Predictors for People's Response to aTornado Warning: Arkansas,1March1997

Page created by Lois Romero
 
CONTINUE READING
Disasters, 2000, 24(1): 71^77

Predictors for People's Response to aTornado
Warning: Arkansas,1 March 1997

                   Lina Balluz                                       Laura Schieve
                   Centers for Disease Control                       Centers for Disease Control

                   Talmage Holmes                Stephanie Kiezak
                   Arkansas Department of Health Centers for Disease Control
                   Josephine Malilay
                   Centers for Disease Control

On 1 March 1997, powerful tornadoes touched down in Arkansas (USA) on a Saturday
afternoon. Twenty-six fatalities and 400 non-fatal injuries were reported. We
performed a population-based cross-sectional study to determine factors associated
with appropriate responses to tornado warnings.
   Of 146 survey participants, 140 (96 per cent) knew the difference between ‘torna do
watch’ and ‘torna do warning’ and were aware of when the warning was announced.
Of those 140 participants, 64 (45.7 per cent) responded to the warning by seeking
shelter, and 58 (90.6 per cent) of those 64 acted within five minutes of hearing the
warning. Four factors were positively associated with those seeking shelter: having
graduated from high school (OR = 4.2, 95 per cent CI =1.1ÿ15.5); having a basement
in one=s house (OR = 3.8, 95 per cent exact CI=1.1ÿ17.1); hearing a siren (OR =
4.4, 95 per cent CI = 1.3ÿ18.9); and having prepared a household plan of response
when tornadoes occur (OR=2.6, 95 per cent CI = 1.1ÿ6.3).
   On the basis of these findings, we recommend: first, that people who live in tornado-
prone areas have a personal plan of action to help them respond immediately to warnings;
second, public-health education officials in areas with frequent tornadic activity should do
more to educate the public about what they can do to protect themselves from a tornado;
and third, that emergency-management officials planning protection measures for
vulnerable communities should consider that most people have limited time (our study
documented five minutes) in which to respond to a tornado warning. Thus, shelters in
tornado-prone areas should be quickly accessible by residents.

Keywords: USA, tornadoes, mitigation methods, population surveys.

Introduction

Approximately nine powerful tornadoes hit Arkansas from the south-west to the
north-east part of the state (Figure 1) during the afternoon of 1 March 1997

ß Overseas Development Institute, 2000.
Published by Blackwell Publishers, Oxford OX41JF, UK and 350 Main Street, Malden, MA 02148, USA.
72      L. Balluz, L. Schieve, T. Holmes, S. Kiezak and J. Malilay

(National Weather Service, 1997). Damage indicated that two counties (Saline and
Clark) had been hit by tornadoes of F4 intensity on the Fujita-Pearson tornado scale
(National Weather Service, 1997). These tornadoes were estimated to be one-half to
one mile wide, with wind speeds in excess of 200 miles per hour (National Weather
Service, 1997). Twenty-six fatalities and 400 non-fatal injuries were reported
(CDC, 1997).
   Tornadoes are the most frequent adverse weather event likely to result in a disaster
in the United States (Noji, 1997). Other countries, such as Canada, China, Russia,
Australia, India, Bangladesh, and France, have also recorded significant tornado-
related disasters (Grazulis, 1993). The path of destruction that accompanies a tornado
is highly localised. Therefore, while the infrastructure of an affected community
usually remains intact, these storms can result in mass-casualty situations for
emergency and local medical facilities (Harris, 1992; Glass et al., 1980; CDC, 1985).
   Tornadoes pose a great health threat as they move through heavily populated areas
(CDC, 1992). Because inadequate warning for the population at risk is one of the main
contributing factors to tornado-related deaths and injuries, the National Weather
Service continues to modernise its adverse weather warning systems (Noji, 1997).
However, there is little information available about the public’s response to these
warnings, such as the percentage of people seeking appropriate shelter because of
hearing a tornado warning.
   In this study, we present the results of a post-tornado survey of community
residents in Arkansas to determine factors associated with appropriate responses to
tornado warnings.

Methods

On 13 March 1997, we conducted a population-based cross-sectional study of
residents of Clark and Saline counties. These counties were chosen because they
experienced the highest wind intensities, sustained the most severe property damage
and experienced the most fatalities and injuries as a result of the tornadoes.
    We used random-digit dialling to select a sample of residents from these counties.
An eligible participant was defined as a resident of Saline or Clark county who was at
home while a tornado warning was in effect for the county; and who was present in the
county when the tornado struck the county; and who agreed to participate in the
survey. The random list of telephone numbers in the two counties was generated so as
to be proportional to the population of the two counties: 76 per cent from Saline
county (population 71,143) and 24 per cent from Clark county (population 22,128). On
the basis of discussions with the Arkansas Department of Health and its experience
with random-digit telephone surveys, we estimated that if we attempted to call each
number on the list four times, we would obtain eligible participants from only 20–25
per cent of the numbers dialled. Therefore, we used a list of 1,000 telephone numbers
in an attempt to obtain 200 participants.
    We dialled 1,000 telephone numbers from the list; 156 (17 per cent) of the people
contacted refused to participate. Other results of our attempts are listed in Table 1.
We administered a comprehensive questionnaire to the 146 participants who met the
criteria for eligibility. The questionnaire addressed demographic characteristics,
awareness of the tornado activities and warnings, tornado preparedness, behavioural
Predictors Associated with Response to Tornado Warnings          73

              Table 1 Results of telephone dialling attempts; Arkansas,1997

Results of telephone attempts                                 Number contacted

Disconnected                                                    263
Businesses                                                       85
No response                                                     177
Not at home during the tornado                                   89
Refused to participate                                          156
Participants                                                    146
Other (fax machine, duplicate number)                            84

Total                                                         1,000

responses, characteristics of the residential structure, property damage, injuries
sustained, past experience with tornadoes and resources used following the disaster.
    We used EPI Info software for univariate and bivariate analysis and SAS software for
multivariate logistic-regression analysis. Odds ratios (ORs) were calculated as measures
of the strength of the association between variables. Differences in frequency distribution
were tested by the Mantel-Hanzel chi-square test or Fisher’s exact test when the expected
cell size was under five. The dependent variable for the logistic-regression model was
whether the person responded to the tornado warning by seeking shelter (yes/no).
Independent variables included in the model were all factors that were statistically
associated (p < 0:05) with response to warning in the bivariate analysis. Independent
variables with more than two levels were treated as dummy variables.

Results

Of the 146 participants, none was injured during the tornado, and five (3.4 per cent)
reported damage to their homes. Respondents were predominantly female (73.3 per
cent), and had a mean age of 45 years; 4.9 per cent did not graduate from high school,
32.1 per cent completed high school and 63 per cent had an education above high-
school level. In addition, 140 (95.9 per cent) understood the terms ‘tornado watch’
which indicates that weather conditions are favourable to the formation of a tornado
(Belville, 1987), and ‘tornado warning’ which indicates that a tornado has been
spotted (Belville, 1987). Of those 140 participants, 64 (45.7 per cent) responded to the
warning by seeking shelter inside or outside their house. Seventy-five participants
(53.6 per cent) ignored the warning and did not seek shelter. One participant refused to
provide his response to the warning.
    Television bulletins and warning sirens were the most common means by which
respondents reported being warned about tornadic activity in the area. Of the 88
participants who heard sirens, 75 per cent reported that the signals were clear and loud.
The most common sources from which participants learned about tornadic activity are
listed in Table 2.
    Because response to tornado warnings is critical for preventing or reducing injuries
and deaths, we looked at the characteristics of people who responded to tornado
warnings in Saline and Clark counties by actively seeking shelter. Of the 64 people
74      L. Balluz, L. Schieve, T. Holmes, S. Kiezak and J. Malilay

Table 2   Sources of tornado warnings heard by survey participants who did and did not seek
                                 shelter; Arkansas,1997

Sources                                  Participants who            Participants who
of tornado warning                       sought shelter              did not seek shelter
                                         Number        %             Number      %

Watching television                      52           81.2           59          78.6
Hearing siren                            46           71.8           42          56.0
Noticing weather changes                 32           50.0           30          40.0
Hearing commercial radio                 16           25.0            8          10.6

who responded to warnings, 58 (90.6 per cent) had more than a high-school education,
26 (40.6 per cent) lived in wooden-frame houses, and 12 (18.8 per cent) had basements
in their homes. Of the 75 participants who did not respond to warnings (that is, they
ignored warnings and did not seek shelter), 34 (45.3 per cent) had more than a high-
school education, 26 (34.7 per cent) lived in wooden-frame houses, and only five (6.7
per cent) had basements, namely a room below ground inside the house.
    Factors that were positively associated with people responding to warnings
included: having at least a high-school education; having a basement in their homes;
living in an area that was in the tornado path and receiving a tornado warning; having
heard sirens; and having a prepared plan of action. People with at least a high-school
education were more likely to respond to warning (OR = 4.2, 95 per cent CI =
1.1ÿ15.5). People living in a house with a basement were more likely to respond than
those without a basement (OR = 3.8, exactly 95 per cent CI = 1.1ÿ17.1). In addition,
people who responded were more likely to live in an area that was in the path of the
tornado (OR= 7.2, 95 per cent CI=1.940.6), to have heard sirens (OR=4.4,
CI=1.3ÿ18.9) and to have prepared a plan of action (OR=2.6, 95 per cent
CI=1.1ÿ6.3) than were those who did not respond to the warning.
    Results of the multiple logistic regression modelling to predict people’s response to
a tornado warning were similar to those from the univariate analysis (Table 3).
    We also looked at when participants first learned of tornado activity and the amount
of time they had to respond. Of the 146 participants, 82 (56 per cent) reported first
being aware of a tornado threat at least 10 minutes before it passed through the area,
and of the 64 who responded to the warnings, 58 (90.6 per cent) reported a response
time (time between hearing a warning and seeking shelter) of five minutes.
    Forty (62.5 per cent) of the 64 participants who responded to the warning reported
seeking shelter inside their homes, and 14 (21.8 per cent) reported seeking shelter
underground.
    The areas of the house most frequently used as shelter were hallways (reported by
25, or 39 per cent, of those who sought shelter) and basements (reported by 11, or 17
per cent). The use of other forms of protection — such as standing behind a table, desk
or bench, or covering oneself with a blanket or mattress — was uncommon. Only 17
(12 per cent) of the 146 participants in the study reported having an external storm
cellar within 300 feet of their residence.
    Most survey respondents indicated some degree of pre-tornado preparedness. Of
the 146 participants, 77 (52.7 per cent) had experienced previous tornado events or
Predictors Associated with Response to Tornado Warnings             75

Table 3 Predictors for seeking shelter derived from multiple logistic regression; Arkansas,1997

Independent variables                        Odds ratio       Confidence          P-value
                                             (OR)             intervals (CI)

Intercept                                                                         0.001
Education level                               4.2             1.14–15.5           0.031
Hearing a siren                               9.2             1.63–54.1           0.012
Being in path of a tornado                   14.9             2.71–81.6           0.002
Having a prepared plan of action              3.7             1.41–9.9            0.009

other severe weather conditions; 107 (73 per cent) had received safety information
regarding tornadoes but 51 per cent did not use the information appropriately.

Discussion

Unlike other epidemiologic studies that identified risk factors for morbidity and
mortality during tornado disasters, this investigation identified factors associated with
seeking shelter. These factors were hearing warning sirens, having a basement in one’s
home, having a plan of action and having at least a high-school education.
   The decline in tornado-related deaths since 1950 has been attributed to
improvements in tornado-warning systems (Noji, 1997). We found that the warning
system in the study area was adequate. People heard warnings via radio and television
channels, which issued them as early as 11.34 am on 1 March, more than 90 minutes
prior to the first occurrence of severe weather in Arkansas. Between 12.04 pm and
6.52 pm, 33 of the 55 counties where severe weather occurred received tornado
warnings and the remaining 22 received severe thunderstorm warnings.
   Although most of the participants heard loud sirens or received warnings from the
radio, television or a telephone call, adequate warning is not effective if people at risk
have no access to shelter (National Climatic Data Center, 1989). We found that the
lack of a basement in the house or the inability to gain access to a nearby storm cellar
were important factors for responding to tornado warnings. Most participants who did
not respond did not have access to such shelters. Although several studies recommend
the use of an interior room in the house as an alternative for people without basements
(Grazulis, 1993; Simin et al., 1996; Sellers et al., 1986; CDC, 1985), only a few people
sought shelter in such a room (that is, a room inside the house without windows)
which may indicate a lack of public knowledge about the protection that an interior
room can offer. In addition, we showed that among those who responded by seeking
shelter in a basement or a hallway, 26 (41 per cent) did not use further protection,
although current recommendations are for people to protect themselves with a mattress
or similar covering in order to reduce the risk for injuries (Duclos and Ing, 1989).
These results indicate that areas with a high tornadic activity need increased public-
health education about how residents can protect themselves after hearing a tornado
warning.
   Most people reported hearing a tornado warning at least 10 minutes before the
tornado passed through the area. Those who responded typically sought shelter within
five minutes of first becoming aware of the warning. Thus, storm shelters in tornado-
76     L. Balluz, L. Schieve, T. Holmes, S. Kiezak and J. Malilay

prone areas should be quickly accessible by residents. Only a few participants in our
study reported that they had access to cellars close to their residence.
   Public safety during tornado disasters could be improved through public-education
programmes about appropriate protective actions (Brenner and Noji, 1993). In
Arkansas, although one such programme provides this information via television, we
found that 51 per cent of the people did not use the information appropriately.
   Selection bias was a possible factor because we included only those people with
telephones still in working order, and excluded those who were displaced by the
tornadoes, such as residents of mobile homes. None the less, this survey addresses the
experience of many people who were at risk of sustaining injury, death or property
damage during the tornado. Recall bias among respondents also may have occurred
because of their estimates of the time between first hearing a tornado warning and the
impact of the tornadoes.

Conclusions
Our findings are consistent with other post-disaster investigations that identified
factors specifically associated with reducing mortality and injury. We showed that
those factors were also associated with seeking appropriate shelters after hearing
tornado warnings.
   We recommend that people who live in tornado-prone areas without underground
shelters should prepare a plan of action to help them respond immediately to tornado
warnings in an effective manner. Elements of this plan may include engaging in tornado
drills to familiarise members of the household with safety procedures, sources of
shelters in the home or outdoors, and keeping specialised weather radios such as those
from the National Oceanic and Atmospheric Administration (NOAA) available in the
home. Because the frequency of tornado watches may de-sensitise people to the dangers
of tornadoes, we also recommend increased public-health education about how to
respond to tornado watches and warnings in areas with high tornadic activity. This
education should focus on eliciting immediate action from residents who hear a tornado
warning and helping them to understand why certain actions and behaviours are
recommended. When planning protective actions — such as constructing accessible
tornado-resistant shelters and developing tornado-preparedness and tornado-response
programmes for at-risk populations — emergency-management officials should keep in
mind that people have limited time to find shelters. Preparedness programmes should
also include the establishment of local observer networks and warning devices that are
not dependent on radio and television broadcasts.

Acknowledgement
The authors thank Eric Cregson, Dr Donna Barre and the staff of the Arkansas
Department of Health for their co-operation and assistance in conducting the survey.

References
Belville, J.D. (1987) The National Weather Service Warning System. Ann Emerg Med 16:
    1078–80.
Predictors Associated with Response to Tornado Warnings       77

Brenner, S.A. and E.K. Noji (1993) Risk Factors for Death or Injury in Tornadoes. An
    Epidemiologic Approach. American Geophysical Union.
Centers for Disease Control and Prevention (1985) Tornado Disaster, North Carolina,
    South Carolina. MMWR 34: 205–12.
(1985) Tornado Disaster, North Carolina, South Carolina, March 28, 1984. MMWR 34:
    205–13.
(1992) Tornado Disaster, Kansas, 1992. MMWR 41: 181–3.
(1997) Tornado-associated Fatalities; Arkansas, 1997. MMWR 46 (no.19).
Duclos, P.J. and R.T. Ing (1989) Injuries and Risk Factors for Injuries from the 29 May
    1982 Tornado; Marion, Illinois. Int J Epidemiol 18: 213–19.
Glass, R.I., R.B. Carven, D.J. Bergman et al. (1980) Injuries from the Wichita Falls
    Tornado: Implications for Prevention. Science 207: 734–8.
Grazulis, T.P. (1993) Significant Tornadoes 1680–1991. The Tornado Project of
    Environmental Film, St Johnsbury.
Harris, L.F. (1992) Hospitalized Tornado Victims. Ala Med 61: 12–16.
National Weather Service report (1997) Service Assessment of the March 1, 1997 Arkansas
    Tornado Outbreak.
National Climatic Data Center (1989) National Summary of Tornadoes 1989. National
    Climatic Data Center, 1013 (storm data; vol. 31), Asheville.
Noji, E.K. (1997) The Public Health Consequences of Disaster. Oxford University Press,
    New York.
Sellers, F., D. Read, A.B. Rakow, et al. (1986) Tornado disaster- Pennsylvania. MMWR
    35: 233–5.
Simin, L., L.E. Quenemoen, J. Malilay, E. Noji, T. Sinks, J. Mendlein (1996) Assessment
    of a Severe Weather Warning System and Disaster Preparedness. Calhoun County,
    Alabama, 1994. Am J Publ Health 86: 87–9.

Address for correspondence: Lina Balluz, Centers for Disease Control and Prevention,
Health Studies Branch, 4770 Buford Highway, N.E., Mail stop F-46, Atlanta, Georgia
30341-3724, USA. E-mail: >
You can also read