Human and Pet-related Risk Factors for Household Evacuation Failure During a Natural Disaster

Sebastian E. Heath ) 2 3 Philip H. Kass 1 3 Alan M. Beck 0 3 Larry T. Glickman 0 3 0 Department of Veterinary Pathobiology, Purdue University School of Veterinary Medicine, West Lafayette, IN. Reprint requests to Dr. Sebastian E. Heath , 1650 Harvard Street, NW, Apt. 210, Washington, DC 20009-3727 ( 1 Department of Population Health and Reproduction, School of Veterinary Medicine, University of California , Davis, CA 2 Department of Veterinary Clinical Sciences, Purdue University School of Veterinary Medicine , West Lafayette, IN 3 Received for publication December 8 , 1999, and accepted for publication July 18, 2000. Abbreviations: CI, confidence interval; OR, odds ratio; OR This study characterized risk factors for household evacuation failure. A random digit dial telephone survey was conducted of 397 households in Yuba County, California, in July 1997, 6 months after residents had been under evacuation notice due to flooding. Case households failed to evacuate, whereas control households evacuated. The cumulative incidence of household evacuation failure was 19.4%. Fewer households with children (25.8%) failed to evacuate than households without children (45.9%, p < 0.01). More households with pets (20.9%) than households without pets failed to evacuate (16.3%, p = 0.11). With multivariate logistic regression, the risk of household evacuation failure was lower in households with children (odds ratio = 0.4, 95% confidence interval: 0.2, 0.8) compared with households without children. The risk of household evacuation failure increased in pet-owning households without children (odds ratio = 1.3, 95% confidence interval: 1.1, 1.5) compared with pet-owning households with children; the more pets a household owned, the higher the risk of household evacuation failure was. Impediments to pet evacuation, including owning multiple pets, owning outdoor dogs, or not having a cat carrier, explained why many households that owned pets failed to evacuate. Predisaster planning should place a high priority on facilitating pet evacuation through predisaster education of pet owners and emergency management personnel. Am J Epidemiol 2001;153:659-65. aged; cats; child; disasters; dogs; natural disasters - Household evacuation failure rates have been as high as 3070 percent in cresting floods (1), 510 percent along the shorefront where hurricanes were predicted to make landfall and up to 70 percent inland (2), and up to 12 percent in hazardous materials spills (3). The most consistent factor associated with household evacuation appears to be the presence of children (4, 5), while the factor most often associated with evacuation delay or failure has been the increasing age of persons in households (6). Pet owners may not evacuate from a disaster area if they cannot take their animals with them (710). Particularly relevant to the evacuation of pet owners is the strength of the human-animal bond (a term used to describe the close relationship between humans and domestic species) (11), the perceived risk to the owners and their pets, the time and resources needed to evacuate animals, and facilities for the housing of evacuated animals (8, 9, 12). Public health regulations typically do not allow animals into public shelters (13). If owners anticipate being separated from their pets or being turned away from public shelters because of their pets, they may refuse to evacuate (14, 15). Because a large proportion (>50 percent) of US households owns pets, even a small effect of pet ownership may have a large impact in disasters. The objective of this study was to characterize risk factors for household evacuation failure by comparing characteristics, attitudes, and resources of households that evacuated from an impending disaster with those that did not. Households in this study had at least 36 hours advance notice to evacuate. MATERIALS AND METHODS Residents of Yuba County, California, were issued a voluntary evacuation notice on the morning of January 1, 1997, because of flooding (16). After a levee broke the following day, a mandatory evacuation order was issued and remained in effect for 2 days. A random digit dial population survey of residential telephones in Yuba County where residents were known to have been under evacuation notice was conducted in July 1997. A commercial contractor (Chilton Research Services, Radnor, Pennsylvania) obtained information through a structured interview. First, four screening questions were used to determine whether the respondent could represent the entire household, whether the household had been under an evacuation order, whether the household had evacuated, and whether there were pets of any type in the household. Second, on the basis of the responses to the screening questions, households under evacuation notice were placed into one of four groups. Questions were designed to ascertain the evacuation behavior of each member of the household. Two components of the human-animal bond, attachment (17) and commitment (18), were measured in households that owned pets; pets included dogs and cats only. Household evacuation failure was designated as the primary outcome of interest (dependent variable), while predictor variables for household evacuation failure included sociodemographic factors, household composition, perceived level of risk from the flood, pet-related factors, and pet attachment and commitment scores. The unit of study was defined as a household because all human members of a household usually exhibit the same evacuation behavior (4, 19, 20). On the basis of screening questions, four household groups were defined: 1) evacuated and owned pets (n 203); 2) evacuated and did not own pets (n 50); 3) did not evacuate and owned pets (n 100); and 4) did not evacuate and did not own pets (n 44). The measure of association between evacuation failure and each putative risk factor was expressed as the odds ratio and 95 percent confidence interval. Confidence intervals for the regression coefficients were estimated by using the maximum likelihood estimators (21). All statistical analyses were conducted using SAS version 6.12 (22) and Epi-Info 6.04 (23) statistical software programs. The representativeness of the sample population was determined by comparing sociodemographic information for the same area obtained from the US Bureau of Census. Two-tailed t tests were used to compare continuous data, and the 2 test for homogeneity or Fishers exact test was used for categorical variables. The 2 test for trend was used to analyze ordered categorical variables. Results were considered statistically significant at p < 0.05. Potential risk factors for household evacuation failure were initially examined in univariate logistic regression models for all households and separately for those with and those without dogs or cats. The risk of household evacuation failure was also estimated in analyses stratified by the presence of children (those aged <19 years) (...truncated)