European Psychiatry 37 (2016) 6369

Contents lists available at ScienceDirect

European Psychiatry
journal homepage: http://www.europsy-journal.com

Original article

How to early recognize mood disorders in primary care: A nationwide,

population-based, cohort study
G. Castellini a, S. Pecchioli b, I. Cricelli b, F. Mazzoleni c, C. Cricelli c, V. Ricca d, J.J. Hudziak e,f,g,
O. Brignoli c, F. Lapi b,*
a
Psychiatric Unit, Department of Neuroscience, Psychology, Drug Research and Child Health Sexual Medicine and Andrology Unit,
Department of Experimental, Clinical and Biomedical Sciences, University of Florence, Florence, Italy
b
Health Search, Italian College of General Practitioners and Primary Care, Florence, Italy
c
Italian College of General Practitioners and Primary Care, Florence, Italy
d
Psychiatric Unit, Department of Neuroscience, Psychology, Drug Research and Child Health, Sexual Medicine and Andrology Unit,
Department of Experimental, Clinical and Biomedical Sciences, University of Florence, Florence, Italy
e
Vermont Center for Children, Youth, and Families, University of Vermont, Burlington, VT, United States
f
Erasmus University, Sophia Childrens Hospital, Rotterdam, The Netherlands
g
Department of Child Psychiatry, Washington University, Saint-Louis, MO, United States

A R T I C L E I N F O A B S T R A C T

Article history: Background: Mood disorders are managed predominantly in primary care. However, general
Received 5 January 2016 practitioners (GPs) ability to detect and diagnose patients with mood disorders is still considered
Received in revised form 31 March 2016 unsatisfactory. The aim of the present study was to identify predictors for the early recognition of
Accepted 2 April 2016
depressive disorder (DD) and bipolar disorder (BD) in general practice.
Methods: A cohort of 1,144,622 patients (605,285 women, 539,337 men) was investigated, using the
Keywords: Health Search IMS Health Longitudinal Patient Database. Predictors of DD or BD were identied at
Depressive disorder
baseline encompassing somatization-related features, lifestyle variables, medical and psychiatric
Bipolar disorder
General practice
comorbidities. Patients were followed up as long as the following events occurred: diagnoses of DD or
Public mental health BD, death, end of the registration with the GP, end of the study period.
Somatization Results: We found an incidence rate of DD or BD of 53.61 and 1.5 per 10,000 person-years, respectively.
Primary care For both the conditions, the incidence rate grew with age. Most of the lifestyle variables and medical
comorbidities increased the risk of mood disorders. The strongest effect was found for migraine/
headache (HR [95% CI] = 1.32 [1.261.38]), fatigue (1.32 [1.251.39]) irritable bowel syndrome
(1.15 [1.08-1.23]), and pelvic inammation disease (1.28 [1.181.38]).
Conclusions: Several predictors, in particular somatic symptoms, could be interpreted as an early sign of a
mood disorder, and represent a valid indication for the GPs diagnostic process of mental disorders.
2016 Elsevier Masson SAS. All rights reserved.

1. Abbreviations ISTAT Italian National Statistical Institute

ESEMeD European Study of the Epidemiology of Mental Disorders
DD depressive disorder
BD bipolar disorder
2. Introduction
GPs general practitioners
DSM Diagnostic and Statistical Manual of mental disorders Mood disorders are the most prevalent, disabling, and costly
ICD International Classication Disease mental disorders worldwide. They include depressive disorder
LPD Longitudinal patient database (DD) and bipolar disorder (BD), with a 6.9% [1,2] and 1% [3] yearly
SIMG Italian College of General Practitioners prevalence, respectively. They account for more than 30% of the
total costs of all mental disorders which is mostly due to social and
occupational difculties, as well as suicide attempts [13]
* Corresponding author. Health Search, Italian College of General Practitioners.
psychiatric and medical comorbidities [4].
Via Sestese, 61 50141, Florence, Italy. Tel.: +39 55 494 900; fax: +39 55 779 311x66. Mood disorders are managed predominantly in primary care,
E-mail address: lapi.francesco@simg.it (F. Lapi). with general practitioners (GPs) providing the initial clinical

http://dx.doi.org/10.1016/j.eurpsy.2016.04.003
0924-9338/ 2016 Elsevier Masson SAS. All rights reserved.
64 G. Castellini et al. / European Psychiatry 37 (2016) 6369
contact and following most of the clinical course [5]. Primary care
is considered to be more accessible, less stigmatizing, and more
comprehensive as compared to specialized care [6]. General
practitioners have therefore a key role in facilitating diagnosis,
accessing specialist care, and providing continuing monitoring and
support [4]. However, GPs ability to detect and diagnose patients
with mood disorders is still considered unsatisfactory. It has been
reported that 3070% of GPs patients with mental disorders go
undetected [7], and that the delay in diagnosis leads to a more
severe or chronic form of depression and poorer outcomes for the
patient [8].
So far, the information regarding factors that enable or hinder
GPs diagnosis of mental disorders are still scarce. The delay in
receiving care was attributed to insight and motivation issues. For
example, a consistent rate (almost 50%) of bipolar patients do not
seek help, and in most of the cases, they are misdiagnosed and do
not receive an adequate pharmacological treatment [3,4].
A controversial role in the delayed diagnostic process has been
attributed to somatic symptoms. Many patients with depression
initially present with painful physical symptoms (e.g., headache,
musculoskeletal pain) [9,10], with reported co-occurrence rates of
30%50% [9,11]. Conversely, affective disturbances involving
alterations of mood, anxiety and irritability may be early
symptoms of medical illnesses [12].
Some authors reported that painful symptoms might negatively
affect the recognition of depression in primary care and represent a
possible reason for poor detection or misdiagnosis of mood
disorders [10]. The reason could be that GPs often fail to look for
symptoms of depression in patients presenting with pain.
Accordingly, it has been estimated that about half of all patients
with depression may go undiagnosed because they present with
somatic or physical symptoms, which are misinterpreted [13]. Fur-
thermore, it is important to note that medical conditions represent
signicant predisposing factors for development of mood disorders
[14] and about 50%80% of patients with depression initially have
somatic symptoms [15]. There is some evidence that concurrent
pain and depression are associated with increased use of health
care resources and costs compared with either condition alone
[16].
2.1. Aims of the study
Few data are available regarding the incidence rates and early
recognition of mood disorders in primary care. We therefore
performed a study in attempt to estimate the incidence rates and
identify predictors of mood disorders in general practice.
3. Materials and methods
3.1. Data source
We used the Health Search (HS) IMS Health Longitudinal
Patient Database (LPD), which was set up by the Italian College of
General Practitioners and Primary Care in 1998. It is the Italian
largest computerized database of electronic medical records of
patients in primary care. The database was build up from a
network of 1000 GPs homogeneously distributed across Italy, who
agreed to provide information in an electronic database referring
to almost one-and-a-half million of patients. Each GP undergoes
formal training for data entry and uses standard software to record
data. A patients encrypted identier links demographic and
prescription information, clinical events and diagnoses, laboratory
data, specialist referrals, hospital admissions, and date of death.
The diagnosis of certain diseases, such as DD and BD, which need a
specialists consultation, are expectedly coded by GPs whether
conrmed by a specialists referral.
Within this general practice population, 700 GPs were selected
on the basis of the accuracy (completeness and consistency) of the
information registered in order to perform epidemiological
studies. It has been demonstrated that HS IMS Health LPD is
representative of the Italian population after comparison with data
from the Italian National Statistical Institute (ISTAT), where no
signicant difference was observed in terms of demographics (i.e.,
age and sex distributions; www.healthsearch.it, last accessed
October 2014). The role of GPs in Italy should allow inferential
hypotheses on the psychiatric population, as they generally act as
gatekeepers to psychiatric services, even though patients self-
referral to private specialty services is also possible.
HS IMS Health LPD has been extensively used in previously
published epidemiological studies that conrmed its research
validity [1719].
3.2. Study population
A cohort of patients (aged 18 years or older) attending the
selected GPs was enrolled between January, 1, 2002 and June, 30,
2012. The index date was the rst contact with the GP in the
enrolment period. To be eligible, patients required to have at least
one year of medical history prior to or on the index date. Exclusion
criteria were subjects diagnosed with DD or BD preceding the
index date. They were followed up as long as the following events
occurred whichever came rst: diagnoses of DD (event date) or BD
(event date), death, end of the registration with the GP, end of the
study period (June, 30, 2013).
3.3. Event denition
The event under study was the rst diagnosis of DD including
ICD9CM diagnoses of DD, major depression single or recurrent
episode (ICD9 codes: 311, 296.2, 296.3) or BD (referred to both
type I and II) including ICD9CM diagnoses of manic disorder,
single (ICD9 code: 296.0) or recurrent episode (ICD9 code: 296.1),
bipolar affective disorder manic (ICD9 code: 296.4) or depressed
bipolar affective disorder (ICD9 code: 296.5), mixed bipolar
disorder (ICD9 code: 296.6), and unspecied bipolar disorder
(ICD9 code: 296.7) -.
3.4. Predictors of mood disorders
Potential predictors of DD and BD were identied prior or on the
index date. They included age and sex, along with somatization
and lifestyle variables, medical and psychiatric comorbidities.
Somatization variables included somatic conditions, such as
weight loss, joint pain, dizziness, migraine/headache, back pain,
neck pain, amnesic syndrome, fatigue, psychalgia (i.e., chronic pain
without a medical reason), irritable bowel syndrome, constipation,
abdominal pain, pelvic inammatory disease.
Lifestyle variables included smoking, alcohol abuse and related
diseases, and obesity (body mass index higher  30 kg/m2). For
these variables, we adopted the last measurement recorded prior
to or on the index date.
Medical comorbidities included congestive heart failure,
ischemic heart disease, peripheral vascular disease, chronic
respiratory disease, rheumatic diseases, peptic ulcer disease, mild
liver disease, severe liver disease, diabetes with and without
complications, hemiplegia or paraplegia, kidney disease (also
referred to isolated proteinuria or hematuria; or proteinu-
ria > 200 mg/24 hours or > 150 mg/L, the last measurement prior
or on the index date); albuminuria or microalbuminuria ( 30 mg/
24 hours or  3 mg/mmol, the last measurement prior or on the
index date); any value of GFR < 60 mL/min/1.73 m2 (last mea-
surement prior or on the index date), dialysis and/or kidney
 G. Castellini et al. / European Psychiatry 37 (2016) 6369 65

transplantation or free text dialysis or kidney transplant, any
form of cancer, metastatic cancer, stroke, transient ischemic attack,
celiac disease, and senile dementia.
Psychiatric comorbidities included neurotic disorders, stress-
related disorders, substance abuse, anorexia and bulimia nervosa,
and eating disorder not otherwise specied.
3.5. Statistical analysis
Continuous variables were reported as mean  standard devia-
tion, whereas categorical variables were reported as percentage.
Incidence rates for both DD and BD was calculated as the ratio
between number of new cases of the disorders in the follow up period
and person-times cumulated by patients. Disorder-specic inciden-
ces were also calculated.
Cox proportional hazards regression models were used to
estimate the effect of somatization and lifestyle variables as well as
medical and psychiatric comorbidities on time to diagnosis of DD
or BD. Hazard ratios (HRs) with 95% condence intervals (CI) were
calculated. The multivariable model was estimated by a stepwise
backward (P values < 0.10 for entering and < 0.15 for exiting
variables) approach.
We also performed a sensitivity analysis to verify the
robustness of the results. Given that HS IMS Health LPD did not
specically record diagnoses made by specialist physicians, the
outcome denition might be prone to some misclassication. We
therefore rerun the multivariable model by limiting cases of DD
and BD to those with a neurology or psychiatry referral within
3 months before or after the date of diagnosis.
4. Results
4.1. Incidence rates
The nal sample was composed of 1,144,622 persons, including
605,285 women, and 539,337 men. The number of cases diagnosed
prior to the or on the index date was 25,518 for DD and 1728 for BD.
We estimated an incidence rate of DD or BD of 53.61 and 1.5 per
10,000 person-years, respectively. Table 1 depicts the incidence
rates of DD and BD as a unique category. Women show a higher
incidence rate for both DD and BP, as compared to men. This result
is mainly related to the higher number of DD in women for most of
the age intervals, as compared with men (Table 2). For both
DD and BD, the incidence rate rises with the increasing of age
(Tables 3 and 4).
4.2. Predictors of mood disorders
Table 4 depicts patients features which resulted associated with
a diagnosis of DD or BD. Most of the variables entered into the nal
model showed a signicant association with the outcome, with an
overall explained variance of 15.3%. Namely, the adjusted HRs for

Table 1
Incidence rates (*10,000 person-years) of depressive disorder.

Incidence rate (CI 95%) Incidence rate (CI 95%) Incidence rate (CI 95%)
Men (n = 539,337) Women (n = 605,285) Total (n = 1,144,622)

Total 36.93 (36.3637.52) 68.74 (67.9969.5) 53.61 (53.1354.1)

Age categories
1834 20.18 (19.420.99) 39.45 (38.3740.56) 30.02 (29.3530.71)
3544 26.86 (25.827.96) 57.38 (55.8458.96) 42.37 (41.4343.33)
4554 35.50 (34.1736.88) 69.26 (67.4271.16) 52.53 (51.3953.70)
5564 43.09 (41.5244.72) 78.50 (76.4180.66) 61.15 (59.8262.50)
6574 65.17 (62.8867.55) 102.68 (100.09105.34) 85.97 (84.287.78)
7584 94.55 (90.4498.84) 121.52 (117.94125.21) 111.59 (108.86114.40)
 85 85.70 (75.9696.69) 101.49 (94.64108.84) 97.00 (91.3103.04)

Table 2
Incidence rates (*10,000 person-years) for bipolar disorder.

Incidence rate (CI 95%) Incidence rate (CI 95%) Incidence rate (CI 95%)
Men (n = 539,337) Women (n = 605,285) Total (n = 1,144,622)

Total 1.35 (1.251.47) 1.64 (1.521.75) 1.5 (1.421.58)

Age categories
1834 1.2 (1.021.41) 1.38 (1.191.60) 1.29 (1.161.44)
3544 1.59 (1.351.88) 1.88 (1.622.17) 1.74 (1.561.94)
4554 1.30 (1.071.59) 1.95 (1.672.29) 1.63 (1.451.85)
5564 1.37 (1.111.69) 1.61 (1.331.93) 1.49 (1.301.71)
6574 1.52 (1.211.92) 1.73 (1.432.10) 1.64 (1.411.90)
7584 1.08 (0.721.63) 1.29 (0.971.71) 1.21 (0.961.53)
 85 0.64 (0.162.54) 1.00 (0.502.00) 0.90 (0.481.67)

Table 3
Incidence rates (*10,000 person-years) of depressive disorder and bipolar disorder.

Incidence rate (CI 95%) Incidence rate (CI 95%) Incidence rate (CI 95%)
Men (n = 539,337) Women (n = 605,285) Total (n = 1,144,622)

Total 38.21 (37.6238.81) 70.34 (69.5771.11) 55.06 (54.5755.55)

Age categories
1834 21.29 (20.4922.13) 59.11 (57.5660.72) 31.24 (30.5631.95)
3544 28.29 (27.229.42) 71.21 (69.3473.14) 43.95 (42.9944.93)
4554 36.72 (35.3738.12) 80.08 (77.9782.26) 54.11 (52.9555.30)
5564 44.44 (42.8446.10) 104.40 (101.79107.08) 62.61 (61.2763.98)
6574 66.65 (64.3369.05) 122.89 (119.29126.60) 87.58 (85.7989.40)
7584 95.71 (91.58100.04) 102.55 (95.66109.94) 112.89 (110.14115.71)
 85 86.37 (76.5997.40) 59.11 (57.5660.72) 97.95 (92.23104.03)
66 G. Castellini et al. / European Psychiatry 37 (2016) 6369

Table 4
Predictors of mood disorders.

Variable n (%) HR (CI 95%) crude P HR (CI 95%) P

adjusteda

Somatization-related features
Weight loss 1423 (0.1) 1.29 (1.011.65) 0.04 1.2 (0.941.54) 0.14
Joint pain 14,571 (1.3) 1.08 (11.17) 0.05
Insomnia 3974 (0.4) 1.99 (1.762.25) < 0.001 1.16 (0.971.39) 0.10
Hipersomnia 6715 (0.6) 1.82 (1.662) < 0.001 1.2 (1.041.37) 0.01
Migraine/headache 37,081 (3.2) 1.34 (1.291.4) < 0.001 1.32 (1.261.38) < 0.001
Back pain 68,207 (6) 1.1 (1.061.13) < 0.001 1.04 (11.08) 0.03
Amnesic syndrome 6 (0)
Fatigue 25,839 (2.3) 1.29 (1.221.36) < 0.001 1.32 (1.251.39) < 0.001
Psychalgia 3692 (0.3) 1.16 (1.011.33) 0.04 1.14 (0.991.31) 0.07
Irritable bowl syndrome 15,478 (1.4) 1.27 (1.191.35) < 0.001 1.15 (1.081.23) < 0.001
Constipation 10,720 (0.9) 1.48 (1.371.6) < 0.001 1.06 (0.981.15) 0.14
Abdominal pain 23,091 (2.0) 1.31 (1.241.39) < 0.001 1.21 (1.141.28) < 0.001
Neck pain 21,552 (1.9) 1.2 (1.131.27) < 0.001 1.11 (1.051.18) < 0.001
Pelvic inammation disease 266 (0.02) 0.88 (0.51.55) 0.66 1.28 (1.181.38) < 0.001
Dizziness 27,745 (2.4) 1.39 (1.331.46) < 0.001 1.08 (1.031.13) < 0.001
Demographics
Gender
Women 605,285 (52.9) 1.84 (1.811.88) < 0.001 1.79 (1.761.83) < 0.001
Age categories (1834)
35-44 230,441 (20.1) 1.41 (1.371.45) < 0.001 1.4 (1.361.45) < 0.001
45-54 184,604 (16.1) 1.75 (1.691.80) < 0.001 1.73 (1.671.78) < 0.001
55-64 158,709 (13.9) 2.02 (1.962.08) < 0.001 1.97 (1.92.03) < 0.001
65-74 130,811 (11.4) 2.81 (2.732.90) < 0.001 2.63 (2.552.71) < 0.001
75-84 86,621 (7.6) 3.55 (3.443.67) < 0.001 3.16 (3.063.28) < 0.001
 85 26,086 (2.3) 2.95 (2.773.15) < 0.001 2.55 (2.392.72) < 0.001
Lifestyle
Smoking (No)
Current smoker 64,767 (5.7) 1.02 (0.971.06) 0.53 1.35 (1.291.41) < 0.001
Ex smoker 36,949 (3.2) 0.98 (0.931.04) 0.56 1.09 (1.031.16) < 0.001
Missing 938,209 (82.0) 0.94 (0.910.97) < 0.001 1.14 (1.11.18) < 0.001
Obesity (No)
Obesity 33,654 (2.9) 1.27 (1.211.34) < 0.001 1.13 (1.071.19) < 0.001
Missing 992,284 (86.7) 0.97 (0.951) 0.07 1.02 (0.991.06) 0.25
Alcohol consumption 5163 (0.5) 1.52 (1.371.7) < 0.001 1.48 (1.321.65) < 0.001
Medical comorbidities
Congestive heart failure 6238 (0.5) 1.92 (1.742.13) < 0.001 1.12 (1.011.24) 0.03
Ischemic heart disease 10,207 (0.9) 1.48 (1.361.6) < 0.001 1.15 (1.061.24) < 0.001
Stroke 9767 (0.9) 2.04 (1.892.19) < 0.001
Transient ischemic attack 5235 (0.5) 1.94 (1.762.14) < 0.001 1.22 (1.11.34) < 0.001
Peripheral vascular disease 27,509 (2.4) 1.31 (1.251.38) < 0.001 1.1 (1.051.16) < 0.001
Chronic respiratory disease 68,619 (6) 1.4 (1.351.44) < 0.001 1.18 (1.141.22) < 0.001
Reumatic disease 6007 (0.5) 1.94 (1.782.12) < 0.001 1.31 (1.21.43) < 0.001
Peptic ulcer disease 24,860 (2.2) 1.45 (1.381.53) < 0.001 1.19 (1.131.25) < 0.001
Celiac disease 1112 (0.1) 1.38 (1.071.78) 0.01 1.58 (1.232.03) < 0.001
Mild liver disease 8698 (0.8) 1.67 (1.551.81) < 0.001 1.35 (1.251.47) < 0.001
Severe liver disease 431 (0) 1.94 (1.322.84) < 0.001
Diabetes without complications 53,530 (4.7) 1.62 (1.561.68) < 0.001 1.15 (1.11.19) < 0.001
Diabetes without complications 326 (0) 1.36 (0.852.15) 0.20
Hemiplegia or paraplegia 972 (0.1) 1.67 (1.292.15) < 0.001 1.29 (11.67) 0.05
Kidney disease 86,064 (7.5) 1.57 (1.521.61) < 0.001
Cancer 38,441 (3.4) 1.68 (1.611.75) < 0.001 1.2 (1.151.25) < 0.001
Metastatic cancer 470 (0) 1.08 (0.631.87) 0.77
Senile dementia 2100 (0.2) 1.02 (0.771.34) 0.92 0.52 (0.390.69) < 0.001
Psychiatric comorbidities
Insomnia 3974 (0.4) 1.99 (1.762.25) < 0.001 1.16 (0.971.39) 0.10
Hipersomnia 6715 (0.6) 1.82 (1.662) < 0.001 1.2 (1.041.37) 0.01
Neurotic disorders 60,112 (5.3) 1.1 (1.061.14) < 0.001 0.82 (0.790.85) < 0.001
Stress-related disorders 3452 (0.3) 0.76 (0.640.9) < 0.001 0.68 (0.570.81) < 0.001
Substance abuse 9919 (0.9) 1.11 (1.021.21) 0.02
Anorexia nervosa 955 (0.1) 1.63 (1.282.08) < 0.001 1.74 (1.372.23) < 0.001

HR: hazard ratio; CI: condence intervals.

a
According to stepwise backward (P values < 0.10 for entering and < 0.15 for exiting variables).

the predictors rises with the increasing of age, and it results higher
in women than men. Most of the somatic-related features, lifestyle
variables and medical comorbidities increased the risk of mood
disorders. The strongest effect was found for migraine/headache,
fatigue, irritable bowel syndrome and pelvic inammation disease.
In contrast, senile dementia, neurotic and stress-related disorders
seemed to be associated with a reduced risk of mood disorders.
According to the Cox models, a signicant effect of presence of
somatization was observed for time to DD (Fig. 1; crude HR; 95%
CI = 1.24; 1.221.27; adjusted HR; 95% CI = 1.21; 1.181.23) but
not for BD (Fig. 2; crude HR; 95% CI = 1.00; 0.871.15; adjusted HR;
95% CI = 0.96; 0.791.05).
When we reran the nal model using the number of somatic-
related features and medical conditions as covariates, the risk of
 G. Castellini et al. / European Psychiatry 37 (2016) 6369
Fig. 1. Survivial function of time to diagnosis of depressive disorder and presence of
somatization.
Fig. 2. Survivial function of time to diagnosis of bipolar disorder and presence of
somatization.
mood disorders increases with the number of concurrent
conditions. The highest number of conditions did not show
signicant associations likely because of lack of statistical power
[n = 32 and 25 exposed patients for somatization-related features
(6) and medical comorbidities (78), respectively (Supplemental
Table 1)].
With the sensitivity analysis, when predictors were estimated
by restricting cases of DD and BD coupled with specialist referral,
the results were generally consistent with those obtained for the
primary analysis (Supplemental Table 2).
5. Discussion
To the best of our knowledge, this is the rst study, which
provides information about incidence rate and predictors of both
DD and BD in the primary care setting. The main results of this
study were as follows:
 a 53.61 and 1.5 per 10,000 person-years were the incidence rates
of DD and BD, respectively, being found in a large cohort of
primary care;
 several patients clinical features were found to be associated
with development of mood disorders, especially with several
somatic symptoms.
Our results showed some differences or similarities with those
reported in previous population-based studies. Namely, the
67
European Study of the Epidemiology of Mental Disorders
(ESEMeD), a cross-sectional survey carried out on representative
samples of 21,425 individuals report a 12-month prevalence of
depressive disorders of 4.4% [20]. Each year in the UK, an estimated
6% of adults experience an episode of depression, and more than
15% of the population will experience an episode of depression
during their life [5]. Wittchen et al. [21] reported a BD cumulative
lifetime incidence risk for community subjects up to age 30 of 2.6%,
while other community studies estimated a lifetime prevalence of
mania ranging from 2.0% to 6.0% [3,4]. Differences in rate could
largely be explained by design and assessment differences, given
that HS IMS Health LPD provides information regarding a larger
number of persons mostly based on GPs reports. In general, the
present study conrmed an under-recognition of BD, considering
that the estimate for incidence of DD is 36-fold as compared with
that of BD.
According to recent data from other general practice research
databases, the incidence of mood disorders varies according to age
and sex [1]. The gender difference was mostly determined by a
higher incidence rate of DD in women, thus conrming previous
observations, which report a women proportion of life expectancy
with mental disorders more than double those for the male
population [22,23]. Female gender is one of the most consistent
and stable factors associated with depression. An overall 12-month
female/male ratio of 2/1 has been described in the literature [23].
As far as age effect is concerned, our results conrm those from
the Spanish group of ESEMeD study, which showed an increased
risk from the youngest to the 5064 age cohort [24]. This is in
contrast with data from the northern European countries of
ESEMeD network, where the highest rates of mood disorders were
found in the youngest age groups (1824 years old) with a
signicant decline with age [25]. Though some of the differences
described may be due to different methodological approaches, it is
possible that differences in the prevalence of mood disorders
between countries and cultures might really exist. Cross-cultural
meaningful contrasts between Mediterranean countries and the
more secularized and economically developed northern Europe [26]
have been documented and they might have differential effects on
mental health of populations [24]. A possible protective effect on
later age of the more traditional Mediterranean or Latin culture has
been proposed [27]. In southern Europe, a strong social support is
offered by family cohesion or the extended social networks [28].
Within lifestyle variables, smoking, and alcohol abuse were
found to be associated with increasing risk for mood disorders.
These addictive behaviors preceding the onset of depression can be
considered as behavioral markers of a predisposing personality to
mood disorders. Indeed, addictive behaviors are adopted as a way
to manage their negative emotions. Therefore, it is possible that
persons at high risk for mood disorders show a decient emotional
regulation personality trait, and substance abuse represents a
dysfunctional coping strategy to cope with dysphoric mood (i.e.,
self-medication).
Somatic symptoms were strongly and independently associated
with diagnosis of mood disorders. The strongest effect was found for
migraine/headache, fatigue, irritable bowel syndrome and pelvic
inammation disease. These results can be interpreted in different
ways. It is true that the separate HRs seem to suggest a modest
association between individual somatic conditions and mood
disorders. However, there was a growing risk of mood disorders
by increasing the number of somatic symptoms. This nding was
consistent with those yielded with the survival functions. From one
side, somatic symptoms can be considered as early signs of
depression, thus conrming previous observations reporting that
many patients with depression initially present with painful
physical symptoms (e.g., migraine/headache, musculoskeletal
pain). In this case, the association of physical symptoms with an
68 G. Castellini et al. / European Psychiatry 37 (2016) 6369
earlier mood disorders diagnosis seems to support the hypothesis
that somatization represents a handle for depressed patients to
formulate a generic request of help in the primary care setting. As a
further support to this hypothesis, it has been demonstrated that
painful physical symptoms were a signicant predictors of help-
seeking for emotional problems [28]. From a different perspective, it
is possible that the painful physical condition characterized by a
generic cluster of somatic symptoms, such as headache, musculo-
skeletal pain, fatigue, irritable bowel is a strong predisposing
factor for an imminent onset of depression or possibly BD. In any
case, our results seem to contradict previous studies reporting that
somatic symptoms negatively affect the recognition of depression
in primary care and represent a possible reason for poor detection
or misdiagnosis of mood disorders [8,10]. Rather they seem to
represent a valid tool for an early recognition of depression and for a
rst request of help to GPs.
Regarding a possible mechanism underlying the relationship
between pain and depression, a previous research has shown that
painful symptoms from physical conditions are rmly associated
with reductions in role functioning and thus leading to depression
[8]. Indeed, chronic pain due to age-related physical conditions, like
arthritis and diabetes mellitus is common in older people as well as
in mood disorders [29]. Therefore, when pain and depression are
both present, they impact on each other [13], and it may reect
common underlying neurobiological mechanisms, associated with
inammatory mechanisms and ageing [1,8].
It can be argued that somatic symptoms can be attributed to
medical comorbidities, which also associated with depression
onset. However, the performed multivariate models, including
somatic symptoms and medical conditions together, conrm a
direct effect of non-specic somatic symptoms on depression,
independently from medical comorbidities. This represents a
strength of our analyses, as compared with previous studies which
assessed multiple physical conditions without adjusting for the
substantial co-morbidity [30]. Indeed, our data showed that
patients with heart disease and chronic medical conditions, such as
diabetes and chronic respiratory disease represent a specic
population at high risk for developing DD or BD. These ndings are
in line with the psychosomatic literature, which showed that
medical conditions were often found to be associated with
depression [14]. Several studies demonstrated a strong relation-
ship between psychological distress and age-related diseases, such
as osteoporosis, metabolic syndrome [31], cardiovascular disease
[32], and premature all-cause mortality [33]. The pathogenic
processes of depression and medical morbidity are interlinked and
often recursive. Indeed, poor physical health is considered an
important risk factor for depression in later life [34]. It has been
suggested that a chronic condition, by reducing general health,
may lead to the development of depressive symptoms, through an
overall deterioration of quality of life [35].
The strength and contribution of the present study was the
collection of both medical and psychiatric information from a
computerized general practice database. Whereas in Italy, national
statistics on morbidity are based on self-reported information, and
only few researchers have investigated risk factors of mood
disorders using data from medical records [18].
The adjusted HR of 0.82 (0.790.85) for anxiety disorders
should not be interpreted like a protective effect on the diagnosis
of unipolar DD and BD, considering the well-known high co-
morbidity between mood and anxiety disorders. Rather, it is
possible that patients previously diagnosed for anxiety disorders
generally received a more intensive care, including treatment with
antidepressants, which could exert a protective effect on the onset
of mood disorders. Furthermore, the results can be interpreted also
as many depression and bipolar disorder symptoms are usually
mistaken for other psychiatric disorders, such as anxiety and
personality disorders [36]. Misdiagnosis and consequent prescrib-
ing of contra-indicated (e.g., antidepressant monotherapy) or
unnecessary medications for persons with BD can exacerbate the
frequency and severity of mood symptoms as well as promote
treatment resistance to appropriate medications [37].
The rst important limitation of the present study is the
homogeneity and the validity of GPs psychiatric diagnoses. This
may account for the differences with the other population-based
studies. However, the diagnosis of DD and BD, which need a
specialists consultation, are expectedly coded by GPs whether
conrmed by a specialists referral. Indeed, when we estimated
predictors by restricting cases of DD or BD coupled with specialist
referral, the results were consistent with those obtained in the
primary analyses. Second, it can be argued that psychiatric patients
generally do not seek treatment from health care. In particular, our
data might be biased because psychiatric patients referring to GP
could be more represented within middle-aged or older persons.
However, the representativeness of the database is demonstrated
by the fact that demographics of registered population are quite
similar to those of general population of Italy (www.healthsearch.it,
last accessed October 2014). Furthermore, age was a signicant
predictor for all age categories when compared with 1524
subgroup, along with variables whose association with mood
disorders (e.g. heart failure) has been largely demonstrated in
literature [1]. As such, the bias due to outcome underestimation
should be minimal. Furthermore, middle-aged persons are the most
easily referred to the GPs attention for psychiatric diagnoses. Third,
the present study was planned adopting ICD9 criteria, and we did
not convert diagnoses into ICD10 to avoid sources of possible errors.
However, the results are coherent with what we expected, given the
consolidated use of this thesaurus among GPs part of HS IMS Health
LPD network. Finally, we were unable to include the socio-economic
status, which has been reported as an important factor associated
with mood disorders [38,39]. However, our nal model included
37 predictors, with an explained variance of 15.3%, and given the
relevance of multimorbidity in patients with mood disorders [39],
we believe that our ndings maintain their clinical utility.
6. Conclusions
The clinical implication of the present study is that GPs should
consider a cluster of predisposing factors for mood disorder in their
clinical practice. They should be aware of the development of DD or
BD when patients experience substantial disability from chronic
medical conditions, or reporting smoking, obesity and alcohol
abuse. In particular, the reporting of somatic symptoms may not
reect an unwillingness or inability to acknowledge emotional
distress. Instead, it could represent a rst sign of masked mood
disorders, and patients may believe that the reporting of somatic
symptoms is a more appropriate route for seeking help from a GP.
Algorithms based on an accurate description of predisposing
factors for identication of DD and BD in primary care may
improve the clinical practice. Therefore, high quality care for mood
disorders in primary care settings would minimize their clinical
and economic impact on patients and society.
Disclosure of interest
The authors have not supplied their declaration of competing
interest.
Appendix A. Supplementary data
Supplementary data associated with this article can be found, in
the online version, at http://dx.doi.org/10.1016/j.eurpsy.2016.04.
003.
 G. Castellini et al. / European Psychiatry 37 (2016) 6369 69

depressive disorders in the general population: results with three alternative

References
scoring methods. Value Health 2013;16:56473.
[21] Wittchen HU, Mhlig S, Pezawas L. Natural course and burden of bipolar
[1] Reed C, Hong J, Novick D, Lenox-Smith A, Happich M. Health care costs before
disorders. Int J Neuropsychopharmacol 2003;6:14554.
and after diagnosis of depression in patients with unexplained pain: a retro-
[22] Brugha TS, Matthews R, Alonso J, Vilagut G, Fouweather T, Bruffaerts R, et al.
spective cohort study using the United Kingdom General Practice Research
Gender differences in mental health expectancies in early- and midlife in six
Database. Clinicoecon Outcomes Res 2013;5:3747.
European countries. BJP 2013;202:294300.
[2] Wittchen HU, Jacobi F, Rehm J, Gustavsson A, Svensson M, Jonsson B, et al. The
[23] Fryers T, Brugha T, Morgan Z, Smith J, Hill T, Carta M, et al. Prevalence of
size and burden of mental disorders and other disorders of the brain in Europe
psychiatric disorder in Europe: the potential and reality of meta-analysis. Soc
2010. Eur Neuropsychopharmacol 2011;21:65579.
Psychiatry Psychiatr Epidemiol 2004;39:899905.
[3] Pini S, de Queiroz V, Pagnin D, Pezawas L, Angst J, Cassano GB, et al. Prevalence
[24] Gabilondo A, Rojas-Farreras S, Vilagut G, Haro JM, Fernandez A, Pinto-Meza A,
and burden of bipolar disorders in European countries. Eur Neuropsychophar-
et al. Epidemiology of major depressive episode in a Southern European
macol 2005;15:42534.
Country: results from the ESEMeD-Spain Project. J Affect Disord 2010;120:
[4] Mitchell PB, Ball JR, Best JA, Gould BM, Malhi GS, Riley GJ, et al. The manage-
7685.
ment of bipolar disorder in general practice. Med J Aust 2006;184:56670.
[25] Alonso J, Bruffaerts R, Gabilondo A, Haro JM, Kovess V, Vilagut G. ESEMED/
[5] Vedavanam S, Steel N, Broadbent J, Maisay S, Howe A. Recorded quality of care
MHEDEA Investigators. European Commission Task Force on Major and
for depression in general practice: an observational study. Br J Gen Pract
Chronic Diseases of DG SANCOs Health Information Strand Major and Chronic
2009;59:327.
Diseases Report 2007. Luxembourg: European Commission. Depression 2008;
[6] Fleury MJ, Imboua A, Aube D, Farand L, Lambert Y. General practitioners
10116.
management of mental disorders: a rewarding practice with considerable
[26] Saroglou V, Delpierre V, Dernelle R. Values and religiosity: a meta-analysis of
obstacles. BMC Fam Pract 2012;13:19.
studies using Schwartzs model. Pers Individ Dif 2004;37:72134.
[7] Walters P, Tylee A, Goldberg D, Murray RM, Kendler KS, McGufn P, et al.
[27] Fernandez-Cordon JA, Sgritta GB. The southern countries of the European
Psychiatry in Primary Care. Essential Psychiatry, 4th ed., UK: Cambridge
Union: a paradox?In: Paper presented at the Annual seminar of The European
University Press; 2008. p. 47997.
Observatory on the Social Situation, Demography and Family. Low Fertility,
[8] Bair MJ, Robinson RL, Eckert GJ, Stang PE, Croghan TW, Kroenke K. Impact of
families and public policies. 2000. p. 156.
pain on depression treatment response in primary care. Psychosom Med
[28] Levi F, LaVecchia C, Lucchini F, Negri E, Saxena S, Maulik PK, et al. Trends in
2004;66:1722.
mortality from suicide, 196599. Acta Psychiatr Scand 2003;108:3419.
[9] Bair MJ, Robinson RL, Katon W, Kroenke K. Depression and pain comorbidity: a
[29] Stegmann ME, Ormel J, de Graaf R, Haro JM, de Girolamo G, Demyttenaere K,
literature review. Arch Intern Med 2003;163:243345.
et al. Functional disability as an explanation of the associations between
[10] Ohayon MM, Schatzberg AF. Chronic pain and major depressive disorder in the
chronic physical conditions and 12-month major depressive episode. J Affect
general population. J Psychiatr Res 2010;44:45461.
Disord 2010;124:3844.
[11] Kroenke K, Wu J, Bair MJ, Krebs EE, Damush TM, Tu W. Reciprocal relationship
[30] Patten SB. Long-term medical conditions and major depression in a Canadian
between pain and depression: a 12-month longitudinal analysis in primary
population study at waves 1 and 2. J Affect Disord 2001;63:3541.
care. J Pain 2011;12:96473.
[31] McIntyre RS, Soczynska JK, Konarski JZ, Woldeyohannes HO, Law CW, Miranda
[12] Cosci F, Fava GA, Sonino N. Mood and anxiety disorders as early manifestations
A, et al. Should depressive syndromes be reclassied as metabolic syndrome
of medical illness: a systematic review. Psychother Psychosom 2015;84:229.
type II? Ann Clin Psychiatry 2007;19:25764.
[13] Garcia-Cebrian A, Gandhi P, Demyttenaere K, Peveler R. The association of
[32] Schulz R, Beach SR, Ives DG, Martire LM, Ariyo AA, Kop WJ. Association
depression and painful physical symptoms a review of the European
between depression and mortality in older adults: the Cardiovascular Health
literature. Eur Psychiatry 2006;21:37988.
Study. Arch Intern Med 2000;160:17618.
[14] Pasquali R. The biological balance between psychological well-being and
[33] Joynt KE, Whellan DJ, OConnor CM. Depression and cardiovascular disease:
distress. Psychother Psychosom 2006;75:6971.
mechanisms of interaction. Biol Psychiatry 2003;54:24861.
[15] Simon GE, VonKorff M, Piccinelli M, Fullerton C, Ormel J. An international
[34] Ormel J, Rijsdijk FV, Sullivan M, van SE, Kempen GI. Temporal and reciprocal
study of the relation between somatic symptoms and depression. N Engl J Med
relationship between IADL/ADL disability and depressive symptoms in late
1999;341:132935.
life. J Gerontol B Psychol Sci Soc Sci 2002;57:33847.
[16] Aguera L, Failde I, Cervilla JA, Daz-Fernandez P, Mico JA. Medically unex-
[35] Eckhardt MD, van Venrooij GE, van Melick HH, Boon TA. Prevalence and
plained pain complaints are associated with underlying unrecognized mood
bothersomeness of lower urinary tract symptoms in benign prostatic hyper-
disorders in primary care. BMC Family Pract 2010;117.
plasia and their impact on well-being. J Urol 2001;66:5638.
[17] Cricelli C, Mazzaglia G, Samani F, Marchi M, Sabatini A, Nardi R, et al. Preva-
[36] Stensland MD, Schultz JF, Frytak JR. Depression diagnoses following the
lence estimates for chronic diseases in Italy: exploring the differences be-
identication of bipolar disorder: costly incongruent diagnoses. BMC Psychi-
tween self-report and primary care databases. J Public Health Med 2003;25:
atry 2010;1039.
2547.
[37] Smith DJ, Thapar A, Simpson S. Bipolar spectrum disorders in primary care:
[18] Lapi F, Simonetti M, Michieli R, Pasqua A, Brandi ML, Frediani B, et al. Assessing
optimising diagnosis and treatment. Br J Gen Pract 2010;60:3224.
5-year incidence rates and determinants of osteoporotic fractures in primary
[38] Smith DJ, Court H, McLean G, Martin D, Langan Martin J, Guthrie B, et al.
care. Bone 2012;50:8590.
Depression and multimorbidity: cross-sectional study of 1,751,841 patients in
[19] Triro` G, Morabito P, Cavagna L, Ferrajolo C, Pecchioli S, Simonetti M, et al.
primary care. J Clin Psychiatry 2014;75(11):12028.
Epidemiology of gout and hyperuricaemia in Italy during the years 20052009:
[39] Smith DJ1, Martin D, McLean G, Langan J, Guthrie B, Mercer SW. Multimorbi-
a nationwide population-based study. Ann Rheum Dis 2013;72:694700.
dity in bipolar disorder and under-treatment of cardiovascular disease: cross
[20] Vilagut G, Forero CG, Pinto-Meza A, Haro JM, de Graaf R, Bruffaerts R, et al. The
sectional study. BMC Med 2013;11(1):263.
mental component of the short-form 12 health survey (SF-12) as a measure of