Articles
Clinical diagnosis of depression in primary care:
a meta-analysis
Alex J Mitchell, Amol Vaze, Sanjay Rao
Summary
Background Depression is a major burden for the health-care system worldwide. Most care for depression is delivered
by general practitioners (GPs). We assessed the rate of true positives and negatives, and false positives and negatives
in primary care when GPs make routine diagnoses of depression.
Methods We undertook a meta-analysis of 118 studies that assessed the accuracy of unassisted diagnoses of depression
by GPs. 41 of these studies were included because they had a robust outcome standard of a structured or semistructured interview.
Findings 50 371 patients were pooled across 41 studies and examined. GPs correctly identified depression in 47·3%
(95% CI 41·7% to 53·0%) of cases and recorded depression in their notes in 33·6% (22·4% to 45·7%). 19 studies
assessed both rule-in and rule-out accuracy; from these studies, the weighted sensitivity was 50·1% (41·3% to 59·0%)
and specificity was 81·3% (74·5% to 87·3%). At a rate of 21·9%, the positive predictive value was 42·0% (39·6% to
44·3%) and the negative predictive value was 85·8% (84·8% to 86·7%). This finding suggests that for every
100 unselected cases seen in primary care, there are more false positives (n=15) than either missed (n=10) or identified
cases (n=10). Accuracy was improved with prospective examination over an extended period (3–12 months) rather
than relying on a one-off assessment or case-note records.
Interpretation GPs can rule out depression in most people who are not depressed; however, the modest prevalence of
depression in primary care means that misidentifications outnumber missed cases. Diagnosis could be improved by
re-assessment of individuals who might have depression.
Lancet 2009; 374: 609–19
Published Online
July 28, 2009
DOI:10.1016/S01406736(09)60879-5
See Editorial page 587
See Comment page 589
Leicestershire Partnership
Trust, Leicester General
Hospital, Leicester, UK
(A J Mitchell MRCPsych,
A Vaze MRCPsych,
S Rao MRCPsych); and
Department of Cancer and
Molecular Medicine, Leicester
Royal Infirmary, Leicester, UK
(A J Mitchell)
Correspondence to:
Dr Alex J Mitchell, Leicester
General Hospital,
Leicester LE5 4PW, UK
Alex.Mitchell@leicspart.nhs.uk
Funding None.
Introduction
The burden of depression is considerable in terms of
missed workdays1 and disability.2 The WHO study on
psychological problems in general health care (PPGHC)
across 14 countries found that 14% of individuals suffered
from major depression.3 Similar rates were reported from
the longitudinal investigation of depression outcomes in
primary care (LIDO) study4 and the Depression 2000 study
from Germany.5 However, rates are usually higher in
urban than in rural settings.6,7 Most care for depression is
delivered by general practitioners (GPs) and individually
many GPs have considerable expertise in managing
depression.8 Yet, under-recognition of depression in
primary care has been extensively described.9,10 Evidence
shows that clinicians in all medical specialties have
difficulty recognising mental disorders.11 Recognition rates
vary by practitioner, study, and country. The WHO primary
care study3 found that 54·2% of those who met the criteria
for depression were judged by their treating physician as
having a psychological illness, although rates of accurate
diagnosis of depression ranged from 19·3% in Nagasaki
(Japan) to 74·0% in Santiago de Chile (Chile).
Over-detection—ie, the generation of false positives—is
less discussed but nevertheless important.
Clinicians can overestimate or underestimate levels of
distress of their patients.12 Under-detection could lead to
not enough treatment. Studies suggest that active treatwww.thelancet.com Vol 374 August 22, 2009
ment takes place in 15–60% of those identified as
depressed.13,14 Conversely, over-detection (misidentification)
can lead to too much treatment. For example, in the
European study of the epidemiology of mental disorders
(ESEMeD) about 13% of individuals presenting to the GP
with symptoms of depression did not have any mental
disorder.15 In the national comorbidity replication study,
almost 40% received an intervention for depression
without a concurrent diagnosis.16
Whether under-recognition or over-recognition of
depression is most problematic in routine clinical practice
is not clear. We hypothesised that the absolute number of
false positives (over-detections or misidentifications)
would outnumber that of false negatives (underdetections). Our aim was to assess, with a quantitative
analysis, the rate of true positives, true negatives, false
positives, and false negatives in primary care when GPs
make routine diagnoses of depression.
Methods
Literature search
We did a systematic literature search, critical appraisal, and
pooled analysis. We searched Medline from January, 1966,
to April, 2009; PsycINFO from January, 1887, to April, 2009;
Embase from January, 1980, to April, 2009; and Scopus
from January, 1980, to April, 2009. The search terms were
“general practi$ or primary care or family practice$” and
609
Articles
157 physician diagnosis of depression
in primary care
39 no unassisted assessment
by primary care physicians
118 physician diagnosis of depression
in primary care
4 non-validated comparison
61 diagnostic accuracy of
depression in primary care
by severity scale
5 duplicate data
7 insufficient data
41 interview-based diagnostic accuracy of depression in primary care
Sample size*
33 mid-life sample
8 late-life sample
22 sensitivity alone
19 specificity data
Outcome measure
94 to 199 (n=13)
9 CIDI
200 to 499 (n=12)
5 SCID
≥500 (n=7)
10 DIS
≥1000 (n=9)
2 CIE
Outcome criterion
Case ascertainment
21 DSMIII, IIIR, or IV
6 cumulative notes or
electronic record
9 ICD
5 cross-sectional notes or
electronic record
11 other standardised interview
or expert diagnosis
2 cumulative prospective
questionnaire
28 contemporaneous
prospective questionnaire
2 SCAN/PSE
12 others
Figure 1: Quorum diagram of studies
CIDI=composite international diagnostic interview. CIE= Canberra interview for the elderly. DIS=diagnostic interview schedule. DSM=Diagnostic and Statistical
Manual. ICD=International Classification of Diseases. SCAN/PSE=schedules for clinical assessment in neuropsychiatry/present state examination. SCID=structural
clinical interview for DSM. *Refers to raw data extracted.
“diagnos$ or detect$ or case-finding or recogni$” and
“depress$ or mood or affect$”. Refinement for psychiatric
interviews was achieved with “ICD”, “ICD10”, “ICD9”,
“DSM”, “DSMIV”, “DSMIII”, “SCID”, “CIDI”, “PSE”,
“DIS”, “interview”, or “structured”. In full-text collections
including Science Direct, Ingenta Select, Ovid Full text,
and Wiley-Blackwell Interscience, we used the same
search terms but as a full-text search and citation search.
We also searched the abstract database Web of Knowledge
(4.0, ISI) with the above terms as a text word search and
with key publications in a reverse citation search.
610
Inclusion and exclusion criteria
The main inclusion criterion was the unassisted
diagnostic ability of GPs to identify depression.
Unassisted means that the practitioner is trying to
diagnose depression without specific help from severity
scales, diagnostic instruments, education programmes,
or other organisational approaches. Of those studies
with a randomised intervention to improve detection
(eg, an education package), we only included data from
GPs without (or before) intervention. To define robust
studies, we included those with a minimum sample size
www.thelancet.com Vol 374 August 22, 2009
Articles
of 50 cases and excluded all vignette studies. We used a
robust outcome definition in the form of interview-based
diagnoses, using a psychiatric expert diagnosis or
validated structured or semi-structured interviews
applied by a research interviewer. If a study assessed
detection rates of different types of depression, we used
data for major depression alone.
Data extraction
Data were extracted with a standardised spreadsheet by
two authors (AV and SR) and re-examined by another
author (AJM) independently. We followed the review
guidelines for diagnostic tests recently outlined.17 Questions
for each report included the sample (eg, age of patients),
data integrity, choice of reference criterion, method of case
ascertainment, and duration of clinician assessment of
depression (cumulative or cross-sectional). At least two
authors rated each publication on a 5-point scale.
Statistical analysis
To assess diagnostic accuracy, we extracted and pooled
raw data with proportion meta-analysis to correct for
variations in sample size.18 Considerable heterogeneity
was present. I² was 94·0% (95% CI 93·0% to 94·8%)
from the full sample of 41 studies and 99·5% (99·5% to
99·5%) in the 19 studies that assessed both rule-in and
rule-out accuracy; therefore, we used random-effects
meta-analysis. We adjusted positive predictive values and
negative predictive values for variations in prevalence
from a Bayesian curve of all post-test probabilities.19 We
calculated overall accuracy with fraction correct, which is
the number of correctly identified individuals in every
100 patients screened. The number of screening
applications to rule in or rule out without error is known
as the number needed to screen (NNS),20 where
NNS=1/(fraction correct–fraction incorrect)
We also undertook a bivariate meta-analysis. This
method fits a two-level model, with independent
binomial distributions for true positives and true
negatives subject to sensitivity and specificity in each
study, and a bivariate normal model for the logit
transformations of sensitivity and specificity between
studies.21 We then constructed a summary Receiver
Operator Characteristic curve (sROC), in which each
datapoint represents a separate study, using the bivariate
model to produce a 95% confidence ellipse within ROC
space. We also examined predictors of accuracy
according to duration of assessment (cumulative vs
cross-sectional), method of case ascertainment (case
notes vs contemporaneous form), country of origin, and
patient age.
Role of the funding source
This study did not have any sponsor. The corresponding
author had full access to all the data in the study and had
www.thelancet.com Vol 374 August 22, 2009
final responsibility for the decision to submit for
publication.
Results
118 studies assessed the ability of GPs to make an
unassisted diagnosis of depression, but only 41 satisfied
criteria for valid studies measured against a robust
outcome standard (figure 1 and table). One study that
used the 10th International Classification of Diseases
(ICD10) criteria was excluded because both recognition
and criterion interview were done by the same GPs. One
study explored anxiety alone. 26 were excluded because
they examined broadly defined distress or mixed mental
disorders. Two studies included diagnosis by nursing
colleagues working in primary care. One study that
examined outcome after recognition but not detection
sensitivity was excluded. We excluded five reports
containing data reported in previous publications. Thus,
19 studies assessed diagnostic specificity and had both
rule-in (case-finding) and rule-out (exclusion) accuracy.
13 of these 19 studies examined22–33 general adult patients
(including one of adolescents34) and six examined elderly
(>65 years) people.35–40 The remaining 22 studies only
assessed diagnostic sensitivity. Of these, two were done
in elderly (>65 years) people,41,42 leaving 20 that were done
in adult patients.43–62
The total pooled sample was 50 371 patients and the
mean sample size per study was 1228 (median 318), but
with a large SD of 4004·5. The mean number of
depressed patients was 135 (median 70; SD 179·4) and
the mean number of GPs was 53·7 (median 26·5;
SD 75·5). Criterion methods of diagnosis of depression
included a composite international diagnostic interview
(n=9), diagnostic interview schedule (n=10), structural
clinical interview for DSMIIIR (n=5), Canberra interview
for the elderly (n=2), schedules for clinical assessment
in neuropsychiatry/present state examination (n=2),
schedule for affective disorders and schizophrenia
(SADS) (n=2), and one study for each of the MINI
international neuropsychiatric interview and geriatric
mental state examination. Eight studies used research
interviews of unspecified design to diagnose depression.
Regardless of method of interview, an ICD-based
diagnosis of depression was used in nine studies, and a
diagnosis based on Diagnostic and Statistical Manual
(DSM) in 21 studies. In all 41 studies, all severities of
depression were included. 30 studies used a simple
questionnaire for GPs to assess whether they could
diagnose depression (defined as contemporaneous
ratings). Of these studies, two asked GPs to rate their
clinical impression over the previous 6–12 months;42,43
11 examined the notes (or electronic records) entered by
GPs (defined as case-note method). In six of 11 case-note
studies, depression was assessed over a period of
3–12 months (cumulative recognition).
The overall prevalence of depression (adjusted) was
19·5% (95% CI 15·7% to 23·7%). The corresponding
611
612
Methodological summary
Statistical summary
Participating Mean
patient
GPs
age
(number)
(years)
Description of case ascertainment
Gold standard
Rating* Analysed Depressed Rate
(I–V)
sample
with GS
Sensitivity Specificity PPV
NPV
Fraction NNS
correct
Prospective case ascertainment (adults)
Aben et al22
Netherlands
39
Aragones et al23
Spain
23
Balestrieri et al24
Italy
25
Borowsky et al25
USA
349
Christensen
et al26
Denmark
Gerber et al27
USA
4
··
Gledhill et al28
UK
10
15·6
38
··
Blind recognition of depression by GP
SCID DSMIV
III
58
29
50·0%
37·9%
89·7%
78·6%
59·1%
63·8%
3·63
Prospective questionnaire of patient
depression
SCID DSMIV
IV
306
120
39·2%
71·7%
59·7%
53·4%
76·6%
64·4%
3·48
··
Prospective questionnaire of patient
depression
ICD10
I
2093
283
13·5%
38·9%
93·7%
49·1%
90·7%
86·3%
1·38
··
Prospective questionnaire about the past
12 months (cumulative)
DIS for DSMIII
II
661
661
··
35·2%
··
··
··
··
··
Prospective questionnaire
SCAN for ICD10
II
301
70
23·3%
45·7%
89·2%
56·1%
84·4%
79·1%
1·72
Prospective GP recording of any depression
SADS
IV
268
84
31·3%
57·1%
84·2%
62·3%
81·2%
75·7%
1·94
Prospective depression checklist
K-SADS
V
38
10
26·3%
20·0%
85·7%
33·3%
75·0%
68·4%
2·71
48·8
38·8
Henkel et al29
Germany
18†
Blind prospective recognition form
CIDI
III
431
73
16·9%
64·4%
74·0%
33·6%
91·1%
72·4%
2·23
Jones et al30
USA
20
44·0
Structured physician interviews and chart
audits
DIS for DSMIII
V
94
21
22·3%
33·3%
··
··
··
··
··
Klinkman et al31
USA
50
39·6
Prospective clinician rating form with yes/no
for presence of depression
SCID (DSM-III-R)
IV
425
52
12·2%
34·6%
78·8%
18·6%
89·6%
73·4%
2·14
Lecrubier32
France
Menchetti et al33 Italy
··
164
··
Uncertain
MINI
III
2419
238
9·8%
26·1%
··
··
··
··
··
191
49·1
Prospective GP rating form
ICD10
III
1896
299
15·8%
73·9%
··
··
··
··
··
··
Prospective GP rating form
PSE
III
1994
184
9·2%
65·8%
··
··
··
··
··
CIDI (primary
health care
version) (ICD10,
current
depression or
dysthymia)
IV
189
66
34·9%
71·2%
68·3%
54·6%
81·6%
69·3%
2·59
148
··
62·8%
··
··
··
··
··
Ormel et al34
Netherlands
25
Pini et al35
Italy
46†
44·5
Prospective physician encounter form for
psychiatric caseness (5-point scale)
Ronalds et al36
UK
13
35·0
GP rated the reason for the index consultation Psychiatric
and medical records were examined
assessment
schedule for
DSMIIIR
IV
··
Prospective depression tool
SCID for DSMIIIR
III
1580
92
5·8%
31·5%
··
··
··
··
··
Prospective depression tool
CIDI
III
970
64
6·6%
64·1%
··
··
··
··
··
www.thelancet.com Vol 374 August 22, 2009
Schwenk et al37
USA
50
Simon and
VonKorff38
USA
··
38·9
Simon et al39
International
··
40·3
Prospective depression tool on a 5-point scale CIDI for ICD10
III
25 916
948
3·7%
42·5%
··
··
··
··
··
Strauss et al40
USA
2
42·0
Retrospective contemporaneous form
symptoms and syndrome of depression
SCID (DSMIV) by
psychiatrist
II
327
52
15·9%
61·5%
··
··
··
··
··
Tiemens et al41
Netherlands
10
40·5
Prospective physician encounter form for
distress
CIDI-PC for ICD10
IV
518
136
26·3%
70·6%
··
··
··
··
··
Tiemens et al42
International
61
··
Prospective physician encounter form for
distress
CIDI-PC for ICD10
II
709
174
24·5%
40·2%
85·8%
47·9%
81·5%
74·6%
2·03
Tylee et al43
UK
36
38·4
Prospective encounter form
GHQ30 and then
combined clinical
interview
V
1756
430
24·5%
41·6%
··
··
··
··
··
Van WeelBaumgarten
et al44
Netherlands
46·0
Routine (ICHPPC-2 criteria) applied
consecutively
CIDI for DSMIV
MDD
III
99
33
33·3%
78·8%
89·4%
78·8%
89·4%
85·9%
1·39
4†
··
(Continues on next page)
Articles
Country
www.thelancet.com Vol 374 August 22, 2009
Methodological summary
Country
Statistical summary
Participating Mean
patient
GPs
age
(number)
(years)
Description of case ascertainment
Gold standard
Rating* Analysed Depressed Rate
(I–V)
sample
with GS
Sensitivity Specificity PPV
NPV
Fraction NNS
correct
(Continued from previous page)
Case-notes ascertainment (adults)
Von Korff et al45
USA
45
Füredi et al46
Hungary
12†
40·5
··
Kirmayer et al47
Canada
2†
··
Miller and
McCrone48
USA
··
Munitz et al49
USA
··
Nuyen et al50
Netherlands
Rost et al51
USA
Schmaling and
Hernendez52
USA
Schulberg et al53
USA
Williams et al54
USA
Crawford et al55
UK
Volkers et al56
Netherlands
59·8
··
Prospective rating of emotional problems
DIS for DSMIIIR
V
49
49
3·9%
75·5%
··
··
··
··
··
Medical notes
DIS for DSMIV
I
1211
60
5·0%
6·7%
··
··
··
·
··
Cumulative diagnosis of depression or anxiety DIS for DSMIIIR
in the medical chart for index presentation or
subsequent 12 months (cumulative)
II
685
65
9·5%
33·8%
··
··
··
··
··
Retrospective case notes
Interview by
mental health
clinicians
III
··
48
Unknown
72·9%
··
·
··
··
··
Blind cumulative 3 months medical notes
(retrospective) (cumulative)
ICD10 checklist
I
1197
134
11·2%
21·6%
··
··
··
··
··
195
45·4
Electronic case records from a 12 month
period
CIDI
I
··
191
··
28·8%
··
··
··
··
··
NR
46·3
Recognition from medical notes together
with prescribing habits over 1 year
(cumulative)
DIS for DSMIV
III
··
98
··
67·3%
··
··
··
··
··
2†
45·5
Retrospective notes review for 6 month
prevalence of depression
DSMIV
II
447
146
32·7%
20·5%
90·4%
50·8%
70·1%
67·6%
2·85
3†
33·4
Recognition from medical notes over
6 months
DIS for DSMIII
II
294
27
9·2%
25·9%
98·1%
58·3%
92·9%
91·5%
1·20
80
56·0
Recognition from medical notes
DIS for DSMIIIR
II
276
38
13·8%
28·9%
··
··
··
··
··
28
··
Retrospective notes examination and GP
interview for 6 months (cumulative)
Short CARE
II
318
62
19·5%
51·6%
71·9%
30·8%
86·0%
67·9%
2·79
Prospective electronic medical record over
12 months (cumulative)
CIDI for DSMIV
II
237
53
22·4%
20·8%
··
··
··
··
··
Prospective clinician questionnaire (4-point
scale)
DIS for DSMIIIR
IV
101
15
14·9%
60·0%
62·8%
22·0%
90·0%
62·4%
4·04
Blind structured questionnaire 1 year after the PRIME-MD for
initial screen (cumulative)
DSMIV
III
··
73
··
32·9%
··
··
··
··
··
4·13
··
63·6
Prospective case ascertainment (elderly)
Bowers et al57
Australia
Licht-Strunk
et al58
Netherlands
11
··
··
64·0
O’Conner et al59
Australia
30
78·2
Prospective depression tool
CIE
III
995
61
6·1%
75·4%
61·2%
11·3%
97·4%
62·1%
Pond et al60
Australia
13
82·5
Prospective questionnaire
CIE for ICD10
IV
105
12
11·4%
50·0%
77·0%
21·9%
92·3%
73·9%
2·09
Turrina et al61
Italy
14
72·8
Prospective contemporaneous form on
4-point Likert scale
Geriatric mental
state
examination for
DSMIIIR
III
255
89
34·9%
71·9%
71·1%
57·1%
82·5%
71·4%
2·34
Van Marwijk
et al62
Netherlands
13
73·6
Prospective depression tool
DIS for DSMIII
III
580
46
7·9%
26·1%
92·7%
23·5%
93·6%
87·4%
1·34
Table: Studies of GPs’ ability to diagnose depression
Articles
CARE=comprehensive assessment and referral evaluation. CIDI=composite international diagnostic interview. CIE=Canberra interview for the elderly. DIS=diagnostic interview schedule. GDS=geriatric depression scale. GHQ=general health
questionnaire. GPs=general practitioners. GS=gold standard. K-SADS=Kiddie-Sads diagnostic interview. NNS=number needed to screen. NPV=negative predictive value calculated from data. PPV=positive predictive value calculated from data.
PRIME-MD=primary care evaluation of mental disorders. PSE=present state examination. *Ratings: I=blind, sample >1000; drop-outs accounted for. II=blind, sample >99; minor design issues. III=blind but underpowered (<100) or with substantial
design issues or unblind or unclear allocation concealment, but otherwise minor issues. IV=not blind or unclear with sample <1000. V=not blind or unclear with sample <100 and/or other substantial methodological weaknesses. †Refers to number of
GP practices (not individuals).
613
Articles
Van Weel-Baumgarten et al (2000)
0·79 (0·61–0·91)
Von Korff et al (1987)
0·76 (0·61–0·87)
Menchetti et al (2009)
0·74 (0·69–0·79)
Miller et al (2005) [CN]
0·73 (0·58–0·85)
Aragones et al (2004)
0·72 (0·63–0·80)
Pini et al (1997)
0·71 (0·59–0·82)
Tiemens et al (1999)
0·71 (0·62–0·78)
Rost et al (1998) [CN]
0·67 (0·57–0·76)
Ormel et al (1990)
0·66 (0·58–0·73)
Henkel et al (2003)
0·64 (0·52–0·75)
Simon and VonKorff (1995)
0·64 (0·51–0·76)
Ronalds et al (2003)
0·63 (0·55–0·71)
Strauss et al (1995)
0·62 (0·47–0·75)
Gerber et al (1989)
0·57 (0·46–0·68)
Christensen et al (2003)
0·46 (0·34–0·58)
Simon et al (1999)
0·43 (0·39–0·46)
Tylee et al (1996)
0·42 (0·37–0·46)
Tiemens et al (1999)
0·40 (0·33–0·48)
Balestriere et al (2004)
0·39 (0·33–0·45)
Aben et al (2003)
0·38 (0·21–0·58)
Borowsky et al (2000)
0·35 (0·32–0·39)
Klinkman et al (1997)
0·35 (0·22–0·49)
Kirmayer et al (1993) [CN]
0·34 (0·23–0·47)
Jones et al (1987)
0·33 (0·23–0·57)
Schwenk et al (1996)
0·32 (0·22–0·42)
Williams et al (1999) [CN]
0·29 (0·15–0·46)
Nuyen et al (2005) [CN]
0·29 (0·22–0·36)
Lecrubier (2007)
0·26 (0·21–0·32)
Schulberg et al (1985) [CN]
0·26 (0·11–0·46)
Munitz et al (2000) [CN]
0·22 (0·15–0·30)
Schmaling and Hernendez (2005) [CN]
0·21 (0·14–0·28)
Gledhill et al (2003)
0·20 (0·03–0·56)
Füredi et al (2003) [CN]
0·07 (0·02–0·16)
O’Connor et al (2001) [E]
0·75 (0·63–0·86)
Turrina et al (1994) [E]
0·72 (0·61–0·81)
Bowers et al (1990) [E]
0·60 (0·32–0·84)
Crawford et al (1998) [CN] [E]
0·52 (0·39–0·65)
Pond et al (1994) [E]
0·50 (0·21–0·79)
Licht-Strunk et al (2008) [E]
0·33 (0·22–0·45)
Van Marwijk et al (1996) [E]
0·26 (0·14–0·41)
Volkers et al (2004) [CN] [E]
0·21 (0·11–0·34)
Combined
0·47 (0·42–0·53)
0·0
0·2
0·4
0·6
Proportion (95% CI)
0·8
1·0
Figure 2: Meta-analytic sensitivity of unassisted diagnosis of depression in primary care
Proportion meta-analysis plot (random effect). CN=case-note method. E=elderly sample.
prevalence was 18·4% (13·5% to 23·9%) in studies
recruiting strictly defined adult patients (aged
18–65 years); 27·6% (17·4% to 39·1%) for mixed adults
and older people (≥65 years); and 15·9% (8·8% to 24·5%)
for people aged 65 or more. Prevalence varied by country
of origin, being highest in the Netherlands (22·6%,
13·5% to 33·3%) and lowest in Australia (10·2%, 5·0%
to 17·0%). In 17 DSM-based studies, depression
614
prevalence was 17·3% (12·2% to 23·1%) and for
ICD-based studies it was 17·2% (10·4% to 25·4%).
Across 41 studies, GPs correctly identified depression
in 2514 out of 5534 true cases (45·4% uncorrected
identification). With the random-effects weighted proportion meta-analysis, this detection rate was corrected to
47·3% (41·7% to 53·0%) (figure 2). However, diagnostic
sensitivity varied greatly between individual studies,
ranging from 6·6% to 78·8%.
Contemporaneous ratings were more accurate than
case-note methods (52·6% [46·5% to 58·6%] vs 33·6%
[22·4% to 45·7%]; χ²=24·7, p<0·001). Diagnostic
sensitivity was greater in case-note studies using
cumulative recognition than in those using cross-sectional
methods (36·8% [21·1% to 54·0%] vs 29·9% [14·3% to
48·3%]; χ²=11·9, p<0·001). Diagnostic sensitivity seemed
to be slightly greater when GPs attempted to diagnose
depression in elderly people (>65 years) than in younger
people (49·6% vs 45·1%; χ²=2·97, p=0·08), which was
significant when restricted to prospective non-cumulative
studies (61·4% vs 49·3%; χ²=12·3, p=0·001). Diagnostic
sensitivity did not differ for ICD-defined or DSM-defined
depression, but differed according to country of origin. To
assess other possible predictors, we analysed the
association between diagnostic sensitivity and sample
size, effect of blinding, year of publication, prevalence,
practice size, and mean patient age. No significant
associations were found. Data were insufficient to analyse
the effect of severity of depression.
Across 19 studies reporting full data for rule-in and
rule-out accuracy, GPs could correctly exclude 5408 out
of 6560 non-depressed individuals, giving a raw detection
specificity of 82·4% (81·5% to 83·3%). The adjusted
meta-analytic (random-effects) specificity was 81·3%
(74·5% to 87·3%; figure 3). In this restricted dataset,
re-calculation of the detection sensitivity showed a
meta-analytic sensitivity of 50·1% (41·3% to 59·0%).
The overall fraction correct was 74·2% (69·2% to 78·9%).
Figure 4 shows the results of the bivariate meta-analysis
with the 95% CIs and a 95% prediction line (the region
within which we could expect the results of a future
study to fall).
In this smaller sample, data were insufficient to analyse
the effect of cumulative recognition. Regarding method
of case ascertainment, studies using contemporaneous
ratings had inferior specificity compared with those
using the case-note method (79·7% vs 88·8%), although
this difference was not significant. When comparing
accuracy in the dataset of 19 studies for individuals of
different ages, no difference existed in GPs’ ability to
diagnose depression in elderly compared with nonelderly samples (sensitivity elderly 56·3% vs adult 47·5%;
specificity elderly 73·5% vs adult 84·5%). The overall
fraction correct was 70·7% for depression in elderly
people and 78·6% for that in younger people. Accuracy
varied by country of origin, being highest in Italy and
lowest in the UK (fraction correct 82·5% vs 50·0%).
www.thelancet.com Vol 374 August 22, 2009
Articles
From the findings of the 19 studies (sensitivity 50·1%,
specificity 81·3%, prevalence 21·9%), the corrected
positive predictive value was 42·0% (39·6% to 44·3%) and
negative predictive value was 85·8% (84·8% to 86·7%).
The positive likelihood ratio was 2·37 (2·21 to 2·54) and
the negative likelihood ratio was 0·64 (0·61 to 0·67).
To test our hypothesis, we calculated the misclassification rates for each 100 consecutive presentations
(figure 5). In general, a motivated GP in an urban setting
(where the rate of depression is 20%) would correctly
diagnose ten out of 20 cases, missing ten true positives.
The GP would correctly reassure 65 out of 80 non-depressed
individuals, falsely diagnosing 15 people as depressed.
Thus, the fraction correct would be 75%, with a
misclassification rate of 25% and a number needed to
screen of 2·0. Out of every five cases judged as depressed,
only two would be true cases but six out of seven
non-depressed individuals would be correctly reassured.
With a Bayesian analysis to apply this accuracy to a low-risk
sample (such as a rural setting) in which the prevalence is
10%, a GP would correctly identify five out of ten cases,
missing five true positives. A typical GP would correctly
reassure 73 out of 90 non-depressed individuals, but might
falsely diagnose 17 people as depressed. Thus, the fraction
correct would be 78%, with a misclassification rate of 22%
and a number needed to screen of 1·77. Therefore, every
five cases judged as depressed, only one is a true case
(positive predictive value 22·9%), and every 20 cases judged
as healthy, about 19 are correctly reassured (negative
predictive value 93·6%).
A
www.thelancet.com Vol 374 August 22, 2009
0·98 (0·96–0·99)
0·94 (0·92–0·95)
0·90 (0·86–0·93)
0·90 (0·73–0·98)
0·89 (0·79–0·96)
0·89 (0·84–0·93)
0·86 (0·83–0·89)
0·86 (0·67–0·96)
0·84 (0·78–0·89)
0·79 (0·74–0·83)
0·74 (0·69–0·78)
0·68 (0·59–0·76)
0·60 (0·52–0·67)
0·93 (0·90–0·95)
0·77 (0·68–0·85)
0·72 (0·66–0·77)
0·71 (0·64–0·78)
0·63 (0·52–0·73)
0·61 (0·58–0·64)
0·81 (0·75–0·87)
0·5
B
Van Weel-Baumgarten et al (2000)
Aragones et al (2004)
Pini et al (1997)
Henkel et al (2003)
Gerber et al (1989)
Christensen et al (2003)
Tiemens et al (1999)
Balestriere et al (2004)
Aben et al (2003)
Klinkman et al (1997)
Schulberg et al (1985) [CN]
Schmaling and Hernendez (2005) [CN]
Gledhill et al (2003)
O’Connor et al (2001) [E]
Turrina et al (1994) [E]
Bowers et al (1990) [E]
Crawford et al (1998) [E] [CN]
Pond et al (1994) [E]
Van Marwijk et al (1996) [E]
Combined
C
Schulberg et al (1985) [CN]
Balestriere et al (2004)
Van Weel-Baumgarten et al (2000)
Christensen et al (2003)
Gerber et al (1989)
Tiemens et al (1999)
Klinkman et al (1997)
Henkel et al (2003)
Pini et al (1997)
Gledhill et al (2003)
Schmaling and Hernendez (2005) [CN]
Aragones et al (2004)
Aben et al (2003)
Van Marwijk et al (1996) [E]
Pond et al (1994) [E]
Turrina et al (1994) [E]
Crawford et al (1998) [E] [CN]
Bowers et al (1990) [E]
O’Connor et al (2001) [E]
Combined
Discussion
We identified 41 studies that assessed the unassisted
ability of GPs to diagnose depression against a robust
outcome standard of psychiatric interviews. The overall
prevalence of depression was 19·5% in various mainly
urban primary care practices across more than ten
countries. 19 studies examined the ability of GPs to
accurately rule in a diagnosis of depressed people and rule
out a diagnosis of non-depressed people. We found a
diagnostic sensitivity of 47·3–50·1%, suggesting that GPs
can generally identify about half of true cases. However,
documentation of depression in medical notes occurred
for only one in three depressed individuals; indeed, when
case notes were examined retrospectively, identification
was as low as one in five true cases.49,52 GPs might be
reluctant to regard patients as depressed even if the criteria
for the disorder are met.63,64 At the same time, GPs could
accurately exclude 81·3% of non-depressed individuals.
In this meta-analysis, sensitivity is higher but specificity
lower than that in large-scale studies that used severity
scales to rate depression.65 Our hypothesis of misidentifications being frequent was confirmed by our data
that suggest that, in a typical urban practice with motivated
staff, out of 100 unselected people the rate of false positives
would outnumber that of true positives by about 50%. In
a rural practice, the rate of false positives would
Schulberg et al (1985) [CN]
Balestriere et al (2004)
Schmaling and Hernendez (2005) [CN]
Aben et al (2003)
Van Weel-Baumgarten et al (2000)
Christensen et al (2003)
Tiemens et al (1999)
Gledhill et al (2003)
Gerber et al (1989)
Klinkman et al (1997)
Henkel et al (2003)
Pini et al (1997)
Aragones et al (2004)
Van Marwijk et al (1996) [E]
Pond et al (1994) [E]
Crawford et al (1998) [E] [CN]
Turrina et al (1994) [E]
Bowers et al (1990) [E]
O’Connor et al (2001) [E]
Combined
0·0
0·5
0·7
0·9
1·1
0·79 (0·61–0·91)
0·72 (0·63–0·80)
0·71 (0·59–0·82)
0·64 (0·52–0·75)
0·57 (0·46–0·68)
0·46 (0·34–0·58)
0·40 (0·33–0·48)
0·39 (0·33–0·45)
0·38 (0·21–0·58)
0·35 (0·22–0·49)
0·26 (0·11–0·46)
0·21 (0·14–0·28)
0·20 (0·03–0·56)
0·75 (0·63–0·86)
0·72 (0·61–0·81)
0·60 (0·32–0·84)
0·52 (0·39–0·65)
0·50 (0·21–0·79)
0·26 (0·14–0·41)
0·50 (0·41–0·59)
0·2
0·4
0·6
0·8
1·0
0·91 (0·88–0·94)
0·86 (0·85–0·88)
0·86 (0·77–0·92)
0·79 (0·74–0·84)
0·76 (0·70–0·81)
0·75 (0·71–0·78)
0·73 (0·69–0·78)
0·72 (0·68–0·77)
0·69 (0·62–0·76)
0·68 (0·51–0·82)
0·68 (0·63–0·72)
0·64 (0·59–0·70)
0·64 (0·50–0·76)
0·87 (0·84–0·90)
0·74 (0·65–0·82)
0·71 (0·65–0·77)
0·68 (0·62–0·73)
0·62 (0·52––0·72)
0·62 (0·59–0·65)
0·74 (0·69–0·79)
0·7
0·9
Proportion (95% CI)
1·1
Figure 3: Meta-analytic specificity of unassisted diagnosis of depression in primary care
(A) Specificity. (B) Sensitivity. (C) Fraction correct. Proportion meta-analysis plot (random effect).
CN=case-note method. E=elderly sample.
615
Articles
1·0
Study estimate
HsROC curve
95% prediction region
95% confidence region
Summary point
0·8
Sensitivity
0·6
0·4
0·2
0
1·0
0·8
0·4
0·6
0·2
0
Specificity
Figure 4: Bivariate meta-analysis of unassisted diagnosis of depression in primary care
HsROC=Hierarchical summary Receiver Operator Characteristic.
Prevalence 27·3%
False negatives (%)
True positives (%)
True negatives (%)
13·6
False positives (%)
Non-depressed
Depressed
13·7
59·1
13·6
Prevalence 20%
10·0
Prevalence 10%
Fraction correct
5·0 5·0
10·0
15·0
65·0
Non-depressed
Depressed
73·2
16·8
Depressed
0
5·0
10·0
Non-depressed
15·0
20·0
25·0
30·0
35·0
40·0
45·0
50·0
55·0
60·0
65·0
70·0
75·0
80·0
85·0
90·0
95·0
100·0
Figure 5: Real-world proportions correctly and incorrectly diagnosed per 100 consecutive primary care
assessments
outnumber that of true positives by three to one. The rate
of false positives is high because even a modest rule-out
error rate becomes significant when the number of nondepressed participants is large. Assuming these detection
rates hold true, a tipping point occurs at a prevalence of
27·3%. Below this value, the number of false positives
outweigh the number of false negatives (figure 5).
616
The highest risk of creating more false positives is in
those with known risk factors for depression.66 Indeed,
the clinical significance of a high false-positive rate
depends on the action taken after a diagnosis of
depression—eg, whether the next step is active treatment,
watch and wait, referral, or re-assessment. The risk of
both false positives and false negatives diminishes if few
people are offered help. A simple way of managing a
high false-positive rate on initial screening is a multistep
assessment process applied to routine clinical
examinations. Assuming 50% sensitivity and 80%
specificity, if each suspected case is re-assessed then the
overall accuracy of the clinician would increase from
77·2% to 89·5% if two assessments were offered in a
rural setting, because a second assessment allows reexamination of not just true positives but also false
positives. Additionally, cumulative assessment over a
period of time was associated with a 35% increase in
detection sensitivity, although data were inadequate to
comment on specificity. This is consistent with previous
work.61 For example, in the MAGPIE study,64 80·2% of
cases seen five or more times during the previous year
were correctly identified compared with 28·8% of those
not seen in the previous year.
Why do clinicians have difficulty diagnosing
depression in routine practice? One issue is the modest
rate of depression, which is between 10% and 20% in
primary care.3–5 A low rate of depression favours
identification of non-depressed cases, whereas a high
rate favours diagnosis of depression. Not all depressed
individuals
spontaneously
express
emotional
symptoms,67–69 but those depressed individuals who
volunteer psychological complaints receive higher
detection rates.5,70 Another issue is the complexity of
fitting the continuous variation in depression severity
into a categorical diagnosis. When allowed to choose
between definite depression, possible (subthreshold)
depression, and unclear cases, clinicians rate over a
third of their decisions as not definitive.14 Many errors
are incorrect estimates of severity—for example, when
symptoms are diagnosed but judged clinically
insignificant.31,64,66 For example, non-depressed individuals in primary care thought to have a psychiatric
disorder were twice as likely to develop depression in
the following year compared with non-depressed
individuals who were not diagnosed as depressed.71 In
the Hampshire depression project,65 72% of people who
were missed had very mild depression, according to the
hospital anxiety and depression scale (HADS). Thus,
more-severe cases of depression are diagnosed more
reliably than less-severe forms.22,50,59 Similarly, many
false positives have related disorders, such as anxiety or
adjustment disorders or subthreshold mood disorder,
and in some cases re-assessment allows revision of the
incorrect diagnosis or simply improvement of the
therapeutic alliance. A further factor is consultation
time. Short appointments seem to compromise
www.thelancet.com Vol 374 August 22, 2009
Articles
diagnosis in difficult cases.72 Patients with depression
might feel under time pressure during consultations to
the extent that they might often be inhibited from fully
disclosing their problems.73 Another issue is whether
GPs prioritise the appropriate signals for major
depression. Stressful life and occupational, financial,
and housing issues were the most influential prompts
to a diagnosis.74 Other predictors of diagnostic sensitivity
(recognition) include a better therapeutic relationship,
presentation with psychological symptoms, clinician
experience, and contact with the patient.14,50,64,75
Our analysis has several limitations. First, we have no
standardised method to rate a health professional’s
opinion as to whether a person is depressed or not. In a
study from van Weel-Baumgarten and colleagues44 in the
Netherlands, GPs were trained to use the criteria for the
International Classification of Health Problems in
Primary Care (ICHPPC-2). Although this was a routine
diagnostic method, training might explain the unusually
high sensitivity in this study. Examination of medical
notes is not optimal because written records might not
indicate the up-to-date opinion of the doctor.76 A
contemporaneous (exit) rating of the preferred diagnosis
of a GP, together with blind re-interview of the patient,
is the most robust method but is subject to a hawthorn
effect. In most cases, allocation concealment existed but
this was not explicitly declared in every study. However,
the two assessments must not occur too far apart to
avoid clinical change contaminating the result. In this
analysis, prospective ratings had better sensitivity but
less specificity than case-note ascertainment. Another
limitation is that a health professional might revise his/
her opinion at a later stage on the basis of further
examination where diagnosis is initially uncertain. Thus,
the accuracy of GPs is likely to vary by the number of
contacts. Few studies have been properly designed to
account for this variation and we found only one study
that used prospective GP case ascertainment in a
cumulative manner.25 Finally, data were insufficient to
adjust for the effect of severity of depression, which has
been strongly associated with detection accuracy in
previous work.
Many interventions have been aimed at improving the
accuracy of the diagnosis of depression in primary care.
Overall, although diagnosis seems to be modestly
improved by use of scales, a positive effect on overall
patient outcomes has not been shown.77 Only when
screening is paired with organised systems of depression
care can outcomes be improved. For example, Christensen
and colleagues26 recruited 38 GPs in Aarhus county,
Denmark. They randomly assigned 900 patients to
severity questionnaires with feedback of results and 885
to routine care. The rate for unassisted diagnosis of
depression was 35% and disclosure of screening
questionnaire increased GPs’ recognition of mental
disorders by only 6·6%, with the greatest effect in those
GPs with moderate or low recognition rates.
www.thelancet.com Vol 374 August 22, 2009
Our results should not be interpreted as a criticism of
GPs for failing to diagnose depression but rather a call
for better understanding of the problems that nonspecialists face. No data suggest that GPs do worse than
other non-psychiatric medical colleagues.11 Depression
is not common in primary care with at least eight out of
ten attendees not suffering serious depression.
Furthermore, depression rarely occurs on its own
(>90% of people with depression have another physical
or mental disorder).78 Finally, depressed patients often
present with physical symptoms and hence their mental
health is often overlooked.49 In primary care, time and
resources are limited, and psychological or even
structured self-help programmes are often not available.
Antidepressants are often the treatment of choice for
clinicians, but frequently not the first choice for
patients, and hence management of depression can be
difficult or unfulfilling. Only about 50% of true cases
are diagnosed and 15% treated.61 Conversely, about 80%
of non-depressed individuals are correctly reassured.
Because one-off brief assessments only facilitate
identification of about half of those with depression, we
suggest that additional consultation time should be
available for those likely to have depression. Repeated
assessments by the GP or other professional in a
collaborative model with a case manager might help to
reduce diagnostic errors and improve overall quality of
care.79,80
Contributors
AJM did data extraction, critical appraisal, data analysis, data
interpretation, and writing of the report. AV and SR did data extraction,
critical appraisal, and data interpretation.
Conflicts of interest
We declare that we have no conflicts of interest.
Acknowledgments
We thank Alex Sutton for advice on bivariate meta-analysis. We also
thank the staff of the postgraduate library, Leicester General Hospital
(Leicester, UK).
References
1
Greenberg P, Stiglin L, Finkelstein S, Berndt E. Depression:
a neglected major illness. J Clin Psychiatry 1993; 54: 419–24.
2
Broadhead WE, Blazer DG, George LK, Tse CK. Depression,
disability days, and days lost from work in a prospective
epidemiologic survey. JAMA 1990; 264: 2524–28.
3
Ustun TB, Von Korff M. Primary mental health services.
In: Ustun TB, Sartorius N, eds. Mental illness in general health
care: an international study. Chichester, UK: John Wiley & Sons,
1995: 347–60.
4
Herrman H, Patrick DL, Diehr P, et al. Longitudinal investigation
of depression outcomes in primary care in six countries: the LIDO
study. Functional status, health service use and treatment of people
with depressive symptoms. Psychol Med 2002; 32: 889–902.
5
Wittchen HU, Pittrow D. Prevalence, recognition and management
of depression in primary care in Germany: the Depression 2000
study. Human Psychopharmacol 2002; 17 (suppl 1): S1–11.
6
Gilchrist G, Gunn J. Observational studies of depression in primary
care: what do we know? BMC Fam Pract 2007; 8: 28.
7
Friedman B, Conwell Y, Delavan RL. Correlates of late-life major
depression: a comparison of urban and rural primary care patients.
Am J Geriatr Psychiatry 2007; 15: 28–41.
8
Harman JS, Veazie PJ, Lyness JM. Primary care physician office
visits for depression by older Americans. J Gen Intern Med 2006;
21: 926–30.
617
Articles
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
618
Hirschfeld R, Keller MB, Panico S, et al. The National Depressive
and Manic-Depressive Association consensus statement on the
undertreatment of depression. JAMA 1997; 277: 333–40.
Depression Guideline Panel. Depression in primary care.
Treatment of major depression. Rockville, MD: US Department of
Health and Human Services, Agency for Health Policy and
Research, 1993.
Cepoiu M, McCusker J, Cole MG, Sewitch M, Belzile E, Ciampi A.
Recognition of depression by non-psychiatric physicians—
a systematic literature review and meta-analysis. J Gen Intern Med
2008; 23: 25–36.
Zastrow A, Faude V, Seyboth F, Niehoff D, Herzog W, Löwe B. Risk
factors of symptom underestimation by physicians.
J Psychosom Res 2008; 64: 543–51.
Maginn S, Boardman AP, Craig TKL, Haddad M, Heath G, Stott J.
The detection of psychological problems by general practitioners.
Influence of ethnicity and other demographic variables.
Soc Psychiatry Psychiatr Epidemiol 2004; 39: 464–71.
Wittchen HU, Hofler M, Meister W. Prevalence and recognition
of depressive syndromes in German primary care settings: poorly
recognized and treated? Int Clin Psychopharmacol 2001; 16: 121–35.
Alonso J, Codony M, Kovess V, et al. Population level of unmet
need for mental healthcare in Europe. Br J Psychiatry 2007;
190: 299–306.
Druss BG, Wang PS, Sampson NA, et al. Understanding mental
health treatment in persons without mental diagnoses: results from
the National Comorbidity Survey replication. Arch Gen Psychiatry
2007; 64: 1196–203.
Pai M, McCulloch M, Enanoria W, Colford JM. Systematic reviews
of diagnostic test evaluations: what’s behind the scenes? ACP J Club
2004; 9: 101–03.
Irwig L, Macaskill P, Glasziou P, Fahey M. Meta-analytic methods
for diagnostic test accuracy. J Clin Epidemiol 1995; 48: 119–30.
Macaskill P, Glasziou P, Irwig L. Meta-analysis of diagnostic tests
In: Armitage P, ed. Encyclopedia of Biostatistics, 2nd edn, 2005.
Mitchell AJ. Predictive summary index and other summary
measures of diagnostic accuracy. In: Kattan MW, ed. Encyclopedia
of Medical Decision Making, 2009.
Reitsma JB, Glas AS, Rutjes AWS, Scholten RJPM, Bossuyt PM,
Zwinderman AH. Bivariate analysis of sensitivity and specificity
produces informative summary measures in diagnostic reviews.
J Clin Epidemiol 2005; 58: 982–90.
Aben I, Verhey F, Beusmans G, Lodder J. Recognition and treatment
of post-stroke depression in general practice. Huisarts en Wetenschap
2003; 46: 487–92.
Aragones E, Pinol JL, Labad A. The overdiagnosis of depression in
non-depressed patients in primary care. Fam Pract 2006; 23: 363–68.
Balestrieri M, Carta MG, Leonetti S, Sebastiani G, Starace F,
Bellantuono C. Recognition of depression and appropriateness of
antidepressant treatment in Italian primary care.
Soc Psychiatry Psychiatr Epidemiol 2004; 39: 171–76.
Borowsky SJ, Rubenstein LV, Meredith LS, Camp P,
Jackson-Triche M, Wells KB. Who is at risk of nondetection of
mental health problems in primary care? J Gen Intern Med 2000;
15: 381–88.
Christensen KS, Toft T, Frostholm L, Ornbol E, Fink P, Olesen F.
The FIP study: a randomised, controlled trial of screening and
recognition of psychiatric disorders. Br J Gen Pract 2003;
53: 758–63.
Gerber PD, Barrett J, Barrett J, et al. Recognition of depression by
internists in primary care A comparison of internists and ”gold
standard” psychiatric assessments. J Gen Intern Med 1989; 4: 7–13.
Gledhill J, Kramera T, Iliffe S, Garrald ME. Training general
practitioners in the identification and management of adolescent
depression within the consultation: a feasibility study. J Adolesc
2003; 26: 245–50.
Henkel V, Mergl R, Kohnen R, Maier W, Möller HJ,
Ulrich Hegerl U. Identifying depression in primary care: a
comparison of different methods in a prospective cohort study. BMJ
2003; 326: 200–01.
Jones LR, Badger LW, Ficken RP, Leeper JD, Anderson RL. Inside
the hidden mental health network. Examining mental health care
delivery of primary care physicians. Gen Hosp Psychiatry 1987;
9: 287–93.
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
Klinkman MS, Coyne JC, Gallo S, Schwenk TL. Can case finding
instruments be used to improve physician detection of depression
in primary care. Arch Fam Med 1997; 6: 567–73.
Lecrubier Y. Widespread underrecognition and undertreatment of
anxiety and mood disorders: results from 3 European studies.
J Clin Psychiatry 2007; 68 (suppl 2): 36–41.
Menchetti M, Belvederi M, Murri B, Bertakis K, Bortolotti B,
Berardi D. Recognition and treatment of depression in primary care
Effect of patients’ presentation and frequency of consultation.
J Psychosom Res 2009; 66: 335–41.
Ormel J, Van Den Brink W, Koeter MWJ, et al. Recognition,
management and outcome of psychological disorders in primary
care: a naturalistic follow-up study. Psychol Med 1990; 20: 909–23.
Pini S, Berardi D, Rucci P, et al. Identification of psychiatric
distress by primary care physicians. Gen Hosp Psychiatry 1997;
19: 411–18.
Ronalds C, Creed F, Stone K, Webb S, Tomenson B. Outcome of
anxiety and depressive disorders in primary care. Br J Psychiatry
1997; 171: 427–33.
Schwenk TL, Coyne JC, Fechner-Bates S. Differences between
detected and undetected patients in primary care and depressed
psychiatric patients. Gen Hosp Psychiatry 1996; 18: 407–15.
Simon GE, VonKorff M. Recognition, management, and outcomes
of depression in primary care. Arch Fam Med 1995; 4: 99–105.
Simon GE, Goldberg D, Tiemens BG, Ustun TB. Outcomes of
recognized and unrecognized depression in an international
primary care study. Gen Hosp Psychiatry 1999; 21: 97–105.
Strauss PR, Gagiano CA, van Rensburg PH, de Wet KJ, Strauss HJ.
Identification of depression in a rural general practice. S Afr Med J
1995; 85: 755–59.
Tiemens BG, Ormel J, Jenner JA. Training primary-care physicians
to recognize, diagnose and manage depression: does it improve
patient outcomes? Psychol Med 1999; 29: 833–45.
Tiemens BG, VonKorff M, Lin EH. Diagnosis of depression
by primary care physicians versus a structured diagnostic
interview. Understanding discordance. Gen Hosp Psychiatry 1999;
21: 87–96.
Tylee AT, Freeling P, Kerry S. Why do general practitioners
recognize major depression in one woman patient yet miss it in
another? Br J Gen Pract 1993; 43: 327–30.
van Weel-Baumgarten EM, van den Bosch WJ, van den Hoogen HJ,
Zitman FG. The validity of the diagnosis of depression in general
practice: is using criteria for diagnosis as a routine the answer?
Br J Gen Pract 2000; 50: 284–87.
Von Korff, Shapiro S, Burke JD, et al. Anxiety and depression in a
primary care clinic. Arch Gen Psychiatry 1987; 44: 152–56.
Füredi J, Rózsa S, Zámbori J, Szádóczky E. The role of symptoms in
the recognition of mental health disorders in primary care.
Psychosomatics 2003; 44: 402–06.
Kirmayer LJ, Robbins JM, Dworkind M, Yaffe MJ. Somatization and
the recognition of depression and anxiety in primary care.
Am J Psychiatry 1993; 150: 734–41.
Miller MJ, McCrone S. Detection of depression in primary care.
Military Med 2005; 170: 158–63.
Munitz H, Valevski A, Weizman A, et al. Recognition and
treatment of depression in primary care settings in 6 different
countries: a retrospective file analysis by WHO. Eur J Psychiatry
2000; 14: 85–93.
Nuyen J, Volkers AC, Verhaak PFM, et al. Accuracy of diagnosing
depression in primary care: the impact of chronic somatic and
psychiatric co-morbidity. Psychol Med 2005; 35: 1185–95.
Rost K, Zhang M, Fortney J, et al. Persistently poor outcomes of
undetected major depression in primary care. Gen Hosp Psychiatry
1998; 20: 12–20.
Schmaling K, Hernendez D. Detection of depression among lowincome Mexican Americans in primary care.
J Health Care Poor Undeserved 2005; 16: 780–90.
Schulberg HC, Saul M, McClelland M, Ganguli M, Christy W,
Frank R. Assessing depression in primary medical and psychiatric
practices. Arch Gen Psychiatry 1985; 42: 1164–70.
Williams JW, Mulrow C.D, Kroenke K, et al. Case-finding for
depression in primary care: a randomized trial. Am J Med 1999;
106: 36–43.
www.thelancet.com Vol 374 August 22, 2009
Articles
55
56
57
58
59
60
61
62
63
64
65
66
67
68
Crawford MJ, Prince M, Menezes P, Mann AH. The recognition
and treatment of depression in older people in primary care.
Int J Geriatr Psychiatry 1998; 13: 172–76.
Volkers A, Nuyen J, Verhaak P, Schellevis F. The problem of
diagnosing major depression in elderly primary care patients.
J Affect Disord 2004; 82: 259–63.
Bowers J, Jorm AF, Henderson S, Harris P. General practitioners’
detection of depression and dementia in elderly patients. Med J Aust
1990; 153: 192–96.
Licht-Strunk E, Beekman ATF, de Haan M, van Marwijk HWJ. The
prognosis of undetected depression in older general practice
patients. A one year follow-up study. J Affect Disord 2009;
114: 310–15.
O’Conner DW, Rosewarne R, Bruce A. Depression in primary
care 2: general practioners’ recognition of major depression in
elderly patients. Int Psychogeratrics 2001; 13: 367–74.
Pond CD, Mant A, Kehoe L, Hewitt H, Brodaty H. General
practitioner diagnosis of depression and dementia in the elderly: can
academic detailing make a difference? Fam Pract 1994; 11: 141–47.
Turrina C, Caruso R, Este R, et al. Affective disorders among elderly
general practice patients: a two-phase survey in Brescia, Italy.
Br J Psychiatry 1994; 165: 533–37.
van Marwijk HWJ, de Bock GH, Hermans J, Mulder JD,
Springer MP. Prevalence of depression and clues to focus
diagnosis: a study among Dutch general practice patients 65+ years
of age. Scand J Primary Health Care 1996; 14: 142–47.
Rost K, Smith R, Matthews DB, Guise B. The deliberate misdiagnosis
of major depression in primary care. Arch Fam Med 1994; 3: 333–37.
Bushnell J. Frequency of consultations and general practitioner
recognition of psychological symptoms. Br J Gen Pract 2004;
54: 838–42.
Thompson C, Ostler K, Peveler RC, Baker N, Kinmonth A-I.
Dimensional perspective on the recognition of depressive
symptoms in primary care: the Hampshire depression project 3.
Br J Psychiatry 2001; 179: 317–23.
Balestrieri M, Baldacci S, Bellomo A, et al. Clinical vs structured
interview on anxiety and affective disorders by primary care
physicians. Understanding diagnostic discordance.
Epidemiol Psichiatr Soc 2007; 16: 144–51.
Prior L, Wood F, Lewis G, Pill R. Stigma revisited, disclosure of
emotional problems in primary care consultations in Wales.
Soc Sci Med 2003; 56: 2191–200.
Cape J, McCullough Y. Patients’ reasons for not presenting
emotional problems in general practice consultations.
Br J Gen Pract 1999; 49: 875–79.
www.thelancet.com Vol 374 August 22, 2009
69
70
71
72
73
74
75
76
77
78
79
80
Stromberg R, Wernering E, Aberg-Wistedt A, Furhoff A-K,
Sven-ErikJohansson S, Backlund LG. Screening and diagnosing
depression in women visiting GPs’ drop in clinic in primary health
care. BMC Fam Pract 2008; 9: 34.
Aragones E, Pinol JL, Labad A, et al. Detection and management of
depressive disorders in primary care in Spain. Int J Psychiatry Med
2004; 34: 331–43.
Barkow K, Maier W, Üstün TB, Gänsicke M, Wittchen H-U,
Heun R. Risk factors for new depressive episodes in primary health
care: an international prospective 12-month follow-up study.
Psychol Med 2002; 32: 595–607.
Hutton C, Gunn J. Do longer consultations improve the
management of psychological problems in general practice?
A systematic literature review. BMC Health Serv Res 2007; 7: 71.
Pollock K, Grime J. Patients’ perceptions of entitlement to time in
general practice consultations for depression: qualitative study. BMJ
2002; 325: 687.
Saltini A, Mazzi MA, Del Piccolo L, Zimmermann C. Decisional
strategies for the attribution of emotional distress in primary care.
Psychol Med 2004; 4: 729–39.
Hickie IB, Davenport TA, Scott EM, Hadzi-Pavlovic D, Naismith SL,
Koschera A. Unmet need for recognition of common mental
disorders in Australian general practice. Med J Austr 2001;
175 (suppl): S18–24.
Rhodes A, Fung K. Self-reported use of mental health services
versus administrative records: care to recall?
Int J Methods Psychiatr Res 2004; 13: 165–75.
Gilbody S, Sheldon T, House A. Screening and case-finding
instruments for depression: a meta-analysis. CMAJ 2008;
178: 997–1003.
Vuorilehto M, Melartin T, Isometsa E. Depressive disorders in
primary care: recurrent, chronic, and co-morbid. Psychol Med 2005;
35: 673–82.
Gilbody S, Whitty P, Grimshaw J, et al. Educational and
organizational interventions to improve the management of
depression in primary care—a systematic review. JAMA 2003;
289: 3145–51.
Christensen H, Griffiths KM, Gulliver A, et al. Models in the
delivery of depression care: a systematic review of randomised and
controlled intervention trials. BMC Fam Pract 2008; 9: 25.
619