Warwick-Edinburgh Mental Well-being Scale

(WEMWBS)

User Guide
Version 1

Professor Sarah Stewart-Brown

Professor of Public Health
sarah.stewart-brown@warwick.ac.uk
&
Dr Kulsum Janmohamed
Academic Clinical Fellow
k.janmohamed@warwick.ac.uk
Warwick Medical School
University of Warwick

Edited by Dr Jane Parkinson

Public Health Adviser
NHS Health Scotland

June 2008

Acknowledgements
The Warwick-Edinburgh Mental Well-being Scale was funded by the Scottish
Governments1 National Programme for Improving Mental Health and Wellbeing, commissioned by NHS Health Scotland, developed by the University of
Warwick and the University of Edinburgh, and is jointly owned by NHS Health
Scotland, the University of Warwick and the University of Edinburgh.
Acknowledgements go to the following:

Members of the research team involved in the development and validation of

WEMWBS: Ruth Fishwick (University of Warwick), Louise Hiller (University of
Warwick ), Professor Stephen Joseph (University of Nottingham), Professor
Stephen Platt (University of Edinburgh), Professor Sarah Stewart-Brown
(University of Warwick) and Ruth Tennant (University of Warwick).

Members of the research Advisory Group: Professor Glynn Lewis (University

of Bristol), Dr Jane Parkinson (NHS Health Scotland), Professor Jenny
Secker (Anglia Ruskin University), Professor Stephen Stansfeld (University of
London) and Professor Scott Weich (University of Warwick).

Those who commented on this manual: Dr David Gordon (NHS Health

Scotland), Emma Hogg (NHS Health Scotland), Kate OHara (CSIP West
Midlands), and Professor Stephen Platt (University of Edinburgh).

Audience
It is anticipated that the audience for this manual includes researchers and
practitioners who are familiar with the use of scales in evaluations. This manual
does not seek to answer questions relating to what to consider for evaluation
purposes. For this, the evaluation guides of NHS Health Scotland provide the
required information and should be referred to in the first instance
(www.healthscotland.com/mental-health-publications.aspx).

Update revisions
This manual will be updated and revised as necessary as further validation and
data on WEMWBS become available. For the current version at any time see
http://www.healthscotland.com/scotlands-health/population/Measuring-positivemental-health.aspx.

Previously known as the Scottish Executive

Summary
The Warwick-Edinburgh Mental Well-being Scale (WEMWBS) was developed by
researchers at the Universities of Warwick and Edinburgh, with funding provided
by NHS Health Scotland, to enable the measurement of mental well-being of
adults in the UK.
WEMWBS is a 14 item scale of mental well-being covering subjective well-being
and psychological functioning, in which all items are worded positively and
address aspects of positive mental health. The scale is scored by summing
responses to each item answered on a 1 to 5 Likert scale. The minimum scale
score is 14 and the maximum is 70. WEMWBS has been validated for use in the
UK with those aged 16 and above. Validation involved both student and general
population samples, and focus groups.
People participating in studies of face validity found the scale clear, unambiguous
and easy to complete. They volunteered the opinion that the scale measured
mental well-being.
Population scores on WEMWBS approximate to a normal distribution with no
ceiling or floor effects, making the scale suitable for monitoring mental well-being
in population samples. The scale is not designed to identify individuals with
exceptionally high or low positive mental health, so no cut off has been
developed (analogous to a mental illness cut-off on for example the GHQ 12
scale). The provisional Scottish population mean score is 50.7 with a 95%
confidence interval of 50.3 to 51.1, obtained from a combined national dataset
comprising data from the Health Education Population Survey 2006 (wave 12)
and the Well? What do you think? 2006 survey.
Scores derived from the student and population samples show a single
underlying factor, interpreted to be mental well-being, with low levels of social
desirability bias and expected moderate correlations with other scales of wellbeing. Scores for individuals are stable over a one week period.
In general population samples, significant differences in WEMWBS scores were
found by certain factors such as tenure, employment status, and marital status.
Non-significant trends were found between mental well-being and social grade
(with lowest scores among those in the most deprived groups), a u-shaped
relationship was found for age and small but non-significant differences were
found for sex (male scores were slightly higher).
Further research on WEMWBS is ongoing. This includes: establishing
WEMWBSs sensitivity to change; assessing its scaling properties and the
potential to reduce the number of items; and validation to determine whether
WEMWBS can be used with children aged 13 to 15 years of age. Other research
still required includes assessing the extent to which it is appropriate to use
WEMWBS to assess mental well-being in English speaking ethnic minority
populations in the UK.
ii

As a short and psychometrically robust scale, with no ceiling effects in population

samples, WEMWBS offers promise as a tool for monitoring mental well-being at
a population level. It is freely available but prospective users should register with
Dr Kulsum Janmohamed K.janmohamed@warwick.ac.uk or Professor Sarah
Stewart-Brown sarah.stewart-brown@warwick.ac.uk. If the scale is reproduced it
must remain unaltered and include the copyright statement which appears with it
(Appendix ii).

iii

Contents
1.

Introduction ....................................................................................................1

2.

A word about mental well-being.....................................................................2

3.

What is WEMWBS and how was it developed?.............................................3

4.

Validation of WEMWBS .................................................................................4

Student populations.................................................................................4
Focus groups...........................................................................................7
Scottish population samples ....................................................................8

5.

Distribution of WEMWBS scores ...................................................................9

6.

Comparison between WEMWBS scores and scores on the GHQ 12 ..........11

7.

Variation across demographic & social groups ............................................13

8.

Using WEMWBS..........................................................................................15

9.

Current usage of WEMWBS ........................................................................18

10.

Further validation research on WEMWBS................................................19

Bibliography ........................................................................................................21
Appendix i

Conditions of using WEMWBS.......................................................22

Appendix ii

Description of scales used to assess the construct validity of

WEMWBS ..................................................................................... 24

Appendix iii WEMWBS median scores across demographic groups ................26

iv

1. Introduction
Practitioners of mental health promotion and public mental health have for many
years recognised the need to focus their efforts on improving mental health as
well as preventing mental illness. Because of confusion relating to use of the
term mental health to describe services for people with mental illness, terms like
positive mental health and mental well-being have been adopted to describe
these initiatives Positive mental health and mental well-being are used
interchangeably in this manual).
Efforts to promote mental well-being have been hampered by a lack of valid
instruments which are suitable for measuring these attributes in the general
population. The monitoring of population mental well-being and the evaluation of
interventions to promote positive mental health has therefore had to be
undertaken using instruments designed primarily to detect mental illness. There
are two problems with such an approach. First, mental illness measures tend to
have significant ceiling effects in general population samples, meaning that
people with only moderately good mental health can achieve the highest possible
score. As a result the instrument cannot show improvements in mental health in
the healthier portion of the population distribution. Second, participants who are
involved in the evaluation of interventions to promote mental health may develop
the erroneous impression that the interventions are designed only to help people
with mental health problems and in this way the evaluation can affect the impact
of interventions.
To overcome these problems NHS Health Scotland commissioned the
development of the Warwick-Edinburgh Mental Well-being Scale (WEMWBS) as
part of the Mental Health Indicators Programme.2
This manual is for those who want to use WEMWBS for monitoring and research
purposes as well as for evaluations. Those who require information on what to
consider for evaluation are referred to the NHS Health Scotland evaluation
guides in the first instance (www.healthscotland.com/mental-healthpublications.aspx).

Ceiling and floor effects these occur when many people score the maximum or
minimum score on a scale. Improvements or deteriorations in the assessed variable
being measured cannot therefore be identified. For example, significant ceiling effects in
a mental health scale used in a general population sample may mean that people who
possess only moderately good mental health can achieve the highest possible score. As
a result the instrument cannot show improvements in mental health in the healthier
portion of the population distribution.

NHS Health Scotland was commissioned by the Scottish Governments National Programme for
Improving Mental Health and Well-being (www.wellscotland.info) to establish a core set of
national, sustainable mental health and well-being indicators for adults in Scotland
(www.healthscotland.com/understanding/population/mental-health-indicators.aspx).

2. A word about mental well-being

A necessary starting point for the development of a new instrument is a clear
understanding of the concept which it is designed to measure. In the past there
has been considerable discussion and debate about the nature of positive mental
health and well-being. Recently a reasonable level of consensus has emerged
among both academics and among the public.
Mental well-being is now largely accepted as covering two perspectives: (1) the
subjective experience of happiness (affect) and life satisfaction (the hedonic
perspective); and (2) positive psychological functioning, good relationships with
others and self realisation (the eudaimonic perspective). The latter includes the
capacity for self development, positive relations with others, autonomy, self
acceptance and competence. Those wanting to understand more about this
subject are referred to the large literature, clearly described in Ryan and Deci
(2001).
There has been some discussion in the academic literature as to whether mental
well-being and mental illness represent two ends of a single spectrum (single
continuum model) or two separate dimensions (two continua or dual continua
model). The two continua model allows for the possibility that people who have
mental illnesses can experience mental well-being. It reflects the finding that
analysis of instruments covering both positive and negative mental health often
suggests two correlated but independent underlying factors. Possible
explanations for these findings include issues relating to how psychiatric
conditions are defined, the fluctuating nature of mental illness, and individuals
interpretations and responses to positively and negatively worded items on
mental health measurement scales.

Mental well-being relates to a persons psychological functioning, life-satisfaction and

ability to develop and maintain mutually benefiting relationships. Psychological wellbeing includes the ability to maintain a sense of autonomy, self acceptance, personal
growth, purpose in life and self esteem. Staying mentally healthy is more than treating
or preventing mental illness.
Mental illness is a term to encompass mental disorders these are illnesses which
affect mood, affect and the ability to function effectively and appropriately.
Hedonic perspective of well-being focuses on the subjective experience of happiness
(affect) and life satisfaction.
Eudaimonic perspective of well-being focuses on psychological functioning, good
relationships with others and self realisation. This is the development of human potential
which when realised results in positive functioning in life, and covers a wide range of
cognitive aspects of mental health.

3. What is WEMWBS and how was it developed?

The Warwick-Edinburgh Mental Well-being Scale (WEMWBS) comprises 14
items that relate to an individuals state of mental well-being (thoughts and
feelings) in the previous two weeks (see Appendix i). Responses are made on a
5-point scale ranging from none of the time to all of the time. Each item is
worded positively and together they cover most, but not all, attributes of mental
well-being including both hedonic and eudaimonic perspectives. Areas not
covered include spirituality or purpose in life. These were deemed to extend
beyond the general populations current understanding of mental well-being and
their inclusion was thought likely to increase non-response.
WEMWBS aims to measure mental well-being itself and not the determinants of
mental well-being, which include resilience, skills in relationship, conflict
management and problem solving, as well as socioeconomic factors such as
poverty, domestic violence, bullying, unemployment, stigma, racism and other
forms of social exclusion.
WEMWBS was developed through research that was conducted at Warwick and
Edinburgh Universities. The starting point for the research was a pre-existing
scale called the Affectometer 2, developed in the 1980s in New Zealand
(Kammann & Flett, 1983). Affectometer 2 consists of 20 statements and 20
adjectives relating to mental health in which positive and negative items are
balanced. It proved to have a broad measure of intuitive appeal to practitioners
and researchers working in this area in the UK. While it had been used in a
number of countries, there was no UK validation of the scale and so this
validation was conducted as the first step in this research project.
Validation of the Affectometer 2 in both population and student samples
suggested that whilst it performed adequately, it was longer than need be and
subject to an unacceptable level of bias due to desirable responding
(respondents answering in a way they thought was likely to be approved of)
(Tennant et al., 2006; Tennant, Joseph & Stewart-Brown, 2007). A focus group
study involving participants from a wide range of socio-economic backgrounds
found that although in general the scale was viewed favourably, some of the
items were considered to be difficult, and in spite of the balance of positive and
negative items, the instrument was viewed predominantly as a measure of
mental illness (Tennant et al., 2006).
These results were reviewed by a multidisciplinary research advisory group
familiar with epidemiological research as well as the academic literature relating
to concepts of positive mental health. The research team drafted a set of items
derived partly from Affectometer 2, but taking into account the findings of the
qualitative focus group research relating to difficult and potentially redundant
items, whilst at all times referring to current literature on positive mental health.
Working iteratively with members of the advisory group this new scale was
refined to the 14 item scale WEMWBS.

4. Validation of WEMWBS
Validation to date has been performed in the UK with those aged 16 and above.
WEMWBS was initially validated in student samples recruited at the universities
of Warwick and Edinburgh in 2006, and subsequently discussed by two minifocus groups in Scotland and England (Tennant et al., 2006; Tennant et al.,
2007). WEMWBS was then included in two national Scottish population surveys
in 2006 allowing validation using population data.
Table 1 below lists whether or not the psychometric tests involved in validating a
scale have been performed on WEMWBS and if so the sample(s) used. Details
of the results are given on the following pages.
Table 1: Psychometric testing of WEMWBS
Psychometric test

Tested

Principal components factor analysis

Construct validity
Internal consistency
Test-retest reliability
Response Bias
Face (or content) validity
Rasch analysis
Sensitivity to change

Criterion validity

Cross-cultural validity

Sample
Student population samples &
Scottish general population samples
Student population samples &
Scottish general population samples
Student population samples &
Scottish general population samples
Student population samples
Student population samples
WEMWBS research advisory group &
Focus groups
Scottish general population samples
Currently being assessed
Gold standard measure to assess
WEMWBS against does not currently
exist
Interest has been expressed in using
WEMWBS in other countries. An
Icelandic version has been created by
translation and back-translation.

Student populations (n = 348)

Principal components factor analysis
The main aims in conducting these analyses were:
to determine whether the number of variables in the scale can be reduced
to determine the relationships between variables
The test considers how much variance is added by each factor the scale
considers. The variance that each additional factor contributes is expressed by
eigenvalues. A scree plot of eigenvalues against component numbers can be
used to illustrate the amount of variance that a single factor contributes.
4

Figure 1: WEMWBS Scree Plot for student samples (n = 348)

Factor analysis confirmed a single underlying factor to the scale, shown in the
sharp elbow of the screen plot. This underlying factor is interpreted to be mental
well-being.
Construct validity
Considers the extent to which there are logical relationships between the scale
and other scales or factors known to affect the concept being measured (such as
age or sex). It is assessed by correlations between the scale under review and
other scales measuring similar concepts (convergent validity) or different
concepts (divergent validity) and by determining statistically significant
differences in scale scores between different groups.
For the validation of WEMWBS, this was assessed by testing correlations
between WEMWBS and other scales that measure aspects of mental health, as
well as scales that measure general health and emotional intelligence (Appendix
ii), and also the extent to which it follows anticipated patterns for age and sex.
Table 2: Correlation of WEMWBS to other scales

Scale
WHO-Five Well-being Index
Short Depression Happiness Scale
Positive and Negative Affect Scale
Positive Subscale
Positive and Negative Affect Scale
Negative Subscale
Satisfaction With Life Scale
Global Life Satisfaction Scale
Scale of Psychological Well-being
EQ-5D Thermometer
Emotional Intelligence Scale

Pearsons correlation coefficient

* => significant at 0.05 level
** => significant at 0.01 level

n
79
71
63

Correlation with WEMWBS

0.77**
0.76**
0.73*

63

-0.55**

79
77
63
72
67

0.72**
0.55**
0.73**
0.42**
0.51**

Correlations were moderately high between WEMWBS and the: Scale of

Psychological Well-being; Satisfaction with Life Scale; Short Depression
Happiness Scale; Positive and Negative Affect Scale positive subscale; and the
WHO-Five Well-being Index. These results were similar to those found between
Affectometer 2 and these scales, which is as expected, given that Affectometer 2
was the starting point for research on the WEMWBS scale. These results
indicate that WEMWBS covers both hedonic and eudaimonic aspects of mental
well-being.
WEMWBS showed moderate to low correlations with the EQ-5D thermometer (a
measure of overall physical and emotional health) and the Emotional Intelligence
Scale (a measure of the ability to accurately assess ones own and others
emotions). This is expected because these two scales measure concepts that
are separate from (but not unrelated to) positive mental health.
Internal consistency
Considers whether the scale describes a consistent underlying theme in this
case, it considers the extent to which WEMWBSs items are focused on
assessing mental well-being. Scores range from 0 to 1 and are measured by
Cronbachs alpha coefficient. The higher the co-efficient, the more highly
correlated the items in the scale. A coefficient of 0.7-0.8 is ideal (Nunnally,
1978), and higher coefficients may suggest that some degree of item redundancy
exists in the scale.
Cronbachs alpha coefficient = 0.89 (n = 348).
This high coefficient suggests that, while there is a good level of internal
consistency, there may be scope to reduce even further the number of items in
the scale (analyses are currently underway to explore the potential for a
shortened scale, see section 10)
Test-retest reliability
Considers the stability of responses over a period of time. Test-retest reliability is
determined by calculating the correlation between two sets of scores for the
same group of people who repeat the test after a set period of time. For
WEMWBS, the time period was one week.
Correlation = 0.83 after one week (n = 124)

Intra-class correlation coefficient

The test-retest reliability score was high for WEMWBS after one week. This
suggests that the transient fluctuations that a person may experience from one
day to the next are not reflected in the scores, and these scores remain robust
over a short period of time.
Response Bias
Considers the extent to which an individual may tailor his or her responses in
order to be perceived in a certain light, a phenomenon known as impression
management. And also the extent to which an individual remains unaware of
6

their true state of mental well-being known as self deception bias. These two
aspects of social desirability responding are measured using the Balanced
Inventory of Desirable Responding (BIDR).
Correlations between the two subscales of the Balanced Inventory of Desirable
Responding and WEMWBS, and between the two subscales and other mental
health scales including Affectometer 2 are shown below:
Table 3: Correlation of WEMWBS to BIDR
Scale

WEMWBS
Affectometer 2
WHO-Five Well-being Index
Positive and Negative Affect Scales
Positive subscale
Positive and Negative Affect Scales
Negative subscale
Satisfaction with life scale
Global life satisfaction scale

115
115
62
52

Impression
Management
0.18*
-0.25**
-0.39**
0.02

51

0.03

-0.16

62
62

0.34**
0.26*

0.40**
0.13

Self-Deception
0.35**
0.55**
-0.20
0.50**

Pearsons correlation coefficient

* => significant at 0.05 level
** => significant at 0.01 level

WEMWBS showed a low correlation with both subscales of the Balanced

Inventory of Desirable Responding. This contrasts with Affectometer 2, where
self-deception bias was a major disadvantage of the scale. WEMWBS also
performed better than three comparison mental health scales on impression
management and better than two on self-deception.
These findings suggest that both impression management and self-deception
response biases, whilst still an issue (as they are with all mental health scales),
are acceptable for monitoring and evaluation purposes at the group/population
level.

Focus groups
Face validity
Face validity assesses whether the items in the scale are suitable for the overall
concept being measured. For WEMWBS this was tested in two mini focus
groups with members of the general population in England and Scotland,
selected on the basis of socioeconomic background, age and sex. Groups
included mental health service users and non-users. Individuals were asked to
complete WEMWBS and discuss their impressions of the scale. The aim of
these investigations was to test what people thought WEMWBS was designed to
measure and to determine its user-friendliness. Participants were asked to
identify any items which they thought irrelevant or confusing. Results of these
7

focus group discussions suggested that WEMWBS was clear, user-friendly and
unambiguous. Unlike the Affectometer 2, no suggestions were made to modify
the scale or to clarify it in any way. Importantly, participants recognised that
WEMWBS measured positive mental health rather than mental illness.

Scottish population samples (n =1749)

WEMWBS was included in the Autumn wave of the Scottish Health Education
Population Survey (HEPS) 2006 (wave 12), which collected data from a random
sample of the Scottish population aged 16 to 74 on a wide range of aspects of
health and health related lifestyles (Gosling et al., 2008), and also in the
population survey Well? What do you think? (Well?) 2006, conducted on a
random sample of the Scottish population aged 16 and above to collect data on
public attitudes to mental health, mental well-being and mental health problems
(Braunholtz et al., 2007).
Analysis of combined data from these two population surveys (n = 2075 for the
combined datasets, with complete WEMWBS scores for n = 1749 and complete
GHQ 12 scores for n = 1239) have confirmed the findings of the student
validation (Tennant et al., 2007):

verification of a pre-hypothesised single underlying factor (n = 1749)

Cronbachs alpha = 0.91 (n = 1749), again indicating that while there is a

good level of internal consistency, there may be scope to reduce the
number of items in the scale even further

good performance against accepted criteria, discriminating population

groups largely as expected and in a way consistent with other population
surveys (see section 7 and Appendix iii)

significant moderate negative correlation to the General Health

Questionnaire 12 (GHQ 12) (see section 7).

5.

Distribution of WEMWBS scores

In both the student and population samples, WEMWBS scores followed a roughly
normal distribution with only a slight left-skew (Figure 2). WEMWBS can be used
to calculate mean scores for different groups of people or for the same people at
different time periods.3 Mean scores can be compared using standard deviations
and 95% confidence intervals.
Figure 2: Distribution of WEMWBS scores for the combined HEPS (wave
12) and Well? 2006 datasets (n = 1749)

WEMWBS Score
Because WEMWBS scores show a roughly normal distribution, WEMWBS can
be expected to capture the full spectrum of positive mental health without floor or
ceiling effects and be suitable both for monitoring trends over time and evaluating
the effect of mental health promoting programmes or interventions. However,
although several studies are now in progress, it is important to note that, at the
time of writing this manual, WEMWBSs sensitivity to change has not been
demonstrated. As Affectometer 2 is sensitive to change there is no reason to
think that WEMWBS will not be.
3

Median scores should be used if data collected are not normally distributed, and mean scores if
the data are. WEMWBS scores followed a roughly normal distribution with a slight left-skew. As
the distribution is so close to normal it is considered appropriate to use mean scores, although
some statisticians may decide that median scores should be used.

As well as not being designed to identify people who have or probably have a
mental illness, WEMWBS does not a have a cut off level to divide the population
into those who have good and those who have poor mental well-being in the
way that scores on other mental health measures, for example the GHQ 12 do
(see section 6).

10

6. Comparison between WEMWBS scores and scores on

the GHQ 12
Both the HEPS and Well? surveys included the GHQ 12 measure alongside
WEMWBS, allowing the two to be compared in the same group of people. As
Figure 3 shows, GHQ 12 scores are heavily skewed in population samples with
the majority of people scoring 0 (no evidence of possible mental illness). This
contrasts with the distribution of WEMWBS scores (Figure 2). Because of this
distribution, GHQ 12 scores are more appropriately analysed in groups
representing different levels of mental illness. A cut off score of 4+ is often used,
with the 4+ group being more likely to have a diagnosable mental illness.
Figure 3: Distribution of GHQ 12 scores for the combined HEPS (wave 12)
and Well? 2006 datasets (n = 1,239)

GHQ 12 score
WEMWBS scores showed a significant moderate negative correlation with GHQ
12 scores in this population (r = -0.53, p <0.01, Spearmans rank correlation),
which persisted when a dichotomous scoring method, (with the four GHQ 12
response categories being scored 0, 0, 1, 1) was used (p < 0.01) (see Figure 4
scatterplot and box and whisker plot). The scatterplot (left) shows that
respondents scoring the same on the GHQ 12 had a wide range of WEMWBS
scores, so although lower WEMWBS scores tend to be associated with higher
GHQ 12 scores (right), one is not simply the inverse of the other. The two scales
are therefore not measuring the same thing.
11

WEMWBS total score

WEMWBS total score

Figure 4: WEMWBS score vs. GHQ 12 score, scatter plot and box and 90%
confidence interval whisker plot: population sample

GHQ 12 total score

GHQ 12 total score

Tennant et al. Health and Quality of Life Outcomes 2007 5:63 doi:10.1186/1477-7525-5-63

12

7. Variation across demographic & social groups

Analysis of data from these two population surveys has also provided provisional
population norms for WEMWBS across different socio-demographic groups.
Table 4 below shows mean WEMWBS scores along with the lower and upper
95% confidence intervals and the number of responses on which these estimates
are based (Appendix iii shows the same analysis but for median scores and also
contains significance p values).
The provisional population mean score is 50.7 with 95% confidence interval 50.3
51.1 from the combined HEPS (wave 12) and Well? 2006 datasets.
Table 4: WEMWBS mean scores across demographic groups: population
sample combined HEPS (Wave 12) and Well? 2006 datasets (n= 1,749)
Variable
Total
Sex

Male
Female
Age in years
16 24
25 34
35 44
45 54
55 64
65 74
75+
Tenure
Own outright
Own with a mortgage
Rent
Self-perceived health status
Very good
Good
Fair
Poor
Very poor
Employment Status
In work
Student
Retired
Unemployed
Other
Marital Status
Single
Married/Living as couple
Widowed/Divorced/Separated
13

n
1749

Mean (95% CI)

50.7 (50.3 - 51.1)

783
966

51.3 (50.6 - 51.9)

50.3 (49.7 - 50.8)

176
245
353
306
334
274
61

51.7 (50.6 - 52.8)

50.1 (49.1 - 51.1)
49.7 (48.8 - 50.7)
49.5 (48.4 - 50.5)
51.4 (50.4 - 52.4)
52.4 (51.3 - 53.4)
51.2 (48.9 - 53.4)

523
705
519

52.3 (51.5 - 53.0)

51.1 (50.5 - 51.7)
48.6 (47.8 - 49.4)

563
753
319
84
29

53.8 (53.1 - 54.5)

50.9 (50.4 - 51.9)
47.6 (46.6 - 48.6)
43.5 (41.3 - 45.6)
40.9 (37.1 - 44.6)

968
82
465
154
79

51.4 (50.9 - 51.9)

51.8 (50.2 - 53.4)
50.6 (49.8 - 51.4)
48.4 (47.0 - 49.8)
46.1 (43.5 - 48.8)

188
418
155

49.4 (48.2 - 50.7)

51.7 (50.9 - 52.5)
47.8 (46.1 - 49.5)

Gross household income, pa

<5000
5000 14999
15000 29999
30000+
Terminal Education Age
<16
16 18
19+
Chief Income Earner Social Grade
A
B
C1
C2
D
E

55
198
180
173

48.3 (46.0 - 50.6)

48.3 (46.9 - 49.7)
52.3 (51.1 - 53.5)
50.6 (49.3 - 51.8)

228
355
181

50.7 (49.5 - 51.9)

49.2 (48.2 - 50.2)
51.8 (50.7 - 52.9)

38
84
217
193
101
124

52.7 (49.5 - 55.9)

50.68 (48.8 - 52.5)
51.5 (50.5 - 52.6)
51.0 (49.8 - 52.2)
49.5 (47.7 - 51.3)
46.8 (45.0 - 48.7)

95% CI = 95% confidence interval of the mean

In this large dataset small differences reach statistical significance (meaning that
the differences are likely to reflect real differences in the population). Significant
differences in mental well-being were found for each of the five categories of self
perceived health status, ranging from very good to very poor. For tenure, those
who rent were found to have significantly lower mental well-being scores from
those who own outright and own with a mortgage. Those who were unemployed
had significantly lower mental well-being scores than those who were in work or
studying, although no significant differences were found between those who were
retired compared to each of the other 4 employment categories. For marital
status, those who were married or living as a couple had significantly higher
mental well-being then those who were categorised as single or as
widowed/divorced/separated. No real pattern was found for mental well-being
with respect to gross household income per annum or terminal education age.
There were no significant differences found either for chief income earner social
grade, gender or age, although there appears to be a trend towards lower mental
well-being for lower social grades and a U-shaped relationship for age.
These are the first results for WEMWBS and larger surveys are required before
population norms are fully established. The availability of data on WEMWBS
from, for example, the Scottish Health Survey (n = 6,000) from 2008 will help in
this respect.

14

8. Using WEMWBS
WEMWBS is free to use but permission needs to be sought. Further information
is included in Appendix i.
Data Collection
To date, WEMWBS has been administered in a self-completion format. This has
been either via CASI (computer assisted self interviewing) whereby respondents
are invited to enter their responses directly into the CAPI (computer assisted
personal interview) machine (Well? survey and HEPS) or by the self-completion
of paper formats of the scale (student samples and focus groups). WEMWBS can
be assumed to be robust using either of these methods.
WEMWBS has not been tested in interview situations where an interviewer reads
out the items to respondents and fills in their responses for them. We do not
therefore know if WEMWBS is robust in these situations.
Scoring
Each of the 14 item responses in WEMWBS are scored from 1 (none of the time)
to 5 (all of the time) and a total scale score is calculated by summing the 14
individual item scores (Table 5). The minimum score is 14 and the maximum is
70.
Table 5: Example: Scoring of WEMWBS - with responses highlighted in green
Statements
Ive been feeling optimistic
about the future
Ive been feeling useful
Ive been feeling relaxed
Ive been feeling interested
in other people
Ive had energy to spare
Ive been dealing with
problems well
Ive been thinking clearly
Ive been feeling good about
myself
Ive been feeling close to
other people
Ive been feeling confident
Ive been able to make up
my own mind about things
Ive been feeling loved
Ive been interested in new
things
Ive been feeling cheerful
Scores

None of Rarely
the time

Some of
the Time

Often

All of the
time

1
1

2
2

3
3

4
4

5
5

1
0

2
0

3
4 x 3 = 12

4
4 x 4 = 16

5
6 x 5 = 30

Total Score = 0 + 0 + 12 + 16 +30 = 58

15

Presenting the results

WEMWBS results should be presented as a mean score for the population of
interest with either a standard deviation or 95% confidence interval. The latter
both provide a measure of variance of the scores in the population studied (either
as a whole or for sub-groups within it). The range of scores within a sample can
also be presented. Scores will vary between 14 and 70.
Interpreting the results
Table 4 (page 14) shows that the average population mean is around 51 and that
this varies according to the population group studied. The mean score for the
population under study can be compared with these provisional population norms
to assess whether the level of mental well-being is above or below this level.
Differences between the scores of different groups or between the scores of the
same group of people at two points in time, for example, before and after an
intervention, need to be tested statistically using students t-Test or equivalent to
assess how likely the differences are to have arisen by chance. At any given
level of difference results are more likely to be significant if the groups being
compared are large and less likely if the groups are small. A sample size
calculation needs to be carried out to assess how big a group should be to show
statistical significance of a specific difference. Table 6 gives examples of this for
different sample sizes using WEMWBS data from the HEPS and Well? surveys
combined.
Table 6: Examples of sample size required
Population size
10,000
50,000
100,000

Difference in WEMWBS scores between two groups

1 points
1082
1184
1199

2 points
294
301
302

3 points
133
134
135

5 points
48
48
49

Sample size (per group) based on difference in mean scores of two groups using a power of 0.8,
a significance level of 0.05 and population sample combined HEPS (Wave 12) and Well? 2006
datasets (n = 1,749).

If groups within the sample are to be compared, then the sample size calculation
needs to be based on these groups, for example, men separately from women,
and not on the total sample size ie the men plus women.
Dealing with missing data
For the WEMWBS validation, HEPS and Well? responders were deleted if they
were not full-responders (ie they did not answer all items of WEMWBS). This
harsh method was appropriate as the vast majority of responders were fullresponders and thus loss of sample size was minimal. However, it may be too
harsh an approach to adopt in other surveys.
Views differ on how to deal with missing data and none of the possible methods
have been assessed for WEMWBS. The problem of missing data in multi-item
16

scales is curiously under-discussed in the methodological literature. Some

researchers use estimation to '
fill in'missing values, thus retaining their original
sample size. The following are noted as alternative methods, to deleting
respondents who are not full-responders, that have been suggested in literature:

calculating the mean value of responses to items that a respondent has

answered, and then using that mean score as the score for those questions
which that respondent did not answer.

using the midpoint of the range of possible responses

using the mean response for the particular item from all respondents

However, using estimations to fill in missing values should only be done in

situations where at least a certain proportion of items are answered. If less than
this proportion has been answered the respondent'
s score should be set to
missing. Researchers do not agree on what the proportion should be. For
WEMWBS it can be anticipated that estimations for more than three missing
items is unlikely to be robust. In such cases, the WEMWBS score should
therefore not be calculated and should be set as missing. It will also be important
to check the '
randomness'of the missing data to ensure that certain items are not
being systematically missed. Overall, however, when dealing with missing
WEMWBS data it is important to note that the effect of using estimations for
WEMWBS scores has not been tested.
Using WEMWBS in individuals
WEMWBS provides robust results for populations and groups. It has not yet
been validated for monitoring mental well-being in individuals.

17

9. Current usage of WEMWBS

WEMWBS is currently being used in numerous surveys and intervention studies,
for example:
Surveys
Scottish Health Survey from 2008
Scottish Prison Service 2007 Annual Prison Survey
British Social Attitudes Survey 2007
HEPS Spring wave (wave 13) 2007 (8-item version in Autumn wave (wave
14) 2007)
NHS Grampian population surveys
A large population survey in Iceland
National Childhood Development Study 2008 sweep
Under consideration for the Scottish Household Survey 2009
Under consideration for the UK Household Longitudinal study 2009
Under consideration for the Health Survey for England
Local evaluations
Evaluations of local Arts on Prescription Services
Assessments of social prescribing projects
Evaluating the impact of parenting programmes on parents'mental wellbeing in the Parenting Interventions Evaluation of Pathfinders projects
Evaluation of Lottery funded projects in England
Monitoring interventions
Monitoring mental well-being among patients attending psychiatric day
hospital
Occupation therapy interventions at a day hospital
WEMWBS is also being used as a national indicator in the:
Scottish Governments Strategic Outcome Indicators 2008 for the Scottish
Governments performance framework to monitor the spending review
(Scottish Health Survey data from 2008) (Scottish Government, 2007)
Scottish mental health and well-being indicator set (Scottish Health Survey
data from 2008)
And is being considered by the Department for Environment, Food and Rural
Affairs (DEFRS) for one of its indicators of well-being, specifically positive mental
health, in its sustainable development indicators (Defra 2007).

18

10. Further validation research on WEMWBS

Although the basic establishment of WEMWBS is now completed, further
research is being undertake in a number of areas as indicated below.
The scaling properties of WEMWBS
Data from the HEPS (wave 12) 2006 and Well? 2006 population surveys are
being used to establish the extent to which WEMWBS fits the Rasch model. This
is a statistical procedure used to determine how the intervals in an ordinal scale
relate to one another. This analysis permits an answer to the following question:
Is a score of 60 twice as good as a score of 30?. A good fit to the Rasch model
indicates that the scale has good scaling properties. This means that a mean
score of, for example, 44 can be taken to be twice as good as a mean score of
22. Rasch analysis can also be used to determine potential item redundancy in a
scale and to assess whether a reduction in the number of scale items may be
appropriate.
Initial indications suggest that it may be possible to develop a shortened (seven
or eight item) version of WEMWBS which has more robust scaling properties
than the full scale.
WEMWBSs sensitivity to change
Several studies are ongoing to assess the sensitivity to change of WEMWBS. For
example, WEMWBS is being used to audit the mental well-being of people
attending a psychiatric day hospital. Changes on WEMWBS will be compared
with patient and clinical assessments of mental health at the beginning and end
of each patients admission. If WEMWBS scores change significantly in the
direction indicated by clinical assessment, sensitivity to change will be confirmed.
WEMWBS is also being used to assess mental well-being in parents attending a
range of parenting programmes in the UK before they embark on the programme
and at the end of the programme. These programmes have previously been
shown to have a positive impact on parents mental health and demonstration of
changes in WEMWBS scores will confirm that the measure is sensitive to
change.
Validation of WEMWBS with secondary school children aged 13 to 15 years
Research began in March 2008 to establish whether WEMWBS can be used to
assess the overall mental well-being of children of secondary school age (13 to
15 years of age). This is the Warwick-Edinburgh Mental Well-being Scale
Acceptability and Validation in English and Scottish Secondary School Students
project (The WAVES Project) being undertaken by Warwick and Edinburgh
Universities. This will report October 2011.
Other research required
Other research still required includes assessing the extent to which it is
appropriate to use WEMWBS to assess mental well-being among different ethnic

19

minority populations in the UK, and other cross-cultural validation for use of
WEMWBS in countries other than the UK.
Further ahead
As understanding of mental well-being develops over the next decade, it is likely
that measurement scales will also need to evolve. Whilst WEMWBS fulfils
criteria for monitoring mental well-being at present and represents a very
significant step forward in terms of other currently available measures, it is likely
that it will need to undergo further development in the future.
This manual will be updated as results of the continuing validation of WEMWBS
are known.

20

Bibliography
Braunholtz S, Davidson S, Myant K and O'
Connor R (2007). Well? What do you
think? (2006): The third national Scottish survey of public attitudes to mental
health, mental wellbeing and mental health problems. Scottish Executive:
Edinburgh. To access the dataset contact Angela Hallam
Angela.Hallam@scotland.gsi.gov.uk
Defra (2007). Sustainable development Indicators in your pocket 2007 An
update of the UK Government Strategy indicators. Department for
Environment, Food and Rural Affairs: London.
Kammann R, and Flett R (1983). Affectometer 2: A scale to measure current
level of general happiness. Australian Journal of Psychology, 35 (2), 259265.
Gosling R, Bassett C, Gilby N, Angle H, and Catto S (2008). Health Education
Population Survey: Update from 2006 survey. NHS Health Scotland:
Glasgow. [database on the Internet]. Colchester: UK Data Archive; 2007.
Available from: http://www.data-archive.sc.uk
Nunnally JC (1978). Psychometric Theory. Second Edition. McGraw-Hill: London.
Ryan RM, and Deci EL (2001). On happiness and human potential: a review of
research on hedonic and eudaimonic well-being. Annual Review of
Psychology, 52, 141-166.
Scottish Government (2007). Scottish budget spending review 2007. The
Scottish Government: Edinburgh.
Tennant R, Fishwick F, Platt S, Joseph S., and Stewart-Brown S (2006).
Monitoring Positive Mental Health in Scotland: Validating the Affectometer 2
Scale and Developing the Warwick-Edinburgh Mental Well-being Scale for
the UK. NHS Health Scotland: Glasgow.
Tennant R, Hiller L, Fishwick R, Platt S, Joseph S, Weich S, Parkinson J, Secker
S, and Stewart-Brown S (2007). The Warwick-Edinburgh Mental Well-being
Scale (WEMWBS): development and UK validation. Health & Quality of Life
Outcomes, 5 (63), doi:10.1186/1477-7525-5-63.
Tennant R, Joseph S, and Stewart-Brown S (2007). The Affectometer 2: a
measure of positive mental health in UK populations. Quality of Life
Research, 16 (4), 687-695.

21

Appendix i
Conditions of Using WEMWBS
We welcome the use of WEMWBS. It is free to use but is copyrighted to NHS
Health Scotland and the Universities of Warwick and Edinburgh. Permission is
required for use. Dr Kulsum Janmohamed K.Janmohamed@Warwick.ac.uk,
working with Professor Sarah Stewart-Brown at the University of Warwick, is
maintaining a register of use and is the person to contact when seeking such
permission.
When you seek permission for use you should indicate how you are planning to
use WEMWBS. We ask that after use you feed back to Dr Janmohamed on how
WEMWBS has performed. Dr Janmohamed is also the person to contact should
you have more questions regarding the scale and its use.
If the scale is reproduced, it must include the copyright statement which appears
below it and no changes to its wording, response categories or layout must be
made.
Any report regarding use of WEMWBS should include the following text:
"The Warwick-Edinburgh Mental Well-being Scale was funded by the Scottish
Government National Programme for Improving Mental Health and Well-being,
commissioned by NHS Health Scotland, developed by the University of Warwick
and the University of Edinburgh, and is jointly owned by NHS Health Scotland,
the University of Warwick and the University of Edinburgh."

22

The Warwick-Edinburgh Mental Well-being Scale (WEMWBS)

Below are some statements about feelings and thoughts.
Please tick the box that best describes your experience of each over the last 2 weeks
STATEMENTS

Rarely Some of
the time

Often All of
the
time

Ive been feeling optimistic 1

about the future

Ive been feeling useful

Ive been feeling relaxed

Ive been feeling

interested in other people

Ive had energy to spare

Ive been dealing with

problems well
Ive been thinking clearly

Ive been feeling confident

Ive been able to make up

my own mind about things

Ive been feeling loved

Ive been interested in

new things
Ive been feeling cheerful

Ive been feeling good

about myself
Ive been feeling close to
other people

None of
the time

Warwick Edinburgh Mental Well-Being Scale (WEMWBS)

NHS Health Scotland, University of Warwick and University of Edinburgh,
2006, all rights reserved.

23

Appendix ii
Description of scales used to assess the construct validity of WEMWBS
Scales of affect/feelings
WHO-Five Well-being Index (WHO-5)

Short Depression Happiness Scale

(SDHS)
Positive and Negative Affect Scale
(PANAS):
Positive subscale (PANAS P)
Negative subscale (PANAS N)

Scales of subjective well-being

Global Life Satisfaction Scale (GLSS)

Satisfaction with Life Scale (SWLS)

Psychological functioning
Scales of Psychological Wellbeing
(SPW)

Emotional Intelligence
Emotional Intelligence Scale (EIS)

Five item scale of statements covering

key mental affect states (e.g. I felt
cheerful, calm, vigorous and interested)
with 5 response categories. All items
positively worded.
Six item scale with 4 response
categories focused on affect with
balanced positive and negative items
(e.g. I felt happy, I felt cheerless)
Twenty item scale with 5 response
categories comprising a list of positive
and negative adjectives covering a
wider variety of feelings than is usual in
mental health scales (e.g. ashamed,
attentive, proud, guilty, and excited).
Single item scale with 4 point response
category. On the whole are you
satisfied with your life? Most
commonly used measure of subjective
well-being.
Five item scale with 7 response
categories. Items cover positive
statements e.g. in most ways my life is
close to ideal. The prototype measure
of well-being.
Fifty four item scale with 6 response
categories assessing psychological
functioning with subscales measuring
autonomy, self acceptance,
environmental mastery, purpose in life,
personal growth and positive relations
with others.
Thirty three item scale with 5 response
categories. Consists of statements
covering appraisal, expression, and
regulation of emotion in self and others,
and the utilisation of emotions in
24

problem solving.
Psychiatric Morbidity
General Health Questionnaire 12 (GHQ Twelve-item scale with 4 response
12)
categories. A well-established
screening instrument designed to
detect possible psychiatric morbidity in
the general population. Respondents
are asked to respond to questions
relating to their recent experience of
anxiety, self-confidence ability to
concentrate, decision-making capacity,
enjoyment of day-today activities, sleep
disturbance and stress etc.
General Health
EQ-5D thermometer

Response Bias
Balanced Inventory of Desirable
Responding (BIDR)

A measure of health in general where

respondents rate their overall health
(physical and mental) on a 0-100 scale.
Responses to this scale tend to reflect
physical more than mental health.
Forty-item scale, split into two subscales. The first sub-scale measures
self-deception (SD) (the tendency to
exaggerate certain responses or
behaviours) and the second sub-scale
measures impression management
(IM) (the tendency to over-report
desirable behaviours and under-report
undesirable behaviours).

25

Appendix iii
WEMWBS median scores across demographic groups: population sample
Combined HEPS (Wave 12) and Well? 2006 Datasets
Variable
Total
Sex

Median (95% CI)

783
966

52 (51-52)
51 (50-52)

<0.05

176
245
353
306
334
274
61

53 (52-53)
51 (50-53)
51 (49-52)
50 (49-51)
52 (51-53)
52 (51-54)
51 (49-54)

<0.01KW

523
705
519

52 (52-53)
52 (51-52)
50 (49-51)

<0.01KW

563
753
319
84
29

54 (54-55)
51 (51-52)
47 (46-49)
44 (40-46)
41 (36-47)

<0.01J

968
82
465
154
79

52 (51-52)
52 (50-54)
51 (50-52)
49 (47-51)
46 (43-50)

<0.01KW

188
418
155

51 (49-53)
52 (51-53)
49 (46-51)

<0.01KW

55
198
180
173

48 (44-53)
49 (47-51)
53 (51-54)
51 (49-53)

<0.01J

228
355
181

52 (50-53)
50 (49-51)
53 (51-54)

<0.05KW

38
84
217
193

55 (51-57)
50 (48-53)
51 (50-53)
53 (51-54)

<0.01J

1749

Male
Female
Age in years
16 24
25 34
35 44
45 54
55 64
65 74
75+
Tenure
Own outright
Own with a mortgage
Rent
Self-perceived health status
Very good
Good
Fair
Poor
Very poor
Employment Status ^
In work
Student
Retired
Unemployed
Other
Marital Status *
Single
Married/Living as couple
Widowed/Divorced/Separated
Gross household income, pa *
<5000
5000 14999
15000 29999
30000+
Terminal Education Age *
<16
16 18
19+
Chief Income Earner Social Grade *
A
B
C1
C2

26

51 (51-52)

D
E

101
124

50 (47-52)
47 (44-51)

* Tests conducted on a reduced set of individuals. Variable only recorded in the HEPS survey.
95% CI = 95% confidence interval of the median
KW
= p-value generated from a Kruskal-Wallis test.
J
= p-value generated from a Jonckheeres tests for ordered alternatives.
^ = test conducted excluding the Other category

27