Priviledge and Place: An Exploratory Study About Healthcare Bypass Behavior

.................................................................................................................................... i ACKNOWLEDGEMENTS ............................................................................................................ ii TABLE OF CONTENTS ............................................................................................................... iii Community Context .................................................................................................................... 2 Individual Privilege and Ability.................................................................................................. 4 Place-based Constraints .............................................................................................................. 5 Summary and Expectations......................................................................................................... 5 METHODS ..................................................................................................................................... 6 Measurements ............................................................................................................................. 7 Dependent Variable: ............................................................................................................... 7 Independent Variables: ........................................................................................................... 8 Analytic strategy ....................................................................................................................... 10 RESULTS ..................................................................................................................................... 11 DISCUSSION ............................................................................................................................... 14 REFERENCES ............................................................................................................................. 17 TABLES ....................................................................................................................................... 20 Table 1: Descriptive Statistics .................................................................................................. 20 Table 2: Logistic regression of Bypass ..................................................................................... 21 Table 3: Logistic regression of Bypass by SoVI ...................................................................... 22

ii ACKNOWLEDGEMENTS I wish to show my gratitude to my Chair, Dr. Michael R. Cope, whose mentorship, beginning in my undergraduate years, was crucial for the completion of this research. He instructed me through all phases of the research process, believed in me, and sent me the most encouraging memes. He also helped me through a faith crisis, mission trauma, anger at the church, and multiple existential crises. Additionally, I am grateful for the feedback and  Bypass behavior occurs when people choose non-local rather than local healthcare (Chan, Hart, and Goodman 2006;Liu et al. 2008;Radcliff et al. 2003). For urban residents, with a wider selection of healthcare, bypass has a limited effect on the community, but for rural residents, bypass can have significant consequences not only on their individual health but on their community (Chan, Hart, and Goodman 2006). Because fewer healthcare providers exist in rural areas, bypassing can lead to under-use, diminished quality, and subsequent closure of rural clinics, ultimately creating healthcare deserts (Chan, Hart, and Goodman 2006;Liu et al. 2008;Liu et al. 2007;Radcliff et al. 2003;Sanders et al. 2015).
Much of the literature on bypass behavior use hospital records, which limits understanding of individual preferences, constraints, and other socioeconomic factors that influence decisions about healthcare (Chan, Hart, and Goodman 2006;Radcliff et al. 2003). It is crucial for public health workers to understand why rural residents bypass their primary healthcare provider (Borders et al. 2000;Sanders et al. 2015). Current research indicates that bypass behavior is used to obtain better healthcare (Liu et al. 2008;Radcliff et al. 2003;Yao and Agadjanian 2018), and assumes that residents are willing to travel further distances because the further away healthcare is higher quality. Bypass is also associated with a perceived limited selection of local doctors or individuals that already use an out-of-town specialist (Borders et al. 2000;Liu et al. 2007). However, a better understanding of why people bypass their primary care provider (PCP) can help policymakers and practitioners provide better healthcare to rural residents and ensure adequate healthcare resources for rural and remote places.
Existing literature describes the variation in bypass behavior by state (Radcliff et al. 2003), insurance type (Radcliff et al. 2003), and on a rural-urban spectrum (Chan, Hart, and Goodman 2006). It also predicts the likelihood of bypassing according to individual characteristics (Liu et al. 2008;Yao and Agadjanian 2018), community characteristics (Sanders et al. 2015;Sanders et al. 2016), frequency of care (Liu et al. 2007), distance to hospitals, and hospital characteristics (Roh and Moon 2005). This study expands on these by measuring not only community ties and individual characteristics, but by creating an index to measure and rank the social vulnerability of each town included in the sample. Just as healthcare selection varies depending on various household and individual socioeconomic characteristics, this study tests the effect that socioeconomic characteristics of communities have on resident's healthcare selection. This study measures social vulnerability of place by including an index of characteristics of place (e.g. percent of the population renting, per capita income, percent of residents with insurance, median home value, and percent of healthcare workers in the area) that better explains rural living conditions. A social vulnerability of place measure will better explain rural healthcare selection behavior, and uses The Social Vulnerability Index (SoVI), a wellestablished measure of social vulnerability of place which takes into account several socioeconomic factors that contribute to the resident's vulnerability (Cutter 1996;Cutter, Boruff, and Shirley 2003;Cutter et al. 2008). This research expands on existing bypass literature by including the SoVI in order to understand the likelihood of individual bypass behavior according to the social vulnerability of the town.

Community Context
A community of place is not only geographic boundaries, it is "filled up by people, practices, objects, and representations" (Gieryn 2000:465). Other research concurs and conceptualizes community as "a particular way of organizing society in which the interactions essential to daily life remain embedded within primary ties in local solidarities", rather than just a geographic space, (Cope et al. 2016:5). Thus, local amenities such as shopping, restaurants, and healthcare, as well as social ties and connections, are an integral part of the community experience, and essential elements of studies attempting to understand healthcare selection.
Geographic mapping alone cannot capture how residents' choices are shaped by the place they live.
Satisfaction with amenities and strength of community ties impact residents' healthcare selection behaviors-previous healthcare research finds community patterns (Sanders et al. 2015;Sanders et al. 2016). Research found that negative perceptions of local healthcare (Borders et al. 2000: Sanders et al. 2015 and dissatisfaction with community amenities such as shopping and restaurants (Brown 1993;Sanders et al. 2015) can push rural residents to seek healthcare outside their community. Additionally, residents who already travel outside their communities for shopping or who work out of town are more likely to consolidate their travel time by bundling errands or commune with healthcare, saving both time and money by scheduling healthcare appointments during the same out-of-town travel (Brown 1993;Sanders et al. 2015).
Alternatively, other community factors can encourage, or pull residents to stay in town for healthcare. Strong community ties, increased number of friends, and friends who have interests in common decrease the incidence of healthcare bypass (Sanders et al. 2015;Sanders et al. 2017). Thus, healthcare decisions are not exclusively an individual's privilege or constraints, but these decisions are influenced by their community of residence.
This study replicates parts of Sanders et al. by including the same community push and pull factors for healthcare selection (2015), as well as testing additional community pull factors: dissatisfaction with local roads and length of residence. People who are dissatisfied with local roads would be less inclined to drive further distances for their healthcare, and people who have lived in the community longer would be more accustomed to local amenities which would pull them to use local healthcare. Community amenities and ties matter for healthcare. Thus, this study explains how a number of community variables influence residents' healthcare selection while controlling for individual privilege, as well as place-based vulnerability.

Individual Privilege and Ability
Decisions to stay or bypass are made within socioeconomic privilege constraints. After all, regardless of how far away services are, less privileged people may not have the ability to bypass. In fact, the research shows that various socioeconomic factors influence the likelihood of bypassing, including age (Sanders et al. 2016;Sanders et al. 2017), income (Sanders et al. 2017), education (Sanders et al. 2017;Yao and Agadjanian 2018), living in remote locations (Yao and Agadjanian 2018), and being a recent migrant (Sanders et al. 2016). This study considered these various individual demographic factors along with the intersection of community privilege and community vulnerability. By breaking down healthcare selection by not only individual demographics, but also how various privileges and disadvantages are exacerbated by community vulnerability, this study provides valuable findings for people who make macro-level healthcare decisions such as public healthcare workers, county health departments, and healthcare administrators.
This study answers not only who is bypassing, but why they are choosing to bypassinformation useful for clinic administrators, doctors, practitioners, and other rural healthcare workers. While previous research has examined who can bypass, the literature has largely overlooked the reasons for doing so. Because a mixed-method approach is needed to understand why people bypass, this study contributes to the existing literature by providing a qualitative assessment of the people's responses explaining why they bypass. Understanding these reasons can help rural practitioners better meet local healthcare needs.

Place-based Constraints
Previous healthcare research examined how distance from services affects healthcare access (Yao and Agadjanian 2018); however, this study measured the influence of place by incorporating not only a distance measure but including a place-based vulnerability measure that captured the influence of place on individual healthcare selection. This study introduced a wellestablished measure-used in disaster literature for measuring characteristics of place-called the Social Vulnerability index, or SoVI, into healthcare literature The SoVI index is composed of town-level socio-economic variables from census data that contribute to the resident's vulnerability (Cutter 1996;Cutter, Boruff, and Shirley 2003;Cutter et al. 2008).

Social vulnerability is partially the product of social inequalities-those social factors
that influence or shape the susceptibility of various groups to harm and that also govern their ability to respond. However, it also includes place inequalities-those characteristics of communities and the built environment, such as level of urbanization, growth rates, and the economic vitality, that contribute to the social vulnerability of places (Cutter, Boruff, and Shirley 2003:243).
The SoVI has previously been used to understand resilience to natural disasters, migration (Myers, Slack, and Singelmann 2008), and mental health (Cope and Slack 2017). Individual healthcare selection decisions are made within various constraints: individual privilege and ability constraints, as well as community and place constraints. Understanding how place-based vulnerability affects healthcare selection will help public health workers better prioritize community sensitivity.

Summary and Expectations
Bypass decisions are made within privilege and place constraints. Previous research has established that various community push/pull factors and individual abilities affect whether a person was likely to bypass. This study introduced a measure of place-based vulnerability to the healthcare literature to illustrate how place privileges and constraints affect individual healthcare selection decisions.
This study contributes to the existing literature on healthcare selection and bypass behavior by 1) illustrating which privileges and abilities enable rural people to bypass their healthcare and which limitations discourage people to bypass, 2) showing which community ties influence individual healthcare selection, 3) measuring place-based social vulnerability level for each town via the SoVI, and 4) using qualitative data to understand the reasons people give for bypassing healthcare available in their community.

METHODS
This study employed survey data collected from residents of 25 rural towns in Utah during the summer of 2017. All of the towns had populations between 2,500 and 5,000. Sample frames were obtained from postal data and respondents were selected using systematic random sampling. Respondents received three rounds of mail surveys using the Dillman approach (Dillman, Smyth, and Christian 2009)-towns with lower response rates received hand-delivered surveys. The adjusted response rate (which accounts for surveys mailed back undeliverable or marked 'return to sender') was 51.44%, making the total sample size 1,309. After accounting for missing data using listwise deletion, the total sample size was 1,061.
Survey data provided information about the respondent's community sentiment, satisfaction with local amenities, self-reported health, healthcare-seeking behaviors, and demographic information. Additionally, qualitative responses from this survey helped us to understand why respondents made their healthcare selection decisions.
Supplementary data from the American Community Survey and the US Census were used to create a Social Vulnerability Index (SoVI). Variables in the SoVI included, for example, per capita income, percent of renters, mean rent cost, percent spent on food stamps, and percent rural. These variables were drawn for each town. The SoVI was created using factor analysis, and after the index was created, each town was assigned a SoVI score. Towns were then sorted into high, mean, or low vulnerability according to their SoVI score. If towns scored greater than one standard deviation above the mean, they were considered highly socially vulnerable, while those with scores less than one standard deviation below the mean were considered low socially vulnerable. Towns with scores that fell within one standard deviation above and below the mean were considered socially vulnerable. For descriptive statistics about the 25 variables drawn from census data that measure social inequalities, as well as place inequalities, included to create the SoVI, see Appendix 1. For SoVI factor loadings, see Appendix 2.

Measurements
Dependent Variable: Primary Care Provider Bypass: Previous measurements of bypass rely on zip code and distance cut-offs to determine if people sought healthcare outside their community (Roh and Moon 2005). Such studies are limited because a person's concept of community usually extends beyond their zip code. Other studies looked at if people traveled outside their county (Borders et al. 2000) or traveled long distances to hospitals (Radcliff et al. 2003) to determine if respondents bypass care close to their residence; and still others mapped the exact location of residents and their healthcare providers (Sanders et al. 2015). However, in ancillary analysis, phone conversations with the county health departments and google maps revealed the nearest clinic and hospital locations for each towns in the sample. Next, researchers called local healthcare providers and clinics and asked where the nearest places patients in their community could go for specific services. Many rural providers reported that their patients had to go to the nearest metropolitan area, rather than a closer hospital or clinic, showing (with follow up questions) that in many cases, providers were unaware of the nearest healthcare to their communities. Such residents sent to metropolitan areas did not intentionally bypass closer local care because they are unaware of its existence. Thus, a self-reported healthcare selection variable was used in this study to account for the said measurement error.
The primary care provider bypass variable was created using survey questions that first ask residents to self-report if they were seeking primary healthcare within or outside their community. Thus, this measurement took into account the resident's perception of the boundaries of their community rather than a geographic range. If they reported seeking outside care, space was provided for them to explain the reasons for their decision. Responses such as "No healthcare available in my town" were coded as 0 for not bypassing. To make the bypass measure more conservative and precise, vague responses such as "availability" or "that's where the doctor is" were also coded as 0 (for not bypassing) because it was unclear whether the respondent had the option to seek local care. Responses that clearly indicated that the respondent had access to local care but decided to choose other care were coded as 1, which means that the respondent deliberately bypassed local healthcare. See Table 4 for various bypass responses.

Independent Variables: Community Push factors: Community push variables include
dissatisfaction with local shopping and dissatisfaction with local healthcare-residents who were dissatisfied with their community amenities were theoretically less likely to shop locally.
Both of these variables were measured on a 1-7 scale, 1 being very satisfied and 7 being very dissatisfied. Missing data was accounted for by mean substitution.

Community Pull factors: One community pull factor is Dissatisfaction with local roads.
If the roads in and surrounding the community are of poor quality, people would be less likely to leave their communities, especially during inclement weather. Dissatisfaction with local roads was measured using a 1-7 scale, 1 being very satisfied and 7 being very dissatisfied. Percentage of friends in the community was measured ordinally. The categories were 0-25%, 26-50%, 51-75%, and 76-100% and indicated that the more friends people have in their communities, the more connected they felt to the community. Finally, length of residence, which measured the proportion of life respondent resided in the community, was considered a pull factor because those who have lived in the community for some time should feel more satisfied with its services and amenities.
Individual privilege: Because bypass behavior theoretically stems from an ability and privilege to bypass, this study included various demographic characteristics to get a better idea of which kinds of privilege enabled individuals to bypass. Age was coded ordinally as 18-34, 35-49, 50-64, and 65+ as found in other healthcare literature (Sanders et al. 2015). Dummy variables were included for sex (male=1) and marital status (married=1). Number of children is a continuous variable and was included because the more children a person has, the more difficult it would be to bypass because it is time-consuming and resource-draining to travel longer distances. Education is categorical and was coded as no college, some college, and college or more, as found in other healthcare literature (Sanders et al. 2015). Employment status was also included as a dummy variable (working=1). Income was treated as a continuous variable, with responses ranging from "$1-10,000" to "$150,000+" USD. This study also included a self-reported health variable from the RUCS survey by asking, "On a scale of 1 to 7 how would you rate your health?" Answer options were on a scale with 1 being very poor, 7 being excellent.
Missing data was accounted for with mean substitution. Distance to the metro was calculated for each town using Google Maps to calculate the miles between each town and the closest metropolitan area.
See Table 1 for descriptive statistics about resident's community ties and demographics.

Social Vulnerability of Community (SoVI):
To expand on other placed-based healthcare selection literature, this study introduces a measure of the social vulnerability of place (SoVI), which is well established in disaster literature (Cutter 1996;Cutter, Boruff, and Shirley 2003;Cutter et al. 2008). The SoVI was calculated using factor analysis, and 42 variables that measured different socioeconomic characteristics of each town, such as percent of renters in town, per capita income, percent of children under 5 below poverty, etc. (see Appendix 1 for a comprehensive description of all variables used in this SoVI). After calculating the index, each town was assigned a SoVI score. Towns one standard deviation above the mean are considered highly vulnerable, and those one standard deviation below the mean are low vulnerable. All towns selected for this study had small populations, were rural and remote, and were thus socially vulnerable. The SoVI shows, however, that there was variation in the degree of placebased vulnerability, even among relatively homogeneous residents.

Analytic strategy
To address the study's research questions, logistic regression models were generated that predicted the likelihood of bypass. Respondents with missing data for community measures, income, and self-reported health were accounted for using mean substitution. Respondents with missing data from age, sex, marital status, number of children, education, and employment were dropped from these analyses using listwise deletion. Table 2 shows a pooled model of all 25 towns, and Table 3 shows the same model with towns separated according to SoVI.

RESULTS
[ Table 2 about here] Logistic regression estimates predicting odds of bypass are reported in Table 2. Model 1 (Table 2) shows a logistic regression predicting the odds of bypass for various independent variables measuring aspects of community and community sentiment. This study found that those who are dissatisfied with local healthcare are more likely to bypass local providers (OR = 1.38). However, dissatisfaction with shopping, with roads, the percent of friends in the community, and length of residence in the community have an insignificant effect on bypass behavior. Thus, dissatisfaction with local amenities, lack of community attachment, and lack of community ties are not significant push factors for the residents.
Because bypass behavior does not seem to be a community phenomenon, this study also brings in individual demographic characteristics in Model 2 (Table 2). Controlling for age, sex, marriage, children, education, employment, income, self-reported health, and distance to metro, the findings in Model 1 are consistent: dissatisfaction with local healthcare increases the odds of bypassing local healthcare (OR=1.37), but no other community characteristics drive local healthcare-seeking behaviors. Additionally, of all the individual demographic variables, men are less likely to bypass local care than women (OR=0.64), but no other variable significantly explains bypass behavior. Model 3 (Table 2) includes a measure of the social vulnerability of place. Divided by one standard deviation from the mean, this model compares low socially vulnerable towns (low SoVI) and high socially vulnerable towns (high SoVI) to mean socially vulnerable towns (mean SoVI). These results show that, compared to the mean, residents in low SoVI towns are significantly more likely to bypass (OR=1.88).
[ Table 3 about here] To further investigate the relationship between bypass behavior and SoVI, Table 3 shows a comparison of bypass behaviors of residents in low, mean, and highly socially vulnerable towns. The likelihood of bypass can be explained with a combination of the level of town social vulnerability and individual privilege. Model 1 (Table 3) shows the community and individual characteristics that influence bypass behavior in low socially vulnerable rural towns. These residents live in less vulnerable places compared to the rest of the residents in the sample, which means that the characteristics of their community are advantageous. Interestingly, this study showed that those who were dissatisfied with shopping were significantly less likely to bypass (OR=0.68), while those who were dissatisfied with local healthcare were significantly more likely to bypass (OR=1.35) (Model 1). Additionally, married individuals were significantly less likely to bypass than single individuals (OR=0.22), and those with a college degree were 4 times more likely to bypass local healthcare compared to those who did not go to college (OR=4.05).
Thus, people living in the most privileged places, with privileged positions (single, more education) are more likely to bypass local care and opt for better care elsewhere. Model 2 (Table   3) shows that people living in areas of average social vulnerability, dissatisfaction with local healthcare have significantly higher odds of bypass (OR=1.66), and men in these towns are significantly less likely to bypass than women (OR=0.56). Model 3 (Table 3) shows that people living in highly vulnerable areas and those who have lived longer in the town are 3 times more likely to bypass (OR=2.97). Additionally, older residents have significantly lower odds of bypass compared to young residents (age 50-64 OR=0.28; age 65+ OR=0.17), and people who are employed have significantly lower odds of bypass compared to unemployed people (OR=0.47) (it is worth noting, most of the "unemployed" respondents are homemakers and retired people, both of which are privileged positions). However, no other individual characteristic or privileged position significantly predicts the likelihood of bypass for people in highly vulnerable towns.
Even so, unlike low SoVI residents, residents in high SoVI towns are more likely to bypass the further they live from metropolitan areas (OR=1.01).
[ Table 4 about here] This research contributes to the bypass literature by answering the following questions: "Who is more likely to bypass?" and "Why are rural people bypassing?" This research applies a mixed-method approach to understand people's reasons for doing so. Table 4 shows the distribution of qualitative responses, with reasons for healthcare bypass, falling into six main categories-better quality healthcare, greater selection, consistency with provider, lower cost or insurance network, one-stop-shop, and confidentiality. Residents who said they bypass local care for better quality care elsewhere offered comments such as "No physicians here, only PAs," "Because it's scary going to local hospital + physicians. Too many mistakes made + lack of knowledge + common sense. I have to tell them what to do," and "I don't want to die." Perceived low-quality local care, as well as negative experiences with care, pushed people to bypass their care. Rural clinics can focus on increasing the quality of their services to pull people to use local clinics. Respondents who cited a greater selection of providers as their reason for bypass said, "Very limited options for healthcare in my community" and "choice (female GYN) quality & privacy". Increasing the number of providers as well as increasing the types of specialist providers could pull people to use local services rather than bypass. Those who wanted consistency with their primary care provider were willing to bypass local care: "Home town with doctors I know" and "Family doctor for 30 years." Rural providers that can build trusting relationships with new residents might be able to encourage them to use local care rather than continue their care in their previous community. Those who mentioned costs as reasons for bypass commented: "Better Healthcare and much cheaper!" and "Affordability with regards to insurance providers." Healthcare administrators who broaden the number of insurance networks accepted at their clinics could appeal to residents who have obscure insurances. Additionally, any financial assistance or payment plans should be made known to prospective patients.
Interestingly, respondents mentioned concerns about confidentiality with their local care, saying that "People know your health issues (they) come out of the doctor's office" and "I don't want locals knowing my business." Clinics need to ensure that all staff keep patient information confidential. Regular HIPPA training could help remind staff of the importance of confidentiality and posted signs around the clinic could help patents feel more trusting towards providers, as well as remind staff to keep patient details private. Others said that they had already bypassed for other reasons, including "Because other family members have doctors there" and "Better care and visit family." These responses indicate that they bundle personal and family healthcare or other social needs and could be more difficult for existing rural providers to address. New rural clinics, however, should choose their location carefully, selecting real estate close to town centers or other important amenities.

DISCUSSION
The SoVI helps to outline healthcare vulnerabilities-not just disaster vulnerability as in previous research (Cutter 1996;Cutter, Boruff, and Shirley 2003;Cutter et al. 2008). The SoVI index was created from publicly available census data, making the index relatively accessible and straightforward. Thus, healthcare policymakers, county health departments, and other stakeholders can create SoVIs for towns in the region to identify areas that need higher investment in healthcare infrastructure. While supplementary healthcare data is needed to understand local needs better, the SoVI alone can help officials target healthcare resources when policymakers or county health departments lack the resources to gather the data.
People living in low SoVI towns are significantly better off compared to medium and high SoVI places. However, our study found that residents were more likely to bypass compared to the residents living in mean SoVI towns (see Table 2, Model 3). Considering which people in low SoVI towns are more likely to bypass healthcare, our study found that the privileged are most likely to do so (see Table 3, Model 1). Increased bypass amongst privileged people, however, can create healthcare wastelands and negatively impact less privileged people living in low SoVI towns. Thus, policymakers should ensure that healthcare services in low SoVI areas meet the needs of less privileged residents.
For residents in high SoVI towns, those who are long-term residents are 3 times more likely to bypass local care. Long-term residents may, compared to new residents, know the quality of healthcare is poor, or are concerned with confidentiality (see Table 4). Additionally, this study shows that, although no amount of individual privilege helps people to bypass in highly vulnerable areas (see Table 3, Model 3), living further away from metropolitan areas makes those in such areas more likely to bypass. This finding has implications for policymakers-highly socially vulnerable areas, which are also remote, should be targeted to improve healthcare quality, selection, cost-effectiveness, and confidentiality for those living there.
Because this survey was not asking about healthcare exclusively, but also included sections on community, education, employment, etc., there were fewer questions about healthcare compared to datasets used in previous healthcare literature. The RUCS survey lacks information about the respondents' health insurance, frequency of use, and personal illness or injury. Additionally, as this survey was limited to rural Utah, these findings are not generalizable. Other state-wide or regional studies should be conducted to understand rural healthcare selection in those areas. National-level research is needed to understand bypass in the US as a whole.
However, the SoVI does work for researchers looking at healthcare selection processes.
Privileged people-those who live in low vulnerable areas and have individual privileges such as higher education-often seek healthcare elsewhere. This study shows that, for rural healthcare, those who live in low vulnerable places and are better educated are able to bypass-although education does not help those in medium and highly vulnerable places to seek better care. The information provided by the SoVI helps providers and policymakers to use a more targeted approach to public health. Rather than targeting only those with low educational attainment for all rural areas, stakeholders can broaden their reach to highly vulnerable areas while focusing on those with lower educational attainment in low vulnerable areas. Additionally, this study shows that aging populations in highly vulnerable rural areas need additional help to access better healthcare. Healthcare selection is best understood as a multifaceted process that includes not only community ties and individual ability but also the area's social vulnerability.
Qualitative responses help us understand why people bypass. These written responses will help rural providers address some of the concerns that people express with local care (for example, by increasing the quality of their services, expanding their insurance networks, or ensuring confidentiality) fewer residents would feel the need to bypass, and healthcare deserts could be avoided.    From the original 25 variables, an eight-dimensional factor structure emerged that accounts for 82.7% of total variance among these items. The emergent dimensions of social vulnerability include 1) industry and spatial disadvantages, 2) socioeconomic status, 3) housing distribution, 4) wealth and income, 5) family poverty, 6) remoteness and rurality, 7) elderly disadvantages, and 8) elderly poverty. Factors are named based on the characteristics of included variables, specifically a dominant variable. Each of these eight components of social vulnerability is discussed in greater detail below.

Industry and spatial disadvantages
The first factor represents a hybrid of vulnerability arising from industry and spatial disadvantages. While percent of workers in construction and rent cost loads negatively into the factor structure, the following variables load positively: population change (inverse), percent of mobile homes, population density (inverse), and percent of children below poverty. This factor accounts for 16.0% of variance shown among variables in the model.

Socioeconomic status
The second factor that emerged is related to patterns of socioeconomic status. Specifically, these variables loaded positively into the structure: percent without high school degree, percent white (inverse), percent without insurance, percapita income (inverse), and percent of the population in poverty. This factor accounts for 12.4% of variance shown among variables in the model.

Housing distribution
The third factor relates to housing distribution. Three variables load positively on this factor: percent of land area, number of housing per 2 mile, and population density (inverse). This factor accounts for 12.0% of the variance explained.

Wealth and income
The fourth factor shows patterns of wealth and income. Four variables load positively into this factor: percent on social security, household income $75,000 or above (inverse), home value (inverse), and percent unemployed. This factor accounts for 11.8% of the variance explained.

Family poverty
The fifth factor shows family poverty. Three variables loaded in positively: Percent of single parent, percent of children under 5 years old who are very poor, and percent renters. This factor accounts for 9.0% of the variance explained.

Remoteness and rurality
The sixth factor touches on remoteness and rurality. Two variables loaded in positively: number of hospitals (inverse), and percent rural. This factor accounts for 7.4% of the variance explained.

Elderly disadvantages
The seventh factor is less clear, however both variables that load into this factor are disadvantages that would affect the elderly. Two variables loaded in positively to this structure: has nursing home in town (inverse) and percent in mining industry. This factor accounts for 7.1% of the variance explained.
Elderly poverty