Depressive Symptom Severity and Community Collective Efficacy following the 2004 Florida Hurricanes
Depressive Symptom Severity and Community Collective Efficacy following the 2004 Florida Hurricanes
Carol S. Fullerton 0 1 2 3
Robert J. Ursano 0 1 2 3
Xian Liu 0 1 2 3
Jodi B. A. McKibben 0 1 2 3
Leming Wang 0 1 2 3
Dori B. Reissman 0 1 2 3
0 Funding: This study was supported by funds from the Centers for Disease Control and Prevention. The funders had no role in study design , data collection and analysis, decision to publish, or preparation of the manuscript. USUHS Award Number: HU-0001- 04-1-0002
1 Data Availability Statement: Our data are available for public access upon request consistent with federal law. For further information on data availability, please contact the Uniformed Services University of the Health Sciences Institutional Review Board , 4301 Jones Bridge Road, Bethesda, MD 20814
2 Editor: Jon D. Elhai, Univ of Toledo, UNITED STATES
3 1 Department of Psychiatry, Uniformed Services University of the Health Sciences , Bethesda , Maryland, United States of America, 2 Department of Psychology, West Chester University , West Chester, Pennsylvania , United States of America
There is a lack of research investigating community-level characteristics, such as community collective efficacy, mitigating the impact of disasters on psychological health, specifically depression. We examined the association of community collective efficacy with depressive symptom severity in Florida public health workers (n = 2249) exposed to the 2004 hurricane season using a multilevel approach. Cross-sectional anonymous questionnaires were distributed electronically to all Florida Department of Health (FDOH) personnel that assessed depressive symptom severity and collective efficacy nine months after the 2004 hurricane season. Analyses were conducted at the individual level and community level using zip codes. The majority of participants were female (81.9%), and ages ranged from 20 to 78 years (median = 49 years). The majority of participants (73.4%) were European American, 12.7% were African American, and 9.2% were Hispanic. Using multilevel analysis, our data indicate that higher community-level and individual-level collective efficacy were associated with significantly lower depressive symptom severity (b = -0.09 [95% CI: -0.13, -0.04] and b = -0.09 [95% CI: -0.12, -0.06], respectively) even after adjusting for individual sociodemographic variables, community socioeconomic characteristics, individual injury/damage, and community storm damage. Lower levels of depressive symptom severity were associated with communities with high collective efficacy. Our study highlights the possible importance of programs that enrich community collective efficacy for disaster communities.
State and local public health workers play a critical role as first responders. Concern over public
health response to natural disasters has increased in the aftermath of the 2004 Asian tsunami,
Hurricanes Katrina and Rita, and the 2010 earthquakes in Haiti and Chile. Public health
workers living in disaster-affected communities experience the direct effect of disasters, and at the
Competing Interests: The authors have declared
that no competing interests exist.
same time are responsible for providing care to others. Public health workers include
individuals who provide medical care as well as those in traditional state public health positions, such
as epidemiologists and laboratory technicians. Public health workers exposed to disasters have
high rates of disaster-related distress and posttraumatic stress disorder (PTSD) [1–9], as well as
other psychiatric disorders including depression and alcohol abuse following a disaster [10–
13]. Symptoms of major depression were reported by 27% of firefighters 13 weeks after
responding to Hurricane Katrina , and importantly, symptoms of depression increase with
the number of critical work-related exposures in first responders  Although depression is
often co-morbid with PTSD, it is not always present, and has different risk and protective
factors and a different course than PTSD, suggesting the importance of examining depression
Multiple community characteristics influence mental health outcomes [8, 14, 15]. The
majority of studies of disaster mental health, which address neighborhood and social processes,
measure and analyze them as individual-level variables [16, 17]. Collective efficacy, defined as
social cohesion among neighbors and their willingness to intervene for the common good ,
can be both an individual-level perception and a community-level capacity. Individual
perception of collective efficacy refers to the strength of one’s community in two domains, informal
social control and social cohesion/trust in neighbors . Community collective efficacy is a
measure of the community members’ average perception of collective efficacy associated with
their community. At the community level, willingness to intervene for the common good
depends on mutual trust and solidarity among neighbors.
Nearly all studies supported the association of neighborhood characteristics and depressive
symptoms in a review of studies of neighborhoods and depression . Increases in
community collective efficacy are associated with lower levels of depressive symptoms [18, 19] and
may moderate the effects of stressors, such as perceived discrimination, on depression .
The protective effects of community collective efficacy on depression may be due, in part, to
community resources that buffer psychological responses to stressful events [19, 21, 22], and
promote conservation of individual resources . In contrast, exposure to both displacement
and low social cohesion following Hurricane Katrina was associated with higher risk for
depression , highlighting the importance of both the individual- and community-level
social environment in psychological functioning. Similarly, perception of neighborhood-level
income inequality (characterized by poor quality or lack of resources) was related to increased
depression following 9/11, particularly in residents with lower income, who may be more
vulnerable to the effects of disasters, dependent on community resources, and experience lower
social cohesion .
In the United States, the 2004 Florida hurricane season was unprecedented. Four hurricanes
(Charley, Frances, Ivan, and Jeanne) and one tropical storm (Bonnie) made landfall within a
period of seven weeks . Nearly $4 billion in recovery costs was provided to individuals and
communities . Our previous study of Florida Department of Health (FDOH) public health
workers following the 2004 hurricanes has shown a high mental and behavioral health burden
on the public health workers of the disaster [4, 7]. In addition, higher levels of individual and
community-level collective efficacy were associated with lower rates of PTSD . Given the
differences between depression and PTSD, and previous studies indicating the association of
depression to individual and community collective efficacy, it is important to better understand
the association of depression and collective efficacy following disaster. Such studies can identify
prevention and intervention strategies specific to depression for public health workers.
In previous studies using the current sample and survey instrument, we have shown that,
(1) PTSD, depression, alcohol and tobacco use, and sleep and arousal are important mental
health and behavioral outcomes of this disaster [4,7], and (2) collective efficacy is associated
with PTSD outcomes . In the current study, using the same sample and survey, we examine
the association of collective efficacy to depression. Depression is a distinct disorder associated
with disasters with different symptoms and different treatment implications than PTSD.
Understanding the relationship of collective efficacy to depression can suggest new possible
public health planning surveillance and intervention opportunities.
The present study examines the relationship of both community level and individual level
collective efficacy, as well as injury/damage to depressive symptoms in the same population of
FDOH public health workers nine months after the 2004 hurricanes.
The study was conducted in accordance with the ethical standards and approval of the
Institutional Review Board, Uniformed Services University of the Health Sciences, Bethesda, MD.
Participation was voluntary. Approximately nine months after the 2004 hurricane season
questionnaires and a project description were distributed to FDOH employees using the personnel
e-mail distribution lists. All participants indicated agreement to participate by completing a
questionnaire that was transmitted electronically and anonymously.
Participants and Procedures
This study reports on the association of depression and collective efficacy using the same
sample and survey previously published by the authors where we previously examined core
outcomes, sleep, PTSD and collective efficacy [4,7,8]. Two versions of the questionnaire (i.e., A
and B) were distributed randomly so that each potential participant received either version.
Questionnaire versions contained some of the same items and some unique items, with version
A focusing on mental health items. Of an estimated 8564 FDOH personnel who worked during
the 2004 hurricanes and were available at the time of the survey, we were able to contact and
invite 6637 individuals to participate. After reading a description of the study and the informed
consent, 4323 agreed to participate and completed and returned the questionnaire (Version
A = 2249; Version B = 2074), for an estimated response rate of 65.1%. This study included
respondents who completed Version A. Ages of the participants ranged from 20 to 78 years
(median = 49 years). The majority were female (80.4%; n = 1787) and currently married
(65.6%, n = 1242). Also, the majority were White (73.4%, n = 1390), 12.7% (n = 240) were
African American, 9.2% (n = 175) were Hispanic, and 4.7% (n = 88) other. Approximately half of
the participants had less than a BA/BS degree (50.4%, n = 954). Prior trauma exposure only as
a child was reported by 5.7% (n = 128) of participants, 20.8% (n = 464) reported prior trauma
exposure only as an adult, and 14.8% (n = 330) reported prior trauma exposure both as an
adult and as a child. Occupations of the sample were: administration/program management
(54%, n = 1188); medical (27%, n = 588); epidemiologist/outbreak investigator (7%, n = 145);
statistical/IT services (6%, n = 137); and support services/maintenance (6%, n = 105).
Depression. Depressive symptom severity scores were assessed with the nine-item Patient
Health Questionnaire Depression Scale-9 (PHQ-9) [28, 29]. The PHQ-9 lists all symptoms of
DSM-IV Criterion A of Major Depressive Episode . Respondents rated how much they
had been bothered by each symptom in the previous two weeks on a scale ranging from 0, not
at all to 3, nearly every day. Responses are summed to produce a depressive symptom severity
score ranging from 0 to 27.
Collective Efficacy. Collective efficacy was assessed with the 10-item scale employed by
Sampson and colleagues . The scale has five items in each of two domains: informal social
control and social cohesion / trust scored on five-point Likert scales (ranging from very likely
to very unlikely, and strongly disagree to strongly agree, respectively), and summed to produce
a total score for individual level collective efficacy ranging from 10 to 50. Informal social
control includes five items that ask how likely it would be that respondents’ neighbors could be
counted on to intervene if: a) children were skipping school and hanging out on a street corner;
b) children were spray painting graffiti on a local building; c) children were showing disrespect
to an adult; d) a fire broke out in front of their house; and e) if a fire station closest to their
home was threatened with budget cuts. The social cohesion / trust scale has five items that
assess the extent to which participants agreed that in their home neighborhood: a) people are
willing to help their neighbors; b) it is a close-knit neighborhood; c) people can be trusted; d)
people generally get along with each other; and e) people share the same values. Higher scores
indicate greater collective efficacy.
Community level collective efficacy was assessed using zip codes to define the community
units. For each zip code, the sample mean of those individuals in the zip code was obtained and
rescaled as a centered variable about the grand mean of the entire sample. Since a zip code
represents a collection of people and institutions that occupy a unique subsection of a geographic
location, each zip code is sufficiently externally heterogeneous and internally homogeneous to
be used in multilevel analyses. Given this design, 825 zip codes served as the level-two unit in
this study. Sampson and colleagues  demonstrated high between-neighborhood reliability
(ranging from 0.80 to 0.91) across 343 neighborhoods in Chicago, IL. There was a strong
association between social cohesion and informal social control across neighborhoods (r = 0.80,
p < 0001), suggesting that these scales were measuring aspects of the same latent construct.
Individual hurricane injury/damage. Injury/damage at the time of the hurricanes was
assessed as an individual-level variable by asking participants whether they had experienced
any of the following six events during each of the five hurricanes: loss of electrical power;
damage to vehicle; injury or harm to self; injury or harm to spouse/significant other; injury/harm to
children; and injury/harm to pets. Those reporting at least two of the events during the five
hurricanes were considered to have high hurricane-related injury/damage (n = 1093, 58.14%).
Community hurricane damage. Using FEMA county data for all five storms , we
identified the zip code level of FEMA public and individual assistance received. Each zip code
was scored based on its highest community storm damage across the five storms to index the
level of individual and public assistance received. The level of community storm damage was
assessed using FEMA categories A to G (A. debris removal, B. emergency protective measures,
C. roads and bridges, D. water control facilities, E. public buildings and equipment, F. public
utilities, and G. recreational or other) . We combined categories to create five levels of
public assistance and, therefore, community damage, as follows: 0 = no assistance; 1 = individual
assistance only; 2 = public assistance for category B; 3 = public assistance for categories A-B;
and 4 = public assistance for categories A-G. This level-two variable was then centered.
Socioeconomic characteristics. Ten zip code-specific census measures assessed
socioeconomic characteristics . Following Sampson’s model , three community-level factor
scores, concentrated disadvantage, immigrant concentration, and residential stability, were
extracted from the ten zip code-specific census measures. We used a principal factor analysis
with squared multiple correlations (SMC) for the prior communality estimates. Both
orthogonal and oblique rotations were applied. The oblique rotated factor pattern was highly consistent
with those reported by Sampson and associates [8, 14]. Factor 1, concentrated disadvantage,
had an eigenvalue of 3.94, with high loadings for poverty, receipt of public assistance,
unemployment, female-headed families, density of children, percentage of Black residents, and
percentage of owner-occupied homes. Factor 2, immigrant concentration, captured two
variables with high loadings for percentage of Latinos and percentage of foreign-born individuals.
Factor 3, residential stability, had one variable with a high loading for percentage of persons
living in the same house for the past five years. The three factors were constructed as standardized
scores with a mean of 0 and a standard deviation of 1. These factors were used as level-2 control
variables in the multilevel analyses.
Potential individual and community-level risk factors for higher depressive symptom severity
scores at 9 months post-hurricane in FDOH employees were analyzed using a multilevel
modeling approach. The level 1 unit was individuals (n = 1893) and the level-2 unit was zip
code-defined communities (n = 825). Mean collective efficacy was calculated using all available
data (N = 2.249) yielding a mean of 2.73 participants in each zip code for this calculation. All
subsequent analyses excluded missing cases across all covariates (n = 1893). Statistical analyses
were conducted using SAS software Version 9.2 . Specifically, we applied SAS PROC
MIXED procedure that uses empirical Bayesian approach for handling low reliability in some
of the level-2 units [33, 34].
Random coefficient analyses were used to evaluate the associations with depressive
symptoms. The individual-level collective efficacy predictor was considered in the presence of
individual and community-level control variables. The interaction between injury/damage and
individual-level collective efficacy, and the interaction between injury/damage and community
storm damage were included as additional fixed effects. We considered three random effects
for the intercept, for the slopes of injury/damage, and for the participants within communities.
The degree of clustering within zip codes was assessed by the intra-communities correlation
. We applied the same multilevel approach for community-level collective efficacy, except
that individual collective efficacy was replaced by community-level collective efficacy. We
constructed a multilevel model by including all of the aforementioned covariates.
Nine months after the 2004 hurricanes, high levels of individual injury/damage and high levels
of community storm damage were reported in this group of FDOH workers. Specifically,
57.95% (n = 1097) experienced high levels of personal injury/damage, and the average level of
community storm damage was 1.50 (SD = 1.14) (see Table 1). On a scale ranging from 0 to 27,
the average total depressive symptom severity score was 3.39 (SD = 4.35). Approximately 18
percent (18.2%, n = 344) reported mild depressive symptoms (scores ranging from 5 to < 10)
and 8.9% (n = 168) scored in the moderate to severe depressive symptom level range
(scores 10). The average score for individual-level and community-level collective efficacy
was 36.10 (SD = 7.64) and 36.12 (SD = 4.29), respectively. After accounting for missing data
across the predictor variables, 1893 cases remained for all analyses below.
Depressive symptom severity. Two random coefficient effects analyses were conducted to
evaluate the associations between a) individual-level collective efficacy and depressive
symptom severity and b) community-level collective efficacy and depressive symptom severity.
These relationships were considered while adjusting for the aforementioned individual and
community sociodemographic variables, individual injury/damage, community storm damage,
the interaction between injury/damage and collective efficacy, and the interaction between
injury/damage and community storm damage.
Dependent and independent variables
1 Mean 2%
Individual-level collective efficacy. Beginning with a model containing all covariates,
analyses revealed that a higher level of individual-level collective efficacy was associated with a
significantly lower score of depressive symptom severity. In addition, having high
injury/damage was positively related to depressive symptom severity, other variables being equal. Further,
the interaction between individual injury/damage and community storm damage was not
significant (95% CI: -0.06, 0.82). We also examined the model after removing the nonsignificant
two interactions and three socioeconomic characteristics. This modification to the model did
not significantly change the model chi-square and the parameter estimates remained essentially
unchanged. We used the full model because theoretical as well as previous empirical studies
suggest the relevance of these variables to both depression and collective efficacy and similar
constructs . In the selected full model, after adjusting for all other covariates, a one point
increase in individual-level collective efficacy for those with low injury/damage was associated
with a depressive symptom severity score that was 0.09 points lower (95% CI: -0.12, -0.06). On
the other hand, for those with high injury/damage, a one point increase in individual-level
collective efficacy resulted in a depressive symptom severity score that was 0.13 points lower
(-0.09–0.04), other variables being equal (the score difference in this group was not statistically
significant: see Table 2). The intra-communities correlation for the individual-level efficacy
model was 0.038 (Model χ2 = 89.10, p < 0.001).
Community-level collective efficacy. In a model with all covariates included, analyses
revealed that a higher level of community-level collective efficacy was associated with a
significantly lower score of depressive symptom severity. Further, having high injury/damage was
associated with a higher level of depressive symptom severity. We examined the model after
removing the nonsignificant two interactions and three socioeconomic characteristics. Making
this change to the model did not significantly change the model chi-square and the parameter
estimates remained essentially unchanged. Again, we selected the full model because theoretical
and previous empirical studies suggest the relevance of these variables to depression and
collective efficacy or similar constructs . In the full model, including all covariates, a one point
increase in community-level collective efficacy for those with low injury/damage was also
associated with a depressive symptom severity score that was 0.09 points lower, though with a
slightly different 95% confidence interval (95% CI: -0.13, -0.04). For those with high injury/
damage, a one point increase in individual-level collective efficacy resulted in a depressive
Individual-level collective efficacy models
Community-level collective efficacy models
Full model b (95% confidence interval)
Full model b (95% confidence interval)
symptom severity score that was 0.12 points lower (-0.09–0.03), other covariates being equal
(again the score difference in this group was not statistically significant; see Table 2). The
intracommunities correlation for the community-level efficacy model was 0.034 (Model χ2 = 43.80,
p < 0.001). To assess the impact the small average sample size might generate, we performed
an analysis combining the zip codes with small sample sizes with those having a sufficient
number of participants. Specifically, we used the propensity score matching approach to merge
the subjects residing in zip codes with less than five participants into the zip codes of larger
sample sizes that had the closest propensity score. In this matching process, the propensity is
specified as the probability of low zip-specific collective efficacy (cut-point is 33), using 14
ziplevel variables as predictors. The merging resulted in 135 communities. Using this smaller
number of communities did not change the analytic results significantly, and therefore, the
original, actual zip code was used in formal analyses. Technically, low reliability in some level-2
units was adequately handled by the application of the empirical Bayesian approach available
in the SAS PROC MIXED procedure [33, 34].
Community collective efficacy is associated with mental health and in particular, depression
[18, 19], including following a disaster. Most studies assess community resources or
characteristics at the individual level, and do not address the influence of community-level collective
efficacy. This study examined the relationship of depression and collective efficacy at both the
individual level (the perception of collective efficacy) and the community level using zip codes
to define the community units. We found that both higher community-level and
individuallevel collective efficacy were associated with lower depressive symptoms. This similarity may
reflect neighborhoods that are cohesive in collective efficacy, i.e., individual perceptions are
similar within neighborhoods and therefore may be another indicator of a cohesive
community. Future studies should examine this. Alternatively, those who are depressed are more likely
to report lower perceived collective efficacy and to live in the same community. The relatively
small number of individuals in some of our zip codes may have also influenced this finding.
Among those individuals who had low levels of injury/damage, higher community-level
collective efficacy was associated with lower depressive symptoms, even after adjusting for
individual sociodemographics, community socioeconomic characteristics, the individuals’ degree
of injury/damage, and the community level of storm damage. Among those with high
individual injury/damage, higher community-level collective efficacy was also associated with a lower
depressive symptom score, a relatively though not statistically higher effect compared to those
with low injury/damage. These reductions were also true for individual-level (perceived)
collective efficacy. These findings support a possible role of collective efficacy in depression in post
disaster communities, perhaps more so in communities with lower injury/damage, suggesting
a ceiling effect of injury/damage above which collective efficacy may not be associated with
Though fairly minor at face value, such statistically meaningful decreases in symptoms are
important to understanding mechanisms in population health and psychiatric symptoms.
Although we cannot infer causality, and in particular, reverse causality may be possible, these
changes may imply a considerable alleviation of distress when even such a small change is
applied to a large population after a disaster. Our recent study of PTSD and collective efficacy
showed a similar pattern in which the effects of collective efficacy were fairly low but they
translated into a decreasing trend in PTSD when a wide range of collective efficacy is
considered . The regression coefficients of collective efficacy on depressive symptoms are lower
than those reported for PTSD symptom scores (in the case of community level, -0.09 versus
-0.17). However, this reflects an effect on a health score ranging from 0 to 27 for depression
compared to a range from 17 to 85 for PTSD. Therefore, a one-unit decrease in depressive
symptoms is more meaningful than the same decrease for PTSD, as the score varies within a
much narrower score interval. Therefore, future studies should examine the possibility that
community collective efficacy may actually have a stronger impact on depressive symptoms
than on PTSD.
As mentioned earlier, due to the cross-sectional nature of this study, assumptions about
causal direction are challenging. If perceived collective efficacy is causally related to depression
it may be that higher perceived collective efficacy is protective for depression, and lower
collective efficacy leads to higher depression. Alternatively, it may be that depression makes
one perceive collective efficacy in a more negative light, suggesting further research is needed
to better understand the relationship between depression and perceived collective efficacy.
The present findings must be interpreted in terms of several methodological considerations.
Since this is a cross-sectional study, further research using longitudinal designs is needed.
Because the sample was subdivided into zip codes, the sample size may affect the
representativeness of the zip code. While this is a reasonable choice, it is plausible that in some cases, zip
codes cross neighborhoods. It should be noted that this study examined general depression
that was not necessarily related to trauma and possibly had an onset that preceded exposure to
the hurricanes. Although the current injury/damage assessment captured this variable
adequately, future studies should include general items regarding damage to property.
Although assumptions about the causal nature of our findings cannot be assumed, our
study highlights the possible importance of programs that enrich community collective efficacy
that can be integrated into public health disaster planning. Involvement of community
members in disaster planning can foster community cohesion and a sense of working together in
the face of adversity. This is particularly important for public health care workers who have a
responsibility to care for community members while simultaneously managing their own
responses to a disaster. Community-level interventions should make resources available across
members of the community, strengthening social cohesion and individual and collective
response to disaster events. Opportunities to study such interventions will clarify the causal
relations, and inform programs that promote and strengthen individual and community
resources and mental health in public health workers.
Conceived and designed the experiments: CSF RJU DBR. Performed the experiments: CSF RJU
DBR. Analyzed the data: CSF RJU XL JBAM LW. Contributed reagents/materials/analysis
tools: XL JBAM LW. Wrote the paper: CSF RJU XL JBAM DBR.
1. Akbayrak N , Oflaz F , Aslan O , Ozcan CT , Tastan S , Cicek HS ( 2005 ) Post-traumatic stress disorder symptoms among military health professionals in Turkey . Mil Med 170 : 125 - 129 . PMID: 15782832
2. Benedek DM , Fullerton CS , Ursano RJ ( 2007 ) First responders: mental health consequences of natural and human-made disasters for public health and public safety workers . Annu Rev Public Health 28 : 55 - 68 . PMID: 17367284
3. Carson MA , Paulus LA , Lasko NB , Metzger LJ , Wolfe J , Orr SP , et al. ( 2000 ) Psychophysiologic assessment of posttraumatic stress disorder in Vietnam nurse veterans who witnessed injury or death . J Consult Clin Psychol 68 : 890 - 897 . PMID: 11068975
4. Fullerton CS , McKibben JB , Reissman DB , Scharf T , Kowalski-Trakofler KM , Shultz JM , et al. ( 2013 ) Posttraumatic stress disorder, depression, and alcohol and tobacco use in public health workers after the 2004 Florida hurricanes . Disaster Med Public Health Prep 7 : 89 - 95 . doi: 10.1017/dmp. 2013 .6 PMID: 24618140
5. Kerasiotis B , Motta RW ( 2004 ) Assessment of PTSD symptoms in emergency room, intensive care unit, and general floor nurses . Int J Emerg Ment Health 6 : 121 - 133 . PMID: 15481474
6. McCarroll JE , Fullerton CS , Ursano RJ , Hermsen JM ( 1996 ) Posttraumatic stress symptoms following forensic dental identification: Mt . Carmel, Waco, Texas. Am J Psychiatry 153 : 778 - 782 . PMID: 8633689
7. McKibben JBA , Fullerton CS , Ursano RJ ( 2010 ) Sleep and arousal as risk factors for adverse health and work performance in public health workers involved in the 2004 Florida hurricane season . Disaster Med Public Health Prep 4 : S55 - S62 . doi: 10.1001/dmp. 2010 .9 PMID: 23105036
8. Ursano RJ , McKibben JBA , Reissman DB , Liu X , Wang L , Sampson RJ , et al. ( 2014 ) Posttraumatic stress disorder and community collective efficacy following the 2004 Florida hurricanes . PLOS ONE 9: e88467 1-9. doi: 10.1371/journal.pone.0088467 PMID: 24523900
9. van Kamp I , van der Velden PG , Stellato RK ( 2006 ) Physical and mental health shortly after a disaster: first results from the Enschede firework disaster study . Eu J Public Health 16 : 253 - 259 .
10. Committee on Responding to the Psychological Consequences of Terrorism ( 2003 ) Preparing for the psychological consequences of terrorism: a public health strategy . Washington, DC: National Academies Press . PMID: 25057664
11. Grieger TA , Fullerton CS , Ursano RJ , Reeves JJ ( 2003 ) Acute stress disorder, alcohol use, and perception of safety among hospital staff after the sniper attacks . Psychiatr Serv 54 : 1383 - 1387 . PMID: 14557525
12. Tak S , Driscoll R , Bernard B , West C ( 2007 ) Depressive symptoms among firefighters and related factors after the response to Hurricane Katrina . J Urban Health 84 : 153 - 161 . PMID: 17216569
13. Fullerton CS , Ursano RJ , Wang L ( 2004 ) Acute stress disorder, posttraumatic stress disorder, and depression in disaster or rescue workers . Am J Psychiatry 161 : 1370 - 1376 .12. PMID: 15285961
14. Sampson RJ , Raudenbush SW , Earls F ( 1997 ) Neighborhoods and violent crime: a multilevel study of collective efficacy . Science 277 : 918 - 924 . PMID: 9252316
15. Gapen M , Cross D , Ortigo K , Graham A , Johnson E , Evces M , et al. ( 2011 ) Perceived neighborhood disorder, community cohesion, and PTSD symptoms among low-income African Americans in an urban health setting . Am J Orthopsychiatry 81 : 31 - 37 . doi: 10.1111/j.1939- 0025 . 2010 .01069.x PMID: 21219273
16. Benight CC ( 2004 ) Collective efficacy following a series of natural disasters: Anxiety Stress Coping 17 : 401 - 420 .
17. Norris FH , Stevens SP , Pfefferbaum B , Wyche KF , Pfefferbaum RL ( 2008 ) Community resilience as a metaphor, theory, set of capacities, and strategy for disaster readiness . Am J Community Psychol 41 : 127 - 150 .19. PMID: 18157631
18. Mair C , Diez Roux AV , Galea S ( 2008 ) Are neighbourhood characteristics associated with depressive symptoms? A review of evidence . J Epidemiol Community Health 62 ( 11 ): 940 - 6 . doi: 10.1136/jech. 2007.066605 PMID: 18775943
19. Ahern J , Galea S ( 2011 ) Collective efficacy and major depression in urban neighborhoods . Am J Epidemiol 173 ( 12 ): 1453 - 1462 . doi: 10.1093/aje/kwr030 PMID: 21527512
20. Chou KL ( 2012 ) Perceived discrimination and depression among new migrants to Hong Kong: The moderating role of social support and neighborhood collective efficacy . J Affect Disord 138 ( 1-2 ): 63 - 70 .
21. Kim D ( 2008 ) Blues from the neighborhood? Neighborhood characteristics and depression . Epidemiol Rev 30 : 101 - 17 . doi: 10.1093/epirev/mxn009 PMID: 18753674
22. Cutrona CE , Wallace G , Wesner KA ( 2006 ) Neighborhood Characteristics and Depression: An Examination of Stress Processes . Curr Dir Psychol Sci 15 ( 4 ): 188 - 192 . PMID: 18185846
23. Wind TR , Komproe IH ( 2012 ) The mechanisms that associate community social capital with post-disaster mental health: a multilevel model . Soc Sci Med 75 ( 9 ): 1715 - 1720 . doi: 10.1016/j.socscimed. 2012 . 06.032 PMID: 22883254
24. Le F , Tracy M , Norris FH ( 2013 ) Galea S. Displacement , county social cohesion, and depression after a large-scale traumatic event . Soc Psychiatry Psychiatr Epidemiol 48 ( 11 ): 1729 - 1741 . doi: 10.1007/ s00127- 013 - 0698 - 7 PMID: 23644724
25. Ahern J , Galea S ( 2006 ) Social context and depression after a disaster: the role of income inequality . J Epidemiol Community Health 60 : 766 - 770 . PMID: 16905720
26. National Oceanic and Atmospheric Administration NWS , National Hurricane Center ( 2014 ) Atlantic Hurricane Season . Available from: http://www.nhc. noaa.gov/2004atlan .shtml. Accessed: 23 September 2014 .
27. FEMA ( 2014 ) A decade after 2004 storms, FEMA urges hurricane preparedness . Available from: http:// www.fema. gov/news-release/2014/08/22/decade-after-2004-storms-fema-urges-hurricanepreparedness . Accessed 23 September 2014 .
28. Kroenke K , Spitzer RL ( 2002 ) The PHQ-9: a new depression diagnostic and severity measure . Psychiatric Annals 32 : 509 - 521 .
29. Spitzer RL , Kroenke K , Williams JB ( 1999 ) Validation and utility of a self-report version of PRIME-MD: the PHQ primary care study . Primary Care Evaluation of Mental Disorders. Patient Health Questionnaire. JAMA 282 : 1737 - 1744 . PMID: 10568646
30. American Psychological Association ( 1994 ) Diagnostic and Statistical Manual of Mental Disorders ( 4th edition) . Washington, DC: American Psychological Association.
31. FEMA ( 2015 ) Categorizing losses: using the required FEMA categories to properly handle emergencies . Disaster Recovery Today . Available from: http://adjustersinternational.com/publications/disasterrecovery-today/categorizing-losses/. Accessed 08 February 2015 .
32. SAS Institute Inc . ( 2006 ) SAS/STAT Software, Version 9.2 for Windows. Cary, NC: SAS Inst. Inc.
33. Littell RC ( 2006 ) SAS for mixed models . Cary, NC: SAS Institute Inc .
34. Wang J , Xie H , Fisher JH ( 2009 ) Multilevel models: applications using SAS . Beijing, China: Higher Education Press.
35. Raudenbush SW , Bryk AS ( 2002 ) Hierarchial linear models: applications and data analysis methods ( 2nd edition) . Newbury Park, CA: Sage Publications, Inc.