in Pakistan
Scale Sample size Optimal threshold Sensitivity (%) Specificity (%) Authors
GHQ-12 238 2/3 93 88 Minhas and Mubbashar (1996)
SRQ 300 4/5 63 77 Minhas et al. (1995)
SRQ 191 7/8 (women) 78 (women) 81 (women) Saeed et al. (2001)
3/4 (men) 78 (men) 70 (men)
BSI-44 191 21/22 (women) 82 (women) 71 (women) Saeed et al. (2001)
13/14 (men) 59 (men) 62 (men)
BSI-21 62 20/21 80 77 Mumford et al. (1996)
PADQ AD scale 130 5/6 96 88
PADQ D scale 130 5/6 91 94
D.B. Mumford et al. / Journal of Affective Disorders 88 (2005) 175–182 179
whereas a typical Pakistani patient would describe the
disruption of their social functioning. Pakistani
patients are often acutely aware of their failure to
fulfil basic social obligations in the household and
at work. Talking about personal feelings – especially
to a doctor, someone who occupies a much higher
social status than the patient – does not come easily
or naturally.
Some of these distinctively Pakistani symptoms
were eliminated during the validation stages because
of their less favourable psychometric properties. The
discarded symptoms included some religious items
(dDo you feel like saying your prayers?T, dDo you
have blasphemous thoughts about God and the
Prophet?T) and some interpersonal items (dDo you
argue with people a lot?T, dCan you put up with
noise?T). The item referring to libido (dHave your
sexual feelings decreased?T) had to be discarded as
unacceptable to female respondents.
Many of the 30 items in the final version of the
questionnaire are familiar to western psychiatry. The
idiom of expression is nonetheless more functional
and less abstract than in equivalent western question-
naires, and more family-orientated. Two religiously-
based items have proved robust (dDo you feel you
have committed some serious sin?T, dDo you feel you
are being punished for something?T).We found that it was possible to distinguish
between anxiety and depressive disorders based on
their symptomatology. Many symptoms were
reported by depressed patients significantly more
frequently than by patients with anxiety disorders.
For most items, patients with diagnoses of anxiety
plus depression showed higher rates of endorsement
than those with a single diagnosis. This pattern of
responses led to the decision to construct two sub-
scales, dAnxiety and DepressionT and dDepressionT.Depressed patients score highly on both subscales;
patients with anxiety disorders scored highly only on
the AD subscale.
The sensitivity and specificity of the Pakistan
Anxiety and Depression Questionnaire compares
favourably with that of other screening instruments
for common mental disorders in Pakistan (Table 3).
The PADQ has significant advantages: the questions
reflect local idioms of distress, and the questionnaire
is easy to administer to less educated subjects. The
PADQ shows excellent psychometric properties in
differentiating patients and normal controls, and
good differentiation of anxiety and depressive dis-
orders. It is offered to clinicians and researchers as
an innovative culturally-sensitive questionnaire to
detect these common mental disorders in Pakistan.
Acknowledgements
Our warm thanks to many colleagues in Bradford,
Peshawar and Lahore who contributed in different
ways to the project: Yasmin Hussain and Kumud
Bhatnagar (Bradford); Jaffar Hussain, Zubair Khan,
S Farooq and the late Anis-ur-Rehman (Peshawar);
Sohail Ali, Lutfullah, Raheel Aziz, Imtiaz Cheema,
Abbas Ali Cheema and Nauman Khalil (Lahore).
Copyright
The authors retain the copyright of the Pakistan
Anxiety and Depression Questionnaire, but it may be
used free of charge with due acknowledgement of its
source.
D.B. Mumford et al. / Journal of Affective Disorders 88 (2005) 175–182180
Appendix A. Pakistan Anxiety and Depression Questionnaire
English Translation
AD scale
Do you think that you have some mental problems? yes no
Do you feel anxious amongst a lot of people? yes no
Is your mind in peace? no yes
Do you worry over trivial things? yes no
Has your tolerability decreased? yes no
Does one idea come to your mind again and again? yes no
Have you become more irritable? yes no
Do you feel lazy? yes no
Have you lost your self-confidence? yes no
Do you get frightened? yes no
Do you feel that your mind is not working? yes no
Do you feel that you are being punished for something? yes no
Do you sleep well at night? no yes
Do you keep on thinking without any purpose all the time? yes no
Do you feel that you do not understand anything? yes no
Total
D scale
Are you happy these days? no yes
Do you feel sad at heart? yes no
Do you feel like working? no yes
Do you enjoy this world? no yes
Are you excessively hopeless? yes no
Has your interest decreased? yes no
Do you feel like staying in bed all day? yes no
Do you enjoy your food? no yes
Do you feel like crying? yes no
Do you feel fed up with your family members? yes no
Do you feel you have committed some serious sin? yes no
Do you feel like running away from your home? yes no
Are you fed up with life? yes no
Do you have a wish to die? yes no
Do you ever think that death would be better than this life? yes no
Total
AD scale: score of 6 or more = probable anxiety or depressive disorder
D scale: score of 6 or more = probable depressive disorder
Appendix B. Pakistan Anxiety and Depression Questionnaire (Urdu)
D.B. Mumford et al. / Journal of Affective Disorders 88 (2005) 175–182 181
References
Minhas, F.A., Mubbashar, M.H., 1996. Validation of General Health
Questionnaire (GHQ-12) in primary care settings of Pakistan.
Journal of the College of Physicians and Surgeons Pakistan 6,
133–136.
Minhas, F.A., Iqbal, K., Mubbashar, M.H., 1995. Validation of self-
reporting questionnaire in primary care settings of Pakistan.
Pakistan Journal of Clinical Psychiatry 5, 60–69.
Mumford, D.B., Tareen, I.A.K., Bajwa, M.A.Z., Bhatti, M.R.,
Karim, R., 1991a. Translation and evaluation of an Urdu version
of the hospital anxiety and depression scale. Acta Psychiatrica
Scandinavica 83, 81–85.
Mumford, D.B., Bavington, J.T., Bhatnagar, K.S., Hussain, Y.,
Mirza, S., Naraghi, M.M., 1991b. The Bradford Somatic Inven-
tory. A multi-ethnic inventory of somatic symptoms reported by
anxious and depressed patients in Britain and the Indo-Pakistan
subcontinent. British Journal of Psychiatry 158, 379–386.
Mumford, D.B., Tareen, I.A.K., Bhatti, M.R., Bajwa, M.A.Z.,
Ayub, M., Pervaiz, T., 1991c. An investigation of dfunctionalTsomatic symptoms among patients attending hospital medical
clinics in Pakistan: II. Using somatic symptoms to identify
patients with psychiatric disorders. Journal of Psychosomatic
Research 35, 257–264.
Mumford, D.B., Nazir, M., Jilani, F.M., Baig, I.Y., 1996. Stress and
psychiatric disorder in the Hindu Kush: a community survey of
D.B. Mumford et al. / Journal of Affective Disorders 88 (2005) 175–182182
mountain villages in Chitral, Pakistan. British Journal of Psy-
chiatry 168, 299–307.
Mumford, D.B., Saeed, K., Ahmad, I., Latif, S., Mubbashar, M.H.,
1997. Stress and psychiatric disorder in rural Punjab: a commu-
nity survey. British Journal of Psychiatry 170, 473–478.
Mumford, D.B., Minhas, F.A., Akhtar, I., Akhter, S., Mubbashar,
M.H., 2000. Stress and psychiatric disorder in urban Rawal-
pindi: community survey. British Journal of Psychiatry 177,
557–562.
Saeed, K., Mubbashar, S.S., Dogar, I., Mumford, D.B., Mubba-
shar, M.H., 2001. Comparison of self-reporting questionnaire
and Bradford Somatic Inventory as screening instruments for
psychiatric morbidity in community settings in Pakistan. Jour-
nal of the College of Physicians and Surgeons Pakistan 11,
229–231.
World Health Organisation, 1993. The ICD-10 Classification of
Mental and Behavioural Disorders: Diagnostic Criteria for
Research. WHO, Geneva.