The impact of social isolation on the health status
and health-related quality of life of older people
Annie Hawton•Colin Green•Andy P. Dickens•
Suzanne H. Richards•Rod S. Taylor•Rachel Edwards•
Colin J. Greaves•John L. Campbell
Accepted: 5 July 2010/Published online: 25 July 2010
? Springer Science+Business Media B.V. 2010
and older people at risk of social isolation: (1) health status
and health-related quality of life (HRQL); (2) the rela-
tionship between social isolation and health status/HRQL;
(3) the relationship between two alternative measures of
Older people at risk of social isolation (n = 393)
completed the EQ-5D and the SF-12. Multiple regression
analyses were performed to examine the relationship
between levels of social isolation and health status/HRQL,
controlling for demographic/clinical characteristics. The
agreement between EQ-5D and SF-6D (SF-12) scores was
explored using descriptive psychometric techniques.
Health status and health state values were much
lower than UK general population age-matched norms.
After controlling for depression, physical co-morbidities,
age, gender, living alone status, employment and accom-
modation, social isolation was significantly associated, to
a degree that was clinically relevant, with EQ-5D DSI,
SF-6D (SF-12) and SF-12 MCS scores. The potential for
ceiling effects on the EQ-5D with this population was
To investigate for socially isolated older people,
isolation may have on the health and well-being of older
people. The potential HRQL gains from addressing social
isolation may be considerable, with those at risk of social
isolation also a key target group.
This work highlights the burden that social
Health status ? Quality of life
Aged ? Frail elderly ? Social isolation ?
The impact of social isolation, generally defined as the
absence of contact with other people , is a key factor
when exploring the influence of the social environment on
people’s quality of life [2, 3]. Social isolation is consis-
tently related to a negative impact on health and well-being
[2, 4–6], and there is growing evidence of the effects of
social isolation in older people [1, 7, 8]. As the proportion
of older people in the population increases , more are at
risk of social isolation . Recent UK research suggests
that between 11 and 17% of those aged 65 years or over are
socially isolated . In people aged 50 years or over,
those living alone are most likely to experience long
periods of detachment from society (13%) .
Strategies for promoting ‘quality of ageing’ for older
people have become a major component of UK Govern-
ment policy [9, 11, 12]. There is an increasing awareness of
the challenges of an ageing society, and the impact of
social isolation in older people is becoming one of the most
important social, political and health care priorities. Yet,
research to date has been mainly descriptive, and the
independent effect of social isolation on health-related
quality-of-life (HRQL) has not been considered [2, 4–6].
There is a need for further research to explore the
A. Hawton (&) ? C. Green ? R. S. Taylor
Institute of Health Service Research, Peninsula College
of Medicine & Dentistry, University of Exeter,
Veysey Building, Salmon Pool Lane, Exeter,
Devon EX2 4SG, UK
A. P. Dickens ? S. H. Richards ? R. Edwards ?
C. J. Greaves ? J. L. Campbell
Primary Care, Peninsula College of Medicine & Dentistry,
University of Exeter, Smeall Building, St Luke’s Campus,
Magdalen Road, Exeter, Devon EX1 2LU, UK
Qual Life Res (2011) 20:57–67
relationship between social isolation and HRQL and to
inform the development and evaluation of interventions to
address social isolation.
Research has considered the relationship between social
isolation and dimensions of health status, such as self-rated
health, poor physical health, restricted mobility and limi-
tations in activities of daily living, all of which have been
shown to be associated with social isolation [13–15].
However, as these relationships have been investigated in
univariate models, we do not know the extent to which
these associations are mediated by other characteristics
(e.g. age, gender, depressive symptoms and physical mor-
bidity). Victor et al.  have explored the independent
relationship between health status and social isolation for
older people, but only in terms of how health status inde-
pendently relates to social isolation, not in terms of how
social isolation (against a backdrop of other characteristics)
is associated with health status/HRQL. This highlights the
need for further research to explore these relationships
multi-dimensionally, in the context of other demographic
and clinical factors.
Interventions developed to address social isolation will
require evaluation to assess their effectiveness and cost-
The design of appropriate methods for evaluation will be
aided by an understanding of how measures of HRQL per-
form in relation to the construct of social isolation. For
example, measures need to be sufficiently discriminatory to
social isolation, and sufficiently sensitive to respond to
changes in social isolation (e.g. before and after an
Investigating social isolation is a challenge, as it has been
[1, 7, 8]. It has been conceptualised as comprising both
structural and functional elements . Structural refers to
objective aspects of social isolation, such as the size of
people’s social networks or their frequency of contact, and
functional relates to people’s perceptions of the quality of
their interactions. Although the construct is multifaceted, at
its core is an absence of contact from other people [1, 16]
and a number of social ‘indices’ have been developed
[16–19] based on assessments of social contact. No defini-
tive index has emerged, but the measures are characterised
by self-report frequency of contact with family, friends and
neighbours. This measurement approach has been used in a
recent UK survey of a nationally representative sample of
people aged 65 years or over living in the community .
Social isolation was objectively defined in two ways: (1)
those with less than weekly direct contact with family,
friends or neighbours and; (2) those with less than monthly
direct contact with family, friends or neighbours. We
used this approach in the present study to describe the
relationship between social isolation and measures of health
status and HRQL for older people in the context of key
demographic/clinical characteristics. We estimate health
state values for social isolation and compare them with UK
population age-matched norms to explore potential differ-
ences. We also compare findings from an analysis of two of
the most commonly used measures of health status/HRQL,
the EQ-5D and SF-6D (SF-12), for socially isolated older
people and those at risk of social isolation.
Data are from the Devon Ageing and Quality of Life
(DAQoL) study , which explored the effectiveness and
cost-effectiveness of a community mentoring intervention
for socially isolated older people. During 2007–2008,
participants were recruited via community mentoring ser-
vice providers (intervention group) or through screening
surveys sent via general practices in geographical areas
where the intervention was not available (control group).
All participants were aged 50 years or over and were
deemed to be socially isolated or at risk of social isolation
based on service provider assessment (intervention group)
or self-report data (control group). Community mentoring
teams worked to identify older people who were, or were at
risk of becoming, socially isolated, and potentially eligible
clients were identified via referrals from health and social
care professionals, family and friends, or self-referral.
Participants were also identified as at risk of social isola-
tion through a survey via general practice. They were
screened for characteristics of social isolation based on
social activity scores [21, 22] and mental health status .
Participants were visited at home by a researcher for a
baseline assessment. Informed consent was obtained, and
participants completed a battery of self-report measures.
The baseline assessment included information on the
‘Absence of contact with other people’ (structural social
isolation) was self-assessed using an item from the Rand
58 Qual Life Res (2011) 20:57–67
Social Health Battery : ‘How many times a year do you
get together with friends and relatives e.g. going out
together or visiting each other’s homes?’. This is rated on a
seven point scale from ‘every day’ to ‘less than 5 times a
year’. Victor et al.  defined social isolation in two
ways: (1) those with less than weekly direct contact with
family, friends or neighbours and; (2) those with less than
monthly direct contact with family, friends or neighbours.
We applied these definitions to this data in the current
analyses to represent (1) ‘social isolation’ and (2) ‘severe
social isolation’, respectively. Those who did not meet
these criteria (i.e. had weekly or more frequent contact with
family, friends or neighbours) were categorised as being ‘at
risk of social isolation’ on the basis that they had been
identified by community mentoring health/social care
professionals or via a general practice screening survey as
socially isolated or at risk of social isolation.
Two additional questions from the Social Health Battery
 addressed social activity: ‘How many close friends and
family do you have?’ and ‘About how many clubs/groups/
organisations do you belong to?’. Six items from the
Medical Outcomes Study Social Support Survey  were
used to measure the availability of aspects of functional
Participants completed version 2 of the SF-12 . This
gives mental health component (MCS) and physical health
component (PCS) summary scores. Higher scores indicate
better health status, with scores ranging from 0 to 100.
Scores are standardised to a mean of 50 and a standard
deviation of 10. There is a general consensus that changes
of 2–3 points in SF-12 scores are clinically meaningful .
Respondents also completed the EQ-5D  that com-
prises five dimensions of health (mobility, self-care, usual
activities, pain/discomfort and anxiety/depression) with
each dimension having three response levels (1 no prob-
lems, 2 moderate problems and 3 severe problems).
Health-related quality of life
Participant responses to the EQ-5D were converted to a
weighted health state index (derived single index, DSI)
using valuations (tariffs) elicited from a general population
sample for the UK . Health state tariff values for the
EQ-5D DSI range from 1.00 for the best health state to
-0.594 for the worst health state. (A score of 0 is equiv-
alent to death, and scores less than 0 are worse than death).
Participant responses to the SF-12 were converted to the
SF-6D (SF-12), a preference-based measure of health using
tariffs elicited from a representative sample of the UK
general population. Scores on the SF-6D (SF-12) can range
from 0.29 to 1.0, where 0.29 indicates the worst health
state, and 1.0 the best health state .
Depression was assessed using the Geriatric Depression
Scale-10 [27, 28] which includes 10 yes/no response items.
levels of depressive symptoms. A score of four or more has
potential diagnosis of clinical depression .
Participants rated (yes/no) whether they had experienced
nine common physical health conditions (angina, arthritis,
cancer, diabetes, heart failure, high blood pressure, prob-
lems with sight or hearing, stroke, chronic respiratory
conditions) over the previous 6 months . A co-mor-
bidity index was then calculated by summing the number
of co-morbid conditions reported .
Age (years), gender (male/female), living alone status (yes/
no), accommodation type (home owner/not home owner)
and employment status (employed/not employed) were
Ethics approval forthe DAQoL study was granted by Devon
& Torbay NHS Research Ethics Committee (REC number
07/Q2102/9), and research governance approval given by
the relevant NHS and social care Trusts. All participants
gave their informed consent prior to inclusion in the study.
Health status and HRQL were described, and EQ-5D data
considered alongside UK age population norms [32, 33].
(UK norms are not currently available for the SF-12 v2 and
the SF-6D [SF-12]).
SF-12 (MCS and PCS), EQ-5D DSI and SF-6D (SF-12)
scores were compared across the three groups (‘at risk of
social isolation’, ‘socially isolated’ and ‘severely socially
isolated’) using one-way, unrelated ANOVAs. Differences
between each of the groups were determined using a priori-
specified contrasts and the t statistic.
Forward stepwise regression analyses were conducted to
estimate the strength of the association between social
isolation and health status/HRQL in relation to other key
demographic and clinical features. SF-12 MCS, SF-12 PCS,
Qual Life Res (2011) 20:57–6759
EQ-5D DSI and SF-6D (SF-12) scores were the dependent
variables in four separate analyses. Social isolation,
depression, age, gender, living alone status, accommodation
type, employment status and number of physical co-mor-
bidities were included as independent variables.
The regression method used automatically selected the
independent variable that had the largest partial correlation
with the dependent variable, controlling for variables
already in the model. A variable’s partial regression coef-
ficient had also to be significant at 0.05 and 0.01% of its
variance had to be independent of other independent vari-
ables (tolerance value) in order to be selected. Multi-col-
linearity was assessed by checking correlations between
the independent variables prior to inclusion in the regres-
sion analyses, and determining variable tolerances fol-
lowing the analyses. The adequacy of model fit was
reviewed by examining normality plots and scatterplots of
predicted versus residual values.
The comparative analysis of how the EQ-5D (and EQ-
5D DSI) and SF-6D (SF-12) operate for this population
draws on the work of Brazier et al.  and Grieve et al.
. The relationships between dimension scores were
assessed using Spearman’s rank correlations, and the
potential for ceiling or floor effects of the measures was
considered by exploring the spread of responses across the
measure levels. The agreement between the EQ-5D DSI
and SF-6D (SF-12) index scores was estimated using the
intra-class correlation coefficient (ICC) . The ICC can
fall between -1 and 1, where 1 indicates absolute agree-
ment, 0 no agreement and -1 absolute negative agreement.
All data analysis was conducted using SPSS 15.0, and
P values less than 0.05 were considered statistically
Ninety-four (24%) participants were classified as ‘socially
isolated’ and 67 (17%) as ‘severely socially isolated’. The
remaining 232 (59%) were classified as being ‘at risk of
social isolation’. (Two participants did not report their
degree of social isolation and were excluded from sub-
Demographic, clinical and social health characteristics of
the sample are given in Tables 1 and 2. There was a neg-
ative association betweenthe severityof social isolationand
both social network size and functional social support.
Health status and HRQL of socially isolated older
The percentages of respondents reporting any problem on
each of the EQ-5D dimensions are given in Table 3 by age,
alongside age-matched UK population norms . With
just one exception, for each of the dimensions, by each of
the age categories, the percentage reporting any problem
was greater for the three groups (‘at risk of social isola-
tion’, ‘socially isolated’ and ‘severely socially isolated’)
than for UK population norms.
(SF-12) scores are given in Table 4. All the SF-12 scores are
below the standardised mean of 50, and those who were
‘severely socially isolated’ reported more negative physical
isolated’, as assessed by both the EQ-5D DSI and the SF-6D
(SF-12). Statistically significant differences were found
between the three social isolation groups on the SF-12 MCS,
PCS (Table 4). The scores of the ‘severely socially isolated’
group were statistically significantly poorer than those of the
‘at risk’ group and those of the ‘socially isolated’ group for
each of the health status and HRQL measures.
Mean (sd) EQ-5D DSI scores were compared with UK
population norms  (Table 5). For each of the age cat-
egories, the scores of the ‘severely socially isolated’ indi-
viduals were lower, indicating poorer HRQL.
Association between social isolation and older people’s
health status/HRQL, in relation to demographic
and clinical features
Table 6 gives the findings of the final steps of the four
stepwise regression analyses. Each of the models signifi-
cantly accounted for the variance in the dependent variable
(all at P\0.01), with R2adjusted values from 0.34 to 0.41.
When the effects of depression, physical co-morbidity, age,
gender, living alone, accommodation type and employment
status were controlled for, social isolation was still signifi-
cantly (at P\0.01) independently associated with SF-12
MCS, SF-6D (SF-12) and EQ-5D DSI, but not SF-12 PCS,
scores. Social isolation was most strongly related to SF-12
MCS scores (b -0.16, P\0.01), followed by the EQ-5D
Of all the demographic/clinical variables, social isola-
tion had the second strongest association with SF-12 MCS
scores, after depression. Respondents who were ‘severely
socially isolated’ had SF-12 MCS scores an average of 4.73
points lower than respondents who were not ‘severely
socially isolated’. Given the general consensus that chan-
ges of 2–3 points on the SF-12 are clinically relevant ,
this difference in scores between those who were and were
not ‘severely socially isolated’ appears clinically relevant.
model of SF-6D (SF-12) scores. Respondents reporting
60 Qual Life Res (2011) 20:57–67
of 0.04 points lower than those who were not ‘severely
socially isolated’. A minimal meaningful difference on the
SF-6D (SF-12) of 0.041 is suggested , implying that the
difference found here is clinically important.
‘Severely socially isolated’ individuals had EQ-5D DSI
scores an average of 0.09 points less than those who were
not ‘severely socially isolated’. Given that the minimally
important difference on the EQ-5D DSI is thought to be
between 0.03 [39, 40] and 0.075 , this difference would
be viewed as meaningful.
The consistency of the model findings in the extent
to which ‘severe social isolation’ is associated with the
EQ-5D DSI (b = -0.11) and SF-6D (SF-12) (b = -0.10)
scores implies that these measures relate to the construct of
social isolation in a similar way.
The relationship between the EQ-5D and the SF-6D
The Spearman rank correlation coefficients between scores
on the EQ-5D and SF-6D (SF-12) dimensions are given in
Table 7. Brazier et al.  have suggested instances where
a high correlation might be expected because the dimen-
sions appear to measure similar constructs. These are
indicated in the table in bold. All correlations were sig-
nificant at P\0.01. The high correlations between Pain/
Discomfort and Pain (0.71) and Mobility and Physical
functioning (0.63) and Anxiety/Depression and Mental
health (0.62) support the argument that the tools are
measuring health status in similar ways with regards to
Respondents were more likely to report health status at
the poorer levels of the SF-6D (SF-12) than they were when
reporting health status against the EQ-5D. For example,
34.9% of respondents rated the worst level of Physical
functioning (SF-6D [SF-12]), but only 0.3% rated the worst
level of Mobility (EQ-5D). The clearest example is that of
Role limitation (SF-6D [SF-12]) where 52.5% of respon-
dents marked the worst level, when compared to 7.9%
rating the worst level of Usual activities (EQ-5D). These
responses suggest that problems on the EQ-5D are consid-
ered to be worse than problems on the SF-6D (SF-12).
Table 1 Demographic and
clinical features of participants
Demographic/clinical characteristic‘At risk of social
isolation’ (n = 232)
(n = 94)
isolated’ (n = 67)
Age, mean (sd) years71.5 (11.8) 69.7 (12.0)69.8 (12.1)
Gender, n (%) male 64 (27.6)41 (43.6) 33 (49.3)
Lives alone, n (%) yes 115 (49.6)36 (38.3) 34 (50.7)
Accommodation type, n (%)
Home owner 175 (75.4)76 (80.9) 42 (62.7)
Rented/council 54 (23.3) 17 (18.1)25 (37.3)
Residential home3 (1.3) 1 (1.1)0
Employment status, n (%)
Employed30 (12.9) 19 (20.2)8 (11.9)
Unemployed4 (1.7) 2 (2.1)3 (4.5)
Long-term sick or disabled24 (10.3) 10 (10.6)9 (13.4)
Retired174 (75.0) 63 (67.0)47 (70.1)
Physical co-morbidities, n (%)
Angina 30 (12.9)12 (12.8)2 (3.0)
Arthritis 118 (50.9) 30 (31.9) 34 (50.7)
Cancer 12 (5.2)6 (6.4) 7 (10.4)
Diabetes24 (10.3)11 (11.7) 10 (14.9)
Heart failure14 (6.0) 3 (3.2) 4 (6.0)
High blood pressure112 (48.5) 36 (38.3)28 (41.8)
Sight/hearing problems110 (47.4) 41 (46.3) 35 (52.2)
Stroke 10 (4.3)2 (2.1)3 (4.5)
Chronic respiratory conditions 39 (16.8) 14 (14.9) 15 (22.4)
Number of physical co-morbidities,
2.0 (1.3) 1.6 (1.3)2.1 (1.3)
Depression (GDS-10), n (%)
‘Clinically depressed’85 (37.6)32 (34.4)43 (65.2)
Not ‘clinically depressed’ 141 (62.4)61 (65.6)23 (34.8)
Qual Life Res (2011) 20:57–6761
Table 2 Social health characteristics of participants
No. of close friends and family,
‘At risk’6 (4, 10)
‘Socially isolated’4 (3, 6.3)
‘Severely socially isolated’2 (1, 4.3)
No. of clubs/groups/organisations
belong to, median (IQR)
‘At risk’1 (0, 2)
‘Socially isolated’ 1 (0, 2)
‘Severely socially isolated’ 1 (1, 1)
A little of
Have a good time with ‘At risk’24 (10.4%)42 (18.2%) 67 (29.0%)64 (27.7%) 34 (14.7%)
‘Socially isolated’17 (18.1%) 22 (23.4%) 27 (28.7%) 13 (13.8%) 15 (16.0%)
‘Severely socially isolated’28 (41.8%) 22 (32.8%)7 (10.4%) 4 (6.0%)6 (9.0%)
Get together with for relaxation‘At risk’ 27 (11.6%)41 (17.7%)68 (29.3%) 59 (25.4%)37 (15.9%)
‘Socially isolated’ 9 (9.6%)26 (27.7%) 33 (35.1%)12 (12.8%)14 (14.9%)
‘Severely socially isolated’ 28 (41.8%)18 (26.9%)9 (13.4%) 6 (9.0%)6 (9.0%)
Do something enjoyable with‘At risk’ 22 (9.5%) 41 (17.7%)72 (31.0%)61 (26.3%) 36 (15.5%)
‘Socially isolated’ 7 (7.4%)29 (30.9%)28 (29.8%) 16 (17.0%)14 (14.9%)
‘Severely socially isolated’ 23 (34.3)22 (32.8%) 12 (17.9%)4 (6.0%)6 (9.0%)
Do something with to take mind
‘At risk’26 (11.2%)42 (18.1%)66 (28.4%) 65 (28.0%)33 (14.2%)
‘Socially isolated’12 (12.8%) 28 (29.8%)26 (27.7%) 16 (17.0%)12 (12.8%)
‘Severely socially isolated’ 24 (36.4%) 22 (33.3%)10 (15.2%)4 (6.1%) 6 (9.1%)
Confide in ‘At risk’14 (6.0%) 34 (14.7%) 41 (17.7%)75 (32.3%) 68 (29.3%)
‘Socially isolated’ 6 (6.5%)24 (25.8%) 22 (23.7%)15 (16.1%) 26 (28.0%)
‘Severely socially isolated’ 23 (34.3%)16 (23.9%)9 (13.4%)5 (7.5%) 14 (20.9%)
Turn to for suggestions of how to
deal with a personal problem
‘At risk’14 (6.0%)35 (15.1%) 45 (19.4%) 67 (28.9%)71 (30.6%)
‘Socially isolated’9 (9.6%) 24 (25.5%)20 (21.3%) 16 (17.0%) 25 (26.6%)
‘Severely socially isolated’21 (31.3%) 18 (26.9%) 11 (16.4%)4 (6.0%)13 (19.4%)
Table 3 Percentages of individuals reporting any problem on the EQ-5D dimensions by degree of social isolation, when compared to UK
population age norms 
Social isolation category (%)
50–59Norms 21.9 5.221.9 43.7 27.2
‘At risk’ (n = 48) 45.818.854.2 62.5 54.2
‘Socially isolated’ (n = 25) 36.0 16.0 56.060.0 48.0
‘Severely socially isolated’ (n = 18)64.7 17.652.9 64.7 64.7
‘At risk’ (n = 52)38.59.646.259.646.2
‘Socially isolated’ (n = 21)33.3 19.0 38.152.438.1
‘Severely socially isolated’ (n = 12)66.733.3 66.783.375.0
‘At risk’ (n = 66)77.3 19.765.2 72.7 42.4
‘Socially isolated’ (n = 23)60.926.152.273.9 52.2
‘Severely socially isolated’ (n = 24)83.312.570.875.0 58.3
‘At risk’ (n = 66)71.221.2 126.96.36.199
‘Socially isolated’ (n = 24)70.825.0 52.258.329.2
‘Severely socially isolated’ (n = 13)84.646.276.969.2 84.6
62Qual Life Res (2011) 20:57–67
The EQ-5D had a larger proportion of respondents in
the best health status (level 1) categories than the SF-6D
(SF-12) for all dimensions. One person reported full health
on the SF-6D (SF-12), and 64 (16%) respondents had full
health on the EQ-5D. Of those who reported full health on
the EQ-5D, 40% (n = 25) were ‘severely socially isolated’
or ‘socially isolated’. This may imply that ceiling effects
could be found if the EQ-5D is used to assess changes in
health status following an intervention to address social
isolation, meaning that the EQ-5D might be unable to
distinguish between health states close to full health in this
population of vulnerable older people.
The distribution of responses showed a greater, and more
normally distributed, spread of scores on the SF-6D (SF-12)
when compared to the EQ-5D DSI. The ICC was 0.543
(P\0.01), implying that although there was a ‘moderate’
level of agreement between the tools, there were also dif-
ferences . This relationship is shown in Fig. 1.
Table 4 Mean (sd) health
status and HRQL scores by
degree of social isolation
(n = 232),
(n = 94),
isolated’ (n = 67),
SF-12 MCS47.9 (10.2) 47.1 (10.4)40.0 (11.6)
F = 15.00, df = 2, P\0.01
F = 2.47, df = 2, P = 0.09
F = 8.86, df = 2, P\0.01
F = 9.91, df = 2, P\0.01
SF-12 PCS 39.1 (12.4)40.0 (13.4)35.7 (12.6)
EQ-5D DSI0.65 (0.30)0.69 (0.27)0.50 (0.32)
SF-6D (SF-12)0.67 (0.14) 0.67 (0.12)0.59 (0.12)
Table 5 Mean (sd) EQ-5D DSI scores by degree of social isolation,
when compared to UK population norms 
Social isolation category, mean (SD)EQ-5D DSI
55–64Norms 0.80 (0.26)
‘At risk’ (n = 54) 0.66 (0.32)
‘Socially isolated’ (n = 25)0.68 (0.29)
‘Severely socially isolated’ (n = 17)0.65 (0.30)
65–74 Norms 0.78 (0.26)
‘At risk’ (n = 51)0.67 (0.29)
‘Socially isolated’ (n = 17)0.74 (0.23)
‘Severely socially isolated’ (n = 15)0.49 (0.31)
75? Norms 0.73 (0.27)
‘At risk’ (n = 107)0.66 (0.28)
‘Socially isolated’ (n = 38)0.65 (0.30)
‘Severely socially isolated’ (n = 26) 0.52 (0.31)
Table 6 Final step results of
regression analyses of social
isolation and demographic/
clinical factors on measures of
health status and HRQL
(N = 382)
a to e indicates the ordering of
the independent variables into
the model; – not entered into
* P\0.05; ** P\0.01
SF-12 MCSSF-12 PCS EQ-5D DSISF-6D (SF-12)
Standardised regression coefficients (b)
Living alone status––––
Employment status– 0.22**(c) 0.17**(c)0.15**(d)
Regression coefficients (SE)
Social isolation-4.73 (1.25)–-0.09 (0.03)-0.04 (0.02)
Depression-11.09 (0.96)-7.13 (1.11)-0.23 (0.03)-0.14 (0.01)
Physical co-morbidity–-3.42 (0.43)-0.07 (0.01)-0.03 (0.01)
Age0.10 (0.04)– 0.003 (0.001)0.002 (0.001)
Gender2.53 (0.97)-3.16 (1.13)––
Living alone status––––
Employment status–8.04 (1.59) 0.15 (0.04)0.06 (0.02)
Qual Life Res (2011) 20:57–6763
Despite recognition of the potential impact of social iso-
lation on the quality of life of older people, and the drive
for interventions to tackle it, the influence of social isola-
tion on health status/HRQL is poorly understood . Pre-
vious work has considered these relationships, but often
descriptively, or in terms of what may be related to social
isolation, rather than the likely effects of social isolation
itself [1, 14, 15]. We believe this is the first study that has
estimated the health status (SF-12 and EQ-5D) and the
health state values associated with HRQL (SF-6D [SF-12]
and EQ-5D DSI) for socially isolated older people in the
UK and explored the relationship between social isolation
and health status/HRQL using multivariate analyses across
a number of measures.
The health status and health state values (HRQL) of those
who were ‘socially isolated’, and especially those reporting
‘severe social isolation’, were lower than expected in the
UK general population for people of the same age. Those
‘at risk of social isolation’ also had lower scores than
reported population norms for this age group.
A major finding of the current study is that social iso-
lation was significantly, independently related to health
status and HRQL, even when depression, physical co-
morbidity, age, gender, living alone, employment status
and accommodation type were accounted for. The differ-
ences in scores, from the regression analyses, between
those who were ‘severely socially isolated’ and those who
were not, were statistically significant and clinically
Table 7 Spearman rank
correlations between EQ-5D
and SF-6D (SF-12) dimension
Role limitation 0.400.20
Pain0.46 0.28 0.49
Mental health 0.19 0.16 0.34 0.24
Vitality 0.470.220.48 0.26 0.32
1.00 0.90 0.80 0.70 0.600.50 0.40 0.300.20 0.10 0.00 -0.10-0.20 -0.30 -0.40
Fig. 1 Relationship between
EQ-5D DSI and SF-6D (SF-12)
64 Qual Life Res (2011) 20:57–67
relevant for both health status (SF-12 MCS) and HRQL
(EQ-5D DSI and SF-6D [SF-12]).
The analysis of the relationship between the alternative
measures of health status/HRQL demonstrated a difference
in how the measures assessed HRQL by the ‘moderate’
intra-class correlation of 0.54. The data suggested a ceiling
effect might be found on the EQ-5D if it were used to
assess change in health status following a social interven-
tion. These findings are consistent with the results of others
who have compared the tools in different settings [34, 35].
Such a potential ceiling effect might have meant that the
EQ-5D DSI was unable to discriminate between differ-
ences in social isolation. This was not the case in this cross-
sectional analysis. Both the EQ-5D DSI and the SF-6D
(SF-12) discriminated between those who were ‘severely
socially isolated’ and those who were not, and the instru-
ments appeared to relate to the construct of social isolation
in similar ways.
Limitations and strengths of the study
The data used in this study were collected as part of a
controlled clinical trial considering the effectiveness and
cost-effectiveness of a community mentoring intervention
(DAQoL). This needs to be taken into account when con-
sidering the applicability of our findings.
As part of the DAQoL study, demographic and some
social health data were also collected from a community
sample of individuals of the same age as the DAQoL
participants. This showed that a greater proportion of the
DAQoL participants lived alone and had lower levels of
social activity when compared to the community sample
(data not presented), indicating that our sample differed
from an age-matched population in terms of social isolation.
However,it was notpossiblewithin the aims and constraints
of the DAQoL study to collect health status and HRQL data
from such a sample of older people who were not socially
isolated. Therefore, the regression findings are based on a
comparison of those who were ‘severely socially isolated’
and those who were ‘socially isolated/at risk of social iso-
lation’—they were not compared with individuals who were
not socially isolated. This warrants careful interpretation of
the strength of association between social isolation and
health status/HRQL and implies that the figures reported
here may underestimate the strength of the relationship. The
strength of this relationship may also have been underesti-
mated by virtue of the fact that those who took part in the
research were more socially active than individuals who
declined to participate (data not presented), suggesting that
those most socially isolated were not included.
DAQoL was based largely in rural Devon. This raises
the question of the generalisability of our findings to urban
populations. For example, the nature of social isolation
may be different in urban areas, and the associations with
health status and HRQL may differ. This requires further
tested and validated, but the definition of social isolation
has lacked theoretical clarity, and the construct has been
described as notoriously difficult to measure [1, 7].
Although there is no ‘gold standard’ measure of social iso-
lation, assessment approaches consistently have ‘absence of
contact from other people’ at their core, and objectively
categorise self-report data . The approach taken in this
work has adopted these principles and followed the method
of Victor et al.  in their UK nationally representative
Correlational analyses do not enable the determination
of cause and effect, so the directionality of the relationships
between social isolation and health status/HRQL cannot be
deduced from the reported analyses. However, the use of
regression techniques in this sample has demonstrated that
health status/HRQL is independently related to social
Implications and conclusions
Further work is now needed to explore whether the current
findingshold forlarger, nationallyrepresentativesamplesof
older people, and to explore the strength of these relation-
ships when individuals who are not socially isolated are
included in the analyses. However, this study highlights the
potential burden of social isolation on the health and well-
reducing social isolation on mental and physical well-being.
Our analyses have shown that social isolation is inde-
pendently (negatively) associated with the health status/
HRQL of older people. The effect is of a magnitude that is
clinically relevant, and it is independent of depression
levels, physical co-morbidities, age, gender, living alone
status, employment status and accommodation type. This
has implications for health and social care policy makers
New (and existing) social isolation interventions will
require evaluation, and the development of such interven-
tions should be concomitant with the advancement of
methodologies to assess their effectiveness and cost-
effectiveness. We would suggest the inclusion in such
research of a preference-based measure of health status so
both health status and HRQL can be assessed and QALYs
Our results imply that the use of either the SF-6D
(SF-12) or the EQ-5D in cost-effectiveness analyses would
be acceptable, as they relate to the construct of social iso-
lation in similar ways. Both measures may be sensitive to
changes in social isolation, as clinically meaningful
Qual Life Res (2011) 20:57–6765
differences in HRQL were found using both tools between
individuals who were ‘severely socially isolated’ and those
who were ‘socially isolated’/’at risk of social isolation’,
independent of demographic/clinical characteristics. The
SF-6D (SF-12) may be preferred for use with this popula-
tion, as the distribution of responses was more normal than
that of the EQ-5D DSI, and one of the six dimensions spe-
cifically addresses social functioning. There were also some
indications that a ceiling effect might be found if the EQ-5D
was used to assess changes in health status following inter-
ventions to address social isolation in such populations of
vulnerable older people. Alternatively, recently developed
preference-based measures could be considered for use in
estimating the effectiveness and cost-effectiveness of social
interventions. For example, OPUS, a social care outcome
capabilities of older people , may have greater sensi-
tivity/relevance for these client groups.
The growth in interventions to address social isolation is
set to continue , and the potential health gains from it
being addressed may be considerable. For example, our
data suggest that in the United Kingdom approximately 1.3
million older people may be ‘severely socially isolated’
. Addressing and reducing social isolation, and pre-
venting its initial advent, may give a significant and
meaningful improvement in health gain, regardless of
Our findings also suggest that policy and practice should
target individuals ‘at risk’. Those ‘at risk of social isola-
tion’ represent an important treatment group; people need
to be identified early and targeted with interventions before
health and quality of life losses may occur. Considering
and addressing the role of social isolation, as part of a
broader approach to promoting ‘quality of ageing’ in older
people and considering the effect of the social environment
in influencing health and well-being, are key challenges for
health and social care policy makers.
This work was supported by Devon County Council via
funding from the Department of Work and Pensions; and
the Department of Health.
collaboration of Devon County Council and the community mentor-
ing teams across the study area, and the support and enthusiasm of the
We would like to acknowledge the input and
1. Victor, C., Scrambler, S., & Bond, J. (2009). The social world of
older people. Oxford: Oxford University Press.
2. Freyne, A., Fahy, S., McAleer, A., Keogh, F., & Wrigley, M.
(2005). A longitudinal study of depression in old age 1: Outcome
and relationship to social networks. Irish Journal of Psycholog-
ical Medicine, 22(3), 87–93.
3. Seeman, T. E. (2000). Health promoting effects of friends and
family on health outcomes in older adults. American Journal of
Health Promotion, 14(6), 362–370.
4. MuCulloch, A. (2001). Social environments and health: Cross
sectional national survey. BMJ, 323, 208–209.
5. Berkman, L. (1995). The role of social relations in health pro-
motion. Psychosomatic Medicine, 57(3), 245–254.
6. Alpass, F. M., & Neville, S. (2003). Loneliness, health and
depression in older males. Aging & Mental Health, 7(3), 212–216.
7. Barnes, M., Blom, A., Cox, K., Lessof, C., & Walker, A. (2006).
The Social exclusion of older people: Evidence from the first
wave of the English Longitudinal Study of Ageing (ELSA). Final
Report. London: Office of the Deputy Prime Minister.
8. Victor, C., & Scrambler, S. (2000). Being alone in later life:
Loneliness, social isolation and living alone. Reviews in Clinical
Gerontology, 10, 407–417.
9. Department of Health. (2001). National service framework for
older people. London: The Stationery Office.
social isolation and living alone in later life. http://www.growin
golder.group.shef.ac.uk/ChristinaVic_F17.pdf. Accessed October
11. Department of Health. (2000). The NHS Plan: A plan for
investment, a plan for reform. London: The Stationery Office.
12. Department of Health. (1998). Modernising social services.
London: The Stationery Office.
13. de Belvis, A. G., Avolio, M., Sicuro, L., Rosano, A., Latini, E.,
Damiani, G., et al. (2008). Social relationships and HRQL: A
cross-sectional survey among older Italian adults. BMC Public
Health, 8, 348.
14. Townsend, P. (1957). The family life of old people. London:
Routledge & Kegan Paul.
15. Tunstall, J. (1966). Old and alone. A sociological study of old
people. London: Routledge & Kegan Paul.
16. Wenger, G. C., & Burholt, V. (2004). Changes in levels of social
isolation and loneliness among older people in a rural area–a 20-
year longitudinal study. Canadian Journal on Aging, 23(2),
17. Kutner, B., Fanshel, D., Togo, A. M., & Langner, T. S. (1956).
Five hundred over sixty: A community survey on aging. New
York: Russell Sage Foundation.
18. Scharf, T., Phillipson, C., Smith, A., & Kingston, P. (2002).
Growing older in socially deprived areas: Social exclusion in
later life. London: Help the Aged.
19. Scharf, T., Phillipson, C., & Smith, A. E. (2004). Poverty and
social exclusion–growing older in deprived urban neighbour-
hoods. In A. Walker & C. H. Hennessy (Eds.), Growing older:
Quality of life in old age (pp. 81–106). Maidenhead: Open Uni-
20. Dickens, A., Richards, S., Hawton, A., Green, C., Taylor, R.,
Edwards, R., et al. (2009). Devon Ageing and Quality of Life
Study: A report on the effectiveness and cost-effectiveness of
community mentoring in Devon using mentoring clients’ out-
comes. Exeter: Peninsula Medical School, Universities of Exeter
21. Donald, C., & Ware, J. (1984). The measurement of social sup-
port. Research in Community and Mental Health, 4, 325–370.
22. Sherbourne, C., & Stewart, A. (1991). The MOS social support
survey. Social Science and Medicine, 32(6), 705–714.
23. Ware, J., Kosinski, M., & Keller, S. (1995). How to score the SF-
12 physical and mental health summary scales. Boston: The
Health Institute, New England Medical Center.
66 Qual Life Res (2011) 20:57–67
24. The EuroQol Group (1996). EQ-5D user guide. Rotterdam, Download full-text
Netherlands: The EuroQol Group.
25. Dolan, P. (1997). Modeling valuations for EuroQol health states.
Medical Care, 35(11), 1095–1108.
26. Brazier, J., & Roberts, J. (2004). The estimation of a preference-
based measures of health from the SF-12. Medical Care, 42(9),
27. Sheikh, J., & Yesavage, J. (1986). Geriatric Depression Scale
(GDS): Recent evidence and development of a shorter version.
Clinical Gerontologist, 5, 165–173.
28. D’ath, P., Katona, P., Mullan, E., Evans, S., & Katona, C. (1994).
Screening, detection and management of depression in elderly
primary care attenders. I: The Acceptability and Performance of
the 15 Item Geriatric Depression Scale (GDS15) and the Devel-
opment of Short Versions. Family Practice, 11(3), 260–266.
29. Jongenelis, K., Pot, A., Eisses, A., Gerritsen, D., Derksen, M.,
Beekman, A., et al. (2005). Diagnostic accuracy of the original
30-item and shortened versions of the Geriatric Depression Scale
in nursing home patients. International Journal of Geriatric
Psychiatry, 20(11), 1067–1074.
30. Greaves, C., & Farbus, L. (2006). Effects of creative and social
activity on the health and well-being of social isolated older
people: Outcomes from a multi-method observational study.
Journal of the Royal Society for the Promotion of Health, 126(3),
31. Ethgen, O., Vanparijs, P., Delhalle, S., Rosant, S., Bruyere, O., &
Reginster, J. -Y. (2004). Social support and health-related quality
of life in hip and knee osteoarthritis. Quality of Life Research, 13,
32. Kind, P., Dolan, P., Gudex, C., & Williams, A. (1998). Variations
in population health status: Results from a United Kingdom
national questionnaire survey. BMJ, 316, 736–741.
33. Kind, P., Hardman, G., Macran, S. (1999). UK population norms
for EQ-5D. Discussion paper 172. University of York, Centre for
34. Brazier, J., Roberts, J., Tsuchiya, A., & Busschbach, J. (2004). A
comparison of the EQ-5D and SF-6D across seven patient groups.
Health Economics, 13, 873–884.
35. Grieve, R., Grishchenko, M., & Cairns, J. (2008). SF-6D versus
EQ-5D: Reasons for differences in utility scores and impact on
reported cost-utility. The European Journal of Health Economics,
36. Koch, G. (1982). Intraclass correlation coefficient. In S. Kotz &
N. Johnson (Eds.), Encyclopedia of statistical sciences (Vol. 4,
pp. 213–217). New York: Wiley.
37. Altman, D. G., & Bland, J. M. (1983). Measurement in medicine:
The analysis of method comparison studies. Journal of the Royal
Statistical Society. Series D (The Statistician), 32(3), 307–317.
38. Walters, S. J., & Brazier, J. E. (2005). Comparison of the mini-
mally important difference for two health state utility measures:
EQ-5D and SF-6D. Quality of Life Research, 14, 1523–1532.
39. Drummond, M. (2001). Introducing economic and quality of life
measurements into clinical studies. Annals of Medicine, 33(5),
40. Marra, C. A., Woolcott, J. C., Kopec, J. A., Shojania, K., Offer,
R., Brazier, J. E., et al. (2005). A comparison of generic, indirect
utility measures (the HUI2, HUI3, SF-6D, and the EQ-5D) and
disease-specific instruments (the RAQoL and the HAQ) in
rheumatoid arthritis. Social Science and Medicine, 60(7),
41. Brazier, J., Ratcliffe, J., Salomon, J. A., & Tsuchiya, A. (2007).
Measuring and valuing health benefits for economic evaluation.
Oxford: Oxford University Press.
42. Ryan, M., Netten, A., Skatun, D., & Smith, P. (2006). Using
discrete choice experiments to estimate a preference-based
measure of outcome–An application to social care for older
people. Journal of Health Economics, 25, 927–944.
43. Coast, J., Flynn, T., Grewal, I., Natarajan, L., Lewis, J., &
Sprotson, K. (2006). Developing an index of capability for older
people: A new form of measure for public health interventions?.
Cambridge: Paper for the Future of Health Conference.
44. Office for National Statistics. Population estimates. http://www.
statistics.gov.uk/cci/nugget.asp?ID=6. Accessed October 28, 2009.
Qual Life Res (2011) 20:57–6767