Diabetes and Sprawl

1
Diabetes and Sprawl

• Advanced GIS (Spring 2004)
• Project 3. Public Health Analysis
• Feb.26. 2004
• Prepared by Su Young Kang

2
Why Diabetes and Sprawl?

• In US Today (August 2003), sprawl creates an
  unhealthy environment for people.
• People living in sprawling neighborhoods walk
  less and weigh more than those living in compact
  urban neighborhoods.
• People living in sprawling neighborhoods where
  more likely to suffer from obesity could lead to
  a higher risk of cancer, diabetes, and other
  diseases.

3
Purpose

• to look at temporal-spatial clustering of
  diabetes in Alabama from 2000 to 2002 by
  population-adjusted Morans I procedures (Odens
  Ipop Method).
• to think about the relationship between diabetes
  and sprawl according to current data of diabetes
  available.

4
Background

• Alabama and Mississippi have the highest rate of
  prevalence of diagnosed diabetes per 100 adult
  populations in 2001 (State Health Facts) .

5
Background

• The most populated county in Alabama is Jefferson
  County in 2001.

6
Background
7
Background
8
Background
9
Background
10
Background
11
Literature Review (1)

• Public Health, GIS, and Spatial Analytic Tools
• (Rushton, 2003)
• Geographic Information Systems (GIS) is the
  organizing system for health data or the spatial
  analytic tools will likely be incorporated in
  GIS-based analyses in the future.
• Recent advances in the analysis of disease maps
  have been influenced by and benefited from the
  adoption of new practices for georeferencing
  health data and new ways of linking such data
  geographically to potential sources of
  environmental exposures, the locations of health
  resources and the geodemographic characteristics
  of populations.

Source Gerard Rushton (2003), Public Health,
GIS, and Spatial Analytic Tools, Intelligent
Synthesis of the Scientific Literature, Annual
Reviews, May 2003, Vol. 24, pp. 43-56
12
Literature Review (2)

• Time-Space Clustering of Human Brucellosis,
  California, 1973-1992 (Fosgate, 2002)
• A modification of the Morans I technique, the
  population-adjusted Morans I (Ipop), which
  adjust for the underlying population density in
  each area, was used to evaluate spatial
  clustering of reported human brucellosis cases in
  California.
• The Ipop analysis is more powerful than the
  unadjusted Morans I, so the Ipop procedure is
  recommended when both numerator and denominator
  data are available

Source Geoffrey T. Fosgate, Tim E. Carpenter,
Bruno B. Chomel, James T. Case, Emilio E. DeBess,
and Kevin F. Reilly (2002), Time-Space Clustering
of Human Brucellosis, California, 1973-1992
13
Literature Review (3)

• Neighborhood-Based Differences in Physical
  Activity An Environment Scale Evaluation
• (Saelens et al. 2003)
• Residents of high-walkability neighborhoods
  reported higher residential density, land use
  mix, street connectivity, aesthetics, and safety.
  
• They had more than 70 more minutes of physical
  activity and had lower obesity prevalence
  (adjusted for individual demographics) than did
  residents of low-walkability neighborhoods

Source Brian E. Saelens, PhD, James F. Sallis,
PhD, Jennifer B. Black, BA and Diana Chen, BA,
(2003), Neighborhood-Based Differences in
Physical Activity An Environment Scale
Evaluation, American Public Health Association,
September 2003, Vol 93, No. 9 American Journal of
Public Health 1552-1558
14
Data Description

• USA Prevalence of Diagnosed Diabetes per 100
  Adult Population, state data 2001, U.S. 2000 ----
  State Health Facts
• Alabama diabetes data ---- Department of Public
  Health of Alabama
• USA Boundary ---- Census Bureau Tigerline
• Alabama Boundary ---- Collins Software Company

15
Problem Statements

• Identifying relationship between sprawl and
  diabetes needs more information and data, and
  more time for the research
• Problems in using Odens Ipop Method and
  analyzing the result, e.g. Rook and Queen, Z- and
  P- value, Ipop and E (I)
• Significance Testing for Small Samples uses Monte
  Carlo, Automatic calculation of the Monte Carlo
  p-Value for each analysis

Source Rooks Case (http//www.uottawa.ca/academ
ic/arts/geographie/lpcweb/newlook/data_and_downloa
ds/download/sawsoft/rooks.htm)
16
Approach and Methodology

• Look at the diabetes cases in USA
• Find one state having the highest diabetes rates
  in USA
• Look at the diabetes rates and population of the
  state
• Choose the highest population county of the state
  and near adjacent counties
• Look at the population and diabetes cases of the
  counties
• Run Cluster Seer2 program with total diabetes
  cases and populations of the counties using
  Odens Ipop method (2000-2002)
• Run the program for male diabetes cases and
  female diabetes cases (2000-2002)
• Discuss the results

17
Note for Analysis (Odens Ipop Method)

• The Ipop statistic will be large when there is
  clustering within a region or among adjacent
  regions.
• The expectation of Ipop under the null hypothesis
  (no clustering) approaches zero for large total
  population.
• Positive z-score indicates tendency towards
  clustering, negative value dispersion.
• Within , Among Percentage of estimated
  spatial clustering attributed to cases in the
  same counties (Within ) and in adjacent counties
  (Among ).
• All identified clustering attributed to cases in
  the same counties. Negative value in
  demonstrates dispersion of cases in adjacent
  counties.
• Significance level 0.05
• Number of Monte Carlo randomization runs 99
• Geographic meter
• Polygon Contiguity Queen

18
Analysis (1)

• Total Cases

• Significant clustering was examined in 2000 and
  2002 year
• Total cases in 2001 were not clustered (random
  event)
• The clustering effect in all tests resulted from
  the number of cases within counties.
• Cases in adjacent counties during 2000-2002 were
  dispersed, driving the test statistic in negative
  clustering effect.

19
Clustering 2000 Total Cases
20
Random Event - 2001 Total Cases
21
Analysis (2)

• Male Cases

• Male cases during 2000-2002 were not clustered
  (random event).
• Results also show that most clustering was due to
  the number of cases within counties, rather than
  adjacent counties.
• Cases in adjacent counties during 2000-2002 were
  dispersed

22
Analysis (3)

• Female Cases

• Woman cases during 2000-2002 were not clustered
  (random event).
• Results also show that most clustering was due to
  the number of cases within counties.
• Cases in adjacent counties during 2000-2002 were
  dispersed.

23
Conclusion

• In total cases during 2000-2002, significant
  clustering was found in 2000 and 2002 years,
  except for 2001 year.
• Significant clustering in male and female cases
  was not examined during 2000-2002 year.
• In all three cases, results show that most
  clustering was due to the number of cases within
  counties, rather than adjacent counties.
• Also, cases in adjacent counties during 2000-2002
  were dispersed, driving the test statistic in
  negative clustering effect.

24
Discussion (1)

• High diabetes rate happens in southern part
  rather than northern part.
• It occurs in the eastern part rather than the
  western part.
• Why? Was it related to living places? Was it
  related to the race or other factors?

25
Discussion (2)

• Although Jefferson County was the most populated,
  some cases in other counties were higher rate
  than Jefferson County, especially, Walker County
• Female rates were higher than male rates
• Why? Was it related to sprawl? Do women like
  living in sprawling neighborhood?