dunlap plan b

Evaluation of a Health Indicator Used to Identify Populations Vulnerable to Heat and Air Quality

Emergencies in the Twin Cities Using GIS

Master of Public Health Candidate Sara Dunlap

Submitted to the University of Minnesota School of Public Health

June 11, 2009

Academic Advisor: Ian Greaves, B Med Sci, MB BS, FRACP, FAAAS University of Minnesota Associate Professor

Division of Environm s ental Health Science

Rese PhD Minnesota D Disease and

arch Advisor: Jean Johnson, epartment of Health Chronic Environmental Epidemiology

Sara Dunlap June 11, 2009 MPH Thesis

Abstract

2

Growing research has focused on morbidity and mortality during heat and air quality

emergencies amid concern over climate change. To help plan for public health responses during these

emergencies, researchers in the State Environmental Health Indicators Collaborative (SEHIC) developed

an indicator model to identify both elderly individuals who live alone and individuals who live below the

poverty line according to census tract. This study piloted the SEHIC template using 2000 census data for

a seven county metropolitan area surrounding the Twin Cities, Minnesota. A case study also examined

potential relationships between the vulnerable populations and Chronic Lower Respiratory Disease

(CLRD). Two key research questions were addressed in this paper. First, does the SEHIC indicator

accurately account for demographic and vulnerability inequities? Second, does the addition of CLRD

hospitalization data to the existing SEHIC template provide evidence of errors in the SEHIC calculation

methods?

Results of the piloted SEHIC template indicator showed an overrepresentation of below poverty

line populations among the vulnerable census tracts. The overrepresentation is a serious concern after

literature identified elderly to be more at risk during emergencies and to be outnumbered four times by

low income groups. Four areas of concern were identified and potential recommendations made to

adjust for errors. First, the Minnesota state averages were established as standard vulnerability to

evaluate each census tract against. Second, use of percentages for vulnerability scores showed to be

inaccurate. Recommendations include changing percentages to a standard error model and calculate

vulnerability score by Z‐Score. Additionally, calculating individual vulnerable population z‐scores then

creating a composite score resulted in a more equal representation of both vulnerable populations.

Third, no risk threshold had been established for choropleth maps. Recommendations include using Z‐

score increments in a five color gradient. Fourth, the existing template is not compatible with other

health data. To adjust, percentages were changed to rates of vulnerable individuals per 10,000. Results

of overlaid SEHIC and CLRD data also showed over‐representation in below poverty line populations.

When combined with the recommended SEHIC calculations a more equal distribution of vulnerable

individuals was identified.


3

Table of Contents

Introduction: ..................................................................................................................................................4

Statement of the Problem: ........................................................................................................................4

Study Objectives:........................................................................................................................................5

Study Population Demographics:...............................................................................................................6

Literature Review: ......................................................................................................................................6

Conceptual Framework: Public Health Geography..................................................................................17

Research Questions..................................................................................................................................19

Methodology:...............................................................................................................................................19

Study Design.............................................................................................................................................21

Data Collection:........................................................................................................................................22

Results: .........................................................................................................................................................23

Individual Vulnerable Population Analysis: .............................................................................................32

Recommendations’ to SEHIC: Adjusted SEHIC Indicator:.......................................................................35

Case Study: Chronic Lower Respiratory Disease: ....................................................................................38

Discussion:....................................................................................................................................................44

Study Limitations:.....................................................................................................................................44

Bias: ..........................................................................................................................................................46

Conclusions: .................................................................................................................................................47

References....................................................................................................................................................50


4

Introduction:

Statement of the Problem:

Climatologists and public health researchers agree that climate change causes serious health

effects in the population. Scientists predict a 70% increase in hazardous ozone concentration days and

heat waves extending from 3‐8 days in some cities (Ebi, 2006). Public health officials in urban areas fear

a combination of growing populations and increasing frequency and severity of air quality and heat

emergencies have the potential to be extreme public health hazards. To help plan for public health

responses during these emergencies, researchers have developed indicator models to help identify the

location of vulnerable populations in order to most effectively distribute education resources, staff, and

preventative care.

The Council of State and Territorial Epidemiologists (CSTE) is a collaborative of state and local

public health agency epidemiologists that encourage interdisciplinary research and project

development. As a subgroup of CSTE, the State Environmental Health Indicators Collaborative (SEHIC)

developed an indicator model to identify populations most likely to be effected by air quality and

elevated heat emergencies. The SEHIC indicator model identifies both elderly individuals who live alone

and individuals who live below the poverty line. These groups are at risk because of their reduced

abilities to alter behaviors to adapt to rapidly changing environments (Ebi, 2006), (Luber, 2006). The

SEHIC indicator model identifies at risk geographic areas by mapping vulnerable populations using

Geographic Information Technology (GIS). The SEHIC indicator template is currently in pilot stage. If the

indicator can be established as an accurate method, it could become a standard for inter‐state air

quality and heat emergency preparedness.


5

Study Objectives:

The purpose of this study is to validate the SEHIC indicator model’s ability to identify vulnerable

populations. The study method included three major components. First, to identify the vulnerable

populations in the Twin Cities according to the SEHIC template. The study will apply 2000 census data to

seven counties surrounding Minnesota’s Twin Cities including Anoka, Carver, Dakota, Hennepin, Ramsey,

Scott, and Washington Counties (Figure 1). The second component is to map the results of the

vulnerability scores and to make recommendations based on the accuracy of the SEHIC template. The

third component is to add health data to the existing SEHIC template. The addition of Chronic Lower

Respiratory data to create a new ‘Disease Indicator,” provides the opportunity to identify overlapping

geographic areas of disease and at risk populations.

Figure 1: Twin Cities Metropolitan Area


6

Study Population Demographics:

Demographic expert’s project that the entire U.S. population will increase by 130% from 2005‐

2035; individually, Minnesota’s state’s population is expected to increase by 122% from 2005 to 2035

(U.S. Census Bureau, 2008). Of the targeted vulnerable populations, the 2000 census found 12.37% of

the U.S. population lived below poverty line (BPL). In Minnesota the 2000 census confirmed that 7.94%

of the population reported to be surviving below the poverty line, with 47% of all below poverty line

individuals in Minnesota living in the Twin Cities. By 2007 the number had expanded by 136% to nearly

244,000 individuals (Bureau, 2007).

In the U.S., elderly individuals living alone (ELA) accounted for 6.15% while Minnesota’s number

was slightly lower at 5.74% in 2000. The elderly population is projected to increase by 213% during

2005‐2035; 233% in the seven county Twin Cities study area alone. Combined, over 26,300 vulnerable

individuals were identified in 2000. This is expected to rise to nearly 350,000 by 2035.

Literature Review:

Review of the literature justifies SEHIC’s development of the indicator model to identify

vulnerable populations at risk during heat and air quality emergencies. Information gathered includes

the health outcomes related to climate change, air quality, heat waves, urban areas, and vulnerable

populations. An extensive review of current and relative literature defines study methods, research

gaps, statistically significant studies, and conceptual developments.

Included literature was identified using the University of Minnesota and Minnesota Department

of Health’s library systems, in addition to Pubmed, and Google Scholar internet search engines.

Database searches within the University of Minnesota’s Biomedical Library returned results of literature

based on set criteria. Limitations of data included those articles in English, published after 2000, and


7

selected by keyword entries. Keyword search examples included “urban health,” “vulnerable elderly,”

“health indicator models,” “air quality and health outcomes,” and “heat related morbidity.” Literature

was excluded if the content was outside the scope of the project or did not have significant statistical

relevance to the topic.

Bias is acknowledged based on electronic availability of resources, journals the University of

Minnesota subscribes to, and those fitting within the set criteria. However, literature included covers a

wide breath of authors and viewpoints and should be considered representative of the extensive

literature available on the topics.

Climate Change and the Health of Vulnerable Populations:

Scientists show that climate change is being accelerated by decisions in the built environment. Research

shows an association between uncapped industrial productions, agricultural growth, unlimited urban

development, and resulting climate change symptoms (Control, Climate Change and Public Health, 2007).

Climate change symptoms include include coral reef beds bleaching, wildlife migration changes, rising sea

levels, changes in precipitation, frequency of severe weather, and reduced crop output (Administration

N. A.). Air quality and elevated heat both have drastic health outcomes with broad supporting

literature.

Air Quality and Health Outcomes

Reduced air quality and increasing temperatures are intimately related to Climate Change.

Greenhouse gasses (carbon dioxide, methane, nitrous oxide, and chlorofluorocarbons) are a factor in

climate change. Scientists believe a combination of greenhouse gases are trapping the earth’s rays near

the surface, preventing them from disseminating back out into the atmosphere (Administration, 2008).

Although the largest hole in the ozone layer is located above Antarctic stratosphere, there is a rapid


8

depletion of ozone in a band stretching across the United States. Scientist point to the U.S. extravagant

traffic use, industrial production, and lenient restrictions (Agency, 2009). Rising temperatures have

serious public health consequences. Extreme heat waves cause large populations to become ill straining

health care facilities.

Ozone is associated with the greenhouse effect and is a public health concern because it

increases the intensity of the sun and temperatures in urban areas. Multiple studies have compared

ozone levels and mortality rates using cities with stratified data sets. The stratified data defines

historical community mortality rates to compare against deaths during elevated ozone levels. Results

indicate cities with generally low ozone levels saw a greater spike in mortality during high ozone days

suggesting the greater variance in ozone, the greater probability for deaths in the vulnerable

populations (Ren, 2008), (Abelsohn, 2002).

Particulate Matter is a concern to public health officials because it aggravates existing respiratory

diseases and leads to greater health care needs. Air quality data is a surveillance tool used to link health

outcomes with environmental factors. The Environmental Protection Agency’s (EPA) National Ambient

Air Quality Standards (NAAQS) regulate air quality for six pollutants including Ozone (O3), Particulate

Matter (P.M. 10) and (P.M. 2.5), Carbon Monoxide (CO), Sulfur Dioxide (SO2), Nitrogen Dioxide (NOx), and

Lead (Pb). Standards are based on adverse health effects in vulnerable populations including asthmatics,

children, and the elderly (Agency M. P., 2003). According to the EPA, P.M.2.5 and ozone are the highest

concern in Minnesota (Agency M. P., 2003).

Heat and Health Outcomes

According to the National Oceanic and Atmospheric Administration, “heat is the number one

weather related killer” claiming over 1,500 lives each year and climate change experts predict death


9

rates to rise with increases in temperature (Administration, 2008). A report by the EPA listed

Minneapolis as the 10th highest city with heat related mortality. Researchers estimated a mortality rate

of 2.32 per 100,000 according to 1990 populations records (Agency, 2006). Models from a prospective

study of U.S. northeastern cities estimate summer temperatures to rise by 3 to 80F degrees by 2050,

increasing the potential for deadly heat waves (O'Neill, 2009). Among them, Minneapolis is expected to

increase from 8 to 16 “Mortality Days, ” defined as days with a temperature over 1040F and resulting in

an addition 35 heat related deaths in Minneapolis each summer (Agency, 2006). In the United Kingdom,

scientists project a 250% increase in heat related morbidity and mortalities by 2050 from increasing

average summer temperatures. The potential for more frequent and severe heat waves is a major

concern for urban health officials who need to coordinate education and public health response efforts

(Knowlton K. L., 2007).

Elevated heat emergencies or heat advisories are issued by the national weather service within

12 hours of the incoming weather conditions. Criteria for health alerts include a “heat index of at least

105°F but less than 115°F for less than three hours per day, or night time lows above 80° for two

consecutive days” and are broadcast by television and radio networks (Administration, 2008).

Epidemiological studies use data from past heat emergencies to project health effects for future

elevated temperature events. A time‐series analysis measures health outcomes before, during, and

after the heat event. In a specific study, Kovats reviewed time‐series data from multiple heat waves

concluding that temperature‐death relationships are usually U or V shaped. Kovats reported the highest

death rates occur during very low and very high temperatures during the year (Kovats, 2008). Given

Minnesota’s variety of temperatures, researchers may find an example of Kovat’s U shaped mortalities.

Furthermore, studies examined 11 eastern United States cities found that cities in generally


10

cooler climates had higher mortalities during high heat events hypothesizing that individuals were not

accustomed to such drastic changes in weather (Curriero, 2000), (O'Neill, 2009), (Figure 2), (Agency

2006). In another time‐series study, Braga measured weather extremes in 12 U.S. cities using a 3‐week

‘lag time’ scale to examine respiratory conditions aggravated by heat. Braga hypothesized that death

rates would climb above the normal rate during the elevated heat days and drop to the normal death

rate afterwards. Braga’s study showed more deaths occurred than were expected to according to

baseline death rates. The study explained if those who had died during the high heat days were those

that were going to die within the next few days, there would have been a below average response in the

death rates following the heat event. However, because no dip below the average occurred researchers

concluded that excess in deaths occurred during the heat event (Braga, 2002).

Figure 2: “Estimated Excessive Heat Related Events‐ Attributed Mortality Rates” (Agency, 2006)


11

Climate Change and Urban Areas

Degrading urban health is concerning on multiple levels, but population explostions in urban

areas pose a serious public health risk. In 2007, U.S. census identified nearly 30% of the U.S. population

residing in urban areas (Bureau, 2007). When 30% of the total population is subject to chronic

environmental health based on their urban living status, it is a major health problem.

Urban areas are the major site of industry and development and are growing in both population

and size to accommodate for population needs. Growing urban populations, unbridled industrial

production, traffic construction, and power generating facilities have spewed gases into the atmosphere

that are having a direct effect on the climate (Bellia, 2007), (Amann, 2006). Research has shown that

urban areas are both expelling the most pollution but also suffering the consequences with increasing

numbers of reduced air quality days and more frequent heat waves. In Minnesota, the Pollution Control

Agency’s (MPCA) Air Quality Index monitors identified 27 reduced air quality days across the state in

2003 and 35 alert days in 2005. 17 of the 35 low air quality days were located in the Twin Cities (Agency,

2005).

Urban areas are at particular risk from rising temperatures. The elements of the built

environment including parking lots, roofs, and reflective buildings that trap hot air and prevent heat

dissemination. Known as the “urban heat effect,” the phenomenon has been linked to poor health

outcomes (Shmaefsky, 2006), (Rosenzweig, 2006), (Figure 3), (Agency, 2006). McGeehin found

convincing data linking morbidity and mortalities to urban heat retention. The study specifically

recognized non‐ air‐conditioned buildings and low socioeconomic populations as having a great risk

during heat events (McGeehin, 2001), (Kovats, 2008).


12

Figure 3: Impact of the urban heat island on ambient temperatures (Agency, 2006)

Vulnerable Populations: Elderly and Below Poverty Line Individuals

The increasing number of vulnerable individuals in urban areas poses a public health challenge

during weather emergencies. Vulnerable populations during heat and air quality events include those

with pre‐existing health conditions, the elderly, children, and individuals living below the poverty line

(Ebersol, 2005), (Braga, 2002).

Vulnerable Elderly Populations:

Extensive evidence identifies the elderly populations as heavily at risk for morbidity and mortality

during weather emergencies. Science has shown that social and clinical vulnerabilities make elderly the

most concerning population during heat and air quality events.

Researcher Thomas used the “The Vulnerability Perspective” framework to identify social causes

of elderly vulnerabilities. He concluded that limited access to health care, reduced social networks,

reduced economic, political, social, and education resources, dangerous living locations, and lack of

disaster preparedness are the basis for elderly vulnerabilities during emergencies (Thomas, 2002).

These results agree with findings from a heat wave analysis in Lyon, France where factors of being


13

confined to bed, not leaving the home daily, medical co‐morbidities, and having no access to cooling

centers were associated with increased deaths from heat stroke (Argaud, 2007). Multiple studies also

remarked on various risk factors including: being male, single marital status, living alone, lack of

transportation, smoking, being overweight, and having depressed moods (Vandentorren, 2006),

(O'Malley, 2007), (Perkins, 2004 ), (Ostro, 2006), (Halonen, 2008), (Youger, 2008), (Gilmour, 2006),

(Bellia, 2007).

The elderly population is also clinically more at risk during emergencies because of pre‐existing

health concerns. Studies identify pre‐existing medical conditions including Cardio Vascular Diseases

(CVD’s), chronic respiratory diseases and psychiatric diseases to be associated with more heat related

deaths (Luber, 2006). A study following a 2006 heat wave in California examined morbidity and

mortality in the elderly by geographic areas. The study correlated deaths with the heat index and

identified an excess of over 16,000 Emergency Room visits and over 1,000 excess hospitalizations

statewide in the elderly population (Knowlton, 2009). A study by O’Neill identified Hyperthermia as the

underlying cause of death in over 54% of deaths during 1999‐2003 during elevated heat events (O'Neill,

2009).

It may be difficult for individuals with chronic disease to distinguish between symptoms of their

chronic disease, side effects of their medications, or symptoms of heat or air quality distress.

Researchers also hypothesize that elderly individuals who experience broad‐spectrum somatic

symptoms of heat exhaustion and heat stroke may not recognize the severity of their health needs.

Some researchers also attribute deaths to unrecognized homeostatic changes caused by anti‐epileptic,

beta‐blockers, diuretics, and anti‐ cholinergic pharmaceutical drugs. These drug classes often have heat

dispersion side effects that could be intensified during a heat event (Conti, 2007), (Naughton, 2002).


14

Additional factors may also discourage elderly from seeking help during emergencies. Elderly

individuals may resist seeking help during heat waves justifying that heat always comes in the summer

and there should be no reason to be concerned over a particularly hot day (O'Malley, 2007), (Conti,

2007), (Jelicic, 1997). As proof, the U.S. National Assessment on Climate Change indicated that only 46%

of elderly individuals modified their behaviors to adapt to changing temperatures after hearing a heat

warning since most individuals did not consider themselves to be ‘at risk’ and therefore not necessary to

take the necessary precautions (O'Neill, 2009). Changing behaviors for elderly individuals may be very

difficult, particularly for those living alone. Making decisions to go to a cooling shelter include finding

direction, navigating traffic, parking, and other major arrangements. Often staying inside may seem like

the best option. Unfortunately, elderly individuals may also resist using the air conditioner to save

money and may lack the strength to open windows and doors for ventilation. Considering these factors,

it is apparent that identifying pockets of elderly who are living alone before emergencies happen is

extremely important.

Low Social Economic Status

Social economic status is a solid predictor of health outcomes during emergencies. The U.S.

census identifies below poverty line (BPL) as those individuals with a yearly income under $8,959

(Bureau, 2008). During emergencies, individuals must react with the resources at hand and individuals

with limited funds often lack the avenues to change their behaviors for a limited time, especially if the

event is not considered an extreme danger.

Health and social economic status (SES) are highly related to the built environment. For instance,

an individual with limited funds will most likely live in urban areas to use public transit, access to

services, and low‐cost housing. Analysis in urban areas show low socioeconomic areas are generally


15

close to high traffic areas resulting in high particuate matter exposure and causing greater health

outcomes (Gold, 2005). Low SES individuals may also not participate in preventative health care and

may be adversely affected by heat and air quality events because they are unaware of an underlying

respiratory condition or lack the medication to treat it.

Health risks are also associated with housing construction year and placement. Dwellings built

before 1975 have shown to be more susceptible to ventilation problems. Advances in airtight homes

have made windows and doors heavier and more difficult to open in emergencies (Thomas, 2002),

(Selgrade, 2006). Homes in the Midwest and northeast of the U.S. are built to retain heat during colder

months but are not adequately designed to disperse heat during summer (Miller, 2007). As a result,

occupants who lived on the top floor or had a bedroom directly under the roof are shown to have little

airflow and a higher risk of heat‐related death (Vandentorren, 2006).

Clinically, BPL and minority racial groups are overrepresented in morbidity and mortality rates

during past heat and air quality emergencies. Thomas examined mortalities in Chicago between 1990

and 1997 and found high correlations between socioeconomic status, racial group, and asthma

hospitalizations. The study found that the urban rate of asthma hospitalizations was more than twice

the national average, and low socio‐economic status individuals were overly represented compared to

the whole population. Thomas hypothesized that factors of living in urban environments, substance

abuse, and low use of preventative care was associated with elevated asthma rates (Thomas S. W.,

1999). Another cross‐sectional study of black and caucasian individuals indicate a lifetime prevalence of

asthma two times as high in black individuals compared to caucasian groups. Researcher Browning

point to socioeconomic factors including barriers to health care, racial discrimination, and differential

housing treatment (Browning, 2006).


16

Case Study: Chronic Lower Respiratory Diseases (CLRD):

Chronic Lower Respiratory Diseases are associated with both vulnerable populations and urban

areas. Urban areas have a higher concentration of particulate matter and ozone that exacerbate

symptoms. Both elderly living alone and individuals below poverty line have shown to have a higher

morbidity and mortality during weather events because of limited resources and pre‐existing conditions.

Thus, CLRD data is used as a case study to examine any spatial overlap of the vulnerable populations

identified by the SEHIC indicator and CLRD data.

Chronic diseases account for 70% of all deaths in the U.S. and cause a tremendous financial and

workforce stress on the health care system. Nearly half of Americans were diagnosed with at least one

chronic condition in 2005. America’s chronic disease pushed medical expenditures to $2 trillion a year,

equivalent to over 75% of total U.S. health care costs. It is estimated that Chronic Obstructive

Pulmonary Disease (COPD) alone cost $30 billion in 2000, over $14 billion from direct care costs

(Control, Chronic Disease Overview, 2008). 2006 data from the Center for Disease Control (CDC) state

over 9.5 million adults were diagnosed with chronic bronchitis within the last year, 4.1 million adults

were diagnosed with emphysema, and 16.1 million adults had been reported to have asthma.

Combined, the CLRD’s were responsible for 130,933 deaths in 2006 (Control, 2008). Noting the

enormous number of people diagnosed with a chronic respiratory disease and the financial cost, treating

individuals with respiratory distress during emergencies alone would require a great deal of medical

resources and staff. The financial cost of treatment is unpredictable.

Chronic Lower Respiratory Diseases (CLRD) are a group of diseases that include asthma,

bronchitis, COPD, and emphysema. The CLRD’s are conditions that restrict or limit the exchange of air

into the lungs and cause shortness of breath, reduced lower limb function, skeletal muscle strength,


17

balance, and basic physical actions (Eisner, 2008). Medication and respiratory interventions are

available for treatment of asthma, while chronic bronchitis, COPD, and emphysema are non‐repairable

conditions (Health, 2006).

Chronic Lower Respiratory Disease morbidities and mortalities are associated with high

temperatures and air quality events. CLRD’s are a concern in urban areas because symptoms are

exacerbated by particulate matter and high heat and humidity. Considering these factors, researchers

are watching CLRD’s as an indicator for climate change health outcomes. The respiratory system is

responsible for oxygen and carbon dioxide transfer through the body and is divided between the upper

and lower respiratory tracts. The upper respiratory tract uses the nose, tonsils, adenoids, and trachea to

eliminate and reduce the particulate matter and allergens pushed into the lungs during respiration. The

lower respiratory function is most commonly associated with symptoms of CLRD’s and involves the

bronchi, bronchioles, alveolar duct, and alveoli (Lewis, 2004). During air quality emergencies, heat

events, stressful situations, or acute illness, the respiratory tract becomes vital to maintain oxygen flow

to the body. A lowered oxygen exchange leads to shortness of breath, elevated heart rates, and

generalized weakness can instill anxiety or panic, and in severe cases impair cognitive function (Lewis,

2004). Additionally, public health scientists are concerned that climate change is predicted to intensify

the pollination cycles in plants and molds that trigger CLRD symptoms (Gilmour, 2006), (Peden, 2002),

(Perkins, 2004).

Conceptual Framework: Public Health Geography

Geographic Information Systems (GIS) tools use spatial data of geographic areas to form maps

representing a variety of topics. Maps can portray physical topographic characteristics, population

demographics, health information, or economic development. In public health, GIS is used to identify


18

risks, assess hazards, and develop health responses. These actions are loosely grouped under the term

‘health geography.’ Health geography is a combination of social epidemiology, geography, and

cartography, and used to develop a more cohesive picture of health outcomes based on a set place and

population (Cutchin, 2007), (Roberts, 2008). A specific aspect of public health geography are health

indicators. Indicator models are a risk evaluation method used to calculate multiple aspects of health,

the social environment, and the physical environment. Indicator often combine many risk factors of the

target populations into one value that represents the overall health of the target population or area.

Multiple studies combine public health indicator techniques with GIS technology. For example,

the CDC’s Geospatial Research, Analysis, and Service Program (GRASP) used a composite indicator based

on four risk categories to calculate a ‘Human Vulnerability Assessment Index (HVA).’ The four categories

include a variety of 15 risk factors, loosely grouped by economic measurements, personal and household

assessments, housing and transit evaluations, and race and ethnicity factors. Mapping techniques used

four separate maps to represent the four major indicator groups for each study area (Keim, 2007). A

study by Knowlton used an indicator model to project hospital admissions associated with increasing

temperature (Knowlton K. L., 2007). The indicator model is significant to planners who would benefit

from knowing where hospital admissions raise the most. Although dated, a study by Dever in 1988 is an

excellent example of how a variety of social factors can be identified, classified, and measured. Dever

developed a social vulnerability index to rank health justice in Alabama using 13 variables. The variables

were converted to five distinct categories and using GIS technology mapped the five indexes including

social pathology, economic resources, education level, access to health care, and health status. The

total indexes were combined and mapped in a scale of ‘poor, fair, good, and excellent.’ Tangible results

from this study helped the Alabama Department of Health identify neighborhoods not reached by health

services (Dever, 1988).


19

Research Questions

There are two key research questions addressed in this paper. First, does the SEHIC indicator

accurately account for demographic and vulnerability inequities?

Rational: First, the SEHIC indicator template calculations consider the vulnerability of both

populations to be equal. However, existing literature suggests that elderly living alone are more at risk

than below poverty line populations during weather events. Second, below poverty line groups far

outnumber the vulnerable elderly in the Twin Cities.

Second, does the addition of CLRD hospitalization data to the existing SEHIC template provide

evidence of errors in the SEHIC calculation methods?

Rational: The SEHIC template is not designed to be compatible with health data and therefore

difficult to relate to in Public Health significance. However, CLRD is related to both elderly living alone

and below poverty line populations. Using the 2000 CLRD hospitalization records as a case study, the

existing SEHIC template can be analyzed by overlapping the CLRD data on the indicator template.

Methodology:

As a health indicator, the SEHIC template is considered an observational epidemiological study.

In theory, an epidemiological study maybe designed to formulate hypotheses about relationships

between health outcomes and environmental exposures. Observational studies generally use

population data sets and can be used for hypothesis generating, testing, and tabulating data.

Specifically, the SEHIC model is an Ecologic study design. Ecological studies are used to identify

exposures in populations but data results are not transferable to individuals.


20

Epidemiological studies use five measurements to verify the content of the study including

consistency, biological plausibility, gradient, temporality, and strength of association (Aschengrau,

2003). First, the consistency of the study is determined by the study’s stability during retesting. This

paper is primarily a challenge to the calculation methods used by the SEHIC indicator. Results of this

paper show inconsistencies and flaws in the indicator that need to be addressed before it be used on a

larger scale.

Second, biologic plausibility refers to how well the study adheres to biological agents and

resulting health outcomes. The literature base shows a positive correlation between temperature and

air quality emergencies resulting in higher morbidity and mortality rates among vulnerable populations.

Biologically, this can be explained by acknowledging that elderly living alone and populations below the

poverty line are at increased risk given their limited resources, reduced access to healthcare,

substandard housing, and depleted social support networks.

Third, the gradient response is a major theme behind the development of vulnerability indexing.

In this study the gradient response examined if higher temperatures or lower air quality result in higher

morbidity and mortalities. Although measuring the extent of morbidity and mortalities in the population

was outside the scope of this project, the literature has shown a significant correlation between weather

and health.

Fourth, temporality is used to confirm that the exposure precedes the health outcome. Specific

to this study, temporality first identifies the environmental exposure as heat and air events, and then

identifies the vulnerable population with increased morbidity or mortality as the result. The

environmental exposures examined in the study have previously been identified by the SEHIC group and

the exposures’ legitimacy confirmed with literature reviews.


21

Fifth, the most statistically significant measure of causality in epidemiological studies is the

strength of the association. These measures calculate the association between exposures and health

outcomes. Although this study is an epidemiological study, it does not have individual health

information, thus the study cannot use strength of association measurements. Therefore the strength

of the study is based on the study design, results, and conceptual recommendations of the SEHIC

indicator.

Study Design

The SEHIC indicator analysis uses biostatistical calculations to examine possible bias in the

calculation methodology. The study involves a three‐step process, plus a case study (Figure 3). First, the

SEHIC vulnerability scores of the Twin Cities areas were calculated according to the published template

(Appendix 1). Second, to test the hypothesis of unequal representation, elderly living alone and below

poverty line groups are calculated and analyzed separately. Third, after identifying the vulnerable

populations separately, recommendations are made to adjust for demographic and vulnerability

inequalities. Fourth, after identifying each population separately and making recommendations to

SEHIC, information from Chronic Lower Respiratory Diseases is used as a case study. The purpose of the

case study was to examine the accuracy of recommendations to SEHIC.


22

Figure 4: Information Collection Procedure

Data Collection:

Information for the SEHIC analysis was gained from the 2000 U.S. Census Bureau. The Census

Summary Files are sample of the population, roughly 1 in 6 complete the survey. Information in the

summary files includes detailed population data about the place of birth, employment, housing

information, education and ethnicities (Bureau, 2008). Based on the template, census date for all seven

counties was downloaded and interpreted using Microsoft Excel software.

CLRD data was accumulated through the Minnesota Hospital Discharge Database, purchased

yearly from the Minnesota Hospital Association. Cases of CLRD include Minnesotans aged 65 and older

in 2000 and hospitalized within the Twin Cities zip codes. Included cases where identified from billing

information from the physician diagnostic codes and records. Tabulated cases were identified by the

International Classification of Diseases‐9 codes of the four disease including 490 (COPD), 491.2

(bronchitis), 492 (emphysema), 493 (asthma), and 496 (chronic airway obstruction) (Organization,

Diseases of the Respiratory System, 2008).


23

Results:

Description of the Original SEHIC Template Calculations:

SEHIC is a health indicator that calculates the vulnerability of two populations, elderly living alone

(ELA) and below poverty line population (BPL) in one measurement. SEHIC’s indicator template analysis

requires two steps. First, establishing a standard to measure each tract against. Second, calculating

each tract’s vulnerability against the set standard. The basic calculations includes:

SEHIC Total vulnerability score % = (MN BPL % ‐ Tract BPL %) + (MN ELA % ‐ Tract ELA %)

Issues of Concern 1: No Set Standard for Measuring Census Tract Vulnerability:

Although the template directs the users to measure the risk of each tract BPL and ELA the

average, the template does not identify which average to use. The significance of choosing an

appropriate average is to make the results as realistic as possible to aid public health planners in

preparing for heat and air quality emergencies. The user could complete vulnerability calculations using

averages from a single county, the seven county metropolitan areas, the state of Minnesota, the

Midwest region of the U.S., or the U.S. (Table 1). Changing the standard vulnerability changes the

amount of vulnerable areas seen in the study area. Differences could be skewed to portray either a

larger or less significant problem depending on the need. Additionally, identifying the level of

vulnerability in urban areas may have financial implications from grants and city allocations to combat

problems.


24

Table 1: ELA and BPL Averages (Census, 2008)

Averages ELA % BPL %

U.S. 6.15% 12.37%Minn. 5.75% 7.94%7 County 4.35% 6.91%

Anoka Co. 3.04% 4.19%

Hennepin Co.

4.93% 8.27

Solution: Minnesota Averages Set as Standard for Establishing Census Tract Vulnerability

Therefore, this study uses averages of below poverty line and elderly living alone individuals

based on the 2000 Minnesota averages. In 2000, the average number of below poverty line individuals

was 7.94% (Government, 2009). Elderly individuals living alone accounted for 5.75% of the population

(Government, 2009). Using the Minnesota averages, the extracted SEHIC equation is:

SEHIC Total vulnerability score %= ∑ 7.94 ‐ [(tract BPL population / total tract population) *100] + 5.75 ‐

[(tract ELA population / occupied housing units) *100]

Calculating Census Tract Vulnerability Scores Using the SEHIC Indicator:

To effectively demonstrate the calculations, each group is calculated and explained separately.

Below Poverty Line Calculations:

Census Tract BPL % = 7.94‐ [(tract BPL population / total tract population) x 100]

The measured below poverty line risk in each census tract equals the tract 1999 BPL population

per tract divided by the total tract population in 1999, multiplied by 100. The result is the census tract’s

percent of below poverty line individuals. The calculation numerator equals the number of Minnesotans

per census tract identified by the 2000 U.S. Census Bureau Summary File 3 data series P087002 as


25

having an income below the poverty line in 1999. The denominator source is the total tract population

accounted for during the 2000 U.S. Census under data series number P087001 in Summary File 3.

Determining the specific vulnerability of each tract is determined by subtracting the census

tract’s BPL percentage from Minnesota’s BPL, 7.94 % (Government, 2009). Negative scores for the tract

equate to greater risk; a positive result equals a less vulnerable census tract during heat or air quality

emergencies.

Elderly Living Alone Calculations:

Census Tract ELA % = 5.75‐ [(tract ELA population / total occupied housing units) x 100]

Calculations to identify the percent of elderly living alone in each tract require the same two step

process as determining the risk associated with BPL in census tracts. First, the calculation identifies the

percentage of elderly living alone in every tract. Second, the tract percentage is subtracted from the

Minnesota average ELA percentage to determine if the tract is above or below the standard level of risk.

A negative calculated value indicates that the tract has more elderly living alone than the state average

equaling a greater risk area, a postive number indicates fewer numbers of elderly living alone and a

lower risk during weather emergencies.

The calculation specifically includes the number of elderly living alone in each census track

divided by the total number of occupied housing units in each census tract multiplied by 100. Elderly

living alone are defined as the number of Minnesotans identified in the 2000 U.S. Census Bureau

Summary File 3 series as male or female householders, who live alone, and are aged 65 or over. The

denominator source is the total occupied housing units during the 2000 U.S. Census, in Summary File 3,

data set H019001. Each tract percentage is subtracted from Minnesota’s percentage of elderly living

alone, roughly 5.75% (Government, 2009) to establish the risk in each tract.


26

SEHIC Issue of Concern 2: Vulnerability Scores Calculated in Percentages

A composite score using percentages is an issue of concern because it does not account for

either demographic or vulnerability differences between the populations. An example from the

database analyzed according to the SEHIC calculations shows inequalities (Table 2). First, the 2000

census showed BPL individuals outnumbering ELA 4:1 (Bureau, 2008). The demographic inequalities

cannot be accounted for by using percentages, and shown in tract 1 (Table 2) where BPL out numbers

ELA by 700 individuals. Since the tract has more BPL that ELA, the composite score lists the census tract

as very high risk though ELA are considered at higher risk during emergencies.

Second, the percentage calculations do not adapt of the vulnerability differences in the

population. Examined literature shows ELA are at higher risk during heat and air quality emergencies

(Argaud, 2007) (Thomas, 2002) (Conti, 2007), (Naughton, 2002). In census tract 2 (Table 2), there are

many elderly living alone but the few number of BPL individuals gives the tract a low risk composite

score.

Table 2: Original SEHIC Calculations by Percentage

Census Tract

Total Pop

Total BPL Pop BPL %

Risk Factor 1

Total ELA

Total Housing Units ELA %

Risk Factor 2: Total Vuln. Score

Total Vuln Score %

BPL pop/total pop

7.94-BPL %

ELA/House Units 5.75-% ELA

R.F 1 + R.F 2

1 1721 703 40.85 -32.91 0 464 0 -5.75 -38.662 2008 109 5.43 2.51 95 1182 8.04 -2.29 0.22


27

Solution: Change Vulnerability Scores to a Standard Deviation Model:

A Standard Deviation model is used to normalize a data set with a wide range of data points

(Figure 5). Standardized models with population data are called Standard Error analysis and a Z‐score

and refers to the position of each piece of data relative to the whole distribution (Gravetter, 2000). The

benefit of using z‐score is that it does not change the vulnerability of the census tracts measured by the

indicator calculations. Instead, it normalizes the data into a standardized distribution so the user can

easily decipher where the data lie in the group (Arjomand, 1999).

Figure 5: Standard Deviation Diagram

(Arjomand, 1999)

Changing the SEHIC vulnerability scores from percents to Z‐scores makes the vulnerability of each

tract easier to understand because they are in a normalized distribution and relative to the standard

model. Calculating the z‐score for each census tract includes four major steps. First, establish the

distance or deviation of each census tract ( ) from the mean (µ). This step is already incorporated into


28

the SEHIC template by subtracting the tract vulnerability percentages for both groups from their

associated standards.

Second, to identify the mean of the deviated scores, sum the squared deviated scores to equal

the sum of variances:

Third, the standard error ( of the seven county population data is calculated by taking the

square root of the sum of variances: The standard error is the same for each census

tract since it a calculation involving all census tracts.

Fourth, to find the corresponding z‐score for each census tract, the equation equals the original

SEHIC score ( minus the mean (µ), divided by the standard error ( :

(Gravetter, 2000). An example of using the Z‐score calculation to determine the vulnerability level of

ELA in a tract includes: Z‐score = ( 5.75 ‐ tract ELA %) / (

The significance of the standard error models is that it relates the vulnerability of each tract

according to the SEHIC template.

Table 3: Original SEHIC Scores Standardized by Z‐Score

Census Tract

Total Pop

Total BPL Pop BPL %

Risk Factor 1

Total ELA

Total Housing Units ELA %

Risk Factor 2: Total Vuln. Score

Total Vuln Score

Vul. Z-Score

BPL pop/total pop

7.94-BPL %

ELA/House Units

5.75-% ELA

R.F 1 + R.F 2

SEHIC σ = 9.52

x µ x µ (µ-x)/σ 1 1721 703 40.85 -32.91 0 464 0 -5.75 -38.66 -4.060922 2008 109 5.43 2.51 95 1182 8.04 -2.29 0.22 0.023109


29

From the original SEHIC database the difference between the percentage vulnerability scores and

vulnerability by Z‐score are easily seen (Table 2). The tracts with the lowest vulnerability scores are

located below the average in the entire data distribution and are given negative numbers indicating they

are more at risk. Although the negative number is counter‐intuitive, because the census tract is

subtracted from the mean, negative numbers represent more risk.

SEHIC Issue of Concern 3: No Established Risk Threshold for Choropleth Maps

The SEHIC indicator template directs the user to create a five‐color gradient choropleth map to

represent tract vulnerability scores. A choropleth map enables each geographic area to be represented

by color according to its risk value. Darkest colors indicate the highest risk areas and lighter colors have

less risk than the Minnesota average. However, the template does not establish the cut‐off points that

distinguish at risk tracts from other. Specifically, the template lacks set cut‐off points for determining

which numbers equal higher risk groups. For example, the outcome of the GIS map would be changed a

great deal by determining that only census tracts with 90% at risk verses tracts with 70% at risk.

Solution: Identify ZScore to Set CutOff Points

The solution to the problem is to use the Z‐score to identify appropriate cut‐off points in the

choropleth maps. Z‐scores below zero are considered to have no elevated risk during emergencies.

Scores above zero are mapped in one z‐score increments with increasingly dark colors to represent

higher vulnerability areas (Figure 5).


30

Figure 5: Original SEHIC Template Results

The analyses of the spatial image of the SEHIC vulnerable scores provide a wealth of information.

According to the census, roughly 224,000 individuals were identified as either below either poverty line

or elderly living alone in the seven county study areas (Bureau, 2008). The SEHIC template identified

more than 127,000 (56%) of vulnerable individuals in 198 of 686 (28.8%) census tracts.

These results are concerning considering the indicator did not identify 100,000 vulnerable

individuals. Within the identified vulnerable census tracts, below poverty line groups made up 82% of

the population while elderly individuals living alone only accounted for 39%. The percentages of tracts

do not add to 100% because some tracts are vulnerable in both elderly and below poverty line


31

populations. The SEHIC map seems to overly represent areas with below poverty line individuals

regardless that the literature indicates elderly have a higher risk during emergencies.

In census tracks where both elderly living alone and below poverty line percents were above the

Minnesota average, the areas are considered at greatest risk. In these areas, the SEHIC indicator

identified 24,321 people (21%) in 43 of 686 tracts with extremely high potential for morbidity and

mortality. Within these small areas, below poverty line individuals account for 80% while elderly living

alone make up only 20%.

Table 4: Results from Original SEHIC Calculations (Row Totals)

Below Poverty Line Elderly Living Alone Total

# Tracts Identified by SEHIC

163/198 (82%) of SEHIC tracts

79/198 (39%) of SEHIC tracts 198 (29%) of 7 Co. tracts

SEHIC Population

Identified

102,949 (90%) of SEHIC Pop

10,822 (10%) of SEHIC Pop 113,71 (51%) of all Vul. Pop.

High Risk Areas

43/198 (22%) of SEHIC Area

19,660 (80%) of High Risk pop.

4,661 (20%) of High Risk pop. 24,321 (21%) of total SEHIC

SEHIC Issue of Concern 4: SEHIC Template is Incompatible with Other Data

As stated, vulnerability scores in percents have a number of limitations. In addition to not

accounting for demographic inequalities, percentages are not compatible with other sources of data.

Although the SEHIC indicator is clearly a public health related measure, most health related analysis use

health outcome/population rates. This is especially relevant in epidemiological studies.


32

Solution: Calculate Vulnerabilities Using Rate per 10,000 People

To make the SEHIC indicator more compatible with other health data, the calculation formula

has been altered to portray the number of vulnerable people per 10,000, expressed in a rate. The

benefit of using a rate is that it retains the original vulnerability of the tract, only represented as a rate.

Changing the calculation methods from using a percentage (1/100) to a rate of 1/10,000 does not

change the relative vulnerability.

Individual Vulnerable Population Analysis:

In order to investigate the inequalities in the SEHIC calculations, each vulnerable population is

mapped separately. Separating the maps into two individual analyses allows the researcher to analyze

the patterns of vulnerability.

Below Poverty Line Groups:

Tract BPL Z‐score = MN average (value 773.44 per 10,000) ‐ [(1999 BPL/total 1999 pop)*10,000] /

(value of 942.76)

According to the 2000 census, 179,316 were listed as living below the poverty line in the Twin

Cities. In addition, the number of at risk BPL individuals was changed from vulnerable population

percentages to a rate of the number of vulnerable people per 10,000 individuals. Additionally, the BPL

scores are changed into Z‐Scores for map scales and mapped by five color gradients associated with risk

levels (Figure 6).


33

Figure 6: Below Poverty Line Areas in the Twin Cities

Visually analyzing the map, the BPL individual map (Figure 6) is very similar to the original SEHIC

template (Figure 5). Individually, the BPL choropleth map designated 204 of 686 total census tracts as

being high‐risk BPL areas, 41 more tracts than the original SEHIC template identified as elevated BPL.

The 204 is also six more than the total number of tracts identified in the original template.

The individual BPL also identified 116,117 BPL individuals, 13,000 more than identified in the

original SEHIC template. The number identified in this individual analysis is also 2,500 more than the

total vulnerable individuals identified in the original analysis.


34

Elderly Living Alone Individual Analysis:

The analysis of the ELA population is the same as the BPL individual choropleth. The ELA

percents are changed to rates per 10,000 people and mapped by five color Z‐score gradients associated

with risk levels. Changes in vulnerable area between the ELA map and the individual SEHIC map identify

under‐represented areas Figure 7).

Tract ELA Z‐score = MN average (value 575.2 per 10,000)] ‐ [(1999 ELA/total 1999 Occupied Housing

units)*10,000 / (value of 343.91)

Figure 7: Individual Analysis of Elderly Living Alone


35

The Twin Cities metropolitan counties account for over 44,000 elderly individuals living alone,

24.9% of all those counted in Minnesota. Statistics from the 2000 census show the number of elderly

living alone in occupied housing in Minnesota to be 862.38 per 10,000, the standard for measuring at

risk areas.

In 2000, 10,822 vulnerable elderly were recognized in the original SEHIC template, measuring just

10% of the total vulnerable population in 39% of the census tracts. However, in the individual ELA

analysis, 139 (70%) of census tracts reported elevated rates of ELA.

The ELA analysis identified 18,635 ELA people, 7,800 more than the original calculations. The

100 tract difference and number of people identified in the separate analysis suggest a drastic

underrepresentation in the original template. Additionally, the spatial patterns show vulnerable elderly

groups residing in suburban areas and not highly associated with below poverty line areas but did not

show up on the original SEHIC indicator map.

Recommendations’ to SEHIC: Adjusted SEHIC Indicator:

Recommendations to SEHIC are based on the four major issues of concern found in the original

calculations that led to errors. To adjust for the bias, an “Adjusted SEHIC indicator” has been developed

to adapt for the demographic inequalities, promote data compatibility, and encourage accurate and

standardized mapping techniques.

The Adjusted indicator calculation equals the sum of the individually calculated BPL and ELA Z‐

scores. By adding the Z‐scores of each vulnerable group together to create an adjusted vulnerability

score, the score accommodates for the drastic overpowering of BPL to ELA people in the Twin Cities

(Table 3). Adjusted SEHIC tract Score Calculations:


36

Tract Vulnerability Score = BPL Rate Z‐Score + ELA Rate Z‐Score

In the table example, each of the vulnerable populations has been individually normalized from

their own z‐scores. Since each of the populations is standardized to their own data, there is no error

based on demographic inequalities. When the two Z‐Scores are added, it creates an adjusted

vulnerability score for the census tract that is more accurate and compatible with other health data.

Table 5: Adjusted SEHIC Calculations

Census Tract

BPL Rate

MN-BPL Rate

BPL Z-Score

ELA Rate

MN-ELA Rate

ELA Z-Score

Adjusted SEHIC Score

MN = 773.40

MN = 575.2

BPL Z + ELA Z

1 670.2 103.2 0.11 646.3 -71.1 -0.21 -0.1 2 1160 -386.6 -0.41 94.4 480.8 1.4 0.99

The table portrays examples of tracts with inequalities in both populations. By using Z‐scores, each

population is normalized and creates a more accurate composite vulnerability score in each census tract.

Adjusted SEHIC Indicator Map Analysis:

The Adjusted SEHIC indicator map is the spatial version of all altered census tracts. All tracts

have been adjusted according to the calculations changes to adjust for issues of concern.


37

Figure 8: Adjusted SEHIC Vulnerable Areas

In a comparison between the original and adjusted SEHIC models, the adjusted calculation

method identified 128,211 at risk individuals. 3,117 more than the original template in an additional 38

census tracts (Figure 8). The original SEHIC map includes 88% BPL population and 12% ELA. The

adjusted SEHIC template changed the BPL population to 83% and ELA to 17%.

Table 6: Analysis of Original and Adjusted SEHIC Results (Row Totals)

Original SEHIC Below Poverty Line Elderly Living Alone Total

# of Tracts Identified



198 (29%) of 7 Co. tracts

SEHIC Population 102,949 (90%) of SEHIC 10,822 (10%) of SEHIC 113,71 (51%) of all V.Pop

High Risk Areas 43/198 (22%)



24,321 (21%) of total SEHIC


38

Adjusted SEHIC (A.S)

Below Poverty Line Elderly Living Alone Total

Tracts 142/235 (60%) of A.S. 139/235 (59%) of A.S 235 (34%) of 7 Co. tracts

Population 106,285 (83%) of A.S 21,926 (17%) of A.S 128,211 (57%) of all V.Pop

High Risk Areas

46/235 (20%)

20,708 (80%) of High Risk 5,058 (20%) of High Risk 25,766 (20%) of total A.S

In the spatial analysis, 235 census tracts were identified as vulnerable in the Adjusted SEHIC

analysis. Of those, 142 were BPL areas, a decrease of 11 from the original template. In contrast, elderly

living alone tracts accounted for 139 of the 235 tracts, an increase of 60 tracts from the original SEHIC

template. The nearly equal number of tracts associated with the vulnerable population is significant

because it equally represents the vulnerable areas and accommodates for the overwhelming burden of

below poverty line individuals against elderly individuals.

Associated with the change in tract distribution the population percentages saw a reduction in

the number of BPL individuals and an increase in ELA. Additionally, an additional 1,500 very high risk

individuals were identified, those who fit census tracts with elevated ELA and BPL scores. Although

more individuals were identified, the distribution did not change. This can be explained from the small

percentage of individuals relative to entire vulnerable population.

Case Study: Chronic Lower Respiratory Disease:

Chronic Lower Respiratory Diseases (CLRD) has historical excess morbidity and mortality levels

among low‐income elderly populations during heat and air quality emergencies. Analyzing CLRD data in

the Twin Cities accomplishes two tasks. First, to incorporate health data into the existing SEHIC


39

indicator. Second, to test the strength of the adjusted SEHIC calculation methods against the original

methods. CLRD data is portrayed as a rate per 10,000 people per zip code. In 2000, 1,474 CLRD

hospitalizations were reported in Twin Cities zip codes, establishing a seven county standard rate of 8.14

hospitalizations per 10,000 people. The CLRD rate for the twin cities was used because the remainder of

the state had very few CLRD hospitalizations and the purpose of the indicator is to identify relative risk

areas.

CLRD Risk by Zip Code = 8.14‐ [CLRD hospitalizations per zip code / (total ZIP code population x 10,000)]

The map of 2000 CLRD cases are limited to those zip code with 20 cases or above to be

considered statistically significant. Additionally, the CLRD rates have been standardized and are mapped

by Z‐score.


40

Figure 9: CLRD Hospitalization Rates by Z‐Score

Map results show excessive CLRD prevalence in 46 of the total 200 (23%) zip codes. 8 zip codes reported

lower than the standard Twin Cities rate, and 146 zip codes had too few hospitalizations to be

statistically significant (Figure 9).

Disease Indicators:

Disease indicators refer to data analysis that examines a specific, vulnerable population and

related health outcomes associated with an exposure. Disease Indicators can be used retrospectively to

investigate sources of exposures that led to disease. Indicators may also be used prospectively, as in this


41

study, to identify the vulnerable populations where diseases are currently overlapping to aid in

preventive public health planning.

Analyzing the Disease indicators requires two changes in spatial analysis. First, to use the rate

calculations. CLRD data is only available by rate, thus the SEHIC calculations must be based in rates to be

accurate. Second, CLRD data is based on Zip code, not by census tract. Therefore, the unit of spatial

analysis is called a ZIP Code Tabulation Areas (ZCTA’s) (Division, 2001) is used as the mapping base.

ZCTA’s are a sort of ‘best‐fit’ polygon of census tracts and ZIP codes, which have overlapping boundaries.

The change to ZCTA’s does not change the vulnerability index, only a slight appearance difference in the

map. The GIS method to combine the census tracts and Zip Codes into ZCTA’s eliminates the ability to

identify the number of people associated with outcomes. Additionally, the number of ZCTA’s drastically

increases to 1316, verses 686 census tracts.

Disease Indicator # 1: Original SEHIC Template and CLRD

Hospitalizations:

The Disease Indicator 1 (DI1) is the combination of the original SEHIC indicator in rates and the

CLRD data in rates.

ZCTA Vulnerability by Z‐Score = (Original SEHIC template scores) + (CLRD Scores)

Z‐scores are especially important to incorporate into this model because the average rate of

SEHIC vulnerability is 1348.60 people per 10,000 while the CLRD rate is only 8.14 people per 10,000.

Even the largest prevalence in CLRD rates would have little effect on a low SEHIC vulnerability rate.

In the spatial analysis, it is important to recognize that darker areas are considered vulnerable by

both SEHIC and CLRD measures. These areas indicate that the vulnerable populations may have


42

environmental exposures that lead to higher rates of CLRD. In the spatial analysis of the combined

original SEHIC rates and CLRD rates by z‐score, the DI1 found 455 of 1316 (34.5%) of overlapping ZCTA’s.

Of the overlapping areas, 69% are SEHIC vulnerability areas and 97% are CLRD vulnerable area (Figure

10). The importance of identifying the content of the ZCTA population is to identify exactly which

component of the formula is being represented. An overrepresentation by either CLRD or the SEHIC

calculations indicate errors. From the result, the researcher can conclude that CLRD areas are being

overly represented. Spatially, the map has very similar visual components to the original SEHIC map. To

recall, the original method had an over representation of BPL population that is evident by comparing

Figures 6 and 10.

Figure 10: Disease Indicator 1: Original SEHIC Template and CLRD Hospitalizations


43

Disease Indicator # 2: Adjusted SEHIC Template Scores and CLRD

Hospitalizations:

As a comparison to the DI1, a Disease Indicator 2 (DI2) has been created from the Adjusted SEHIC

calculations combined with CLRD rates. The purpose of the DI2 is to quantify changes of the Adjusted

SEHIC indicator by looking for better identification of at risk areas and more vulnerable populations

(Figure 11).

DI2 = BPL Rate Z‐Score + ELA Rate Z‐Score + CLRD Z‐ Score Rate

Figure 11: Adjusted Disease Indicator


44

Analysis of the combined adjusted SEHIC template and CLRD hospitalization records revealed vulnerable

areas in 35.7% of ZCTA’s. In the overlapping areas, population components include 77% BPL, 59% ELA,

and 97.6% CLRD.

Table 7: Analysis of Disease Indicator 1 and 2

CLRD Data SEHIC Data Total

DI 1: ZCTA’s Identified 440/455 (97%) of Total 313/455 (69%) of Total 455/1316 (34.5%) of all ZCTA’s

DI 2: ZCTA’s Identified 361/469 (97.6%) of Total

BPL: 361/469 (77%)

ELA: 277/469 (59%)

Each of Total

469/1316 (35.7%) of all ZCTA’s

Reflection on the outcomes of the DI2 map provide excellent grounds to change the SEHIC

calculation methods from percentages to rates and to use Z‐scores to accurately accommodate for both

vulnerable populations.

Discussion:

Discussion of the SEHIC indicator study includes two areas of analysis; the study limitations of the

SEHIC design, and biases that could impact the results and conclusions of the research study.

Study Limitations:

Limitations in the SEHIC study primarily include unresolved issues with the indicator template.

Because the indicator model is under development, there are several weaknesses in the indicator

methods that detract from the effectiveness and reliability of the indicator. The limitations of the SEHIC


45

indicator reduce its potential for interstate use as an accurate and consistent model. Limitations are

loosely grouped into two categories: data and population, and methods

First, limitations in the study related to data include the completion and accuracy of the census.

The populations focused on in this study can be considered relatively transient groups. Below poverty

line populations may be frequently moving to accommodate for jobs or better housing opportunities.

Elderly individuals are also a transient group, possibility moving from family homes to smaller homes or

apartments. Depending on health concerns, they may move to be with their children or into assisted

living centers for more advanced health care.

The census also does not distinguish living standards other than by yearly income. Regardless

that a person makes a below the poverty line income per year, they also may have a life savings that

allows them to live comfortably with excellent health care. In addition, the census also cannot calculate

the clinical age of an individual who may have extensive chronic diseases who would technically be of an

age under 65 though clinically fit descriptions of chronically ill and vulnerable populations.

Data related study limitations include the potential that hospitals could be under or over

reporting CLRD diagnostic codes. Because Chronic Lower Respiratory Diseases have similar symptoms,

doctors may diagnose more than one condition or consider it part of another systematic illnesses.

Additionally, it is impossible to tell if an individual was admitted to the hospital with CLRD more than

once, but the visit would be counted as a separate incident. For example, one person with excessive

CLRD symptoms could potentially skew the CLRD hospitalizations rates by being hospitalized multiple

times.

A methodological study limitation is that this study is a pilot of a template that is under‐

development. Because there is no existing data to compare the results against, it is difficult to weigh the


46

effectiveness of the recommended measures other than using census tract analysis. Additionally,

because there is no individual health data, no strength of association calculations can be completed to

add statically significance to the findings. Therefore, the study must rely on the conceptual models to

provide a better health indicator.

Bias:

Study bias is defined as a systemic flaw resulting in a false association between the exposure and

health outcome (Aschengrau, 2003). Bias is difficult to remove once calculated into the data thus the

study design must reactively identify and account for the bias. Methods for reducing bias in

epidemiological studies include large participation numbers, multiple data collection methods, and using

multiple researchers to compare data analysis outcomes. Specific to the SEHIC study, bias types include

confounding and Ecological bias.

Confounding error is defined as an unknown factor that propels a false association between the

exposure and outcome (Gordis, 2004). Criteria for a confounding factor include being associated with

the exposure, being an independent cause of the disease, and the factors cannot lie in the causal

pathway of the exposure‐health outcome track. Confounding factors in quantitative studies are

measured using crude and adjusted relative risk calculations, but because the analysis of the SEHIC

indicator is an Ecological study the individual confounding errors cannot be quantitatively analyzed.

Specific to the SEHIC study, confounding factors may result from using census data, where

individuals’ responses to the census surveys may not explain the circumstances that might include or

exclude them from the vulnerable population. Confounding factors in CLRD calculations may include

misdiagnosis from physicians, or individuals who chose not to be treated for their symptoms due to

financial strain or personal beliefs.


47

Ecological Fallacy are unique to Ecological studies and occur when the results of a population

based study are imposed on individuals. Specific to the SEHIC indicator, an ecological bias would suggest

that all individuals that are identified by the census as ‘vulnerable’ might have a greater chance of

morbidity or mortality during a heat or air quality event. In addition, when investigating CLRD rates

among the vulnerable populations, the study must not indicate that all patients with CLRD fit the criteria

of a ‘vulnerable population’ or that all vulnerable individuals will eventually develop CLRD.

Conclusions:

Fervent investigation of the SEHIC indicator reveals that the model identified two vulnerable

populations that need public health surveillance, assistance, education, and resources during air quality

and heat emergencies. Reviewed literature supports the associated between climate change, urban

health, and health outcomes in vulnerable populations. Results identified below both poverty line

populations and elderly living alone to have an excess risk during these events, though elderly

individuals have shown to have a greater risk considering their pre‐existing clinical conditions, and

reduced social and economic capacity.

The study first calculated the data according to the indicator template. From the results, four

issues of concerns were found stemming from the conceptual design of the calculation methodology.

First, the indicator model did not address what average to use if more than one county is analyzed. This

is significant because using a different standard of vulnerability will result in a different number of at‐risk

individuals being identified. The Minnesota state averages for both vulnerable populations were chosen

to serve as the standard to provide the most realistic results for the Twin Cities. Second, the original

SEHIC indicator template used vulnerability scores in percents to show at risk areas. Percentages were

shown to be inaccurate and difficult to map. Therefore, the standard error method was incorporated to


48

use Z‐Scores to identify each piece of data in a standardized distribution. The change to Z‐scores made

analysis and mapping easier to complete. Third, the SEHIC template did not establish cut‐off points for

determining high risk individuals from low risk areas. Therefore, Z‐scores computed in the databases

were used as mapping scales and effectively identified at risk areas based on the distribution of the data.

Fourth, the original SEHIC data template is not compatible with other data sources. Therefore, the

vulnerability index was changed to using rates per 10,000 individuals to be well‐matched in other health

data analysis. This change was especially relevant to the CLRD case study.

After discovering the four issues and solutions identified, a new calculation methodology was

developed using the solutions. The new calculation method or ‘adjusted SEHIC indicator’ succeeded in

identifying more vulnerable census tracts, adjusted for demographic and vulnerability inequalities, and

highlighted very high risk areas more than the original SEHIC template.

To investigate the effectiveness of the adjusted SEHIC indicator, Chronic Lower Respiratory

Disease hospitalization was overlaid on the SEHIC template to create two “Disease Indicators.” The DI

that combined the adjusted SEHIC indicator with the CLRD data accomplished identifying more ZCTA

areas, the same proportion of CLRD percentages, and more SEHIC vulnerabilities areas.

Based on the testing of the original SEHIC model and development of an adjusted model, the

researcher feels confident submitting these results to the SEHIC indicator development team.

Additionally, the researcher hopes the team will consider these issues when designing future indicators.


49

Appendix 1: SEHIC “How‐To‐Guide”

Indicator: Susceptible Populations to Health

Measure: Heat Vulnerability Index: Percent of Households of Elderly Living Alone and Percent Below the Poverty Level at the Census Tract

How‐To‐Guide

1. Got to U.S. Census website, http://factfinder.census.gov/home/staff/main.html?_lang=en 2. Select “get data” under “Decennial Census” 3. Select “Census File 3” 4. Select “detailed tables: 5. Select “Census tract” as geographic type; your state and county of interest, and select “all census

tracts” 6. Click on the ‘add’ button. All census tracts in the selected country should appear. Then click

“next” 7. Select “P87 Poverty Status in 1999 by Age” and “H19 Tenure by Housing Type (including living

alone) by Age of Householder.” Click “add” to add each one. Click “show results.” 8. Table of results will appear. Click “print/download” at the top of the page. 9. Select “download” and “Microsoft Excel zip file” 10. Download zip file to a folder you specify on your hard drive. 11. Open the zip file. There will be two excel files. One has geographic information, including census

tract number, and the other file will have the population data. (In the population data file, the census tract number is also embedded in a variable called “geographic identifier.”

12. In the population excel file, erase all columns except “geography identifier, P087001, P087002, H019001, H019037, H019038, H019054, and H019055”.

13. Create a new column and calculate percent of population below poverty level by dividing P087001 by P087002 and multiplying by 100.

14. Compute the mean for the percent of the population below poverty level. 15. Create a new column and center each value of the percent of the population below the poverty

level by census tract by subtracting each value from the mean 16. Add the total male householders living alone (65+) by adding columns H019037 and H019038.

Do the same for females (H019054 and H019055.) Sum the males and females by census tract. 17. Compute the percent of households with elderly living alone by dividing the total in #16 by

H019001. 18. Center the percentages as before in #15 19. Create a final heat vulnerability score by adding the centered values for the two variables. 20. Create a choropleth map by census tract using a standard GIS package. We will need to decide

on standard cut‐points.

http://factfinder.census.gov/home/staff/main.html?_lang=en


50

References

Abelsohn, A. S. (2002). Identifying and managing adverse environmental health effects: Outdoor air pollution. Journal of the

Canadian Medical Association Vol. 166 (9), 1161‐1167.

Accetta, D. (2006, October 30). Asthma. Retrieved August 3, 2008, from National Institute of Health:

http://www.nlm.nih.gov/medlineplus/ency/article/000141.htm#Definition

Administration, N. A. (n.d.). NASA Global Warming. Retrieved April 5, 2009, from NASA:

www.nasa.gov/worldbook/global_warming_worldbook.html

Administration, N. O. (n.d.). Heat Waves. Retrieved February 26, 2009, from National Oceanic and Atmospheric

Agency, E. P. (2007, November 27). Air Quality Guide for Particle Pollution. Retrieved 14 2008, December, from AirNow:

http://airnow.gov/index.cfm?action=tvweather.aqguidepart

Agency, E.P. (2006, June). Excessive Heat Events Guidebook. United States Environmental Protection Agency Office of

Atmospheric Programs. EPA 430‐B‐06‐005

Agency, E. P. (2008, April 8). Six Common Air Pollutants. Retrieved December 19, 2008, from Urban Air Pollution:

http://www.epa.gov/air/urbanair/

Agency, M. P. (2003, May 15). Air Quality Index for Minnesota. Retrieved December 12, 2008, from Minnesota Pollution

Control Agency: http://aqi.pca.state.mn.us/

Alfrsio, L. Z. (2002). The Effect of Weather on Respiratory and Cardiovascular Deaths in 12 U.S. Cities. Environmental Health

Prospectives Vol. 110 (9) , 859‐863.

Amann, M. B. (2006, March). Climate Change and Air Pollution Research and Policy . Global Change Newsletter Vol. 65.

Analysis, D. o. (n.d.). Population Projection: Minneapolis‐St. Paul, MN‐WI . Retrieved February 20, 2009, from Minnesota State

Demographic Center: http://www.lmic.state.mn.us/datanetweb/php/DemProjection/PopPrjReport.php

Argaud, L. F. H. (2007). Short and Long‐term Outcomes of Heatstroke Following the 2003 Heat Wave in Lyon, France. Archives

of Internal Medicine Vol. 167 (20) , 2177‐2183.


51

Aschengrau, A. S. (2003). Essentials of Epidemiology in Public Health. Boston: Jones and Bartlett Publishers.

Basu, R. B. (2008). A Multicounty Analysis Identifying the Populations Vulnerable to Mortality Associated with High Ambient

Temperature in California. American Journal of Epidemiology Vol. 168 (6), 632.

Bauman, J. B. (2000). From Tenements to the Taylor Homes: In Search of an Urban Housing Policy in Twentieth‐Century

America. University Park, PA: The Pennsylvania State University Press.

Bellia, V. P. (2007). Asthma in the Elderly: Mortality Rates and Associated Risk Factors for Mortality. American College of

Chest Physicians Vol. 132 (4), 1175‐1182.

Belmin, J. A. (2007). Level of Dependency: A Simple Marker Associated with Mortality During the 2003 Heat wave Among

French Dependant Elderly People Living in the Community or in Institutions. Journal of Age and Ageing Vol. 36, 298‐303.

Braga, A. Z. (2002). The Effect of Weather on Respiratory and Cardiovascular Deaths in 12 U.S. Cities. Environmental Health

Perspectives Vol. 110 (9), 859‐863.

Browning, C. W. (2006). Neighborhood Social Processes, Physical Conditions, and Disaster‐Related Mortality: The Case of the

1995 Chicago Heat Wave. American Sociological Review Vol. 71, 661‐678.

Bureau, U. C. (2006, August 29). Poverty Thresholds 2000. Retrieved December 10, 2008, from Poverty:

http://www.census.gov/hhes/www/poverty/threshld/thresh00.html

Change, I. P. (2001). Climate Change 2000: Working Group II: Impacts, Adaptations and Vulnerability. San José, Costa Rica:

United Nations Environment Programme.

Choi, M. A. (2006). Geographic Information Systems: A New Tool for Environmental Health Assessment. Journal of Public

Health Nursing Vol. 23 (5), 381‐391.

Conti, S. M. (2007). General and Specific Mortality Among the Elderly During the 2003 Heat Wave in Genoa, Italy. Journal of

Environmental Research Vol. 103, 267‐274.

Control, C. f. (2008, December 17). Chronic Lower Respiratory Disease. Retrieved February 27, 2009, from National Center for

Health Statistics: www.cdc.gov/nchs/fastats/copd.htm


52

Corburn, J. (2004). Confronting the Challenges in Reconnecting Urban Planning and Public Health. American Journal of Public

Health Vol. 94(4), 541‐546.

Curriero, F. H. (2000). Temperature and Mortality in 11 Cities of the Eastern United States. American Journal of Epidemiology

Vol. 155 (1) , 80‐87.

Cutchin, M. (2007, September ). The Need for the "New Health Geography" in Epidemiologic Studies of Environment and

Health. Health Place Vol. 13 (3), 725‐742.

Dever, G. S. (1988). Creation of a Social Vulnerability Index for Justice in Health Planning. Journal of Family Community Health

Vol. 10(4), 23‐32.

Digenis‐Bury, E. B. (2008). Use of a Population‐Based Survey to Describe the Health of Boston Public Housing Residents .

American Journal of Public Health Vol. 98 (1), 85‐91.

Ebersol, R. (2005). Out of Breath. National Wildlife Vol. 43 (3).

Ebi, K. M. (2006). Climate Change and Human Health Impacts in the United States: An Update on the Results of the U.S.

National Assessment. Environmental Health Perspectives Vol .114 (9), 1318‐1324.

Eisner, M. B. (2008). COPD as a Systemic Disease: Impact on Physical Functional Limitations. The American Journal of

Medicine Vol. 121, 789‐796.

Epidemiologist, C. o. (n.d.). The State Environmental Health Indicators Collaborative . Retrieved January 29, 2009, from

Council of State and Territorial Epidemiologist: http://www.cste.org/OH/SEHIC.asp

Epidemiologists, C. o. (2008, October 21). Environmental Public Health Indicators. Retrieved November 1, 2008, from The

State Environmental Health Indicators Collaborative: http://www.cste.org/OH/SEHIC.asp

Fitzpatrick, A. P. (2004). Barriers to Health Care Access Among the Elderly and Who Perceives Them. American Journal of

Public Health Vol. 94 (10), 1788‐1794.

Gilmour, M. J. (2006). How Exposure to Environmental Tobacco Smoke, Outdoor Air Pollutants, and Increased Pollen Burdens

Influences to Incidence of Asthma. Environmental Health Perspectives Vol. 114 (4), 627‐633.


53

Gold, D. W. (2005). Population Disparities in Asthma. Annual Review of Public Health Vol. 26, 89‐113.

Gordis, L. (2004). Epidemiology. Philadelphia: Elsevier Saunders.

Gravetter, F. a. (2000). Statistics for the Behavioral Sciences. Belmont, CA: Wadsworth‐Thomson Learning.

Group, T. A. (2005). Global Initiative for Chronic Obstructive Lung Disease strategy for the diagnosis, management and

prevention of chronic obstructive pulmonary disease: An Asia–Pacific perspective. Respiratory Health Vol. 10.

Gwilliam, J. F. (2006). Methods for Assessing Risk from Climate Hazards in Urban Areas. Municipal Engineer Vol. 159 (ME4),

245‐255.

Halonen, J. L. (2008). Urban Air Pollution, and Asthma and COPD Hospital Emergency Room Visits. Pediatric Asthma, Allergy &

Immunology Vol. 21 (1), 44‐55.

Health, W. V. (2006, September 12). Chronic Lower Respiratory Disease:. Retrieved September 14, 2008, from West Virginia

Heath Statics Center: http://www.wvdhhr.org/bph/oehp/hsc/pubs/clrd/national.htm

Heitgerd, J. (2001). Using GIS and Demographics to Characterize Communities at Risk: A Model from ATSDR. Journal of

Environmental Health Vol. 64 (5), 21‐23.

Hood, E. (2005). Dwelling Disparities: How Poor Housing Leads to Poor Health. Environmental Health Perspectives Vol.113 (5),

A310‐A317.

Howell, E. H. (2005). The Health Status of HOPE VI Public Housing Residents. Journal of Health Care for the Poor and

Underserved Vol.16, 273‐285.

Jackson, R. J. (2003). The Impact of the Built Environment of Health: An Emerging Field. American Journal of Public Health Vol.

93 Issue 9, 1382‐1383.

Jelicic, M. G. (1997). Cognitive Function in Community‐Dwelling Elderly with Chronic Medical Conditions. International Journal

of Geriatric Psychiatry Vol. 12, 1039‐1041.

Keim, M. R. (2007). The CDC/ATSDR Public Health Vulnerability Mapping System: Using a Geographic Information System for

Depicting Human Vulnerabilities to Environmental Emergencies. Atlanta: Center for Disease Control.


54

Kindig, D. S. (2003). What is Population Health? American Journal of Public Health Vol. 93(3), 380‐383.

Knowlton, K. L. (2007). Projecting Heat‐Related Mortality impacts Under a Changing Climate in the New York City Region.

American Journal of Public Health Vol. 97 (11), 2028‐2033.

Knowlton, K. R.‐E. (2009). The 2006 California Heat Wave: Impacts on Hospitalizations and Emergency Department Visits.

Environmental Health Perspectives Vol. 117 (1), 61‐66.

Kovats, R. H. (2008). Heat Stress and Public Health: A Critical Review. Annual Review of Public Health Vol. 29, 41‐55.

Krieger, J. H. (2002). Housing and Health: Time Again for Public Health Action. American Journal of Public Health Vol.92(5) .

Lewis, S. C. (2004). Medical‐Surgical Nursing. New York: Mosby.

Liebers, V. R.‐H. (2008). Health Effects Due to Endotoxin Inhalation (Review). Toxicology Vol. 82, 203‐210.

Luber, G. S. (2006). Heat‐Related Deaths‐ United States, 1999‐2003. Morbidity and Mortality Weekly Report Vol. 55 (29), 796‐

797.

McCoy, C. R. (2005). A Multiple Cause‐of‐Death Analysis of Asthma Mortality in the United States 1990‐2001. Journal of

Asthma Vol. 42, 757‐763.

McGeehin, M. M. (2001). The Potential Impacts of Climate Variability and Change on Temperature‐Related Morbidity and

Mortality in the United States. Environmental Health Perspectives Vol. 109(2), 185‐189.

Medina‐Ramon, M. S. (2008). Who is More Vulnerable to Die from Ozone Air Pollution? Journal of Epidemiology Vol. 19 (5) ,

672‐679.

Moorman, J. R. (2007, October 19). National Surveillance for Asthma ‐ United States 1980‐2004. Retrieved August 18, 2008,

from Center for Disease Control: www.cdc.gov/mmwr/preview/mmwrthml/ss560a1.html

Naughton, M. H. (2002). Heat‐Related Mortality During a 1999 Heat Wave in Chicago. American Journal of Preventative

Health Vol. 22(4), 221‐227.


55

Norback, D. W. (2002). Asthma Symptoms in Relation to Measured Building Dampness in Upper Concrete Floor Construction

and 2‐ethyl‐1‐hexanol in Indoor Air. International Journal of Tuberculosis Vol. 4 (11) , 1016‐1025.

O'Malley, P. (2007). Heat Waves and Heat‐Related Illness: Preparing for the Increasing Influence of Climate on Health in

Temperate Areas. Journal of the American Medical Association Vol. 298 (8), 917‐919.

O'Neill, M. E. (2009). Temperature Extremes and Health: Impacts of Climate Variability and Change in the United States.

Journal of Occupational Environmental Medicine Vol. 51, 13‐25.

Organization, W. H. (2008). Diseases of the Respiratory System. Retrieved October 21, 2008, from International Classification

of Diseases: www.who.org.int/classifications/apps/icd/icd10online.gj40.htm

Organization, W. H. (2006). Preventing disease through healthy environments: Towards an estimate of the environmental

burden of disease. Retrieved March 8, 2008, from Quantifying Environmental Health Impacts:

http://www.who.int/quantifying_ehimpacts/publications/preventingdisease/en/index.html

Ostro, B. B. (2006). Fine Particulate Air Pollution and Mortality in Nine California Counties: Results from CALFINE.

Environmental Health Perspectives Vol. 114 (1), 29‐33.

Parry, M. C. (2008). Key IPCC Conclusions on Climate Change Impacts and Adaptations. Intergovernmental Panel on Climate

Change. WMO Bulletin 57 Vol. 1

Peden, D. (2002). Pollutants and Asthma: Role of Air Toxins. Environmental Health Perspectives Vol. 110 (4).

Perdue, W. S. (2003). The Built Environment and Its Relationship to the Public’s Health: The Legal Framework. American

Journal of Public Health Vol. 93 (9).

Perkins, S. (2004 , July 3). Dead Heat. Science News Vol. 116 (1).

Prevention, C. f. (2008, April 11). Chronic Lower Respiratory Disease. Retrieved September 14, 2008, from National Center for

Health Statistics: http://www.cdc.gov/nchs/fastats/copd.htm

Ren, C. W. (2008). Ozone Modifies Associations Between Temperature and Cardiovascular Mortality: Analysis of the MNMAPS

Data. Journal of Occupational and Environmental Medicine Vol. 65, 255‐260.


56

Roberts, E. E. (2008). Spatially Continuous Local Rate Modeling for Communication in Public Health: A Practical Approach.

Journal of Public Health Management Practice Vol. 14 (6), 562‐568.

Rosenzweig, C. S. (2006). Mitigating New York City's Heat Island With Urban Forestry, Living Roofs, and Light Surfaces. New

York: New York City Regional Heat Island Initiative Final Report.

Rushton, G. A. (2008). Decoding Health Data: The Uses of Geographic Codes in Cancer Prevention and Control, Research, and

Practice. New York: CRC Press.

Saegert, S. K. (2003). Healthy Housing: A Structured Review of Published Evaluations of US Interventions to Improve Health by

Modifying Housing in the United States, 1990–2001. American Journal of Public Health Vol. 93 (9), 1471–1477.

Selgrade, M. L. (2006). Introduction of Asthma and the Environment: What We Know and Need to Know. Environmental

Health Perspectives Vol. 114 (4), 615‐619.

Shankardass, K. M. (2007). The Association between Contextual Socioeconomic Factors and Prevalent Asthma in a Cohort of

Southern California School Children. Social Science & Medicine Vol. 65 (8), 1792‐1806.

Shmaefsky, B. (2006). One Hot Demonstration: The Urban Heat Island Effect . Journal of College Science Teaching Vol. 35 (7) ,

52‐54.

Stafford, M. D.‐W. (2008). Small Area Inequalities in Health: Are We Underestimating Them? Social Science and Medicine Vol.

67, 891‐899.

Thomas, N. S. (2002). Preventable Tragedies: Heat Disaster and the Elderly. Journal of Gerontological Social Work Vol 38 (4),

53‐65.

Thomas, S. W. (1999). Asthma Hospitalizations and Mortality in Chicago: An Epidemiologic Overview. American College of

Chest Physicians Vol. 116(4) , 135‐141.

Vandentorren, S. B.‐B. (2006). August 2003 Heat Wave in France: Risk Factors for Death of Elderly People Living at Home.

European Journal of Public Health Vol. 16 (6), 583‐591.


57

Vandentorren, S. S. (2004). Mortality in 13 French Cities During the August 2003 Heat Wave. American Journal of Public

Health Vol. 94 (9), 1518‐1520.

Vincenzo, B. P. (2007). Asthma in the Elderly. Mortality rate and Associated Risk Factors for Mortality . American College of

Chest Physicians Vol. 132(4).

Whitman, S. G. (1997). Mortality in Chicago Attributed to the July 1995 Heat Wave. American Journal of Public Health Vol. 87,

1515‐1518.

WHO/USAID. (2002). Addressing the links between indoor air pollution, household energy and human health. Geneva: World

Health Organization.

Youger, M. M.‐A. (2008). The Built Environment, Climate Change, and Health Opportunities for Co‐Benefits. American Journal

of Medicine Vol. 35 (5), 517‐526.

dunlap plan b

Documents