See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/30850225

Urban neighborhood socioeconomic status and incidence of depression:

Evidence from a population-based cohort study

Article  in  American Journal of Epidemiology · March 2007

DOI: 10.1093/aje/161.Supplement_1.S124c · Source: OAI

CITATIONS READS

7 58

6 authors, including:

Sandro Galea Jennifer Ahern

Boston University University of California, Berkeley
1,057 PUBLICATIONS   40,398 CITATIONS    192 PUBLICATIONS   9,572 CITATIONS   

SEE PROFILE SEE PROFILE

Melissa Tracy David Vlahov

Columbia University Yale University
102 PUBLICATIONS   5,371 CITATIONS    920 PUBLICATIONS   62,022 CITATIONS   

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

ALIVE study View project

Gene x Childhood Environment Interaction in Psychological Risk and Resilience View project

All content following this page was uploaded by Jennifer Ahern on 26 May 2014.

The user has requested enhancement of the downloaded file.

Urban Neighborhood Poverty and the Incidence of Depression
in a Population-Based Cohort Study
SANDRO GALEA, MD, DRPH, JENNIFER AHERN, MPH, ARIJIT NANDI, MPH,
MELISSA TRACY, MPH, JOHN BEARD, PHD, AND DAVID VLAHOV, PHD

PURPOSE: It has long been suggested that certain characteristics of the urban environment may influ-
ence population mental health. However, evidence from multilevel research addressing the relation
between intraurban environments and depression has been conflicting, and prospective evidence in this
regard has been limited. We assessed the relation between urban neighborhood poverty and incident
depression in a population-based prospective cohort study.
METHODS: Using random-digit-dial telephone surveys, we recruited 1570 adult residents of New York
City (NYC) in 2002. All persons interviewed at baseline were contacted again for follow-up 6 and
18 months after the initial interview. Eighty-one percent of the sample completed at least one follow-up
visit. Analysis was restricted to 1120 persons who could be geocoded to NYC neighborhoods, which
were represented by NYC community districts (N Z 59).
RESULTS: Among persons with no history of major depression at baseline (N Z 820) there were 113
incident cases of major depression during the 18 months of follow-up; cumulative incidence of depression
during the study period was 14.6 per hundred persons (95% confidence interval, 10.9–18.3). In low–
socioeconomic status (SES) neighborhoods, the cumulative incidence of depression was 19.4 per hundred
persons (95% confidence interval, 13.5–25.3), which was greater than that in high-SES neighborhoods
(10.5; 95% confidence interval, 5.9–15.2). In multivariable models adjusting for individual covariates
(sociodemographics, individual SES, social support, stressors, traumas, and history of post-traumatic stress
disorder), the relative odds of incident depression was 2.19 (95% confidence interval, 1.04–4.59) for
participants living in low-SES compared with high-SES neighborhoods.
CONCLUSIONS: SES of neighborhood of residence is associated with incidence of depression indepen-
dent of individual SES and other individual covariates. Additional work needs to characterize the pathways
that may explain the observed association between living in low-SES neighborhoods and risk for depression.
Ann Epidemiol 2007;17:171–179. Ó 2007 Elsevier Inc. All rights reserved.
KEY WORDS: Mental Health, Neighborhood, Context, Socioeconomic Position, Urban.

residents. Subsequently, results from the Midtown Manhat-

INTRODUCTION
tan Study also suggested that features of the urban environ-
The majority of the world’s population will soon reside in ment, such as social disorganization, might influence mental
urban areas (1, 2). Empiric evidence has long suggested health (5). In recent years, several cross-sectional popula-
that certain characteristics of the urban environment may tion-based studies assessed urban–rural or interurban differ-
influence population mental health (3, 4). Faris and xDun- ences in prevalence of mental health problems (6–10).
ham (3) argued that community disorganization in Chicago Results from these prevalence studies were conflicting. For
neighborhoods contributed to poor mental health of urban example, greater prevalences of mental illness were docu-
mented in urban compared with rural areas in the United
Kingdom (9), but urban–rural differences in prevalence of
From the Department of Epidemiology, University of Michigan, School
of Public Health, Ann Arbor, MI (S.G., J.A., M.T.); Center for Urban mental illness were not found in a Canadian study using
Epidemiologic Studies, New York Academy of Medicine (S.G., A.N., similar methods (8).
J.B., D.V.); Department of Epidemiology, Columbia University Mailman Intraurban analyses, typically neighborhood level, pres-
School of Public Health; New York, NY (S.G., D.V.); Department of
Epidemiology, University of California Berkeley School of Public Health; ent an opportunity to assess whether specific characteristics
Berkeley, CA (J.A.); and Department of Epidemiology, Johns Hopkins of a local urban environment are associated with psycho-
Bloomberg School of Public Health; Baltimore, MD (A.N., D.V.). pathologic states. A few recent studies used multilevel ana-
Address correspondence to: Sandro Galea, MD, DrPH, Department of
Epidemiology, University of Michigan, School of Public Health, 1214 S. lytic techniques to consider whether characteristics of urban
University, Ann Arbor, MI 48104. Tel.: (734) 647-9741; fax: (734) 998- neighborhoods were associated with individual mental
0006. E-mail: sgalea@umich.edu. health. The empirical work on neighborhoods and mental
Funded in part by grants DA 017642, MH 066391, and MH 66081 from
the National Institutes of Heath. health to date produced inconsistent results. Some studies,
Received April 20, 2006; accepted July 16, 2006. primarily from the United Kingdom, reported that living

Ó 2007 Elsevier Inc. All rights reserved. 1047-2797/07/$–see front matter

360 Park Avenue South, New York, NY 10010 doi:10.1016/j.annepidem.2006.07.008
172 Galea et al. AEP Vol. 17, No. 3
URBAN NEIGHBORHOOD POVERTY AND DEPRESSION March 2007: 171–179

a computer-assisted telephone interview system. House-

Selected Abbreviations and Acronyms
holds were screened for eligibility by location. If eligible,
NYC Z New York City
SES Z socioeconomic status an adult in each household was selected randomly by choos-
BSI Z Brief Symptom Index ing the adult with the birthday closest to the interview date.
GEE Z generalized estimating equation Up to 10 attempts were made to conduct the interview. The
NWS Z National Women’s Study
PTSD Z post-traumatic stress disorder response rate among those eligible was 56%. Contact infor-
mation was obtained for respondents, their key family mem-
bers, and other important contacts. At 6 and 18 months after
baseline, we conducted follow-up interviews, again using
in socially and economically deprived neighborhoods did telephone surveying (September 25, 2002, to January 31,
not contribute to mental illness independent of individual 2003, and September 25, 2003, to February 29, 2004, respec-
covariates (11–15). Conversely, a number of US reports tively). We were successful in contacting and completing at
indicated increased risk for mental illness in poorer neigh- least one follow-up interview in 81% of baseline
borhoods, even after accounting for individual covariates respondents.
(16–18). For example, Walters et al. (13) found that living This analysis was restricted to residents of NYC itself
in the most deprived areas of Britain was not associated with (N Z 1570). We included in this analysis residents of
depression among older persons after adjusting for individ- NYC who completed at least one follow-up interview
ual characteristics, whereas studies in multiple cities across and could be linked to their neighborhood of residence,
the United States reported that neighborhood disadvantage for a total of 1120 participants. Additional details on sample
was associated with greater rates of depressive symptoms selection are provided elsewhere (20). This work was
independent of individual characteristics (16, 18). reviewed and approved by the Institutional Review Board
Multilevel analyses published to date are predominately of the New York Academy of Medicine.
cross-sectional and hence susceptible to prevalence bias,
limiting causal inference about the relation between neigh-
borhood context and onset of mental health illnesses. We Measures
are aware of only two other studies that assessed the relation The outcome of interest in this analysis is major depression.
between characteristics of neighborhoods and incident men- We assessed major depression by using a modified version of
tal health problems. One study failed to document an asso- the Structured Clinical Interview for Diagnostic and Statisti-
ciation between poverty-area residence and incident cal Manual of Mental Disorders, Third Edition Revised major
depression in adjusted models (15). Another more recent depressive disorder subscale (21), a validated measure that
study using a prospective design suggested that neighbor- captures symptoms of major depression consistent with
hood disorder was associated with depressive symptoms Diagnostic and Statistical Manual of Mental Disorders, Fourth
over time when controlling for baseline depressive symp- Edition criteria (22, 23). To meet criteria for depression,
toms (19). Building on this work, we were interested in respondents had to report five or more symptoms for at least
assessing whether living in a poorer neighborhood was 2 weeks, one of which was depressed mood or loss of pleasure
associated with greater risk for incident major depression or interest. Past-6-month depression was based on reporting
independent of individual covariates by using data from five or more symptoms within the same month during the
a prospective general population survey of New York City past 6 months, and lifetime depression was based on report-
(NYC) adults. ing the required symptoms at any time in the past. Incident
cases were those satisfying criteria for depression at any time
METHODS during follow-up among persons who did not report depres-
sion at any time in the past at baseline. Cronbach a for the
Participants depression scale was 0.79 in our sample (24). Furthermore,
We conducted a random-digit-dial telephone survey of res- in a validation study comparing our instrument with the
idents of the NYC metropolitan area between March 25 and Brief Symptom Inventory (BSI) (25), a widely used depres-
June 25, 2002. The study is designed to document popula- sion scale, the BSI depression scale had sensitivity of 73%
tion mental health in the aftermath of the September 11, and specificity of 87%, with our instrument as the ‘‘gold
2001, terrorist attacks. The sampling frame consisted of all standard’’ (22). In a receiver operating characteristic analy-
noninstitutionalized adults (age >18 years) and over- sis (26), the BSI depression cutoff score of 65 or higher best
sampled residents living in the area closest to the World predicted depression defined by our instrument (area under
Trade Center site. Interviews were conducted in English, the curve Z 0.89) (22).
Spanish, Mandarin, and Cantonese by trained interviewers We considered variables that are known determinants of
using translated and backtranslated questionnaires and depression that could be potential confounders of the
AEP Vol. 17, No. 3 Galea et al. 173
March 2007: 171–179 URBAN NEIGHBORHOOD POVERTY AND DEPRESSION

central relationship of interest to this analysis (27). Informa- Neighborhood units for this analysis were the 59 commu-
tion for individual covariates, including age, sex, race/eth- nity districts in NYC. These neighborhoods initially were
nicity, income, education, and marital status, was obtained defined by a resident consultative process organized by the
from respondents by using a structured questionnaire. We Office of City Planning to reflect residents’ own descriptions
assessed social support by asking about emotional (e.g., of neighborhoods in the 1970s. Therefore, community dis-
‘‘having someone to love you and make you feel wanted’’), tricts delineate meaningful neighborhoods within NYC,
instrumental (e.g., ‘‘someone to help you if you were con- each with an administrative community board, that as
fined to bed’’), and appraisal (e.g., ‘‘someone to give you such have political and social a priori significance for their
good advice in a crisis’’) support in the 6 months before residents. Although community districts are not demo-
the September 11 attacks and summed responses (28). graphically homogenous, they represent neighborhoods as-
The combined social support score was divided into thirds sociated with resident behavior and health (30–33). Using
for analysis. 2000 US Census data (34), we dichotomized neighborhood
September 11, 2001, event experiences also were socioeconomic status (SES) as high or low based on a median
assessed, and respondents were classified by whether they split of neighborhood median ($36,470) household income.
were affected directly by the attacks of September 11,
2001 (in the World Trade Center complex during the
attacks, injured during the attacks, lost possessions or
property, had a friend or relative killed, lost job as a result
of the attacks, or involved in rescue efforts), and/or living Statistical Analyses
within 2 miles of the World Trade Center site on September Sampling weights were developed and applied to the data to
11, 2001. Respondents also were asked about the occurrence correct for potential selection bias relating to number of
of any of 12 traumatic events (natural disaster; serious acci- household telephones, persons in the household, and over-
dent at work, in a car, or somewhere else; assault with sampling. We compared distributions of key demographic
a weapon; assault without a weapon; unwanted sexual con- characteristics, neighborhood poverty, and history of de-
tact; serious injury or illness; other situation involving seri- pression for respondents included in our analysis with those
ous injury or physical damage; situation causing fear of death of respondents excluded because they could not be linked to
or serious injury; seeing someone seriously injured or their neighborhood of residence or were lost to follow-up.
violently killed; death of a spouse or mate; death of a close We also compared included respondents with 2000 US Cen-
family member other than a spouse; or any other extraordi- sus data for NYC (34). We calculated the prevalence of
narily stressful situation or event) in their lifetime, as well as lifetime and past-6-month depression at baseline and
about stressors in the past year (divorce or separation, mar- cumulative incidence of depression during follow-up and
riage, family problems, problems at work, and unemploy- used two-tailed chi-square tests to test for associations be-
ment). For the analysis, prior lifetime traumatic tween covariates of interest and incidence of depression. In-
experiences were categorized according to whether respon- cident depression is defined as depression at either of the
dents experienced zero, one, two, three, or four or more trau- follow-up visits among those with no history of depression
matic events. Stressors in the past year were categorized as at baseline. We used generalized estimating equations
zero or one or more for the analysis. (GEEs) to fit a multilevel multivariable model that assessed
Finally, we used the National Women’s Study post- the relation between neighborhood SES and risk for depres-
traumatic stress disorder (PTSD) module (29) to assess prob- sion at any time during follow-up for persons who did not
able PTSD symptoms in the respondent’s lifetime. The report depression at baseline (35). The GEE model was
National Women’s Study PTSD module is a 17-item mea- specified with nesting by neighborhood, an exchangeable
sure of probable PTSD that evaluates the presence (yes/ correlation structure, and weighting, as described. The
no) of criterion B (reexperiencing, e.g., intrusive memories model was adjusted for sociodemographic characteristics
or distressing dreams), C (avoidance, e.g., efforts to avoid (i.e., income, education, marital status, social support,
thoughts associated with the trauma or loss of interest in sig- stressors, and traumatic events), as well as September 11,
nificant activities), and D (arousal, e.g., difficulty falling 2001, event experiences and lifetime history of PTSD. All
asleep or concentrating) symptoms and determines content analyses were carried out using SUDAAN (Research Trian-
for content-specific symptoms (e.g., content of dreams or gle Institute, Research Triangle Park, NC) and SAS (SAS
nightmares) if symptom presence is endorsed. We assessed Institute, Cary, NC) (36). We replicated regression analyses
lifetime history of probable PTSD based on the presence by using Cox proportional hazards analysis, which allows
of at least one reexperiencing symptom, at least three avoid- consideration of the person-time contributed by each indi-
ance symptoms, and two arousal symptoms reported to have vidual, and results were essentially equivalent. Only GEE
been experienced any time in the past. models are presented here for the sake of brevity.
174 Galea et al. AEP Vol. 17, No. 3
URBAN NEIGHBORHOOD POVERTY AND DEPRESSION March 2007: 171–179

during the 18-month period of follow-up was 14.6 per 100

RESULTS
persons. The cumulative incidence of depression was greater
Overall, of 1570 NYC residents interviewed, 1120 success- in low-SES compared with high-SES neighborhoods (19.4
fully completed at least one follow-up interview and pro- per 100 versus 10.5 per 100 persons, respectively).
vided information allowing us to link them to their In bivariate analysis (Table 3), characteristics signifi-
neighborhood. The remaining 450 respondents were ex- cantly associated with the incidence of depression during
cluded from further analyses. As listed in Table 1, there the 18-month follow-up were residence in a neighborhood
were no significant differences in distributions of key demo- of lower SES (p Z 0.024), female sex (p Z 0.037), living
graphic characteristics or baseline prevalence of past-6- more than 2 miles from the World Trade Center site on Sep-
month depression between the included and excluded tember 11 (p Z 0.022), and a lifetime history of PTSD (p !
groups. There also was no significant difference in neighbor- 0.001). Age, race/ethnicity, individual income, education,
hood poverty status between the included group and those marital status, social support, being directly affected by the
excluded because of loss to follow-up. Additionally, distribu- September 11 attacks, lifetime traumatic events, and
tions of age, sex, and race/ethnicity of the included sample stressors in the past year were not associated significantly
are consistent with those of the general NYC adult popula- with the incidence of depression during follow-up.
tion from the 2000 US Census (34). Mean age was 41.7 A multilevel model assessing the relation between neigh-
years, 55.6% of respondents were women, 38.1% were white, borhood SES and incidence of depression during follow-up
6.4% were Asian, 25.1% were African American, 26.0% also is presented in Table 3. After adjusting for sociodemo-
were Hispanic, and 4.4% were of other race or ethnicity. graphic characteristics (i.e., income, education, marital
In the 59 neighborhoods, there was a mean of 19 respon- status, social support, stressors, and traumatic events),
dents (median, 11 respondents; range, 2 to 246 September 11 experiences, and lifetime history of PTSD, re-
respondents). spondents residing in lower SES neighborhoods had more
At baseline, the prevalence of past-6-month depression than two times the odds of developing depression during
was 12.2%, and 20.8% had a lifetime history of depression follow-up relative to those residing in higher SES
(Table 2). For subjects who had no history of depression at neighborhoods (odds ratio, 2.19; 95% confidence interval,
baseline (n Z 820), the cumulative incidence of depression 1.04–4.59). Female sex, lower social support, four or more

TABLE 1. Sample demographics

Included in sample, Excluded from sample, Chi-square p Chi-square p
weighted % weighted % Census 2000, % (included vs. census) (included vs. excluded)

Total (n) 1120 450

Age, years
18–24 15.2 13.7 13.2 0.77 0.26
25–34 24.6 31.8 22.5
35–44 18.8 21.2 20.8
45–54 18.8 17.4 16.7
55–64 12.0 9.6 11.3
65þ 10.6 6.3 15.5
Sex
Male 44.4 43.3 46.2 0.72 0.81
Female 55.6 56.7 53.8
Race/ethnicity
White 38.1 31.3 38.7 0.78 0.13
Asian 6.4 6.1 10.1
African American 25.1 20.9 23.0
Hispanic 26.0 34.2 24.7
Other 4.4 7.6 3.6
Neighborhood socioeconomic statusa
High 53.3 46.9 0.28
Low 46.7 53.1
Past-6-mo depression
No 87.8 88.0 0.95
Yes 12.2 12.0
a
Among those with neighborhood information (all included subjects, 235 of excluded subjects).
AEP Vol. 17, No. 3 Galea et al. 175
March 2007: 171–179 URBAN NEIGHBORHOOD POVERTY AND DEPRESSION

TABLE 2. Prevalence and cumulative incidence that the relation of neighborhood SES and incident depres-
of major depression sion persisted in multivariable models independent of these
95% Confidence potentially explanatory individual-level covariates. How-
n % interval ever, we did not document an association between stressors
Baseline lifetime depression (n Z 1120) 300 20.8 17.1–24.4
(including unemployment and family problems) and inci-
Baseline past-6-mo (n Z 1120) 162 12.2 9.2–15.3 dent depression. This suggests that mechanisms that explain
Cumulative incidence of depression 113 14.6 10.9–18.3 the observed relation between neighborhood SES and inci-
over 18 mo (n Z 820)a dent depression may be complex, and it is not sufficient or
High neighborhood socioeconomic 45 10.5 5.9–15.2 particularly informative to discuss stressors as a broad undif-
status (n Z 432)
Low neighborhood socioeconomic 68 19.4 13.5–25.3
ferentiated group as mechanisms that explain the relation
status (n Z 388) between neighborhood conditions and psychopathologic
a
states. It is plausible that in different contexts, particular
Among those with no lifetime depression at baseline.
stressors may have different roles in mediating relations
between neighborhood conditions and mental health.
lifetime traumatic events, and lifetime history of PTSD also It also is possible that stressors not measured in this study
were associated significantly with incidence of depression in contribute to the relation between urban neighborhood SES
this multivariable model. and incident depression. For example, additional noise ex-
posure in poorer urban neighborhoods may adversely affect
mental health (41). In the study by Yen and Kaplan (15)
of poverty-area residence and incident depression in
DISCUSSION Alameda County, CA, adjustment for individual health be-
Using data from a population-based prospective cohort haviors (including smoking and alcohol consumption), in
study, we found that the odds of incident major depression addition to other characteristics, such as individual income,
were greater among persons living in poor neighborhoods, attenuated the relation between neighborhood poverty and
independent of individual characteristics. Specifically, mul- depression. Several other studies showed a greater likelihood
tilevel analyses showed that among persons who had never of smoking and heavy drinking in more deprived areas (15,
had depression previously, residents of poorer urban neigh- 42–44), suggesting that these health behaviors themselves
borhoods had more than two times the odds of incident de- may be influenced by neighborhood residence and may
pression during an 18-month period of follow-up relative to mediate the relation between neighborhood poverty and
residents of neighborhoods of higher SES, independent of depression. Additional research should consider the role of
individual-level risk factors for depression, including these and other potential mediators of the relations
individual income. documented here.
Persons living in poor urban neighborhoods may be ex- The quality of the social and built environments of poor-
posed to a greater number of stressors and have less access er urban neighborhoods may contribute directly to the ele-
to salutary resources than persons living in wealthier neigh- vated risk for depression. For example, in relatively poorer
borhoods. For example, residents of socially and economi- neighborhoods, limited social cohesion may diminish
cally deprived neighborhoods may be exposed more community capacity to control group-level processes (45),
frequently to such traumatic event experiences as rape and potentially resulting in manifestations of neighborhood
interpersonal violence and such stressors as unemployment disorder (46–48). Exposure to visible signs of neighborhood
(37), both consistently linked to poorer mental health, in- disorder then may result in psychologic stress and poorer
cluding anxiety and mood disorders (38, 39). Concomi- mental health (18, 37, 49). Consistent with this hypothesis,
tantly, living in poorer neighborhoods may be associated one recent study showed that perceptions of neighborhood
with limited access to sources of material protections and so- characteristics (i.e., vandalism, litter or trash, vacant
cial supports salutary for mental health (37). This mecha- housing, teenagers hanging out, burglary, drug selling, and
nism was called the ‘‘differential vulnerability’’ hypothesis, robbery) predicted depressive symptoms at a 9-month
suggesting that individuals living in deprived neighbor- follow-up interview (19), and another study showed that
hoods may be more likely to experience intermittent trau- persons living in neighborhoods characterized by poorer
matic events and stressors and more vulnerable to their features of the built environment were more likely to report
adverse effects (40). depressive symptoms than persons living in neighborhoods
In our study, we documented an association between ex- characterized by a better built environment (50).
periencing traumatic events and having low social support We show an association between neighborhood SES and
and greater risk for incident depression during follow-up. depression independent of several other variables that are
However, in contrast to other work (13–15), we found known determinants of depression. Depression is comorbid
176 Galea et al. AEP Vol. 17, No. 3
URBAN NEIGHBORHOOD POVERTY AND DEPRESSION March 2007: 171–179

TABLE 3. Bivariate and multivariate associations between key covariates and incident depression
Population with Population with follow-up and
follow-up no lifetime depression Incident depression Multivariable model
Cumulative 95% Confidence
n % n % N incidence pa Odds ratio interval

Total (n) 1120 100.0 820 782

Neighborhood socioeconomic status
High 598 53.3 432 54.1 45 10.5 0.024 1.00
Low 522 46.7 388 45.9 68 19.4 2.19 1.04–4.59
Age, years
18–24 107 15.2 85 14.4 7 6.3 0.260 1.00
25–34 274 24.6 198 25.0 27 15.6 2.62 0.69–9.94
35–44 231 18.8 156 17.8 26 17.3 5.47 1.23–24.36
45–54 211 18.8 144 18.7 23 18.0 5.39 1.16–25.03
55–64 133 12.0 102 12.9 15 18.6 6.73 1.50–30.09
65þ 155 10.6 126 11.3 14 8.9 1.58 0.23–10.98
Sex
Male 505 44.4 379 45.4 38 10.3 0.037 1.00
Female 615 55.6 441 54.6 75 18.2 2.31 1.22–4.39
Race/ethnicity
White 582 38.1 395 35.3 38 13.0 0.216 1.00
Asian 77 6.4 69 7.5 3 4.0 0.30 0.04–2.10
African American 182 25.1 146 26.7 26 15.9 0.70 0.34–1.41
Hispanic 231 26.0 176 25.9 43 19.6 1.55 0.59–4.09
Other 33 4.4 23 4.6 3 14.0 0.93 0.15–5.74
Income, $
30,000þ 617 50.5 428 48.2 44 12.7 0.597 1.00
!30,000 334 33.2 254 34.6 48 15.7 1.26 0.46–3.49
Missing 169 16.3 138 17.2 21 17.7 2.17 1.00–4.70
Education
!High school graduate 133 14.0 107 15.9 21 18.0 0.603 1.00
High school graduate/general 194 22.9 163 24.2 25 11.6 0.67 0.27–1.64
equivalency diploma
Some college 197 22.5 134 21.4 24 19.1 1.75 0.60–5.07
College degree 373 29.1 265 28.2 32 13.3 1.45 0.46–4.57
Graduate degree 220 11.4 148 10.3 11 11.2 1.20 0.30–4.76
Marital status
Married 374 40.4 292 43.0 25 11.9 0.795 1.00
Divorced 121 9.6 77 8.4 15 16.6 0.64 0.24–1.68
Separated 43 3.5 29 3.7 7 16.5 0.86 0.17–4.35
Widowed 81 5.5 67 5.7 12 23.1 2.92 0.80–10.67
Never married 447 37.7 320 36.8 49 15.4 1.88 0.81–4.35
Unmarried couple 48 3.3 32 2.5 5 22.1 1.85 0.38–8.93
Social support
High 405 33.8 288 33.5 27 8.9 0.060 1.00
Medium 317 29.5 232 28.2 25 14.3 1.44 0.65–3.18
Low 380 36.8 286 38.3 59 20.0 3.09 1.64–5.81
Directly affected by September 11b
No 755 70.3 577 72.7 67 13.7 0.411 1.00
Yes 365 29.7 243 27.3 46 17.0 1.15 0.46–2.85
Lived within 2 miles of World Trade Center
No 692 95.2 535 96.1 86 14.9 0.022 1.00
Yes 428 4.9 285 3.9 27 7.3 0.48 0.20–1.15
Traumatic eventsc
0 249 24.5 223 27.8 21 9.4 0.306 1.00
1 237 17.6 186 18.7 22 17.8 2.79 1.00–7.76
2 201 19.5 148 20.0 19 14.1 1.94 0.68–5.53
3 152 15.0 105 15.7 17 13.3 2.49 0.70–8.78
4þ 281 23.4 158 17.9 34 21.0 3.27 1.34–7.94
(Continued)
AEP Vol. 17, No. 3 Galea et al. 177
March 2007: 171–179 URBAN NEIGHBORHOOD POVERTY AND DEPRESSION

TABLE 3. Continued
Population with Population with follow-up and
follow-up no lifetime depression Incident depression Multivariable model
Cumulative 95% Confidence
n % n % N incidence pa Odds ratio interval

Stressorsd
0 810 72.4 635 76.4 80 13.9 0.507 1.00
1þ 310 27.6 185 23.6 33 17.0 1.16 0.54–2.51
Lifetime post-traumatic stress disorderD
No 861 80.5 734 89.5 79 12.3 !0.001 1.00
Yes 259 19.5 86 10.5 34 34.4 3.07 1.43–6.57
a
Two-tailed chi-square tests.
b
Directly affected by September 11 includes persons in the World Trade Center complex during attacks, injured during attacks, lost possessions or property, had a friend or
relative killed, lost a job as a result of the attacks, or involved in rescue efforts.
c
Traumatic events include natural disaster; serious accident at work, in a car, or somewhere else; assault with or without a weapon; unwanted sexual contact; serious injury or
illness; other situation involving serious injury or physical damage; any other situation causing fear of death or serious injury; seeing someone seriously injured or violently killed;
death of a spouse or mate; death of close family member other than spouse; any other extraordinarily stressful situation or event.
d
Stressors include divorce or separation from mate, marriage, family problems with spouse or child, problems at work, and unemployment.

with many other affective and anxiety disorders (51, 52), Second, we used lay-administered telephone interviews
and we found that prior history of PTSD was associated for establishing a probable diagnosis of depression by assess-
with incident depression, even in this group of persons ing major depression episodes. Although telephone and
who had not previously experienced depression. Consistent in-person assessment of Diagnostic and Statistical Manual of
with previous literature, we also found that women had Mental Disorders, Third Edition, Axis I disorders, including
a greater risk for incident depression than men (53, 54). anxiety disorders and affective disorders, were shown to
We note that there are very few published results from pro- result in similar estimates of symptoms (57), depression
spective studies assessing incident psychopathologic states. assessed in this manner cannot be equated to a full diagnosis
Additional work likely is needed to enable definitive com- of major depressive disorder.
ment about the role of race/ethnicity or age as determinants Third, although we used a prospective cohort, we were
of risk for incident depression (27). unable to estimate length of residence of respondents in
The cohort study we report here was started approxi- their particular neighborhoods and hence duration of expo-
mately 6 months after the September 11, 2001, terrorist at- sure to a particular neighborhood context. However, we sug-
tacks and continued in the 2 years thereafter. Therefore, it is gest it is unlikely that persons with incident depression
plausible that the recent mass trauma may have influenced would systematically be more or less likely to move from
the findings documented here. However, in this longitudinal poor to rich neighborhoods (or vice versa) during the brief
analysis, we show that exposure to the events of September (18-month) period included in this assessment. Therefore,
11, 2001, by either proximity or experience, was not associ- any misclassification caused by this limitation likely would
ated with increased risk for major depression. Although de- be nondifferential and unlikely to influence our conclusions.
pression and other psychopathologic states are elevated Future work should assess whether there is a dose–response
soon after a mass traumatic event (55), we previously relation between time or degree of exposure to neighbor-
showed (56) that the prevalence of depression returns to hood poverty and incidence of depression.
baseline in the first 6 months after such traumas, which is Fourth, it is possible that persons who chose not to partic-
when this particular study was started. The absence of ipate in the study could have been different systematically
a strong effect of exposure to the attacks on prospective from those who did not participate, biasing our results.
risk for incident depression in these data suggest that it is Two observations are reassuring in this regard. Using the
unlikely that exposure to the recent September 11 attacks available literature as a guide, recent analyses showed that
explains the relation between neighborhood poverty and response rates are at most weakly associated with bias for
incident depression that we document here. a range of response rates for telephone surveys between
There are several limitations to this study. First, we chose 30% and 70% (58). In one analysis, although a larger differ-
as our units of analysis neighborhoods that are meaningful to ence in response rate was associated with a larger difference
local residents and, as such, may plausibly represent areas in estimates of cigarette-smoking prevalence between the
that can influence population behavior and health. How- telephone-administered Behavioral Risk Factor Surveil-
ever, these neighborhoods are not homogenous, and it is lance System and the in-person Current Population Survey,
possible that smaller urban neighborhoods are more relevant effects were small: a 45 percentage point difference in
for mental health. response rates predicted a difference in smoking prevalence
178 Galea et al. AEP Vol. 17, No. 3
URBAN NEIGHBORHOOD POVERTY AND DEPRESSION March 2007: 171–179

estimates of 1.5 percentage points (58). In an analysis 9. Paykel ES, Abbott R, Jenkins R, Burgha TS, Meltzer H. Urban-rural
mental health differences in Great Britain: Findings from the national
designed to test potential differences associated with differ- morbidity survey. Psychol Med. 2002;30:269–280.
ent response rates obtained from identical surveys, there 10. Romans-Clarkson SE, Walton VA, Herbison GP, Mullen PE. Psychiatric
were very few significant differences across 91 comparisons morbidity among women in urban and rural New Zealand: Psycho-social
comparing data from two surveys with response rates of correlates. Br J Psychiatry. 1990;156:84–91.
61% and 36% (59). Also, in this study, we show (Table 1) 11. Propper C, Jones K, Bolster A, Burgess S, Johnston R, Sarker R. Local
neighborhood and mental health: Evidence from the UK. Soc Sci Med.
no systematic differences between persons enrolled in the 2005;61:2065–2083.
study and census distributions. Importantly, there also was 12. Wainwright NWJ, Surtees PG. Places, people, and their physical and
no systematic difference between persons successfully fol- mental functional health. J Epidemiol Community Health. 2003;58:
lowed up throughout the study and those lost to follow-up 333–339.
with respect to either independent (neighborhood poverty) 13. Walters K, Breeze E, Wilkinson P, Price GM, Bulpitt CJ, Fletcher A. Local
area deprivation and urban-rural differences in anxiety and depression
or dependent (depression) variable of interest. among people older than 75 in Britain. Am J Public Health.
Caveats considered, this study is one of the first to show 2004;94:1768–1774.
by using a prospective design that living in poor neighbor- 14. Weich S, Twigg L, Holt G, Lewis G, Jones K. Contextual risk factors for
hoods may be associated with greater risk for incident de- the common mental disorders in Britain: A multilevel investigation of the
effects of place. J Epidemiol Community Health. 2003;57:616–621.
pression compared with living in wealthier neighborhoods
15. Yen IH, Kaplan GA. Poverty area residence and changes in depression and
while accounting for individual characteristics, including perceived health status: Evidence from the Alameda County study. Int J
individual household income. Additional work is needed Epidemiol. 1999;28:90–94.
to characterize the pathways that may explain the observed 16. Kubzansky LD, Subramanian SV, Kawachi I, Fay ME, Soobader M,
association between living in low-SES neighborhoods and Berkman LF. Neighborhood contextual influences on depressive symp-
toms in the elderly. Am J Epidemiol. 2005;162:253–260.
risk for depression. Elucidation of the particular pathways
17. Ross CE. Neighborhood disadvantage and adult depression. J Health Soc
between neighborhood poverty and depression can help Behav. 2000;41:177–187.
guide interventions that most effectively promote mental 18. Silver E, Mulvey EP, Swanson JW. Neighborhood structural characteristics
health. For example, if the relation between urban neigh- and mental disorder: Faris and Dunham revisited. Soc Sci Med.
borhood poverty and depression is mediated by signs of phys- 2002;55:1457–1470.
ical disorder in a neighborhood, interventions that address 19. Latkin CA, Curry AD. Stressful neighborhoods and depression: A prospec-
tive study of the impact of neighborhood disorder. J Health Soc Behav.
vandalism and trash in urban neighborhoods may be indi- 2003;44:34–44.
cated. Given the preponderance of urban living worldwide, 20. Galea S, Vlahov D, Resnick H, Ahern J, Susser E, Gold J, et al. Trends of
understanding the characteristics of the urban environment probable post-traumatic stress disorder in New York City after the Septem-
that may influence mental health and how these character- ber 11 terrorist attacks. Am J Epidemiol. 2003;158:514–524.
istics exert their influence can provide opportunities for 21. Spitzer RL, Williams JB, Gibbon M. Structured clinical Interview for
DSM-III-RdNon-patient Version. New York: Biometrics Research De-
substantially improving population mental health. partment, New York State Psychiatric Institute; 1987.
22. Boscarino JA, Galea S, Adams R, Ahern J, Resnick H, Vlahov D. Mental
health service and psychiatric medication use following the terrorist at-
tacks in New York City. Psychiatr Serv. 2004;55:274–283.
REFERENCES 23. Kilpatrick DG, Ruggiero KJ, Acierno R, Saunders BE, Resnick HS, Best
CL. Violence and risk of PTSD, major depression, substance abuse/depen-
1. Brockerhoff MP. An urbanizing world. Popul Bull. 2000;55(3):3–4. dence, and comorbidity: Results from the national survey of adolescents.
2. Vlahov D, Galea S, Gibble E, Freudenberg N. Perspectives on urban con- J Consult Clin Psychol. 2003;71:692–700.
ditions and population health. Cad Saude Publica. 2005;21:949–957. 24. Boscarino JA, Galea S, Ahern J, Resnick H, Vlahov D. Utilization of
3. Faris RE, Dunham HW. Mental disorders in urban areas: An ecological mental health services following the September 11th terrorist attacks in
study of schizophrenia and other psychoses. Chicago, IL: University of Manhattan. Int J Emerg Ment Health. 2002;4:143–155.
Chicago; 1939. 25. Derogatis LR. Brief Symptom Inventory 18 (BSI–18) Manual. Minnetonka,
4. White W. The geographical distribution of insanity in the United States. MN: NCS Assessments; 2001.
J Nerv Ment Dis. 1902;30:257–279. 26. Hosmer DW, Lemeshow S. Applied Logistic Regression, 2nd ed. New
5. Srole L, Langner TS, Michael ST, Kirkpatrick P, Opler MK, Rennie TA. York: Wiley; 2000.
Mental health in the metropolis. New York: McGraw-Hill; 1962. 27. Horwath E, Cohen RS, Wiessman MM. Epidemiology of depressive and
6. Blazer DG, George LK, Landerman R, Pennybacker M, Melville ML, anxiety disorders. In: Tsuang MT, Tohen M, eds. Textbook in Psychiatric
Woodbury M, et al. Psychiatric disorders: A rural/urban comparison. Epidemiology. New York: Wiley-Liss; 2002.
Arch Gen Psychiatry. 1985;42:651–656. 28. Sherbourne CD, Stewart AL. The MOS social support survey. Soc Sci
7. Blazer DG, Kessler RC, McGonagle KA, Swartz MS. The prevalence and Med. 1991;32:705–714.
distribution of major depression in a national community sample: The 29. Kilpatrick DG, Resnick HS, Freedy JR, Pelcovitz D, Resick PA, Roth S,
National Comorbidity Survey. Am J Psychiatry. 1994;151:979–986. et al. The posttraumatic stress disorder field trial: Evaluation of the
8. Parikh SV, Wasylenki D, Goering P, Wong J. Mood disorders: Rural/urban PTSD constructdCriteria A through E. In: Widiger TA, Frances AJ, Pin-
differences in prevalence, health care utilization, and disability in Ontario. cus HA, First MB, Ross R, Davis W, eds. DSM-IV Sourcebook, Vol 4.
J Affect Disord. 1996;38:57–65. Washington, DC: American Psychiatric Association; 1998.
AEP Vol. 17, No. 3 Galea et al. 179
March 2007: 171–179 URBAN NEIGHBORHOOD POVERTY AND DEPRESSION

30. Messner SF, Tardiff K. Economic inequality and levels of homicide: An 46. Ross CE, Jang SJ. Neighborhood disorder, fear, and mistrust: The buffering
analysis of urban neighborhoods. Criminology. 1986;24:297–317. role of social ties with neighbors. Am J Community Psychol. 2000;28:401–
31. Marzuk P, Tardiff K, Leon AC, Hirsch CS, Stajic M, Portera L, et al. Pov- 420.
erty and fatal accidental drug overdoses in New York City. Am J Drug 47. Shaw CR, McKay HD. Juvenile delinquency and urban areas: A study of the
Alcohol Abuse. 1997;23:221–228. rates of delinquents in relation to differential characteristics of local commu-
32. Suecoff SA, Avner JR, Chou KJ, Crain EF. A comparison of New York nities in American cities. Chicago, IL: University of Chicago; 1942.
City playground hazards in high- and low-income areas. Arch Pediatr 48. Skogan WG. Disorder and Decline. Berkeley, CA: University of California;
Adolesc Med. 1999;153:363–366. 1990.
33. Galea S, Ahern J, Vlahov D, Coffin P, Fuller C, Leon A, et al. Income dis- 49. Turner RJ, Lloyd DA. The stress process and the social distribution of
tribution and risk of fatal drug overdose in New York City neighborhoods. depression. J Health Soc Behav. 1999;40:374–404.
Drug Alcohol Depend. 2003;70:139–148. 50. Galea S, Ahern J, Karpati A. A model of underlying socioeconomic vulner-
34. Bureau of the Census. Census Summary Tape, File 3A (STF 3A). Wash- ability in human populations: Evidence from variability in population health
ington, DC: US Department of Commerce; 2000. and implications for public health. Soc Sci Med. 2005;60:2417–2430.
35. Zeger SL, Liang KY. Longitudinal data analysis for discrete and continuous 51. Kessler RC, Stang PE, Wittchen HU, Ustun TB, Roy-Burne PP, Walters
outcomes. Biometrics. 1986;42:121–130. EE. Lifetime panic-depression comorbidity in the National Comorbidity
36. Shah B, Barnwell B, Bieler G. SUDAAN User’s Manual, Release 7.5. Re- Survey. Arch Gen Psychiatry. 1998;55:801–808.
search Triangle Park, NC: Research Triangle Institute; 1997. 52. Horwath E, Lish J, Johnson J, Hornig CD, Weissman MM. Depressive
37. Pearlin L. Stress and mental health: A conceptual overview. In: Horowitz symptoms as relative and attributable risk factors for first-onset major
A, Scheid T, eds. A Handbook for the Study of Mental Health. Cam- depression. Arch Gen Psychiatry. 1992;49:817–823.
bridge, UK: Cambridge University; 1999. 53. Kessler RC, McGonagle KA, Zhao S, Nelson CB, Hughes M, Eshleman S,
38. Seedat S, Nyamai C, Njenga F, Vythilingum B, Stein DJ. Trauma et al. Lifetime and 12-month prevalence of DSM-III-R psychiatric disor-
exposure and post-traumatic stress symptoms in urban African schools: ders in the United States: Results from the National Comorbidity Study.
Survey in Cape Town and Nairobi. Br J Psychiatry. 2004;184:169– Arch Gen Psychiatry. 1994;51:8–19.
175. 54. Bijl RV, De Graaf R, Ravelli A, Smit F, Vollebergh WA. Netherlands
39. Sundaram V, Helweg-Larsen K, Laursen B, Bjerregaard P. Physical vio- Mental Health Survey and Incidence Study. Gender and age-specific first
lence, self rated health, and morbidity: Is gender significant for victimiza- incidence of DSM-III-R psychiatric disorders in the general population:
tion? J Epidemiol Community Health. 2004;58:65–70. Results from the Netherlands Mental Health Survey and Incidence Study
(NEMESIS). Soc Psychiatry Psychiatr Epidemiol. 2002;37:372–379.
40. McLeod J, Kessler R. Socioeconomic status differences in vulnerability to
undesirable life events. J Health Soc Behav. 1990;31:162–172. 55. Galea S, Ahern J, Resnick H, Kilpatrick D, Bucuvalas M, Gold J, et al.
Psychological sequelae of the September 11 attacks in Manhattan, New
41. Evans GM. The built environment and mental health. J Urban Health. York City. N Engl J Med. 2002;346:982–987.
2003;80:536–555.
56. Person C, Tracy M, Galea S. Risk factors for depression six months after
42. Kleinschmidt I, Hills M, Elliott P. Smoking behavior can be predicted by disaster. J Nerv Ment Dis. 2006;194(9):659–666.
neighborhood deprivation measures. J Epidemiol Community Health.
1995;49(Suppl 2):S72–77. 57. Paulsen AS, Crowe RR, Noyes R, Pfohl B. Reliability of the telephone inter-
view in diagnosing anxiety disorders. Arch Gen Psychiatry. 1988;45:62–63.
43. Reijneveld SA. The impact of individual and area characteristics on urban
socioeconomic differences in health and smoking. Int J Epidemiol. 58. Mariolis P. Data Accuracy: How Good Are Our Indicators? Presented at:
1998;27:33–40. Proceedings of the 2001 Methodology Symposium, 2001; Canada.
Avilable at: http://www.statcan.ca/english/freepub/11-522-XIE/2001001/
44. Hill TD, Ross CE, Angel RJ. Neighborhood disorder, psychophysiological session2/s2b.pdf.
distress, and health. J Health Soc Behav. 2005;46:170–186.
59. Keeter S, Miller C, Kohut A, Groves R, Presser S. Consequences of reduc-
45. Sampson RJ, Raudenbush SW, Earls F. Neighborhoods and violent ing nonresponse in a national telephone survey. Public Opin Res Q.
crime: A multilevel study of collective efficacy. Science. 1997;5328: 2000;64:125–148.
918–924.

View publication stats