Disaster Planning and Risk Communication With Vulnerable Communities: Lessons From Hurricane Katrina

Abstract

We studied the experience of Hurricane Katrina evacuees to better understand factors influencing evacuation decisions in impoverished, mainly minority communities that were most severely affected by the disaster.
We performed qualitative interviews with 58 randomly selected evacuees living in Houston's major evacuation centers from September 9 to 12, 2005. Transcripts were content analyzed using grounded theory methodology.
Participants were mainly African American, had low incomes, and were from New Orleans. Participants' strong ties to extended family, friends, and community groups influenced other factors affecting evacuation, including transportation, access to shelter, and perception of evacuation messages. These social connections cut both ways, which facilitated and hindered evacuation decisions.
Effective disaster plans must account for the specific obstacles encountered by vulnerable and minority communities. Removing the more apparent obstacles of shelter and transportation will likely be insufficient for improving disaster plans for impoverished, minority communities. The important influence of extended families and social networks demand better community-based communication and preparation strategies.

Full text

Supplement 1, 2007, Vol 97, No. S1 | American Journal of Public Health Eisenman et al. | Peer Reviewed | Research and Practice | S109

RESEARCH AND PRACTICE

Objectives. We studied the experience of Hurricane Katrina evacuees to better
understand factors influencing evacuation decisions in impoverished, mainly mi-
nority communities that were most severely affected by the disaster.
Methods. We performed qualitative interviews with 58 randomly selected evac-
uees living in Houston’s major evacuation centers from September 9 to 12, 2005.
Transcripts were content analyzed using grounded theory methodology.
Results. Participants were mainly African American, had low incomes, and were
from New Orleans. Participants’ strong ties to extended family, friends, and com-
munity groups influenced other factors affecting evacuation, including trans-
portation, access to shelter, and perception of evacuation messages. These social
connections cut both ways, which facilitated and hindered evacuation decisions.
Conclusions. Effective disaster plans must account for the specific obstacles en-
countered by vulnerable and minority communities. Removing the more appar-
ent obstacles of shelter and transportation will likely be insufficient for improv-
ing disaster plans for impoverished, minority communities. The important
influence of extended families and social networks demand better community-
based communication and preparation strategies. (Am J Public Health. 2007;97:
S109–S115. doi:10.2105/AJPH.2005.084335)
Disaster Planning and Risk Communication With
Vulnerable Communities: Lessons From Hurricane Katrina
| David P. Eisenman, MD, MSHS, Kristina M. Cordasco, MD, MPH, Steve Asch, MD, MSHS, Joya F. Golden, BA, and Deborah Glik, ScD
The aftermath of Hurricane Katrina demon-
strated that a large portion of American soci-
ety lives without the social and economic re-
sources necessary to protect themselves and
their families during disasters. More than
100 000 residents of greater New Orleans,
many of whom were poor and African Ameri-
can,
1
did not evacuate before Katrina’s land-
fall.
2
Research findings on the association be-
tween race/ethnicity and evacuation decisions
are consistent with the pattern seen with Kat-
rina: minority communities are less likely to
evacuate and are more affected by disas-
ters.
3–6
The vulnerability of these communi-
ties appears to be mainly because of eco-
nomic status and resources; however,
“cultural ignorance, ethnic insensitivity, racial
isolation and racial bias” potentially also con-
tribute by leading to lower levels of prepared-
ness, fewer resources for evacuation, and dis-
parities in access to relief and recovery.
4(p169)
Evacuation research generally compares
populations who do and do not evacuate,
7–10
so the few studies that address racial/ethnic
variations describe disparities between broad
majority and minority groups, leaving much
about the crucial factors operating within mi-
nority groups unknown.
11 , 12
In-depth investi-
gations of evacuation decisions are needed to
understand why impoverished, urban, minor-
ity communities may be less likely to evacu-
ate. A survey done by Brodie et al
13
of the
predominantly poor and African American
persons residing in Houston’s shelters after
Hurricane Katrina highlights some of the
challenges faced in evacuating vulnerable mi-
nority communities; the most commonly en-
dorsed responses for not evacuating were
lack of transportation and misjudging the
storm’s danger. Still, social psychological the-
ory predicts that decisionmaking is complex,
multifactorial, and socially embedded.
14
Re-
sults of studies, such as that by Brodie et al,
13
invite further questions about the complex
interrelationships of the reasons offered in the
survey for not evacuating and potentially oth-
ers (e.g., family refused to evacuate, inability
to leave work, and credibility of evacuation
recommendations) that were not included in
the survey.
12
Qualitative research can add to the depth
of our understanding by providing a detailed
accounting of the cultural context, social envi-
ronment, and individual cognitions that led to
people’s decisions and abilities to evacuate
before Hurricane Katrina. We interviewed
evacuees from Hurricane Katrina to better
understand, from their perspective, the com-
plex set of factors that influenced evacuation
behaviors. Understanding these experiences
could inform disaster plans for impoverished,
minority, communities in the United States.
METHODS
Qualitative interviews were conducted with
persons living in Houston’s 3 major evacua-
tion centers: the Reliant Center, Reliant As-
trodome, and George Brown Convention Cen-
ter. Eligibility criteria included being aged 18
years or older and residing in Louisiana be-
fore Hurricane Katrina. Recruitment and
interviewing were conducted September 9 to
12, 2005 (Table 1). On September 9, 8623
persons were residing in these shelters.
15
Study participants were randomly selected
from inside and outside the centers from
9:00
AM to 9:00 PM. Inside each shelter, rows
of residents’ cots were counted as were the
cots in the row. Row and cot number were se-
lected from random-numbers tables, and the
selected cot was approached. Outside, line
sampling was performed from lines of people
sitting against the shelters for shade. Random
numbers were used to select potential partici-
pants counting down the line. In both cases,
children were excluded from the counting.
The sampling method was repeated whether
the person who was approached accepted or
declined to be interviewed.
Participation was voluntary and anony-
mous, and all of the participants provided ver-
bal informed consent. Interviews were con-
ducted in the most private setting available,
and no incentives were offered.
With the exception of the demographic
components (Table 2), all of the interviews
were conducted in a semistructured format.
The interviewer asked participants to describe

American Journal of Public Health | Supplement 1, 2007, Vol 97, No. S1S110 | Research and Practice | Peer Reviewed | Eisenman et al.

RESEARCH AND PRACTICE

TABLE 1—Timeline of Events Associated With Hurricane Katrina, 2005
Date Event
Wednesday, August 24 Storm in Bahamas upgraded to Tropical Storm Katrina
Thursday,August 25 Federal Emergency Management Agency National Coordination Center activated; Katrina
upgraded to category 1 hurricane, passes through Florida into Gulf
Friday,August 26 Governor of Louisiana declares state of emergency; National Hurricane Center forecasts Katrina
will strike east of New Orleans
Saturday,August 27 5:00
AM Katrina upgraded to Category 3
9:00
AM Mandatory evacuation orders in St Charles and Plaquemines Parishes and coastal
areas of Jefferson Parish; voluntary evacuation in St Tammany, St Bernard, New
Orleans Parish, and noncoastal areas of Jefferson Parish
President Bush declares federal state of emergency in Louisiana
Sunday,August 28 12:40 AM Hurricane Katrina upgraded to category 4
7:00 AM Hurricane Katrina upgraded to category 5
10:00 AM Mandatory evacuation of New Orleans ordered; New Orleans Superdome named as
shelter of “last resort for evacuees with special needs”
Monday,August 29 6:10 AM Hurricane Katrina makes landfall just east of New Orleans
Lower Ninth Ward and St. Bernard parishes under 6—10 feet of water
2:00 PM Officials confirm breach of 17th Street Canal; 2 other levees also breached
Tuesday, August 31 Governor of Louisiana orders evacuation of all of New Orleans, including 23 000 people
estimated to be at New Orleans Superdome; agreement made to use Houston’s Reliant
Astrodome as shelter
Wednesday, September 1 5000 evacuees arrive in Reliant Astrodome; additional 10000 people come to New Orleans
Superdome seeking evacuation
Thursday, September 2 Houston’s 3 other major shelters begin accepting evacuees (Reliant Center, Reliant Arena, and
George R. Brown Convention Center)
Friday, September 3 Evacuation of New Orleans Superdome and Convention Center completed with a total of
42 000 evacuees
Saturday, September 4 At their peek census, the 4 major Houston shelters house 27 100 evacuees.
Source. Information taken from a variety of sources and timelines (including The New York Times, Times-Picayune, Houston
Chronicle, Federal Emergency Management Agency, Cable News Network, CBS News, Harris County Joint Information Center,
and the Brookings Institution).
TABLE 2—Summary of Participant
Characteristics (N= 58): Houston, Tex,
September 9—12, 2005
Variable N %
Residence
New Orleans 55 95
Outskirts of New Orleans 3 5
Interview site
Reliant Center/Astrodome 40 69
Brown Center 11 19
Outside centers 7 12
Race
African American 47 81
White 6 10
Latino 3 5
Asian–Pacific Islander/Other 2 3
Gender
Female 28 48
Male 30 52
Age, y
18–24 4 7
25–34 5 9
35–44 10 17
45–54 17 29
55–64 10 17
65–74 8 14
≥75 3 5
Missing 1 2
Income, US $
<20 000 29 50
20 000–29 999 18 31
30 000–39 999 5 9
40 000–49 999 1 2
≥50 000 3 5
Refused 2 3
Marital status
Married 16 28
Separated 5 9
Divorced 7 12
Widowed 6 10
Never Married 24 41
Education, highest degree
None, less than high school 26 45
High school diploma or General 23 40
Educational Development
AA or associates degree, 2-y degree 6 10
BA or BS, 4-y degree 2 3
Missing or refused 1 2
Continued
the following: (1) sources and understanding
of information in the time period before the
hurricane; (2) knowledge, perceptions, and re-
sources that influenced their evacuation be-
havior before the hurricane’s landfall; and
(3) reflections on factors that would have al-
tered their behavior. Interviews were re-
corded and professionally transcribed.
We performed a content analysis to iden-
tify core consistencies and patterns in the in-
terviews that we called themes. We used a
grounded theory approach,
16
which empha-
sizes discovering themes that emerge from
the data, while recognizing that themes may
also come from preconceived frameworks.
First, we read through all of the transcripts
and, in team discussions, identified themes
and their subthemes that emerged from the
texts.
17 , 18
Identifying themes was completed
when no new subthemes, properties, dimen-
sions, or relationships within or among them
came out during discussions.
19
Second, each
transcript was reviewed by the investigators
who independently categorized, sorted, and
labeled participants’ statements within
themes and subthemes, a process known as
coding. Third, the “coders” met, and differ-
ences in coding were resolved by consensus
agreement.
20
Initial agreement was 85% re-
quiring 15% of codes to be settled by con-
sensus. Final coded transcripts were managed
using Atlas.ti software version 5.0 (Atlas.ti
Scientific Software Development GmbH,
Berlin, Germany). We report the most salient

Supplement 1, 2007, Vol 97, No. S1 | American Journal of Public Health Eisenman et al. | Peer Reviewed | Research and Practice | S111

RESEARCH AND PRACTICE

TABLE 2—Continued
Employment
Full time 23 40
Part time 5 9
Unemployed/laid off 3 5
Disabled/not working 10 17
Retired/not working 13 22
Other/not working 4 7
Automobile access and ownership
No access 32 55
Yes access 26 45
BRepresentative Statements From Identified Themes and Subthemes
Instrumental Reasons
Shelter (N = 73)
• “They said go to Texas but I didn’t know anybody in Texas.”
• “Really truly, we had cars, but we didn’t know anybody to go to.”
• “People think when you run to the projects you’re more safer there than here, so
that’s what basically we did.”
Transportation (N =112)
• “I mean, if you’ve got 20 people trying to get in 1 car it’s not going to happen.
So some people, you just stay because you have to.”
• “They would have had to send buses like close up to the door so they get [the
elderly] out, because some people ain’t going to go out, you know, walking.”
Money, property, jobs (N = 109)
• “They were already robbing. And my dad, he had to stay behind because we had
a lot of tools and belongings there.”
Health (N = 29)
• “I no healthy to drive too far.”
• “I take so much medication by that time I was like groggy.”
Social networks (N =106)
• “The older people are really hard to leave because they’ve been through [Hurri-
canes] Betsy and Camille. So it’s very hard for you to get older people to leave.”
• “Like my mom said, she’s been through Betsy, Camille, all the hurricanes, the major
hurricanes and she just wasn’t evacuating. So I wasn’t going to leave my mom
to stay there by herself.”
• “I had a 90-year-old mother that I was taking care of and she would not leave that
house for hell or high water.”
• “I could have made it on my own, but it was just my aunt and my uncle. Every few
steps he made . . . she forgot his walker . . . every few steps he made he was
falling down.”
Cognitive reasons
Sources, timing, and understanding of messages (N= 286)
• “Some was telling us that we should evacuate . . . and some of them was telling
us to stand by.”
• “They said we don’t know what’s going to happen, but there’s slight chance every-
body should leave New Orleans.”
• “All we heard was what was broadcast on television telling everybody that if they
have somewhere to go, get out as soon as possible.”
• “In order for them to make a mandatory evacuation, it had to be a bad, very bad
storm coming our way.”
Risk perception (N = 188)
• “And, well, we didn’t take it seriously. Because we thought it was going to be an-
other Betsy. You know, with a little water coming in the house.”
• “The last storm we had there, it was more people got hurt on the highway trav-
eling away from the storm, running out of gas, accidents, than it would have been
if they stayed home.”
• “The water started to rise due to that breech in the levee system. That’s what
caused that water, not the hurricane.”
Sociocultural reasons (N = 102)
• “The mayor, the governor of New Orleans, that run the city of New Orleans they
let the waters go in the poor neighborhoods and kept it out of the rich neighbor-
hoods like that French Quarter where tourists goes at.”
• “It was from them opening flood gates, telling lies about the levee breaking and
stuff . . . I believe they do these things intentionally . . . so they can flood out those
black neighborhoods.”
N=Number of coded statements.
themes and subthemes we found. The se-
lected quotations exemplify the core topics,
problems, and concerns that were the bases
for the themes and subthemes.
RESULTS
Fifty-eight (77%) of 75 persons ap-
proached agreed to participate. Most partici-
pants were African American, reported low
income, and came from New Orleans
(Table 2). Four participants reported they
had evacuated before the storm.
A total of 1182 statements were identified
and coded into themes and subthemes (see the
box 1on this page). The themes were as fol-
lows: (1) instrumental, defined as the re-
sources needed for and practical concerns re-
lated to evacuation; (2) cognitive, defined as
the cognitive components of how evacuation
messages were received, processed, and un-
derstood; and (3) sociocultural, defined as be-
liefs, attitudes, and perceptions of the nature
of disasters or the underlying community
structure or politics, including discrimination.
Instrumental
Shelter. Participants did not recall specific
destinations outside New Orleans prescribed
in the evacuation orders. Said 1 participant,
“There was nothing about where we was sup-
posed to go at.” Participants described ex-
tended family outside New Orleans who of-
fered them shelter before the hurricane,
although they did not evacuate for other rea-
sons. One person did evacuate before the hur-
ricane “because my daughter’s mother-in-law
lives in Appaloosa, Louisiana, and every time
we hear something going on out in the Gulf . . .

American Journal of Public Health | Supplement 1, 2007, Vol 97, No. S1S112 | Research and Practice | Peer Reviewed | Eisenman et al.

RESEARCH AND PRACTICE

it’s like an open invitation.” Others noted that
the absence of friends and family outside New
Orleans hindered evacuation (see the box on
the previous page). Persons who did not evac-
uate often thought that they would be safer if
they moved to “sturdier” dwellings in New Or-
leans such as hospitals, hotels, high-rise apart-
ments, and public housing buildings.
Transportation. Transportation was an ob-
stacle for participants who did not have a
functioning car, could not find a rental car,
or had no insurance, license, or gas. Even if
participants owned a car, 1 car for the entire
family may not have been enough; other
family members had already evacuated with
it or the family was too large for a single car
(see the box on the previous page). Persons
who wanted to evacuate by bus did not
know where to board them, reported no
buses in their neighborhoods, or had elderly
in their homes who could not walk to them
(see the box on the previous page).
Money, Property, and Jobs. Concerns about
money, property, or jobs impeded evacuation in
several ways. First, disposable income was an
obstacle, because, as 1 single mother reported,
“You have to be able to feed your children
when you leave. You have to be able to have a
place to stay, you have to have gas money, you
have to have rental car money. I couldn’t afford
to do that. You need at least $500/$600, and
that’s the least amount of money.”
Second, participants reported obstacles to
leaving imposed by their employers. Three
participants reported not evacuating for fear of
losing their jobs. One reported a conversation
with his employer who said, “‘If you don’t
come around then, you know, I’ll just see you
when I see you.’. . .That means when I see you
you’re going to be fired.” Another woman, dis-
cussing her clients from a home for persons
with HIV/AIDS said, “We had 5 of them placed,
2 of them were not placed, so that means when
we had to evacuate . . . I had to take them with
me.” Last, personal experience of looters or a
fear of looters influenced evacuation decisions
(see the box on the previous page).
Health. The health of participants or mem-
bers of extended family members influenced
evacuation (see the box on the previous
page). For 1 person, it was fear of being far
from their usual providers, “because I’m a
diabetic and I have to be close by to get to
doctors and hospital—I can’t just go any-
where.” Health problems (fatigue and somno-
lence) hindered 2 participants from driving
themselves, although they owned their own
cars (see the box on the previous page).
Social Network. A total of 106 statements
focused on the importance of social networks
(the web of relationships that surround indi-
viduals) as obstacles or facilitators to evacua-
tion. Statements about social networks that
overlapped with statements about transporta-
tion, shelter, and health described how fam-
ily, friends, and neighbors influenced deci-
sionmaking and the ability to evacuate.
Previous experiences of elderly family mem-
bers affected decisions to evacuate (see the
box on the previous page). Obligations to the
elderly also influenced household’s evacua-
tion attempts. One participant said, “My plans
were to leave. Unfortunately we received a
call and we had to come back home. My
mother-in-law had called for us to come back.
. . . You know when they get a certain age
they get confused.”
Participants described an individual’s evac-
uation decision initiating a chain reaction that
impacted an entire family: “My mother-in-law
wouldn’t leave the house. My husband would-
n’t leave her and I’m not going to leave him.”
Alternatively, the extended family was a re-
source that encouraged the decision and ca-
pacity to evacuate: “I started making phone
calls to my children warning them to get out.
And after that, my sister, she had called me.
So I went to pick her and her children up,
and grandchildren, and we just started driv-
ing, heading toward Florida.”
Cognitive
Sources, Timing, and Understanding of Mes-
sages. Participants’ initial sources of information
about the hurricane were television or family
and friends. Participants reported understand-
ing televised warnings about the hurricane but
evacuation messages were recalled as nonspe-
cific or ambiguous, for instance, messages to
“go somewhere” but not where and how to
evacuate (see the box on the previous page).
Participants focused on 2 terms used in
messages: “category 5” and “mandatory”
evacuation. The term “category 5” impressed
people with the storm’s severity, the threat
it posed, and the need to evacuate. Said 1
participant, “Oh, I knew I would die because
they said, category 5, category 5, and I knew
it was the end.” The term “mandatory” in-
creased the perceived severity of the storm
(see the box on the previous page), and par-
ticipants distinguished between responding to
a “mandatory” versus “voluntary” evacuation:
“They wasn’t saying it was mandatory; they
was saying if you can leave, please leave.
They didn’t say it was mandatory because if
they would have I would have left.”
Participants integrated media messages with
the communications of friends, family, neigh-
bors, and church members in deciding
whether or not to evacuate. Emotional ap-
peals from contacts and seeing that “every-
body was in line at the store” swayed persons
from ambivalence to evacuation. A participant
received direct appeals from a social services
worker that changed her mind: “So our clini-
cal manager called back. She says, ‘Stella, the
Lord said get out of that house.’ I said, ‘We’re
on our way out now if you would hang up.’”
Risk Perception. Participants who did not
evacuate thought that their own vulnerability
would be low based on previous experience
(see the box on the previous page). Vulnera-
bility would be minimal because, “I know it’s
a flooding city but the street I live on does
not flood,” or because, “I was on the second
floor,” or because, “the project was the safest
place to be.” Participants reported mistrusting
warnings as “trying to scare us.” Stated 1 per-
son, “They do this every year when its time
for a hurricane.” Finally, participants saw
evacuation as riskier than staying home be-
cause of the dangers of driving on the high-
way (see the box on the previous page).
Sociocultural. Collective memory of past hur-
ricanes led participants to believe that they
rarely bring damage, dangerous flooding, or
personal risk: “I mean, ain’t nothing exciting
about no hurricane. But, you know, living in
New Orleans . . . you’re under the threat of it.
So it’s been gusting, but that’s where you stay.”
Furthermore, distrust of authorities fueled
their belief that the flooding was not because
of the hurricane. Participants thought that the
levees had been “blown” to save wealthy
neighborhoods and businesses at the expense
of poor, black neighborhoods. Said 1 partici-
pant, “They knew they had to sacrifice some-
body, and they all came together and decided

Supplement 1, 2007, Vol 97, No. S1 | American Journal of Public Health Eisenman et al. | Peer Reviewed | Research and Practice | S113

RESEARCH AND PRACTICE

what they were going to do. And the Ninth
Ward is all the crack dens and the lower in-
comes, the poorest of the poor, the helpless,
the criminal element.”
DISCUSSION
Participants described factors influencing
evacuation decisions that were complex, inter-
acted with one another, and most importantly,
were influenced by extended family and other
members of participants’ social networks, ei-
ther facilitating or inhibiting evacuation. State-
ments regarding social networks were the most
common theme spontaneously mentioned and
influenced the outcomes of other themes, such
as transportation, shelter, and health. Although
advice from others or responsibility for others
(usually children) led several participants to
evacuate, friends and relatives just as often
were obstacles to evacuation. Participants who
took responsibility for elderly or frail family
members or sheltered extended family mem-
bers in their homes were subsequently unable
to evacuate. Family members who evacuated
early using the family’s automobile stranded
others without transportation.
Previous studies describing the importance
of family in the decision to evacuate have fo-
cused on households
21, 22
with less attention to
the extended family, which is often more co-
hesive in minority communities.
23
An ac-
cepted finding of disaster research is that peo-
ple tend to evacuate as family units.
22,24
Research with Mexican Americans found that
accounting for extended family members de-
layed responses to disaster warnings.
5
Simi-
larly, elderly persons may be less likely to
evacuate,
10
and their presence in a household
may complicate family evacuation.
22
Our
study not only confirms the importance of
these findings in the largest natural disaster in
the United States in a century but also may
provide clues to the mechanisms behind
them. We found that broad networks of fami-
lies and friends created demands on partici-
pants and more inertia to evacuation among
the household and group. These demands
stretched limited resources, such as trans-
portation, hindering the household’s and
group’s ability to evacuate. Obligations to
the extended family, especially the elderly,
who resisted evacuation, inhibited many
individuals and nuclear families from evacuat-
ing. Decisionmaking occurred among groups
larger than the nuclear family, and group de-
cisionmaking, with its inherent inefficiencies,
sometimes overwhelmed decisionmaking of
individuals.
25,26
In only a few participants did
the balance among the willingness to evacu-
ate, resources for evacuation, and the needs
of the extended family members favor evacu-
ation, although extended families provided
other vital functions, such as assistance during
the storm and the flood afterward.
Although transportation and financial fac-
tors alone did prevent some from evacuating,
more important was the confluence of these
obstacles to evacuation. For example, al-
though almost 45% of the sample reported
that they owned or had access to a car, many
still did not have money for gas, hotels, or
food, which, as estimated by 1 participant,
could cost 2 weeks’ pay. Participants who did
not have family outside of New Orleans often
lacked specific evacuation destinations and
said that messages did not provide specific
destinations. Moreover, although most study
participants did not initially evacuate New Or-
leans, many participants did leave their homes
for other’s homes or places of employment
because they perceived them as safer than
their own houses. Fears that one would lose
their job, additional responsibilities imposed
by employers, and protecting property from
looting, especially if it meant losing the tools
of one’s trade, kept others from evacuating.
The terms “category 5” and “mandatory”
evacuation strongly resonated with partici-
pants. This is consistent with a previous study
that found that the difference between a cate-
gory 3 and category 4 hurricane is the
boundary at which people feel at risk.
27
This
also highlights the value of education on com-
mon disaster terms, such as hurricane cate-
gories 1 to 5 (Saffir–Simpson Hurricane Cate-
gories), as an important aspect of disaster
policy. Education provided by emergency
planners can increase citizens’ knowledge
about hazard risks,
28
and people are more
likely to evacuate if they perceive enough risk
to themselves and loved ones.
3,10,22
For other participants, however, the collec-
tive memory of past hurricanes combined
with distrust of authorities led to the mini-
mization of their perceived risk associated
with Hurricane Katrina. Initially, this under-
estimation of the danger was weighed against
fears of getting “stuck” on the highway, also
derived from collective memory. Studies
show that trusting information received from
inside one’s social network rather than the
information received from outside one’s net-
work (e.g., media and authority figures) is
characteristic of networks composed of
strong family ties.
29
These networks are bet-
ter suited for social support than for ex-
changing new information.
29
During times of
stress, networks composed of intense ties,
such as extended families, may be less adap-
tive, because they are less likely to exchange
new information.
30
In such cases, disaster
planners partnering with organizations that
are part of these networks, such as churches
in the African American community, may
better communicate new information to indi-
viduals and families who value information
received from inside their network.
31, 32
Limitations
There are limitations to our study. Social
response bias or cognitive dissonance may
have influenced recall of evacuation deci-
sions. Also, the target population (persons
living in the 3 Houston evacuation centers)
was intended to represent the most vulner-
able participants who did not evacuate be-
fore Hurricane Katrina and needed evacua-
tion and housing assistance. Participants
were, therefore, not selected from hotels and
private residences where evacuees were liv-
ing, and so may not generalize to other, pos-
sibly less-disadvantaged, evacuees. The pe-
culiarities of the large-scale evacuation from
impoverished sections of New Orleans pro-
vided a skewed population for study, which
was further compounded by the rapidly
changing occupancy of the shelters. We at-
tempted to reduce bias in sample selection
strategy, but the study sample remains a
convenience sample with respect to those
who may have been evacuated to these 3
shelters and who remained in the shelters on
September 9 to 12, 2005. As a result, our
sample was older, poorer, less well educated,
and disproportionately African American
compared with New Orleans residents be-
fore Hurricane Katrina.
33
Also, without de-
mographic data on the overall population in

American Journal of Public Health | Supplement 1, 2007, Vol 97, No. S1S114 | Research and Practice | Peer Reviewed | Eisenman et al.

RESEARCH AND PRACTICE

the 3 centers, we are unable to determine
whether our sample was representative of
the target population living there. Our ran-
domly selected sample is comparable in gen-
der, income, and marital status to another
sample randomly selected for a survey that
ran concurrently in the shelters.
13
The quali-
tative nature of our investigation is best
suited to range finding and evaluating inter-
actions between factors associated with evac-
uation decisions. Without a quantitative ap-
proach, it is difficult to determine the
prevalence of the factors that we identified.
Disaster programs, policies, and research
can address the obstacles that we identified.
For example, emergency food and gas vouch-
ers that are activated when an evacuation is
announced could be provided to eligible fami-
lies. National and state legislation that pro-
vides job protection for those who evacuate
could be modeled after laws such as the Uni-
formed Services Employment and Reemploy-
ment Rights Act of 1994, which provides
reemployment protection for active duty mili-
tary personnel. Education programs are
needed to address the misperceptions that
persist, for instance, that evacuating to a
home that is sturdier but still in the storm’s
path increases one’s safety. Public education
campaigns are needed to educate the public
about the level of protection provided by a
“sturdy” house against a severe hurricane and
the need to travel away from the storm.
Moreover, disaster programs and research
should focus on social units (households, ex-
tended families, and neighborhoods) and the
organizations and groups they interact with,
such as churches, social clubs, schools, and
labor unions.
31
For instance, churches could
be central to disaster planning and communi-
cations in African American communities, be-
cause they provide important links to
families
31
and have a history of successful in-
volvement in social projects and health pro-
motion.
34–36
They could be used to sway in-
dividuals and families whose norms, risk
perceptions, and decisionmaking are highly
influenced by their social networks and to
provide access to resources. Church groups
might organize phone trees to promote evacu-
ations and organize evacuation planning for
the elderly.
37
They could participate in devis-
ing and pretesting communications developed
along risk communication principles,
38,39
which emphasize that effective crisis mes-
sages are consistent, timely, actionable, and
empathic to the manifold and complex situa-
tions in which families may find them-
selves.
40–42
Studies demonstrate that the
clearer and more specific the disaster warning
messages are, for instance, providing simple
evacuation instructions and directions to spe-
cific destinations, the more likely the public
will respond to them.
43,44
Conclusions
As we seek lessons from the evacuation of
Hurricane Katrina, it is important that com-
munications and disaster plans account for
the specific obstacles encountered by urban,
minority communities. Our work provides an
opportunity to listen to the voices of the
evacuees themselves. These voices lead us to
believe that removing the obstacles of shel-
ter and transportation will be insufficient to
ensure safety in future disasters. Policies
must additionally address the important in-
fluence of extended families and social net-
works through better community-based com-
munication and preparation strategies.
About the Authors
David P. Eisenman is with the David Geffen School of Medi-
cine at the University of California, Los Angeles, and the
RAND Corporation, Los Angeles, Calif. Kristina M. Cor-
dasco is with the David Geffen School of Medicine at Univer-
sity of California, Los Angeles. Steve Asch is with the VA
Greater Los Angeles Healthcare System, the David Geffen
School of Medicine at the University of California, Los Ange-
les, and the RAND Corporation, Los Angeles. Joya F. Golden
is with the VA Greater Los Angeles Healthcare System, Los
Angeles. Deborah Glik is with the University of California,
Los Angeles School of Public Health, Los Angeles.
Requests for reprints should be sent to David P. Eisen-
man, David Geffen School of Medicine, University of Cali-
fornia, Los Angeles, Division of General Internal Medicine
and Health Services Research, 911 Broxton Plaza, Los An-
geles, CA 90095-1736 (e-mail: deisenman@
mednet.ucla.edu).
This research article was accepted July 31, 2006.
Contributors
D.P. Eisenman designed and directed the study and led
its implementation, data collection, data analysis, inter-
pretation, and the writing and editing of the article. K.M.
Cordasco worked with D. P. Eisenman to implement the
study and collected and analyzed data. S. Asch co-
designed the study and advised D.P. Eisenman. J. F.
Golden analyzed data. D. Glik advised the team on data
analysis and performed data analysis. All of the authors
helped to conceptualize ideas, interpret findings, and
write and review drafts of the article.
Acknowledgements
Support for this study was provided through funds
awarded to D. P. Eisenman by the Natural Hazards Re-
search and Applications Information Center from the
National Science Foundation (grant CMS 0408499)
and from an award from the Centers for Disease Con-
trol and Prevention (grant K01-CD000049-02).
The authors gratefully acknowledge the advice and
assistance of Dennis S. Mileti in originating the study,
Michele Allen in interpreting the data on social net-
works, and Joseph Rabatin in reviewing the article.
Human Participant Protection
The study was approved by the Human Subjects Pro-
tection Committee of the University of California, Los
Angeles.
References
1. Gabe T, Falk G, McCarty M. Hurricane Katrina:
Social-Demographic Characteristics of Impacted Areas.
Washington, DC: Congressional Research Service;
2005.
2. Nigg JM, Barnshaw J, Torres MR. Hurricane Katrina
and the flooding of New Orleans: Emergent issues in
sheltering and temporary housing. Ann Am Acad Pol Soc
Sci. 2006;604:115–128.
3. Riad J, Norris F, Ruback RB. Predicting evacua-
tion in two major disasters: risk perception, social influ-
ence, and access to resources. J Appl Soc Psychol.
1999;29:918–934.
4. Fothergill A, Maestas E, Darlington J. Race, ethnic-
ity and disasters in the United States: a review of the
literature. Disasters. 1999;23:156–173.
5. Perry RW, Mushkatel AH. Minority Citizens in
Disasters. Athens, Ga: University of Georgia Press;
1986.
6. Perry RW, Lindell MK The effects of ethnicity on
evacuation decision-making. Int J Mass Emerg Disasters.
19 91;9(47–68).
7. Drabek TE. Shall we leave? A study on family re-
actions when disaster strikes. Emerg Manage Rev.
1999;1:25–29.
8. Fischer HW, Stine GF, Stoker BL, Trowbridge ML,
Drain EM. Evacuation behavior: why do some evacu-
ate, while others do not? A case study of Ephata, Penn-
sylvania (USA) evacuation. Disaster Prev Manage. 19 95;
4:30–36.
9. Aguirre BE. Evacuation in Cancun during Hurri-
cane Gilbert. Int J Mass Emerg Disasters. 19 91;9:31–45.
10. Baker EJ. Predicting response to hurricane warn-
ings: a reanalysis of data from four studies. Mass Emer-
gencies. 1979;4:9–24.
11.Drabek TE. Understanding disaster warning re-
sponses. Soc Sci J. 1999;36:515.
12.Tierney KJ, Lindell MK, Perry R. Facing the Unex-
pected: Disaster Preparedness and Response in the United
States. Washington, DC: Joseph Henry Press; 1999.
13.Brodie M, Weltzien E, Altman D, Blendon RJ,
Benson JM. Experiences of Hurricane Katrina Evacuees
in Houston Shelters: Implications for future planning.
Am J Public Health. 2006;96:1402–1408.
14 .Weinstein ND. The precaution adoption process.
Health Psychol. 1988;7:355–386.

Supplement 1, 2007, Vol 97, No. S1 | American Journal of Public Health Eisenman et al. | Peer Reviewed | Research and Practice | S115

RESEARCH AND PRACTICE

15. Harris County Joint Information Center. Hurricane
Katrina relief report. Available at: http://www.hcjic.
org/default.asp. Accessed November 29, 2005.
16. Glaser B, Strauss A. The Discovery of Grounded
Theory: Strategies for Qualitative Research. Chicago, Ill:
Aldine; 1967.
17. Lincoln YS, Guba EG. Naturalistic Inquiry. New-
bury Park, Calif: Sage Publications; 1985.
18.Ryan G, Bernard HR. Techniques to identify
themes. Field Methods. 2003;15:85–109.
19. Strauss A, Corbin J. Basics of Qualitative Research:
Grounded Theory Procedures and Techniques. Newbury
Park, Calif: Sage Publications, Inc; 1990.
20. MacQueen K, McLellan E, Kay K, Milstein N.
Codebook development for team-based qualitative
analysis. Cultural Anthropol Method J. 19 9 8;10:31–36.
21. Howell S, Bonner D. Citizen Hurricane Evacuation
Behavior in Southeastern Lousiana: A Twelve Parish Survey.
July, 2005. Available at: http://www.poli.uno.edu/unopoll/
Summary%20Report%20July%2019%202005%20(2).
pdf. Accessed December 22, 2005.
22. Perry RW. Evacuation decision-making in natural
disasters. Mass Emergencies. 1979;4:25–38.
23. Ruggles S. The origins of African-American family
structure. Am Sociol Rev. 1994;59:136–151.
24.Drabek TE, Boggs KS. Families in disasters: reac-
tions and relatives. J Marriage Fam. 1968;30:
443–451.
25. Adler P, Kwon S. Social capital: the good, the bad,
and the ugly. In: Knowledge and Social Capital: Founda-
tions and Applications, Lesser E, ed. Boston, Mass: But-
terworth-Heineman; 2000;89–115.
26. Buckland J, Rahman M. Community-based disas-
ter management during the 1997 Red River Flood in
Canada. Disasters. 1999;23:174–191.
27. Whitehead JC, Edwards B, Van Willigen M,
Maiolo J, Wilson K. Heading for higher ground: factors
affecting hurricane evacuation behavior. Environ Haz-
ard. 2000;2:133–142.
28. Zhang Y, Prater CS, Lindell MK. Risk area accu-
racy and evacuation from Hurricane Bret. Nat Hazard
Rev. 2004;5:115–120.
29.Granovetter M. The strength of weak ties: a net-
work theory revisited. Sociol Theory. 19 83;1:201–233.
30. Gottlieb BH, Wagner F. Selecting and planning
support interventions. In: Social Support Measurement
and Interventions, Cohen S, Underwood LG, Gottlieb
BH, eds. New York, NY: Oxford University Press;
2000.
31.Peterson J, Atwood JR, Yates B. Key elements for
church-based health promotion programs: outcome-
based literature review. Public Health Nurs. 2002;19:
401–411.
32. Glass T, Schoch-Spana M. Bioterrorism and the
people: how to vaccinate a city against panic. Clin In-
fect Dis. 2002;34:217–223.
33. United States Census Bureau. 2004 American
Community Survey. Available at: http://www.census.
gov/acs/www. Accessed May 19, 2005.
34. DeHaven MJ, Hunter IB, Wilder L, Walton JW,
Berry J. Health programs in faith-based organizations:
are they effective? Am J Public Health. 2004;94:
103 0–1036.
35. Young DR, Stewart KJ. A church-based physical
activity intervention for African American women. Fam
Commun Health. 2006;29:103–115.
36.Resnicow K, Kramish Campbell M, Carr C, et al.
Body and soul: a dietary intervention conducted
through African-American churches. Am J Prevent Med.
2004;27:97.
37. Eng E, Hatch JW. Networking between agencies
and black churches: the lay health advisor model. Prev
Hum Serv. 19 91;10:123–146.
38. Lasker RD. Redefining Readiness: Terrorism Plan-
ning Through the Eyes of the Public. New York, NY: Cen-
ter for the Advancement of Collaborative Strategies in
Health, The New York Academy of Medicine; 2004.
39. Morgan MG, Fischhoff B, Bostrom A, Atman C.
Risk Communication: The Mental Models Approach. New
York, NY: Cambridge University Press; 2001.
40.Reynolds B, Seeger M. Crisis and emergency risk
communication: an integrative approach. J Health Com-
mun. 2005;10:38–41.
41.Kreps G, Alibek K, Neuhausr L, Rowan K, Sparks
L. Emergency/risk communication to promote public
health and respond to biological threats. In: Harden M,
ed. Global Public Health. Sudbury, Mass: Jones and Bul-
wer; 2005:349–362.
42. Covello V, Peters R, Wojtecki J, Hyde R. Risk
communication, the West Nile virus epidemic, and
bioterrorism: responding to the communication chal-
lenges posed by the intentional or unintentional release
of a pathogen in an urban setting. J Urban Health.
2001;78:382–391.
43. Mileti D, Fitzpatrick C. Causal sequence of risk
communication in the Parkfield Earthquake Prediction
Experiment. Risk Analysis. 19 92;12:393–400.
44. Mileti D, Drabek TE, Hass JE. Human Systems in
Extreme Environments. Boulder, Colo: Institute of Be-
havioral Science, University of Colorado; 1975.