Qual Life Res (2010) 19:1105–1113
DOI 10.1007/s11136-010-9671-z
Using the Short Form-36 mental summary score as an indicator
of depressive symptoms in patients with coronary heart disease
Rosanna Tavella • Tracy Air • Graeme Tucker •
Robert Adams • John F. Beltrame • Geoffrey Schrader
Accepted: 21 April 2010 / Published online: 13 May 2010
Ó Springer Science+Business Media B.V. 2010
Abstract
Background Depression is common in patients with cardiac disease; however, the use of depression-specific health
instruments is limited by their increased responder and
analyst burden. The study aimed to define a threshold value
on the Short Form-36 (SF-36) mental component summary
score (MCS) that identified depressed cardiac patients as
measured by the Centre for Epidemiologic Studies
Depression Scale (CES-D).
Methods An optimal threshold was determined using
receiver-operating characteristic (ROC) curves on SF-36
and CES-D data from a large cardiac cohort (N = 1,221).
The performance of this threshold was evaluated in a further two cardiac populations.
Results In the index cohort, an SF-36 MCS score of B45
was revealed as an optimal threshold according to maximal
Youden Index, with high sensitivity (77%, 95% CI = 74–
80%) and specificity (73%, 95% CI = 69–77%). At this
threshold, in a second sample of hospital cardiac patients,
R. Tavella (&) J. F. Beltrame
Cardiology Research Unit, The Queen Elizabeth Hospital,
University of Adelaide, 28 Woodville Rd, Woodville South,
5011 Adelaide, SA, Australia
e-mail: rosanna.tavella@adelaide.edu.au
T. Air G. Schrader
Discipline of Psychiatry, University of Adelaide, Adelaide,
SA, Australia
R. Adams
Discipline of Medicine, University of Adelaide, Adelaide,
SA, Australia
G. Tucker
Health Statistics Unit, Department of Health, Adelaide,
SA, Australia
sensitivity was 93% (95% CI = 76–99%) and specificity
was 64% (95% CI = 49–77%). In a final sample generated
from a community population, specificity was 100% (95%
CI = 85–100%) and sensitivity was 68% (95% CI = 61–
74%) at the cut-off of 45.
Conclusion The SF-36 MCS may be a useful research
tool to aid in the classification of cardiac patients according
to the presence or absence of depressive symptoms.
Keywords Coronary heart disease
Short-Form 36 mental summary score
Centre for Epidemiologic Studies Depression Scale
Depression Optimal threshold
Abbreviations
AUC
Area under the curve
CABG
Coronary artery bypass grafting
CAD
Coronary artery disease
CES-D
Centre for Epidemiologic Studies
Depression Scale
CHD
Coronary heart disease
DM
Diabetes mellitus
HRQoL
Health-related quality of life
HT/Chol
Hypertension and/or high cholesterol
IDACC
Identifying depression as a comorbid
condition
NPV
Negative predictive value
NWAHS
North West Adelaide Health Service
PCI
Percutaneous coronary intervention
PPV
Positive predictive value
ROC
Receiver-operating characteristic
SF-36
Short-Form 36
SF-36 MCS Short-Form 36 mental summary score
SF-36 PCS Short-Form 36 physical summary score
123
1106
Introduction
Depression is an independent risk factor for adverse cardiovascular outcomes in patients with Coronary heart disease (CHD) [1]. Studies have shown that the prevalence of
depression in patients with CHD ranges from 14% to as
high as 47% [2]. Although contemporary data does not
indicate an improvement in cardiovascular events with the
use of antidepressants in patients with CHD, these agents
can be safely used to improve depressive symptoms [3–5].
The gold standard for the diagnosis of depression
requires a clinical assessment with criteria defined by the
Diagnostic and Statistical Manual of the American Psychiatric Association (DSM). This labour-intensive assessment is impractical for both screening procedures and
research studies so that self-rated measures of depression
have been developed [6]. Self-rated depression scales are
among the best established health measurements, and many
of them have similar degrees of validity and reliability to a
clinical interview [7, 8]. The Centre for Epidemiologic
Studies Depression Scale (CES-D) is a purpose-built scale
that has received extensive assessment in numerous studies
[9–11]. It was developed to identify depression in the
general population and has shown strong internal consistency, good reliability and has been validated against
clinical and self-report criteria [12].
Measuring patient-perceived health-related quality of
life (HRQoL) is an important outcome in cardiac care. [13].
Many instruments have been developed to quantify
patient’s HRQoL with the most commonly used general
instrument being the Short Form 36 (SF-36) [14, 15]. The
SF-36 profiles eight health domains that are relevant to
general functional status and well-being. The eight scales
are aggregated into two summary scales, a physical component score (PCS) and mental component score (MCS).
The component scores are empirically valid and provide an
interpretation of physical and mental dimensions of health
status [16]. The SF-36 has been validated extensively [17,
18] and has been established as the best generic measure of
HRQoL among people with CHD due to its sound psychometric properties [19, 20]. Although the Mental Health
(MH-5) subscale of the SF-36 has been shown to be an
indicator of depression [21], and also related to mental
health status of coronary patients [22], a comprehensive
evaluation of depression requires more than a measure of
mental health itself, but an assessment of limitations in
daily roles and social functioning [23]. The SF-36 MCS is a
psychometrically sound measure capturing overall functioning as a result of mental health. In addition, in general
health populations, the SF-36 MCS has exceeded performance of the MH-5 in both tests of self-reported mental
health and the effects of clinical depression in crosssectional and longitudinal studies [24]. The SF-36 MCS
123
Qual Life Res (2010) 19:1105–1113
has also been shown to be a good indicator of depression in
both general and diseased populations [25–27]. However,
there is no data demonstrating the use of the SF-36 MCS
in identifying cardiac patients who exhibit depressive
symptoms.
Hence, defining the effectiveness of the SF-36 MCS in
identifying depressed cardiac patients would be a useful
research tool. This is particularly pertinent as many studies
have utilised the SF-36 but few have employed specific
depression measures; thus, retrospective identification of
depressed CHD cohorts would increase research capabilities in this field. Therefore, the objective of this study was
to define a threshold score on the SF-36 MCS that would
categorise a population with cardiac disease into depressed
and non-depressed as benchmarked by the CES-D.
Methods
In order to establish a depression threshold score for the
SF-36 MCS, three patient cohorts with cardiac disease
were utilised as outlined below. The threshold was defined
in the first study cohort (IDACC) and then assessed in the
remaining two cohorts, each with differing clinical characteristics and varying prevalence of depression. All three
cohorts were primarily recruited from the same geographic
location, namely the north-western suburbs of Adelaide.
Study cohorts
IDACC cohort
The Identifying Depression as a Comorbid Condition
(IDACC) study was a randomised controlled trial assessing
the impact of psychiatric liaison with general practitioners
on depressive symptoms in patients with cardiac conditions. The study protocol has been previously published
[28] and only patients with confirmed CHD were included
in the analysis. This was defined by hospital admission or
documented history of myocardial infarction, unstable
angina, coronary artery bypass grafting or percutaneous
coronary intervention. All patients completed an SF-36 and
CES-D at baseline, which have been obtained for this
analysis.
NWAHS cohort
The North Western Adelaide Health Service (NWAHS)
study was a population-based biomedical cohort investigating the prevalence of a number of chronic conditions
and health-related risk factors along a continuum [29].
Patients in this study were randomly recruited using the
telephone to conduct interviews and the Electronic White
Qual Life Res (2010) 19:1105–1113
Pages as the sampling frame resulting in a representative
population sample for the Northwest region of Adelaide as
detailed in previous studies [29, 30]. A selection of patients
recruited for baseline examination was defined as the cardiac subset. Studies have indicated that self-reported cardiovascular events show good concordance with medical
records and hospital diagnostic codes [31–33]. Therefore,
the inclusion criterion for this group was the self-report of
patients responding to a history of events, either angina or
myocardial infarction, as diagnosed by a doctor. Several
days prior to an appointment for biomedical examination,
participants were mailed a questionnaire pack containing
the SF-36 and CES-D to be self-administered and then
checked for accuracy during the examination.
Coronary angiogram cohort
The final cohort included consecutively recruited patients
undergoing coronary angiography at the Queen Elizabeth
Hospital for the investigation of chest pain. The purpose of
the study was to assess 12-month health outcomes in these
patients. A small subgroup of these patients completed the
SF-36 and CES-D at baseline which were utilised for this
study.
Study definitions
The cardiovascular risk factors were predominantly based
upon self-report. The existence of coronary artery disease
(CAD) was confirmed angiographically in the coronary
angiogram cohort, hospital admission diagnosis or documented history in the IDACC cohort and by self-reported
MI and/or angina in the NWAHS cohort. Consistent with
previous research, current depression and severe depression were defined by CES-D total scoring of C16, and C27,
respectively [34, 35]. In each cohort, demographic data was
obtained via self-report or extracted from hospital records.
Socio-economic status was defined by the Australian
Bureau of Statistics’ Socio-Economic Indexes for Areas
scores, an accepted proxy measure for socio-economic
status based on regional analyses in Australia [36].
Data analysis
Using the above cross-sectional baseline data from the 3
independent cohorts, all analyses were performed with
STATA (Version 10, StataCorp, Texas, USA). Prior to
defining an SF-36 MCS depression threshold, the relationship between the SF-36 MCS and CES-D was determined using Pearson’s correlation coefficients. Receiver
operating characteristic (ROC) curves were then used to
determine the relative sensitivity and specificity of the
SF-36 MCS relative to the CES-D gold standard. As
1107
established in previous studies, a CES-D score of C16 was
used to define the presence of depression [34, 35]. From the
ROC analysis on the IDACC data, a specific threshold
value was derived demonstrating a maximal Youden Index
(sensitivity ? specificity - 1) [37]. This optimal threshold
from the IDACC cohort was then evaluated in the NWAHS
and coronary angiogram cohorts in conjunction with their
respective ROC curves. In addition, positive predictive
values (PPV) and negative predictive values (NPV) were
measured for different threshold scores in the central range
of the SF-36 MCS scores. Independent samples’ t-tests
compared SF-36 MCS mean scores between depressed and
not depressed groups according to CES-D scoring. Clinical
characteristics between the three groups were compared
using Fishers’ exact test for categorical variables and
ANOVA for continuous data.
Power calculation
The prevalence of depression in the overall IDACC population was 46% (674 out of 1,455). Previous studies
determining SF-36 MCS thresholds relative to CES-D data
in chronic spinal pain patients [38] reported 80% sensitivity and 90% specificity. Therefore, in the primary analysis in the derivation cohort (IDACC), and according to the
prevalence inflation method [39], using a maximum
acceptable width of 10% for the confidence intervals (i.e. a
precision of 5%), to achieve 95% confidence intervals in
sensitivity and specificity, 532 and 258 patients would be
required, respectively.
Results
Study cohorts’ patient characteristics
The patient characteristics and HRQoL scores for the
IDACC, NWAHS and coronary angiogram cohorts are
summarised in Table 1. The three cohorts were of similar
gender distribution; however, the NWAHS cohort was
older (P \ 0.01). There was also heterogeneity in the
coronary risk factors with the prevalence of smoking,
diabetes, hypertension and hypercholesterolaemia being
higher in the IDACC cohort compared to the NWAHS
cohort (P \ 0.01). In addition, prevalence of depressive
symptoms and history of mental illness was higher in the
IDACC patients compared to NWAHS (P \ 0.01).
There was a significant negative correlation between
SF-36 MCS and CES-D scores for the IDACC (r = -0.68,
P \ 0.01), NWAHS (r = -0.65, P \ 0.01) and coronary
angiogram cohorts (r = -0.77, P \ 0.01). The SF-36
MCS scores for IDACC, NWAHS and angiogram cohort
were significantly lower in the depressed patients (CES-D
123
1108
Qual Life Res (2010) 19:1105–1113
Table 1 Clinical characteristics and HRQoL scores (mean ± SD) for the study cohorts
Characteristic
IDACC (n = 1,221
NWAHS (n = 222)
Angiogram (n = 80)
Age
62 ± 12*
69 ± 11
61 ± 11
Females
31%
31%
32%
Risk factors
Current smoker
23%*
12%
15%
DM
22%*
31%
31%
HT/Chol
66%*
82%
85%^
HRQoL
SF-36 PCS
35 ± 11*
41 ± 13
43 ± 10^
SF-36 MCS
46 ± 12
47 ± 12
37 ± 10^
CES-D
17 ± 11*
7±9
14 ± 9
Depressed CES-D C 16
Severe depression CES-D C 27
44%*
19%*
13%
5%
38%
13%^
History of mental illness
28%*
14%
15%^
Married/defacto
66%*
59%
73%
Separated/divorced
16%
15%
16%
Widowed
12%*
21%
4%
Never married
5%
5%
7%
Employed
22%
17%
37%
Unemployed
3%
2%
2%
Home duties/retired
67%*
78%
14%^
Pensioner/other
6%*
3%
46%^
Low advantage
28%
24%
23%
Medium advantage
45%*
37%
35%^
High advantage
23%*
17%
38%^
Marital status
Work status
Socio-economic status
DM, diabetes mellitus; HT/Chol, hypertension and/or high cholesterol; CAD, coronary artery disease; SF-36 PCS, SF-36 physical summary
score; SF-36 MCS, SF-36 mental summary score; CES-D, centre for epidemiologic studies depression scale
P values for ANOVA for continuous variables and Fishers exact test for categorical variables
* P \ 0.01 IDACC versus NWAHS, ^ P \ 0.01 IDACC versus coronary angiogram cohort,
cohort
score C 16), (39 ± 11, 30 ± 10, 35 ± 6, respectively)
compared with those not depressed (52 ± 9, 49 ± 10,
48 ± 8, respectively).
ROC characteristics in IDACC sample
Using ROC analysis, overall there was a significant prediction for the CES-D scores as measured by the SF-36
MCS scores (AUC = 0.82 SE 0.01, 95% CI 0.8–0.85).
Utilising a high specificity (90%, 95% CI = 87–92%), at
the SF-36 MCS threshold value of 39, 72% of patients were
appropriately assigned as being depressed as per the CESD scores. However, the sensitivity of this SF-36 MCS
threshold value was 50% (95% CI = 46–54%), indicating
that half of the patients depressed as per the CES-D scores
were identified correctly.
123
P \ 0.01 NWAHS versus coronary angiogram
Using the maximal Youden Index, the optimal SF-36
MCS threshold value for detecting depressive symptoms
was 45 (PPV = 72%, 95% CI = 68–76; NPV = 79%,
95% CI = 75–82. At this threshold, sensitivity was 73%
(95% CI = 69–77%), so that 27% of patients were falsely
identified as experiencing depressive symptoms. Specificity was increased to 77% (95% CI = 74–80%), and 76% of
patients were appropriately classified. The operating characteristics, NPVs and PPVs are summarised in Table 2.
Accuracy of the SF-36 mental summary score
in NWAHS and coronary angiogram cohorts
To establish the usefulness of the SF-36 MCS in detecting
depressive symptoms, the ROC curves were analysed
separately in a cohort with a low prevalence of depression
Qual Life Res (2010) 19:1105–1113
1109
Table 2 Operating characteristics for the central range of SF-36 MCS cut-points in the IDACC cohort
SF-36 MCS cut-off
Sensitivity 95% CI
Specificity 95% CI
Appropriately classified (%)
PPV 95% CI
NPV 95% CI
B39
50%(46–54)
90% (87–92)
72
80% (75–84)
70% (66–73)
B40
55% (50–59)
88% (85–90)
73
78% (74–82)
71% (68–74)
B41
58% (54–62)
87% (84–89)
74
77% (73–81)
72% (69–75)
B42
61% (57–65)
85% (82–87)
74
76% (72–80)
73% (70–77)
B43
65% (61–69)
82% (79–85)
75
74% (70–78)
75% (71–78)
B44
69% (65–73)
79% (76–82)
75
73% (68–76)
76% (73–79)
B45*
73% (69–77)
77% (74–80)
76
72% (68–76)
79% (75–82)
B46
76% (72–79)
75% (71–78)
75
70% (66–74)
80% (76–83)
B47
78% (74–81)
71% (68–75)
74
68% (64–72)
80% (77–83)
B48
80% (78–84)
69% (63–71)
74
69% (65–72)
80% (76–83)
SF-36 MCS, SF-36 mental summary score; PPV, positive predictive value; NPV, negative predictive values
Data obtained from receiver-operating characteristics for 1,221 observations
* Optimal threshold according to maximal Youden Index (sensitivity ? specificity - 1)
Table 3 Screening abilities of the central range of SF-36 MCS cut-points for the NWAHS and angiogram cohorts
SF-36 MCS cut-off NWAHS
Coronary angiogram
Sensitivity
Specificity
PPV
NPV
Sensitivity
Specificity
PPV
NPV
B40
86% (67–95)
79% (73–85) 38% (27–51) 97% (93–99)
77% (57–89) 86% (73–94) 77% (57–89) 86% (73–94)
B41
86% (67–95)
77% (70–82) 36% (25–48) 97% (93–99)
80% (61–92) 86% (73–94) 77% (58–90) 88% (75–95)
B42
86% (67–94)
75% (68–80) 34% (23–46) 97% (93–99)
80% (61–92) 84% (70–92) 75% (56–88) 88% (74–95)
B43
90% (72–97)
73% (66–79) 33% (23–45) 98% (93–99)
90% (72–97) 74% (59–85) 68% (51–81) 93% (79–98)
B44
93% (76–99)
71% (64–77) 33% (23–44) 99% (94–100)
90% (72–97) 70% (55–82) 64% (48–78) 92% (78–98)
B45
100% (85–100) 68% (61–74) 32% (23–43) 100% (96–100) 93% (76–99) 64% (49–77) 61% (45–75) 94% (79–99)
SF-36 MCS, SF-36 mental summary score; PPV, positive predictive value; NPV, negative predictive value
Data obtained from receiver-operating characteristics for 222 observations for the NWAHS sample and for 80 observations in the coronary
angiogram cohort. Sensitivity, specificity, PPV and NPV are presented with 95% CI
(NWAHS cohort) and one with a similar prevalence to the
IDACC cohort (the coronary angiogram cohort). The
overall accuracy for predicting depressive symptoms as
measured by the AUC for the NWAHS and coronary
angiogram populations were 0.91 (SE 0.02, 95% CI 0.87–
96) and 0.89 (SE 0.04, 95% CI 0.8–0.95), respectively.
Using the optimal SF-36 MCS threshold value of 45
derived from the IDACC cohort, ROC analysis of the
NWAHS cohort elucidated a sensitivity of 100% (95%
CI = 85–100%) and specificity of 68% (95% CI = 61–
74%). Positive and negative predictive values were 32%
(95% CI = 23–43%) and 100% (95% CI = 96–100%),
respectively. Similar levels of accuracy were observed for
the coronary angiogram cohort with this SF-36 MCS
threshold value of 45 (sensitivity = 93%, 95% CI = 76–
99%, specificity = 64%, 95% CI = 49–77%; PPV = 61%,
95% CI = 45–75%; and NPV = 94%, 95% CI = 79–99%,
Table 3).
According to maximal Youden Index, the optimal
threshold found in the NWAHS cohort was also 45, and in
the coronary angiogram cohort, the optimal threshold was
41 (sensitivity = 80%, 95% CI 61–92; specificity = 86%,
95% CI = 73–94; PPV = 77%, 95% CI 58–90; and
NPV = 88%, 95% CI = 75–95).
At the optimal threshold of 45, the proportion of patients
appropriately identified as depressed on the SF-36 MCS
according to CES-D scoring for the NWAHS and coronary
angiogram cohort was 71 and 75%, respectively, compared
to 76% in IDACC (Table 4).
Discussion
This investigation has first established that the SF-36 MCS
is correlated with the CES-D scores in CHD populations.
Secondly, the SF-36 MCS threshold value of 45 derived
from ROC analysis according to the maximal Youden
Index could identify a depressed subgroup in the index
(IDACC) cohort (Table 2), as defined by the CES-D, with
sensitivity of 73% and specificity of 77%. Moreover, when
123
1110
Qual Life Res (2010) 19:1105–1113
Table 4 Proportion of patients appropriately classified by the SF-36
MCS for the central range of cut-points for each study cohort
SF-36 MCS
cut-off
Appropriately classified
IDACC (%)
(n = 1,221)
NWAHS (%)
(n = 222)
Angiogram (%)
(n = 80)
B40
73
80
83
B41
74
78
84
B42
74
76
83
B43
75
75
80
B44
75
73
78
B45
76
71
75
Data obtained from receiver-operating characteristic
SF-36 MCS, SF-36 mental summary score
prospectively applying this threshold value to CHD cohorts
with a low and high prevalence of depression (NWAHS
and coronary angiogram cohorts, respectively), similar
levels of accuracy were observed (Table 3). To our
knowledge, this is the first study to establish the accuracy
of the SF-36 MCS as an indicator of depressive symptoms
in a cardiac population. It must be recognised that the use
of the SF-36 MCS threshold has not been confirmed as a
screening or diagnostic tool but as a research tool for group
observations utilising a generic measure and that interpretations of individual scores should not be made. Accordingly, studies where specific depression questionnaires
were not administered to CHD cohorts may utilise the SF36 MCS as a surrogate method to define a subgroup
experiencing depressive symptoms. This will enable further investigation of depression in cardiac patients particularly in studying the relationships between depressive
symptoms and other variables.
The strength of the above findings includes the following: (a) three well-characterised study cohorts, (b) use of an
index cohort to establish the SF-36 MCS threshold criteria
with subsequent validation in independent populations and
(c) variable prevalence of depression within the study
cohorts. This later feature is particularly important as
disease prevalence within a population substantially influences the predictive ability of the instrument. The difference between the prevalence of depression among the
cohorts reflects the recruitment strategy and further illustrates the diversity of the study population. The IDACC
cohort was recruited from in-hospital patients, which may
account for the higher prevalence of depression. In contrast, the NWAHS cohort was recruited from the general
population by telephone interview. The coronary angiogram cohort was recruited from those undergoing angiography with 59% being undertaken electively as day cases.
Hence, the cohorts are representative of the spectrum of
patients with CHD.
123
As illustrated in Table 2, assigning a threshold value for
the SF-36 MCS is difficult considering the trade-off
between sensitivity and specificity. In order to achieve a
balanced threshold, the Youden Index was employed. This
index provides the optimal threshold value for which sensitivity ? specificity - 1 is maximised, representing the
point on the ROC curve farthest from chance. Although
other indices are available to establish threshold values, the
Youden Index was chosen as it is well established and not
dependent on prevalence rates [40] and has been previously
used in cardiovascular studies [41].
The relationship of the SF-36 MCS to depressive
symptoms has been reported previously. The Medical
Outcomes Study (MOS) investigated the functioning and
well-being of depressed patients and showed that depression has an additive adverse effect on patient functional
status and well-being as indicated by scores on the SF-36
[25]. In addition, the SF-36 MCS scale has been evaluated
as a screening tool for clinical depression using a two-stage
screening approach in which positive CES-D results were
followed by a clinical interview. ROC analysis showed that
the best all around cut-off for the SF-36 MCS for detecting
depressive disorder was at a score of 42 or below in a
community sample of MOS patients with chronic conditions [16].
Other studies have investigated the use of the SF-36
MCS in disease-specific samples. Discriminate function
analysis showed that the SF-36 MCS had a positive
predictive value of 98% in differentiating major depression from minor or no depression in patients with chronic
pain after controlling for age, gender and pain diagnosis
[27]. In a sample of patients with asthma, depressive
symptoms had an important role in patient-derived
functional status and health-related quality of life as
measured by the MCS of the SF-36, even after adjusting
for current disease activity [26]. Walsh et al. [38] confirmed the effectiveness of the SF-36 MCS in detecting
depression in a chronic spinal pain population. The CESD was also used a gold standard and a score of 35 on the
MSS was identified as an optimal cut-off. This lower
threshold could be attributed to a higher cut-off of 19 on
CES-D scoring to indicate depression. Thus, the SF-36
has been shown to be a good indicator of depression in
both general and diseased populations. Previous studies
have demonstrated the ability of the various depressive
screening instruments to identify cardiac patients with
depression [42, 43]; however, there are no previous
studies confirming this for the SF-36 as a generic health
measure in cardiac patients.
Our study has evaluated the effectiveness of the SF-36
MCS in detecting depressive symptoms in cardiac patients,
by determining an optimal threshold value on one patient
population and then applying this value to two different
Qual Life Res (2010) 19:1105–1113
1111
Table 5 Optimal threshold in index cohort and validation cohorts
IDACC
NWAHS
Coronary angiogram
SF-36 MCS threshold
45*
45*^
45*
41^
Sensitivity
73% (69–77)
100% (85–100)
93% (76–99)
80% (61–92)
Specificity
77% (74–80)
68% (61–74)
64% (49–77)
86% (73–94)
PPV
72% (68–76)
32% (23–43)
61% (45–75)
77% (58–90)
NPV
79% (75–82)
100% (96–100)
94% (79–99)
88% (75–95)
Appropriately classified
76%
71%
75%
84%
* Optimal threshold in index cohort according to maximal Youden Index
^ Optimal threshold in individual cohort according to maximal Youden Index
SF-36 MCS, SF-36 mental summary score; PPV, positive predictive value; NPV, negative predictive value. Data obtained from receiveroperating characteristic for 1,221 observations for IDACC study, 222 observations for the NWAHS sample and for 80 observations in the
coronary angiogram cohort. Sensitivity, specificity, PPV and NPV are presented with 95% CI
‘‘test cohorts’’ with clinically diverse characteristics. Thus,
we have illustrated the variation in sensitivity, specificity
and predictive value between the different samples, and
also for different optimal thresholds (Table 5). Accordingly, differences in the populations being sampled will
lead to differences in levels of accuracy and misclassification. The prevalence of depression (CES-D C 16) was
similar in the IDACC and coronary angiogram cohorts, but
it was much lower in the NWAHS cohort, reflecting the
lowest PPV. There were also several differences in the
average of SF-36 MCS and CES-D values, and also in the
age and risk factor distribution between the three samples.
Since sensitivity and specificity are independent of disease
prevalence, these differences may contribute to the variability noted in the capabilities of the SF-36 MCS in
detecting depressive symptoms.
The main limitation of this study is the absence of a
structured diagnostic interview to formally confirm the
presence of depression in the study cohorts. However, as
discussed previously, such a comparison is unlikely to be
undertaken in population studies given the extensive
workload and costs that would be involved. As an alternative, we have used the CES-D questionnaire which is a
well-established, reliable instrument that has been
validated against the structured diagnostic interview [35].
The CES-D was chosen as it was available in the cohorts
studied and compares favourably with other depressionspecific instruments (Table 6). We acknowledge that the
threshold presented may vary with an alternative diagnostic
tool, such as the DSM criteria, resulting in varying estimates of sensitivity and specificity. Our results are also
limited by the lack of generalizability since the cohorts
were recruited from a single region. In addition, wide
confidence intervals noted in the NWAHS and angiographic cohorts limit the precision of the estimates.
Another limitation of this study relates to the potential of
the SF-36 MCS to misclassify depression. However, since
depression is a disorder which lies on a continuum scale,
every dichotomous assessment, even when developed from
the gold standard, will contend with inappropriate classification to some degree. In this case, it would introduce a
degree of non-differential misclassification, which would
tend to bias any difference towards the null and therefore
potentially miss relationships that may exist with depression. Lastly, it would have been desirable to confirm the
self-reported CHD in the NWAHS cohort. Future studies
are needed to validate the findings in large and diverse
CHD populations.
Table 6 Performances of self-rating scales in screening for depressive symptoms
Scale
Cut-off
n
Patient population
PR (%)
Sensitivity (%)
Specificity (%)
PPV
Gold standard
Reference
HADS
9/21
501
Outpatients
13
85
76
35%
DSM-IV
Lowe et al. [44]
PHQ
11/27
501
Outpatients
13
98
80
43%
DMS-IV
Lowe et al. [44]
BDI
9/10
298
Community
9
85
86
37%
DSM-III
Oliver and Simmons [45]
CESD
16/60
80
Stroke
32
90
86
80%
DSM-III
Parikh et al. [46]
GDS
11/30
134
Community
26
68
89
49%
SCID
Gerety et al. [47]
SDS
59/60
55
Outpatients
31
76
82
65%
DSM-III
Okimoto [48]
HADS, hospital anxiety and depression scale; PHQ, patient health questionnaire; BDI, beck depression inventory; CESD, centre for epidemiological studies depression scale; SDS, self-rating depression scale; GDS, geriatric depression scale; PR, prevalence of major depressive
disorder; PPV, positive predictive value; DSM, diagnostic and statistical manual of mental disorders; SCID, structured clinical interview for
DSM disorders
123
1112
This study has described a SF-36 MCS threshold that
will enable CHD populations to be classified into depressed
and non-depressed groups relative to CESD scoring,
thereby enabling important analyses to be undertaken in
large cardiac cohorts that have completed the widely utilised SF-36 questionnaire. The data presented defines an
SF-36 MCS threshold value of 45 as being the optimal
threshold; however, the results for other thresholds are also
presented should future investigators wish to adopt alternative thresholds based upon their study sensitivity/specificity requirements.
Qual Life Res (2010) 19:1105–1113
15.
16.
17.
18.
19.
References
20.
1. Penninx, B. W., et al. (2001). Depression and cardiac mortality:
Results from a community-based longitudinal study. Archives of
General Psychiatry, 58(3), 221–227.
2. Wulsin, L. R., & Singal, B. M. (2003). Do depressive symptoms
increase the risk for the onset of coronary disease? A systematic
quantitative review. Psychosomatic Medicine, 65(2), 201–210.
3. Glassman, A. H., et al. (2002). Sertraline treatment of major
depression in patients with acute MI or unstable angina. JAMA,
288(6), 701–709.
4. Berkman, L. F., et al. (2003). Effects of treating depression and
low perceived social support on clinical events after myocardial
infarction: The Enhancing Recovery in Coronary Heart Disease
Patients (ENRICHD) Randomized Trial. JAMA, 289(23), 3106–
3116.
5. Schrader, G., et al. (2005). Effect of psychiatry liaison with
general practitioners on depression severity in recently hospitalised cardiac patients: A randomised controlled trial. Medical
Journal of Australia, 182(6), 272–276.
6. McDowell, I., & Newell, C. (1996). Measuring health: A guide to
rating scales and questionnaires (p. 523). New York: Oxford
University Press.
7. Kavan, M. G., et al. (1990). Screening for depression: Use of
patient questionnaires. American Family Physician, 41(3), 897–
904.
8. Attkisson, C. C., & Zich, J. M. (1990). Depression in primary
care. National institute of mental health (U.S.), University of
California San Francisco. New York: Routledge.
9. Weissman, M. M., et al. (1977). Assessing depressive symptoms
in five psychiatric populations: A validation study. American
Journal of Epidemiology, 106(3), 203–214.
10. Roberts, R. E., & Vernon, S. W. (1983). The center for epidemiologic studies depression scale: Its use in a community sample.
American Journal of Psychiatry, 140(1), 41–46.
11. Schulberg, H. C., et al. (1985). Assessing depression in primary
medical and psychiatric practices. Archives of General Psychiatry, 42(12), 1164–1170.
12. Radloff, L. S. (1977). The CES-D scale: A self-report depression
scale for research in the general population. Applied Psychological Measurement, 1(3), 385–401.
13. Spertus, J. A., et al. (2002). Health status predicts long-term
outcome in outpatients with coronary disease. Circulation,
106(1), 43–49.
14. Dougherty, C. M., et al. (1998). Comparison of three quality of
life instruments in stable angina pectoris: Seattle Angina Questionnaire, Short Form Health Survey (SF-36), and Quality of Life
123
21.
22.
23.
24.
25.
26.
27.
28.
29.
30.
31.
32.
Index-Cardiac Version III. Journal of Clinical Epidemiology,
51(7), 569–575.
Smith, H. J., Taylor, R., & Mitchell, A. (2000). A comparison of
four quality of life instruments in cardiac patients: SF-36, QLI,
QLMI, and SEIQoL. Heart, 84(4), 390–394.
Ware, J. E., Jr., Kosinski, M., & Keller, S. D. (1994). SF-36
Physical, mental health summary scales: A user’s manual. Boston, Massachusettsd: The Health Institute, New England Medical
Center.
Ware, J. E., Jr., & Gandek, B. (1998). Overview of the SF-36
health survey and the international quality of life assessment
(IQOLA) project. Journal of Clinical Epidemiology, 51(11),
903–912.
McCallum, J. (1995). The SF-36 in an Australian sample: Validating a new, generic health status measure. Australian Journal of
Public Health, 19(2), 160–166.
Failde, I., & Ramos, I. (2000). Validity and reliability of the
SF-36 Health Survey Questionnaire in patients with coronary
artery disease. Journal of Clinical Epidemiology, 53(4), 359–365.
Permanyer-Miralda, G., et al. (1991). Comparison of perceived
health status and conventional functional evaluation in stable
patients with coronary artery disease. Journal of Clinical Epidemiology, 44(8), 779–786.
Berwick, D. M., et al. (1991). Performance of a five-item mental
health screening test. Medical Care, 29(2), 169–176.
Failde, I., Ramos, I., & Fernandez-Palacin, F. (2000). Comparison between the GHQ-28 and SF-36 (MH 1–5) for the assessment
of the mental health in patients with ischaemic heart disease.
European Journal of Epidemiology, 16(4), 311–316.
McHorney, C. A., Ware, J. E., Jr., & Raczek, A. E. (1993). The
MOS 36-item short-form health survey (SF-36): II. Psychometric
and clinical tests of validity in measuring physical and mental
health constructs. Medical Care, 31(3), 247–263.
Ware, J. E., Jr., et al. (1995). Comparison of methods for the
scoring and statistical analysis of SF-36 health profile and summary measures: Summary of results from the Medical Outcomes
Study. Medical Care, 33(4 Suppl), AS264–AS279.
Wells, K. B., et al. (1989). The functioning and well-being of
depressed patients. Results from the medical outcomes study.
JAMA, 262(7), 914–919.
Mancuso, C. A., Peterson, M. G., & Charlson, M. E. (2000).
Effects of depressive symptoms on health-related quality of life
in asthma patients. Journal of General Internal Medicine, 15(5),
301–310.
Elliott, T. E., Renier, C. M., & Palcher, J. A. (2003). Chronic
pain, depression, and quality of life: Correlations and predictive
value of the SF-36. Pain Medicine, 4(4), 331–339.
Cheok, F., et al. (2003). Identification, course, and treatment of
depression after admission for a cardiac condition: Rationale and
patient characteristics for the Identifying Depression As a
Comorbid Condition (IDACC) project. American Heart Journal,
146(6), 978–984.
Grant, J. F., et al. (2006) The North West Adelaide Health study:
Detailed methods and baseline segmentation of a cohort for
selected chronic diseases. Epidemiologic Perspectives Innovations 3(4).
Grant, J.F., et al. (2008). Cohort profile: The North West Adelaide Health study (NWAHS). International Journal of
Epidemiology.
Bergmann, M. M., et al. (1998). Validity of self-reported diagnoses leading to hospitalization: A comparison of self-reports
with hospital records in a prospective study of American adults.
American Journal of Epidemiology, 147(10), 969–977.
Heckbert, S. R., et al. (2004). Comparison of self-report, hospital
discharge codes, and adjudication of cardiovascular events in the
Qual Life Res (2010) 19:1105–1113
33.
34.
35.
36.
37.
38.
39.
40.
Women’s Health Initiative. American Journal of Epidemiology,
160(12), 1152–1158.
Okura, Y., et al. (2004). Agreement between self-report questionnaires and medical record data was substantial for diabetes,
hypertension, myocardial infarction and stroke but not for heart
failure. Journal of Clinical Epidemiology, 57(10), 1096–1103.
Ensel, W. (1986) Measuring depression: The CESD scale, in
social support, life events and depression. In Lin, N., Dean, A.,
Ensel, W. M. (Eds.). Academic Press: New York.
Zich, J. M., Attkisson, C. C., & Greenfield, T. K. (1990).
Screening for depression in primary care clinics: The CES-D and
the BDI. International Journal of Psychiatry in Medicine, 20(3),
259–277.
ABS. (2003). Socio-Economic Indexes for Areas, Australia, 2001.
Canberra: Australian Bureau of Statistics.
Youden, W. J. (1950). Index for rating diagnostic tests. Cancer,
3(1), 32–35.
Walsh, T. L., et al. (2006). Screening for depressive symptoms in
patients with chronic spinal pain using the SF-36 Health Survey.
Spine J, 6(3), 316–320.
Schatzkin, A., et al. (1987). Comparing new and old screening
tests when a reference procedure cannot be performed on all
screenees. Example of automated cytometry for early detection
of cervical cancer. American Journal of Epidemiology, 125(4),
672–678.
Fluss, R., Faraggi, D., & Reiser, B. (2005). Estimation of the
Youden index and its associate cutoff point. Biometric Journal,
47(4), 458–472.
1113
41. Schisterman, E. F., et al. (2001). TBARS and cardiovascular
disease in a population-based sample. Journal of Cardiovascular
Risk, 8(4), 219–225.
42. Stafford, L., Berk, M., & Jackson, H. J. (2007). Validity of the
hospital anxiety and depression scale and patient health questionnaire-9 to screen for depression in patients with coronary
artery disease. General Hospital Psychiatry, 29(5), 417–424.
43. Strik, J. J., et al. (2001). Sensitivity and specificity of observer and
self-report questionnaires in major and minor depression following myocardial infarction. Psychosomatics, 42(5), 423–428.
44. Lowe, B., et al. (2004). Measuring depression outcome with a
brief self-report instrument: Sensitivity to change of the Patient
Health Questionnaire (PHQ-9). Journal of Affective Disorders,
81(1), 61–66.
45. Oliver, J., & Simmons, M. (1984). Depression as measured by the
DSM-III and the beck depression inventory in an unselected adult
population. Journal of Consulting and Clinical Psychology,
52(5), 892–898.
46. Parikh, R. M., et al. (1988). The sensitivity and specificity of the
Center for Epidemiologic Studies Depression Scale in screening
for post-stroke depression. International Journal of Psychiatry in
Medicine, 18(2), 169–181.
47. Gerety, M. B., et al. (1994). Performance of case-finding tools for
depression in the nursing home: Influence of clinical and functional characteristics and selection of optimal threshold scores.
Journal of the American Geriatrics Society, 42(10), 1103–1109.
48. Okimoto, J. T., et al. (1982). Screening for depression in geriatric
medical patients. American Journal of Psychiatry, 139(6), 799–802.
123