Content uploaded by Jaap Peen
Author content
All content in this area was uploaded by Jaap Peen on Jul 02, 2018
Content may be subject to copyright.
Review
The current status of urban-rural differences
in psychiatric disorders
Introduction
Generally, social problems and environmental
stressors are more prevalent in cities than in the
country. Areas with high population densities are
characterized, for instance, by higher rates of
criminality, mortality, social isolation, air pollution
and noise (1). As the extent of various social
problems is related to urbanization, it is often
assumed that rates of psychiatric disorders are also
correlated with urbanization. A frequently cited
milestone in the study area of urban–rural differ-
ences in the prevalence of psychiatric disorders is
the study by Dohrenwend and Dohrenwend (2).
Peen J, Schoevers RA, Beekman AT, Dekker J. The current status of
urban–rural differences in psychiatric disorders.
Objective: Reviews of urban–rural differences in psychiatric disorders
conclude that urban rates may be marginally higher and, specifically,
somewhat higher for depression. However, pooled results are not
available.
Method: A meta-analysis of urban–rural differences in prevalence was
conducted on data taken from 20 population survey studies published
since 1985. Pooled urban–rural odds ratios (OR) were calculated for
the total prevalence of psychiatric disorders, and specifically for mood,
anxiety and substance use disorders.
Results: Significant pooled urban–rural OR were found for the total
prevalence of psychiatric disorders, and for mood disorders and
anxiety disorders. No significant association with urbanization was
found for substance use disorders. Adjustment for various confounders
had a limited impact on the urban–rural OR.
Conclusion: Urbanization may be taken into account in the allocation
of mental health services.
J. Peen
1,2
, R. A. Schoevers
1
,
A. T. Beekman
3
, J. Dekker
1,2
1
Research Department, Arkin Mental Health Institute
Amsterdam,
2
Department of Clinical Psychology, VU
University Amsterdam and
3
Department of Psychiatry,
VU University Amsterdam Medical Centre, Amsterdam,
the Netherlands
Key words: meta-analysis; mental illness; prevalence;
rural health; urban health
J. Peen, Research Department, Arkin Mental Health
Institute Amsterdam, PO Box 75848, 1070 AV,
Amsterdam, the Netherlands.
E-mail: jaap.peen@arkin.nl
Accepted for publication June 10, 2009
Summations
•Pooled total prevalence rates for psychiatric disorders were found to be significantly higher in urban
areas compared with rural areas. Specific pooled rates for mood disorders and anxiety disorders were
also significantly higher in urban areas, while rates for substance use disorders did not show a
difference.
•Adjustment for confounders had limited impact on urban–rural odds ratios found, which shows that
urban–rural differences in prevalence rates are only partly explained by population characteristics.
•Urbanization may be taken into account in the allocation of mental health services.
Considerations
•There was heterogeneity in the dataset which might not be explained by urban–rural differences.
However, possible sources of this heterogeneity that were analysed (culture, diagnostic method,
diagnostic variation within diagnostic categories analysed) did not show significant differences in
outcome.
•The meta-analysis was limited to developed countries.
•Schizophrenia was not included as a separate category.
Acta Psychiatr Scand 2010: 121: 84–93
All rights reserved
DOI: 10.1111/j.1600-0447.2009.01438.x
2009 John Wiley & Sons A/S
ACTA PSYCHIATRICA
SCANDINAVICA
84
This review of nine urban–rural comparisons was
based on studies from 1942 to 1969 from quite
diverse countries. The authors concluded that there
was a tendency towards higher total rates of
psychiatric disorders in urban areas. However,
there was a variation in the difference depending
upon the specific diagnostic category. Rates for
neurosis and personality disorders were higher in
urban areas, while rates for functional psychoses
combined and manic-depressive psychoses sepa-
rately were higher in rural areas. There was no
clear trend in the rates for schizophrenia.
Since Dohrenwend and Dohrenwend (2) a
number of reviews have followed (3–7), generally
showing marginally higher overall rates in urban
areas and, specifically, somewhat higher rates for
depression. However, there is no clear trend in the
outcomes, which often lack statistical significance
as the studies were not pooled.
Furthermore, a number of factors may have
complicated the study of a possible association
between urbanization and psychopathology. First
of all, definitions of ÔurbanÕand ÔruralÕmay vary
(4). Generally, ÔurbanÕrefers to large conglomer-
ates of people, usually in a relatively small area,
resulting in relatively high population densities.
The use of the term ÔrelativelyÕmakes it clear that
what some countries define as ÔurbanÕusing defi-
nitions from national statistical institutions or
research may be defined as ÔruralÕin another
country. The United Nations have defined an
Ôurban localityÕas having at least 20 000 people,
and a city as having at least 100 000 people (8).
However, this definition was not used in any study
cited here. Secondly, the concrete manifestation of
urban and rural phenomena varies widely around
the world. The Netherlands, for instance, does not
have any metropolis such as London or New York,
and the Dutch countryside is much more popu-
lated than the countryside of Arkansas.
Thirdly, there may be other cultural differences
between studies and countries. The Dohrenwend &
Dohrenwend review (2) covers a wide variety of
cultures (7), and this may detract from the external
validity of its findings.
Fourthly, there is considerable heterogeneity in
the methods used in the available literature.
Outcome measures vary from self-report psycho-
logical wellbeing scales to case definition by struc-
tured interviews, and prevalence rates may or may
not be adjusted for different types of confounders.
Since 1984, study designs have gradually improved,
enhancing the validity of results. The five reviews
from Dohrenwend and Dohrenwend to Marsella
(2–7) were based partially on older designs, and
partially on more recent, and more sophisticated
designs. In line with this heterogeneity, none of the
previous reviews was able to pool the data and
perform meta-analyses.
Aims of the study
This study sought to investigate the links between
urbanization and psychopathology in a meta-
analysis using only studies of higher methodolog-
ical quality with adjustment for important con-
founders. Bias through cultural and environmental
variation was limited by including only studies
from developed countries. This allowed us to
establish more accurately the existence and mag-
nitude of potential urban–rural differences in levels
of psychopathology. Establishing urban–rural dif-
ferences for psychiatric disorders not only has
scientific value – by extending our models with
factors that affect the onset of mental disorders –
but may also have consequences for the allocation
of mental health resources to areas with higher
levels of urbanization.
Material and methods
Selection criteria
We included population surveys presenting urban–
rural differences in psychiatric disorders since 1985.
We restricted our study to developed countries.
The studies included were all based on reliable
diagnostic processes using standardized structured
interviews.
We present studies dealing with total rates of
psychiatric morbidity, mood disorders, anxiety
disorders and substance use disorders. For Ômood
disordersÕ, rates for major depressive episodes were
used when available. In the absence of rates for
major depressive episodes, rates for combinations of
mood disorder were used. In the area of Ôsubstance
use disordersÕ, rates for alcohol abuse or alcohol
dependence (combined in some cases) were used
when there were no total rates for substance use
disorders. As stated above, there was variation in
the diagnostic content within the diagnostic groups
of which prevalence rates were pooled in this study.
The rationale for this was that we wanted to include
a reasonable number of studies in each diagnostic
group. Furthermore, we have performed additional
analyses if possible, to check for within-group
variation in urban–rural associations due to differ-
ences in diagnostic content.
As reliable rates are generally difficult to estab-
lish for schizophrenia in standard population
surveys due to the low prevalence of schizophrenia
in the non-institutionalized community, we did not
Urban–rural differences
85
include results for schizophrenia. Finally, we
included only studies of adults or of all age groups.
Search strategy
Our database search comprised all publications
from 1985 onwards containing the subject headings
Ômental healthÕor Ômental disordersÕand i) ÔurbanÕ
and ÔruralÕor with ii) Ôcity residenceÕ,Ôcity bornÕ,
Ôcity livingÕ. The databases used were: all EBM
reviews, Embase psychiatry, Medline and Psycinfo.
A selection based on the abstracts was made from
the initial search results (n= 620). Studies con-
cerning less developed countries were also left out.
We were left with 110 studies relating to the
subject. Figure 1 shows the subsequent stepwise
exclusion process.
Data extraction and statistical analysis
All the selected studies provided basic urban and
rural prevalence rates or urban–rural odds ratios
(OR) which had been at least controlled for age
and gender. However, most studies also presented
rates or OR adjusted for a wider range of variables
(these are summed up for each study in Table 1). In
this study, we refer to the first group of rates as
Ôunadjusted ORÕ(controlled for age and gender at
best) and to the second group as Ôadjusted ORÕ
(adjusted for more than age and gender). If
available, a 12-month rate was chosen as the
outcome measure. Another available rate was used
in other cases.
Unadjusted and adjusted OR with 95% confi-
dence intervals were collected for all included
studies. Some OR and confidence intervals could
be calculated from the available numbers, even
though they were not stated in the studies. Some
stated only that there was no significant difference
for urbanization or that urbanization was not a
significant predictor in a logistic model. An OR of
1 is used for these cases in the figures.
When studies provided more than two categories
of urbanization, the most extreme dichotomy –
metropolis vs. rural, for instance – was chosen for
the analysis. In all selected studies, the level of
urbanization concerns the level or urbanization at
the time of measurement.
The Review Manager (RevMan 4.2, Cochrane
IMS, Oxford, UK) was used to perform meta-
analyses. Log OR and their standard errors were
entered in the program. The generic inverse vari-
ance option was used. Pooled ORs were estimated
using random effect modelling as there was a high
level of heterogeneity between included studies.
Two authors (JP and JD) acting independently
were responsible for the reading and the extraction
of data (including cross-checking) from the studies
selected for the meta-analysis. Any differences in
outcome were resolved by discussion.
Findings
Table 1 lists the 20 studies that were included.
Looking at the number of studies per country,
Great Britain, the Netherlands, Canada and the
USA appear to be well represented. As far as the
year of publication is concerned, 12 of the 20
studies were published after 2000, six in the 1990s
and two in the period 1985–1989. Two European
multi-country studies are presented in the table.
The first is the ODIN study of depression covering
Norway, Finland, Great Britain and Ireland (13).
The second is the Esemed study covering France,
Italy, Spain, Belgium, Germany and the Nether-
lands (15). Most studies presented 12-month
prevalence rates (13 ⁄20). The age ranges ‡18 and
18–64 years were most common. Ten studies used
the composite international diagnostic interview as
the diagnostic instrument, three studies used the
general health questionnaire screening instrument
(other n= 7). The distinction between urban and
rural areas was made in different ways. Straight-
forward approaches are Ôinterviewer judgementÕ
(separately for each respondent), Ôpopulation
sizeÕand Ôpopulation densityÕ.ÔConcentration of
addressesÕis a measure of human activity, includ-
ing industrial activity. ÔDemographic characteris-
ticsÕwas also used for area classification.
Most studies used two categories to differentiate
between urban and rural (the maximum number of
Selected from database/literature search: 110
1) No urban–rural comparison related to the subject: 58
2) No population survey (utilization data): 11
3) Restricted to a diagnostic group outside our focus: 10
4) Restricted to a demographic subgroup: 1
5) No clear urban–rural distinction: 1
6) No dichotomous outcome measure: 1
7) Duplicate use of same data in different publications: 8
R
emaining urban–rural comparisons for meta–analysis: 2
0
Fig. 1. The selection process within the initial search result.
Peen et al.
86
Table 1. Population-based prevalence studies included in meta-analysis of urban–rural differences in psychiatric disorders
Studies
Year of
publication Country
Disorder(s)
(unadjusted
rate)*
Disorder(s)
(adjusted
rate)*
Outcome
measure
Age
range
(years)
Screening ⁄
diagnostic
instrument
Classification
system
Sample
size
Urban ⁄rural
categorization
based on
No.
categories Adjusted for
Europe
Madianos & Stefanis
(9)
1992 Greece 2 – Point prev 18–64 CES-D DSM-III-R 3706 Demogr. charact. 4 –
Hodiamont et al. (10) 1992 Netherlands 1 – Point prev 18–65 GHQ ⁄PSE – 3232 Demogr. charact. 2 –
Lewis & Booth (11) 1994 Great Britain 1 1 Point prev >18 GHQ – 6572 Interviewer judgement 3 1,2,4,8,chronic illness
Paykell et al. (12) 2000 Great Britain 1.4 1.4 1 wk prev 16–64 CIS-R ⁄US-NAS – 9777 Interviewer judgement 3 1,2,3,4,5,6,8,life events,prim. supp.
group, perceived soc.
support,tenure,accomm. type
Ayuso-Mateos
et al. (13)
2001 Finland, Great
Britain, Ireland,
Norway
2 – 12 m prev 18–64 BDI ⁄SCAN DSM-IV 7622 Demogr. charact. 2 –
Kovess-Masfety
et al. (14)
2005 France 2 2 12 m prev ‡18 CIDI-S DSM-IV 2628 Demogr. charact. 2 1,2,3,life events
Kovess-Masfety
et al. (15)
2005 Belgium, France,
Germany, Italy,
The Netherlands,
Spain
1,2,3,4 1,2,3,4 12 m prev ‡18 CIDI DSM-IV 21425 Pop. size 2 1,3,8
Weich et al. (16) 2006 Great Britain 1 1 12 m inc 16–74 GHQ – 7659 Pop. density ⁄demogr.
charact.
2 1,2,3,4,6,8,9,curr. health
probl.,housing tenure,
overcrowding,housing probl.,
househ. type
Kringlen et al. (17) 2006 Norway 1,2,3,4 – 12 m prev 18–65 CIDI DSM-III-R 3146 Demogr. charact. 2 –
Peen et al. (18) 2007 The Netherlands 1,2,3,4 1,2,3,4 12 m prev 18–64 CIDI DSM-III-R 7076 Concentration of
addresses
5 1,2,5,9,occup. status,househ.
comp.
Dekker et al. (19) 2008 Germany 1,2,3,4 1,2,3,4 12 m prev 18–64 CIDI DSM-IV 4181 Pop. size ⁄demogr.
charact.
2 1,2,3,4, and interactions with urb.
North America
Blazer et al. (20) 1985 United States 2,3,4 2,3,4 12 m prev ‡18 DIS DSM-III 3798 Demogr. charact. 2 1,2,3,5,7,residential mob.
Kovess et al. (21) 1987 Canada 2 2 12 m prev ‡18 SCL29 ⁄Wellb DSM-III 3080 Demogr. charact. 3 1,2,3,5,life events
Kessler et al. (22) 1994 United States – 1,2,3,4 12 m prev 15–54 CIDI DSM-III-R 8098 Demogr. charact. 3 1,2,3,5,7,living arrangem.,region
Parikh et al. (23) 1996 Canada 2 – 12 m prev ‡15 UM-CIDI DSM-III-R 9953 Demogr. charact. 2 –
Wang (24) 2004 Canada 2 2 12 m prev ‡12 CIDI DSM-III-R 17244 Demogr. charact. ⁄pop.
density
2 3,7,8,immigr. st.
Kessler et al. (25) 2005 United States – 2,4 12 m prev ‡18 CIDI DSM-IV 3199 Demogr. charact. 6 1,2,3,5,6,7,9
Rohrer et al. (26) 2005 United States 1 1 1 m prev ‡18 BRFSS FMD 5757 Demogr. charact. 3 1,2,3,5,7,9,BMI
Other
Lee et al. (27) 1990 South Korea 1,2,3,4 – Lifet. prev 18–65 DIS-III DSM-III 5100 Demogr. charact. 2 –
Andrews et al. (28) 2001 Australia 1,2,3,4 1,2,3,4 12 m prev ‡18 CIDI ICD-10 10641 Pop. size 3 1,2,3,5,8,country of birth
*1 = total rate of psychiatric disorders; 2 = mood disorders; 3 = anxiety disorders; 4 = substance use disorders.
1 = age; 2 = gender; 3 = marital status; 4 = social class; 5 = educational level; 6 = ethnicity; 7 = race; 8 = unemployment; 9 = income.
CIBI, composite international diagnostic interview; GHQ, general health questionnaire.
Urban–rural differences
87
categories used was six). Eighteen of the 20 studies
presented unadjusted OR, while 14 out of 20
presented adjusted ratios (12 presented both). Of
the six studies without adjusted ratios, four dated
from before 2000. Adjusted odds were generally
adjusted for a large number of confounders (up to
a maximum of 14). In Wang (24), the adjusted
odds were not adjusted for age and sex in a logistic
regression model, because these factors were not
found to be a potential confounder in a preceding
bivariate analysis.
In Table 2 the contents of the prevalence rates
used in the pooled analyses are specified. Concern-
ing prevalence rates for Ôany disorderÕsome rates
were based on diagnoses while other rates were
based on cut-off scores. Concerning mood disor-
ders some studies report total prevalence rates for
mood disorders, while other studies report figures
of major depression plus dysthymia or only major
depression. Two of the studies reporting anxiety
disorders only reported prevalence rates of distinct
anxiety categories, as a total of anxiety disorders
was not available. The studies reporting on sub-
stance use disorders can be divided in a group
reporting on both alcohol and drug abuse and
dependence, and in a second group only reporting
on alcohol abuse and dependence.
Figure 2 presents a forest plot of unadjusted OR
for Ôany disorderÕ(16 comparisons), ordered by year
of publication. The number of comparisons from
European countries was much higher (n= 13) than
from outside Europe (n= 3). Of the unadjusted
OR, 56% indicated an urban–rural OR significantly
higher than 1. Thirty-eight per cent of the studies
presented no significant OR and one Belgian study
(6%) found an urban–rural OR significantly less
than 1 (15). Given the heterogeneity of the 14
studies, we used random effect modelling for the
pooled result. The pooled unadjusted OR was 1.38
(1.17–1.64), P< 0.001. The pooled adjusted OR
was slightly lower: 1.21 (1.09–1.34), P< 0.001 (14
comparisons; data not shown).
Figure 3 shows the unadjusted OR for mood
disorders (21 comparisons). By contrast to the
unadjusted odds for Ôany disorderÕ, the propor-
tion of non-European comparisons was higher
(29%; n= 6 non-European and n= 15 Euro-
pean). Thirty-three per cent of the studies found
a significant urban–rural unadjusted OR higher
than 1 for urban areas compared to rural areas,
Table 2. Specific contents of prevalence rates used in the meta-analysis of urban–rural differences in psychiatric disorders
Studies
Unadjusted rates Adjusted rates
Total
Mood
disorders
Anxiety
disorders
Substance
use disorders Total
Mood
disorders
Anxiety
disorders
Substance
use disorders
Europe
Madianos & Stefanis (9) 1a
Hodiamont et al. (10) GHQ-30 ‡10 ⁄PSE >4
Lewis & Booth (11) GHQ-30 ‡5 GHQ-30 ‡5
Paykell et al. (12) CIS-R ‡12 US-NAS-12 ‡3 CIS-R ‡12 US-NAS-12 ‡3
Ayuso-Mateos et al. (13) 1a
Kovess-Masfety et al. (14) 1a 1a
Kovess-Masfety et al. (15) 1a,b,2a,b,c,d,e,g,3a,b 1a,b 2a,b,c,d,e,g 3a,b 1a,b,2a,b,c,d,e,g,3a,b 1a,b 2a,b,c,d,e,g 3a,b
Weich et al. (16) GHQ-12 ‡3 GHQ-12 ‡3
Kringlen et al. (17) 1,2a,b,c,d,e,f,3a,b,c,d,4,5a,6 1a 2c 3a,b
Peen et al. (18) 1,2a,b,c,d,e,f,3a,b,c,d,4,6 1 2a,b,c,d,e,f 3a,b,c,d 1,2a,b,c,d,e,f,3a,b,c,d,4,6 1 2a,b,c,d,e,f 3a,b,c,d
Dekker et al. (19) 1,2a,b,c,d,e,f,3a,b,e,4c,5 1 2a,b,c,d,e,f 3a,b,e 1,2a,b,c,d,e,f,3a,b,e,4c,5 1 2a,b,c,d,e,f 3a,b,e
North America
Blazer et al. (20) 1a 2a 3a,b 1a 2a 3a,b
Kovess et al. (21) 1a,b 1a,b
Kessler et al. (22) 1,2a,b,c,d,e,3a,b,c,d,4,8 1 2a,b,c,d,e 3a,b,c,d
Parikh et al. (23) 1
Wang (24) 1a 1a
Kessler et al. (25) 1a 3a,b,c,d
Rohrer et al. (26) FMD ‡14 van 30 FMD ‡14 van 30
Other
Lee et al. (27) 1,2a,b,c,d,e,f,3a,b,c,d,4a,b,
5a,6a,7,8,9a,b
1 2a,b,c,d,e,f 3a,b,c,d
Andrews et al. (28) 1a,b,2a,b,d,e,f,g,3a,b,c,d 1a,b 2a,b,d,e,f,g 3a,b,c,d 1a,b,2a,b,d,e,f,g,3a,b,c,d 1a,b 2a,b,d,e,f,g 3a,b,c,d
1 = mood disorders; 1a = major depression; 1b = dysthymia; 1c = bipolar disorder; 2 = anxiety disorder; 2a = agoraphobia; 2b = social phobia; 2c = simple phobia;
2d = panic disorder; 2e = GAD; 2f = OCD; 2g = PTSD; 3 = substance use; 3a = alcohol dependence; 3b = alcohol abuse; 3c = drug dependence; 3d = drug abuse; 3e = illicit
drug use; 4 = non-affective psychosis; 4a = schizophrenia; 4b = schizophreniform disorder; 4c = possible psychotic disorder; 5 = somatoform disorder; 5a = somatization
disorder; 6 = eating disorder; 6a = anorexia; 7 = gambling; 8 = antisocial personality disorder; 9a = mild cognitive impairment; 9b = severe cognitive impairment.
GHQ, general health questionnaire.
Peen et al.
88
while 67% of the studies presented no significant
unadjusted OR. None of the studies found a
significant urban–rural OR less than 1. The
pooled unadjusted OR for mood disorders was
1.39 (1.23–1.58), P< 0.0001. The pooled
adjusted OR was somewhat lower: 1.28 (1.13–
Review: Urban rural differences
Comparison: 01 Inverse var
Outcome: 01 Any disorder - unadjusted
Study Odds ratio (random) Weight Odds ratio (random)
or sub-category 95% CI % 95% CI
01 Sub-category
15 - Belgium 4.98 0.60 (0.39, 0.92)
27 - South Korea 7.04 0.95 (0.84, 1.07)
26 - United States 6.30 1.07 (0.84, 1.37)
15 - Italy 6.23 1.07 (0.83, 1.38)
15 - Spain 6.47 1.14 (0.91, 1.42)
28 - Australia 6.99 1.19 (1.04, 1.36)
16 - Great Britain 6.45 1.25 (1.00, 1.57)
15 - The Netherlands 3.49 1.27 (0.66, 2.43)
15 - Germany 6.10 1.31 (0.99, 1.72)
15 - France 6.51 1.54 (1.24, 1.91)
11 - Great Britain 6.87 1.54 (1.32, 1.80)
19 - Germany 6.52 1.57 (1.27, 1.95)
12 - Great Britain 6.51 1.64 (1.32, 2.04)
18 - The Netherlands 6.67 1.77 (1.46, 2.14)
17 - Norway 6.71 2.47 (2.05, 2.97)
10 - The Netherlands 6.16 3.03 (2.32, 3.96)
Subtotal (95% CI) 100.00 1.38 (1.17, 1.64)
Test for heterogeneity: χ² = 153.45, df = 15 (P < 0.00001), I² = 90.2%
Test for overall effect: Z = 3.80 (P = 0.0001)
0.2 0.5 1 2 5
UrbanRural
Fig. 2. Urban–rural comparisons of any disorder, unadjusted OR with 95% CI.
Review: Urban rural differences
Comparison: 01 Inverse var
Outcome: 02 Mood disorders - unadjusted
Study Odds ratio (random) Weight Odds ratio (random)
or sub-category 95% CI % 95% CI
01 Sub-category
15 - Belgium 3.07 0.76 (0.43, 1.35)
13 - Norway 5.91 0.81 (0.62, 1.06)
27 - South Korea 6.08 1.08 (0.84, 1.40)
13 - Finland 4.16 1.15 (0.75, 1.78)
24 - Canada 6.87 1.19 (0.99, 1.43)
15 - Spain 5.46 1.19 (0.87, 1.61)
9 - Greece 5.18 1.24 (0.89, 1.73)
28 - Australia 5.58 1.25 (0.93, 1.68)
21 - Canada 5.34 1.25 (0.91, 1.71)
23 - Canada 6.10 1.28 (1.00, 1.65)
13 - Great Britain 5.13 1.30 (0.93, 1.82)
15 - France 5.51 1.35 (1.00, 1.83)
15 - Italy 4.70 1.37 (0.94, 2.00)
15 - The Netherlands 1.20 1.61 (0.55, 4.72)
19 - Germany 5.36 1.75 (1.27, 2.39)
14 - France 4.18 1.75 (1.14, 2.69)
15 - Germany 3.99 1.90 (1.21, 2.98)
17 - Norway 4.92 2.05 (1.43, 2.93)
18 - The Netherlands 5.39 2.10 (1.54, 2.87)
20 - United States 3.27 2.96 (1.72, 5.08)
13 - Ireland 2.57 3.06 (1.59, 5.89)
Subtotal (95% CI) 100.00 1.39 (1.23, 1.58)
Test for heterogeneity: χ² = 57.37, df = 20 (P < 0.0001), I² = 65.1%
Test for overall effect: Z = 5.08 (P < 0.00001)
0.1 0.2 0.5 1 2 5 10
Rural Urban
Fig. 3. Urban–rural comparisons of mood disorders, unadjusted OR with 95% CI.
Urban–rural differences
89
1.44), P< 0.001 (15 comparisons; data not
shown).
Figure 4 shows the unadjusted OR for anxiety
disorders (12 comparisons). The number of com-
parisons in this figure is lower (n= 12) than those
for Ôany disorderÕ(n= 16) or Ômood disorderÕ
(n= 21). Of these 12, nine were from Europe
and three from outside Europe. The majority of
unadjusted OR indicated no difference (67%).
Thirty-three per cent indicated an urban–rural
OR significantly higher than 1. The pooled
unadjusted OR for anxiety disorders was 1.21
(1.02–1.42), P= 0.03. The pooled adjusted OR
was 1.13 (1.00–1.28), P= 0.06 (11 comparisons;
data not shown).
Figure 5 shows the unadjusted OR for substance
use disorders (13 comparisons). Of the 13 available
comparisons, 10 were from Europe and three from
outside Europe. As was the case with anxiety
disorders, the majority of unadjusted OR indicated
no difference (69%). Three studies found a signif-
icant urban–rural OR higher than 1 (23%) and
Review: Urban rural differences
Comparison: 01 Inverse var
Outcome: 04 Anxiety disorders - unadjusted
Study Odds ratio (random) Weight Odds ratio (random)
or sub-category 95% CI % 95% CI
01 Sub-category
15 - Belgium 5.25 0.62 (0.36, 1.06)
15 - Italy 8.62 0.90 (0.67, 1.21)
27 - South Korea 10.26 0.93 (0.76, 1.12)
15 - Spain 9.00 0.99 (0.75, 1.30)
20 - United States 9.54 1.13 (0.89, 1.44)
15 - Germany 8.67 1.19 (0.88, 1.59)
28 - Australia 9.39 1.22 (0.95, 1.56)
15 - The Netherlands 3.39 1.39 (0.65, 2.97)
15 - France 9.17 1.41 (1.09, 1.83)
18 - The Netherlands 9.49 1.44 (1.13, 1.83)
19 - Germany 8.76 1.48 (1.11, 1.98)
17 - Norway 8.45 2.37 (1.75, 3.22)
Subtotal (95% CI) 100.00 1.21 (1.02, 1.42)
Test for heterogeneity: χ2 = 43.53, df = 11 (P < 0.00001), I2 = 74.7%
Test for overall effect: Z = 2.22 (P = 0.03)
0.2 0.5 1 2 5
Favours treatment Favours control
Fig. 4. Urban–rural comparisons of anxiety disorders, unadjusted OR with 95% CI.
Review: Urban rural differences
Comparison: 01 Inverse var
Outcome: 03 Substance use disorders - unadjusted
Study Odds ratio (random) Weight Odds ratio (random)
or sub-category 95% CI % 95% CI
01 Sub-category
15 - Belgium 5.49 0.56 (0.22, 1.40)
20 - United States 9.58 0.61 (0.42, 0.87)
19 - Germany 8.77 0.90 (0.57, 1.44)
15 - Spain 6.74 0.94 (0.45, 1.96)
27 - South Korea 10.90 0.99 (0.88, 1.12)
28 - Australia 10.32 1.20 (0.94, 1.54)
15 - The Netherlands 2.90 1.38 (0.30, 6.36)
15 - France 7.23 1.41 (0.72, 2.73)
15 - Italy 2.60 1.50 (0.29, 7.77)
12 - Great Britain 9.64 1.60 (1.12, 2.28)
15 - Germany 6.32 2.15 (0.98, 4.75)
18 - The Netherlands 10.02 2.33 (1.73, 3.14)
17 - Norway 9.50 3.71 (2.56, 5.37)
Subtotal (95% CI) 100.00 1.31 (0.97, 1.78)
Test for heterogeneity: χ2 = 87.24, df = 12 (P < 0.00001), I2 = 86.2%
Test for overall effect: Z = 1.77 (P = 0.08)
0.1 0.2 0.5 1 2 5 10
Favours treatment Favours control
Fig. 5. Urban–rural comparisons of substance use disorders, unadjusted OR with 95% CI.
Peen et al.
90
one study (8%) found a significant urban–rural
OR less than 1. The pooled unadjusted OR was
1.31 (0.97–1.78), P= 0.08. The adjusted OR was
1.03 (0.85–1.26), P= 0.74 (13 comparisons; data
not shown).
Heterogeneity
Several possible sources of heterogeneity, apart
from urban–rural variations, can be put forward.
These sources can be differences in culture or
socioeconomic status of the countries involved, but
also differences in the contents of the prevalence
rates used and the way in which they were estab-
lished (see Table 2). Therefore, we made some
additional comparisons within the diagnostic cate-
gories reported in this study (if the available number
of studies was sufficient to do so). To analyse
possible heterogeneity due to culture, we compared
the pooled (unadjusted) prevalence rate for mood
disorders for European studies to the pooled rate for
the North American studies (see Tables 1 and 2). No
difference was found [1.44 (CI: 1.20–1.71) and 1.40
(CI: 1.08–1.82) respectively]. Furthermore, we anal-
ysed possible heterogeneity due to method in which
prevalence rates were established in each study.
Therefore, we compared prevalence rates for Ôany
disorderÕbased on diagnostic instruments to rates
based on cut-off scores (see Tables 1 and 2). No
differences were found in both unadjusted [1.30 (CI:
1.05–1.60) and 1.59 (CI: 1.18–2.13)] and adjusted
rates [1.17 (CI: 1.01–1.35) and 1.29 (CI: 1.16–1.44)].
Subsequently, possible heterogeneity within diag-
nostic groups was analysed. First, within (unad-
justed) rates for mood disorders, studies from which
rates of major depression were used compared with
other studies (mainly containing mood disorders in
general; see Table 2). No difference was found [1.48
(CI: 1.15–1.90) and 1.36 (CI: 1.19–1.56)]. Likewise,
within (unadjusted) rates for substance use disor-
ders, we compared studies presenting rates for
alcohol dependence and abuse to studies also
including drug dependence and abuse. No difference
was found [1.33 (CI: 0.79–2.25) and 1.26 (CI: 0.86–
1.86)].
Discussion
This is the first meta-analysis investigating urban–
rural differences in prevalence rates for common
mental disorders. Using only higher quality studies
performed since 1985 in high income countries, it
was shown for both Ôany disorderÕ(38% higher),
mood disorders (39%) and anxiety disorders (21%)
that the pooled urban prevalence rate was higher
in urban areas compared with rural areas. No
difference was found for substance use disorders.
In addition, when controlling for important
confounders, we found slightly lower, but statisti-
cally significant, pooled OR. While the number of
confounders was generally considerable, this dif-
ference between adjusted and unadjusted ratios
was limited, showing that urban–rural differences
are only partly explained by population character-
istics.
Although both the use of standardized diagnos-
tic instruments and the extent to which findings are
adjusted for potential confounders has significantly
increased since the period before 1985, the current
study thus confirms less systematically evaluated
findings from earlier reviews (2–7).
One could argue that the association with
urbanization presented here is low at 1.21 (1.09–
1.34) for Ôany disorderÕ. Compared to other factors
associated with the prevalence of psychiatric
disorders – such as being unmarried or childhood
abuse – the strength of the association with
urbanization is limited. Nevertheless, it remains
intriguing that, even when controlling for a rela-
tively large number of confounders, the urban
environment seems to be associated with the
prevalence of psychopathology. This association
does not appear to be explained solely by popula-
tion characteristics such as age, gender, marital
status, social class or ethnicity. In line with studies
examining the association between the urban
environment and schizophrenia (29), we found
that the urban environment appears to be associ-
ated with mental health. Further study is needed to
establish whether this association can partly be
explained by gene–environment interactions (30).
Furthermore, the practical implications of 34%
more cases in urbanized areas are significant in
terms of service allocation and healthcare budget.
The allocation of more services to urban areas is
not only desirable because of the prevalence rates,
but also because comorbidity rates tend to be
higher in urban areas (18, 22). Generally, the
distribution of funds does not keep up with the
extra need for services in urban areas. The conse-
quences are, for instance, relatively long waiting
lists and pressure to keep treatments and admis-
sions short, putting the quality of care at risk.
Ideally, a match between the provision of services
and demand for mental health care is the best
option. Based on our findings, urbanization may
be a useful indicator for allocating mental health
funds and services.
When interpreting these findings, a number of
potential limitations should be addressed. Several
possible sources of heterogeneity apart from
urban–rural variation can be mentioned concern-
Urban–rural differences
91
ing this study. As the analysis contains studies in a
period of 20 years from all over the world there is
possible heterogeneity due to diagnostic methods,
culture and socioeconomic status for instance.
Apart from this, also differences in the diagnostic
contents of the prevalence rates used may be a
source of heterogeneity. For instance, rates used
for the analysis of mood disorders containing ÔonlyÕ
major depression may have a different relation to
urbanization compared to rates containing all
mood disorders. In addition, the latter contrast
may also represent a difference in severity. In a
secondary analysis we made some comparisons
concerning possible heterogeneity due to culture
(Europe vs. North America), diagnostic method
(diagnostic instruments vs. cut-off scores) and
diagnostic content (major depression vs. mood
disorders as a whole and alcohol abuse ⁄depen-
dence vs. substance use disorders as a whole).
These comparisons did not show any significant
differences, which may lower concerns about sys-
tematic heterogeneity in this study.
It has to be taken into account that there is
comorbidity between diagnostic groups reported in
this study, for instance between anxiety and mood
disorders. This means that some research subjects
will be present in more than one comparison.
A more or less similar point is that studies which
are included in two or more diagnostic groups
analysed here, have a relatively larger weight
compared to studies which are only included in
one diagnostic group.
A limitation of the study is that schizophrenia
was not included as a separate diagnostic category.
It is difficult to generate reliable prevalence rates
for schizophrenia from general population studies
due to both the low prevalence of schizophrenia in
the non-institutionalized community, and to selec-
tive exclusion of these patients from population
surveys (31). Accordingly, most of the studies in
our analysis did not present rates for schizophre-
nia.
Our review included two multi-country studies
(13, 15) (one deals with mood disorders only), and
we presented the results for each of the individual
countries. As there is a wide variation of outcomes
between countries within these studies, and as the
findings do not systematically differ from other
studies, we believe this is the preferred strategy.
Presenting ratios for the total study area only
would have resulted in the loss of information
about variation between countries within the areas.
The Esemed study, for example, found that
Belgium, which has higher total rural rates com-
pared to urban rates, differs substantially from its
neighbouring countries (15).
One could argue that using dichotomized
measures for urbanization would underestimate
the influence of this factor on levels of psycho-
pathology. Using continuous measures or com-
paring the extremes of more than two categories
of urbanization, would probably yield a signifi-
cant difference more easily. However, most
studies did not provide such data. Furthermore,
this rule applies only to studies of large con-
nected areas (countries, for example). However,
the choice of either one or the other separate
area in a Ôtwin studyÕhas implications for the
possibility of finding differences (7). When one
chooses to compare one typically rural area
with a metropolitan area, the initial differences
in urbanization are probably greater than
between the extremes of a division into five
categories of a whole country. After all, ÔurbanÕ
and ÔruralÕare relative concepts, and their oper-
ationalization will probably always differ between
studies.
To explain inner-city and urban–rural varia-
tions in psychiatric morbidity, there are two main
theoretical concepts, which originated from the
early ecological research of schizophrenia (32) and
from the Chicago School of Sociology (33): the
drift hypothesis and the breeder hypothesis. The
drift hypothesis assumes on the one hand that
sick and vulnerable people are more or less
doomed to remain in socially unstable, deprived
neighbourhoods, while better off people move
away (social residue theory; 34). On the other
hand, socially deprived neighbourhoods can also
have a Ôpull-functionÕon sick and vulnerable
people, as they move to these areas with low
social control and greater tolerance towards
deviant behaviour (social drift hypothesis). Evi-
dence concerning drift processes is still sparse (6,
35). However, concentration of schizophrenic
patients in deprived inner-city areas has been
described in numerous ecological studies (32, 36).
It remains to be seen however, if these supposed
drift processes apply to all psychiatric illnesses.
The second theory, the breeder hypothesis,
assumes that various environmental factors
cause illness. These can be physical factors (air
pollution, small housing, population density) and
also social factors (stress, life events, perinatal
aspects, social isolation). A lot of the stress
factors mentioned above are more common in
urbanized areas (1, 37).Urbanization is modestly
but consistently associated with the prevalence
of psychopathology. This should be further
examined in studies of the aetiology of mood
and anxiety disorders in particular. Levels of
urbanization should also be taken into account
Peen et al.
92
when planning the allocation of mental health
services.
Declaration of interest
None.
References
1. Freeman H. Mental health and the environment, 1st edn.
London: Churchill Livingstone, 1984.
2. Dohrenwend BP,Dohrenwend BS. Psychiatric disorders in
urban settings. In: Caplan G, ed. American handbook of
psychiatry. New York: Basic Books, 1974:424–447.
3. Mueller DP. The current status of urban–rural differences
in psychiatric disorder. An emerging trend for depression.
J Nerv Ment Dis 1981;169:18–27.
4. Marsella AJ. Urbanization, mental health, and social
deviancy. A review of issues and research. Am Psychol
1998;53:624–634.
5. Neff JA. Urbanicity and depression reconsidered. The
evidence regarding depressive symptomatology. J Nerv
Ment Dis 1983;171:546–552.
6. Verheij RA. Explaining urban–rural variations in health: a
review of interactions between individual and environment.
Soc Sci Med 1996;42:923–935.
7. Webb SD. Rural–urban differences in mental health. In:
Freeman H, ed. Mental health and the environment. Lon-
don: Churchill Livingstone, 1984:227–249.
8. United Nations. Global outlook 2000. New York: United
Nations, 1990.
9. Madianos MG,Stefanis CN. Changes in the prevalence of
symptoms of depression and depression across Greece. Soc
Psychiatry & Psychiatr Epidem 1992;27:211–219.
10. Hodiamont PP,Sijben AE,Koeter MW,Oldehinkel AJ.
Psychological problems and degree of urbanization.
[Dutch]. Tijdschrift voor Psychiatrie 1992;34:426–438.
11. Lewis G,Booth M. Are cities bad for your mental health?
Psychol Med 1994;24:913–915.
12. Paykel ES,Abbott R,Jenkins R,Brugha TS,Meltzer H.
Urban–rural mental health differences in great Britain:
findings from the national morbidity survey. Psychol Med
2000;30:269–280.
13. Ayuso-Mateos JL,Vazques-Barquero JL,Dowrick C et al.
Depressive disorders in Europe: prevalence figures from
the ODIN study. Br J Psychiatry 2001;179:308–316.
14. Kovess-Masfety V,Lecoutour X,Delavelle S. Mood dis-
orders and urban ⁄rural settings: comparisons between
French regions. Soc Psychiatry & Psychiatr Epidem
2005;40:613–618.
15. Kovess-Masfety V,Alonso J,De Graaf R,Demyttenaere K.
A European approach to rural–urban differences in mental
health: the ESEMeD 2000 comparative study. Can J Psy-
chiatry – Rev Canadienne de Psychiatrie 2005;50:926–936.
16. Weich S,Twigg L,Lewis G. Rural ⁄non-rural differences in
rates of common mental disorders in Britain: prospective
multilevel cohort study. Br J Psychiatry 2006;188:51–57.
17. Kringlen E,Torgersen S,Cramer V. Mental illness in a
rural area: a Norwegian psychiatric epidemiological study.
Soc Psychiatry & Psychiatr Epidem 2006;41:713–719.
18. Peen J,Dekker J,Schoevers RA,Ten Have M,De Graaf R,
Beekman AT. Is the prevalence of psychiatric disorders
associated with urbanization? Soc Psychiatry & Psychiatr
Epidem 2007;42:984–989.
19. Dekker J,Peen J,Smit F,Schoevers RA. Prevalence of
psychiatric disorders in relation with urbanization in
Germany. BMC Public Health 2008;8:17.
20. Blazer D. Psychiatric disorders: a rural ⁄urban compari-
son. Arch Gen Psychiatry 1985;42:651–656.
21. Kovess V,Murphy HB,Tousignant M. Urban–rural com-
parisons of depressive disorders in French Canada. J Nerv
Ment Dis 1987;175:457–466.
22. Kessler RCP,Mcgonagle KAP,Zhao SP et al. Lifetime and
12-month prevalence of DSM-III-R psychiatric disorders
in the United States: results from the National Comor-
bidity Survey. Arch Gen Psychiatry 1994;51:8–19.
23. Parikh SV,Wasylenki D,Goering P,Wong J. Mood disor-
ders: rural ⁄urban differences in prevalence, health care
utilization, and disability in Ontario. J Affect Disord 1996;
38:57–65.
24. Wang JL. Rural–urban differences in the prevalence of
major depression and associated impairment. Soc Psychi-
atry & Psychiatr Epidem 2004;39:19–25.
25. Kessler RCP,Chiu WTA,Demler OM,Walters EEM.
Prevalence, severity, and comorbidity of 12-month DSM-
IV disorders in the National Comorbidity Survey Repli-
cation. Arch Gen Psychiatry 2005;62:617–627.
26. Rohrer JE,Borders TF,Blanton J. Rural residence is not a
risk factor for frequent mental distress: a behavioral risk
factor surveillance survey. BMC Public Health 2005;5:46.
27. Lee CK,Kwak YS,Yamamoto J et al. Psychiatric epidemi-
ology in Korea. Part II: Urban and rural differences.
J Nerv Ment Dis 1990;178:247–252.
28. Andrews G,Henderson S,Hall W. Prevalence, comorbidity,
disability and service utilisation. Overview of the Austra-
lian National Mental Health Survey. Br J Psychiatry
2001;178:145–153.
29. Pedersen CB,Mortensen PB. Evidence of a dose–response
relationship between urbanicity during upbringing and
schizophrenia risk. Arch Gen Psychiatry 2001;58:1039–
1046.
30. Pedersen CB,Mortsensen PB. Are the cause(s) responsible
for urban–rural differences in schizophrenia risk rooted in
families or in individuals? Am J Epidem 2006;163:971–978.
31. Bijl RV,Ravelli A,Van Zessen G. Prevalence of psychiatric
disorder in the general population: results of The Nether-
lands Mental Health Survey and Incidence Study (NEM-
ESIS). Soc Psychiatry & Psychiatr Epidem 1998;33:587–
595.
32. Faris R,Dunham H. Mental disorders in urban areas. Chi-
cago, IL: University of Chicago Press, 1939.
33. Park RE,Burgess EW. The city. Chicago, IL: Chicago
University Press, 1925.
34. Freeman HL,Alpert M. Prevalence of schizophrenia in an
urban population. Br J Psychiatry 1986;149:603–611.
35. Moorin RE,Holman CDJ,Garfield C,Brameld KJ. Health
related migration: evidence of reduced Ôurban-driftÕ. Health
& Place 2006;12:131–140.
36. Giggs JA,Cooper JE. Ecological structure and the distri-
bution of schizophrenia and affective psychoses in Not-
tingham. Br J Psychiatry 1987;151:627–633.
37. Maas J,Verheij RA,Groenewegen PP,Vries Sde,Spreeu-
wenberg P. Green space, urbanity, and health: how strong
is the relation? J Epidemiol Community Health
2006;60:587–592.
Urban–rural differences
93
