Lecture 5_Sept 10 2009_IntroductiontoGIS1

Lecture 5_Sept 10 2009_IntroductiontoGIS1 - Medical...

Info iconThis preview shows page 1. Sign up to view the full content.

View Full Document Right Arrow Icon
This is the end of the preview. Sign up to access the rest of the document.

Unformatted text preview: Medical Geography and GIS in Public Health Professor Mike Jerrett Goals • Introduce medical/health geography and the rationale for spatial analysis • What is GIS and why GIS in Public Health? • Examine some applications to prevention research What is Medical/Health Geography? • “the application of geographical perspectives and methods to the study of health, disease and health care” • study of spatial variations in human health • study of spatial variations in all kinds of health care, formal and informal Rationale • Why are GIS and spatial analysis used in public health studies? Association of disease with place means: 1. Population living there experiences increased exposure to a risk agent (e.g., air pollution) 2. Population is more susceptible (e.g, elderly, poor) 3. Can also suggest how the population adapts to its environment A. Pollution Exposure and Asthma Symptoms in Hamilton Canada B. Susceptibility Marker: Cluster of Low Income - Getis Statistic Industrial Core Niagara Escarpment z-scores -2.56 - -1.09 -1.09 - 1.1 1.1 - 3.28 Significance No Data N Hamilton Harbour HW 3 40 Y. MAIN ST. KING S T. Lake Ontario ST. Q.E. W . ER J AM ES UPP "LIN C" HW Y . 53 0 2 4 6 Kilometers Smoking Rates Hamilton Harbour Lake Ontario Industrial Core TSP Exceedence Level Niagara Escarpment N 2 0 2 4 Kilometers Legend: Rate / 1000 0 - 150 151 - 300 301 - 360 361 - 400 401 - 450 Spatial Data Defined • Attributes measured in different scales • Nominal (e.g.,disease names), ordinal (e.g., severity of disease), Interval (e.g., temperature in C), ratio (e.g., mortality rates) • Location measured in a coordinate system • Spatial analysis brings these two types of data together # # ## # ## #### # # # ## # ## ## # ## # ## ## # ## # # ## # # ### # # ### ## ##### # # # # # ## # # # ### # # # # # # ## ## # # # # ## ## ### # # ## # ## #### ### # ## # # # ## ## ## # # ## #### ## # ## # # ## # ## ## # # # # # # # # ## ### ## # # ## # # ## # ## # #### # # # # # # # ### # # ## ####### ## ## # # # # # ## ### # # # # # ## ##### # # ## # # ## # ## ## ### ## ## # # # ## # # # # # # ### # # ## #### # # # ## # # # ### ## ## # ## # # ## # ## ## # # # # ### # # ## # ## ## # ## # # ## # # # # # ## # ## # # ## # ## # # # # # # #### # ### # # # # # # # ## # # # # ## # # # ## # # ### ### # ## # ## # ## # # ## # # # ### ### # ## # # # ### # # # ## # ## ## # # # # #### # # ### # # ## ## # # ## # # ## ## ## # ### # ## # # ### ## # # ###### ## # # # ## # ## # ##### # #### # # # # ## ##### # # # # # # ### # # # # ## # ### # #### ## # ### # ## ## # # ## # #### # # ## # # ## # # ## # ## ## # # # ## # # ## # ## # # ## # # # # ## #### # ## # ## # # # # ## # ## ##### # # ## # ## #### ## # # # # ### ### # # #### ### ### ## # ## ## ### # # ## ## # # # # ### # # # # # # ## # # # # ## ## ##### # # ## #### # # ### # # # ### # ## ## ### # # # # ## # ## # # # # ### # # #### # ### # ## ## # # ## # # # ## # ## # # # ## ## ### ## # ## # # # ## # # ## ## ## # # # # # # # # ## # # # # # ## # ## # # ## ## ## # # # ## # # # # # # ### # # # # # # # ## # #### # # # ### ## # ## # ### ### # ## # # # # ## # ## # ## # # # # # ## ## ## # ## ## # # # ## # # # ## ### # # ## ## ## # # ### ## # ## # ## # # # # ## # # ## # # # # ## # ## ### ## # ## # # ## # # ### # # ### # # # ## ## ## # # # ## # # ## ## # # # # ## ## # # # ## # # # # ## # # ## # #### # # ## # # ## # ## # ## # # # # ### # # # ## # # ## ## ## # ## # ## # ### # # ## # # ## # ## # # ## # # # # # # ## # # ## # ## # # ## # # # # ## # ## # ## # ## # ## ## # ## # ## # # # ## # ## # ## #### # ## ## # # ## # ## # ## # #### # ## ## # # ## ## # # # # ## # # # # # # ## # # # ## # # # ### # ## ## # # # ## ## # ## ## # # ## ## # ## ## # ### #### # # ## # ## # ## # # # ## # ## # # ## # # # # # # # ## # # # ## # ### # ## ## ## ## # ## ## # ## # # # # # ## # # # # ## ## # # # ## # ### ## # # # # ## ## # ## # # # ## ## # # # # # ## ## ## ## ## # # ## # ## # # ## ## # # ## ## ## # ## ## # # # ## ## ###### # # # ## # ## # ## # # ### # # # #### # ## # ## # ## # ## # # ## # # # # # ## ## # ## # # # # # ## # # # # # # # # # # ## # # # # ## # ## ## # # # # ## # ## # # # ## # # # ## ### # ### # ## ## ## ### # # # ## # # # ## ### ## # ## # # # # # # # ### # # # ## # # ## # # # # ## # # #### # # # ### ## # # # # # # ## # ## ## # # # # ### # ## # ## ## # # ##### # ## ## ## # ## # # # # # ## # # ## # # ### # # ## # # # # # ## # ## # # ## ## # # ## # # # # ## # # # ### # # # # ## ## # ## # # # # # # ## ### # # ### ## ## # ## # # # # ### # ## # # # ## # # ## ## # # ## # ## # ## # ## # # ## # # ## # # # # # ## # # # ## # # # ## # ## # # # # # ## # ## # # # # ## # ## # ##### ## ### ## # # ## # # # # ## # # # # ## # # ### ## ## # # # # # ## # ## # ## # ## # # ## # # # # # ## ## ## # ### # ## ## # ## ## # # # # # # # ## # ## # # # ## ## ## # ### ## # ## # # ## ### # # # # ## # ## # # # ## # # # ## ## # # ## # # # # ## # # #### ## ## # # # # ## # ## # ## # # # # # # # # # ## # # # # # # ## ## # # ## ### # # ## # # ## ## # ## ## # # ## # # ## ## # ## # # ## # ## # ## # # # #### ## # # ## # # # # # # # # ## ## ## # # # ## # ## # ## # # ## # ## # # # ### ## ## ## # ## # ## # # # # ## ## # ## ## # # ## # # #### # # ## # # # ## ## # ## # # # # Mortality within 2.5 Km of the Taro Site ð % ð % 1985-94 Wind Frequency Distribution (Windroses) for Ontario Ministry of Environment 100 m meteorological tower on Woodward Ave. (STP grounds) NNW NW 5 N 7.5 10 NNE NE WNW 2.5 ENE 0.5 0 0.5 1 1.5 2 2.5 Km at sensor height of 30.5 m (100 ft) W E WSW ESE SW SE SSW S SSE (indicative of wind patterns over the lower eastern parts of the city, from the escarpment to within a few kilometres of the lakeshore / Hamilton Harbour) NNW NW N 7.5 10 NNE NE 5 WNW 2.5 ENE at sensor height of 91.5 m (300 ft) E W WSW ESE (indicative of wind patterns over the escarpment brow and parts of the upper city in proximity of the escarpment, e.g. TARO site) SW SE SSW S SSE # # ## # # # ## # # # # # # # # # ## # # ## # # # # # # # # hwa ## # # y 20 ## ## Mortality Database # # # # # ## # ## # # # ## # ## # # # # ## # # # ## # # # # # # Mud St. W . # H ig TARO - West TARO - East 2.5 km - radius buffer around the TARO-East landfill site Year Reg_no Evntdate Evntgeo Year Reg_no Evntdate EvntgeoSexResgeo Age Sex Resgeo Age 1985 1985 1985 1985 1985 1985 1985 1985 1985 1985 1985 1985 1985 1985 1985 1985 1985 1985 1985 1985 1985 1985 1985 1985 1985 1985 29462 1985/06/06 29462 1985/06/06 41800 1985/08/14 41800 1985/08/14 56494 1985/10/24 56494 1985/10/24 52972 1985/10/08 52972 1985/10/08 59186 1985/11/02 59186 1985/11/02 # 3560 1985/01/10 3560 1985/01/10 10014 1985/02/22 10014 1985/02/22 59354 1985/11/22 59354 1985/11/22 1487 1985/01/13 1487 1985/01/13 28156 1985/05/24 28156 1985/05/24 55306 1985/10/15 55306 1985/10/15 31863 1985/06/16 31863 1985/06/16 45288 1985/08/30 45288 1985/08/30 25181 25181 25181 25181 24021 24021 25181 25181 25181 25181 25181 25181 25181 25181 39361 39361 25181 25181 25181 25181 25181 25181 25031 25031 25181 25181 Dob Dob Icd9_1 Icd9_1 City City NNW NW N 7.5 10 NNE NE ENE 5 WNW 2.5 W 3.1 E WSW SE SSW S SSE ESE SW Wind Frequency Distribution for Hamilton Airport; 1970-90 Normals (indicative of wind patterns over the upper city areas farther away from the escarpment brow) 11 00 11 0 0 1 1 1 1 0 0 0 0 0 0 1 11 11 11 1 25181 25181 25181 25181 25181 25181 25181 25181 25031 25031 25031 25031 25031 25031 25031 25031 25031 25031 25031 25031 25031 25031 25031 25031 25031 25031 81 1903/09/16 81 1903/09/16 88 1897/08/11 88 1897/08/11 79 1906/10/04 79 1906/10/04 66 1918/12/07 66 1918/12/07 47 1938/03/07 47 1938/03/07 23 1984/12/18 23 1984/12/18 6 1985/08/23 6 1985/08/23 19 1965/12/27 19 1965/12/27 51 1933/12/16 51 1933/12/16 74 1910/12/13 74 1910/12/13 85 1900/07/04 85 1900/07/04 72 1913/04/26 72 1913/04/26 72 1913/07/09 72 1913/07/09 410 410 4414 HAMILTON 4414 HAMILTON 4140 HAMILTON 4140 HAMILTON 5538 STONEY CREEK 5538 STONEY CREEK 1911 STONEY CREEK 1911 STONEY CREEK 7455 STONEY CREEK 7455 STONEY CREEK 7707 STONEY CREEK 7707 STONEY CREEK 8151 STONEY CREEK 8151 STONEY CREEK 5712 STONEY CREEK 5712 STONEY CREEK 4029 4029 410 STONEY CREEK 410 STONEY CREEK 1749 1749 4140 STONEY CREEK 4140 STONEY CREEK Evolutions in Spatial Analysis • Spatial analysis of disease not new • Snow’s 1855 study of cholera illustrates many useful concepts • Clustering, spatial interaction, distance decay Interest in Places and Health Dates back to Ancient Greece (Hippocrates 460-377 BC) Extended Conceptual Framework for Spatial Analysis in Epidemiology and Public Health Health Risk Geographic Information Systems SQUAWK! "Who needs a GIS?" SQUAWK! "Who needs a GIS?" GIS: Geographic Information Systems • Spatial database • Relates points (residences), areas (parks) and lines (streets) to one another • Adds layers of data (traffic, topography, crime) • Assesses spatial relationships Adding other layers of risk information on top of land use Why GIS in Spatial Health Analysis? • Seeing the data – important to medically-trained people used to mapping the “body” • Integration of numerous data • Interactivity in the analysis • Ability to use very large data sets: critical when effect sizes are small • Increased speed of delivery • Leads to novel questions and more of them Visual Map of the Human Body Interactive Web Distributed Health GIS Novel Questions Asked in New Journals Featuring GIS and Health • Many other traditional GIS and Geography journals featuring health PubMed Citations on GIS and Health by Year (Search term: 'GIS AND Health') 80 70 60 50 # of articles 40 30 20 10 0 1994 1995 1996 1997 1998 1999 Year 2000 2001 2002 2003 2004 PubMed Citations about GIS and Health Care by Year (Search term: 'GIS AND Health Care') 18 16 14 12 # of articles 10 8 6 4 2 0 1995 1996 1997 1998 1999 Year 2000 2001 2002 2003 2004 Major Exclusions • Dates restricted to last 3 years to capture innovations • Eliminated infectious disease articles from foreign countries (in the 1000s) • Eliminated those with geotechnical risk assessments (landslides, earthquakes, etc) • Eliminated all pure methods or statistical pieces • Eliminated any health care access pieces Classification Schema • Based on the susceptibility, exposure, adaptation, risk from earlier conceptual framework • Then subdivided into major analytical categories: 1. Visualization – mapping otherwise aspatial data (e.g. mortality rates) 2. Exploration – overlay and cluster analysis Abstract Classification (so far about 77 of 110 articles classified) N =77 Exposure Susceptibility Adaptation Risk Total Visualization 2 9 2 7 20 Exploration 3 10 2 6 21 Modeling 6 14 1 15 36 Total 11 33 5 28 77 Trends • Enormous increase in GIS for epidemiology and public health research • Increasing complexity of analysis, with multilevel and Bayesian models becoming more common • Web based distribution growing Types of Spatial Analysis • Stage 1: Visualization of relative risks, pollution, and covariates • Stage 2: Exploration using Boolean overlays • Stage 3: Modelling of spatial dependence and association Visualization • Viewing attribute data in map format • Usually the first step in the spatial analysis (after painful data prep :-) Tabular Data CENSUS MALE ALLTRACT ID# CAUSE CMF 5370001.01 0.298 5370001.02 0.426 5370001.03 0.749 5370002.01 0.591 5370002.02 0.645 5370002.03 5370003.01 0.670 5370003.02 0.586 5370003.03 0.537 5370003.04 0.849 5370004.01 0.973 5370004.02 0.885 5370005.01 0.944 5370005.02 0.909 5370005.03 0.930 5370006.00 0.917 5370007.00 1.015 5370008.00 1.035 95%CONFIDENCE INTERVAL -0.388 0.034 0.552 0.161 0.396 0.423 0.255 0.179 0.657 0.767 0.659 0.767 0.666 0.694 0.755 0.838 0.841 0.984 0.817 0.946 1.021 0.894 0.917 0.916 0.894 1.041 1.179 1.112 1.122 1.152 1.166 1.080 1.192 1.228 What does this table tell you? Can you see patterns or relationships? Figure 1: Male Comparative Mortality Figures (Ages 0-74) ONTARIO # S Hamilton Harbour Toronto Lake Ontario NEWYORK # S Ê Hamilton Ú Lake Erie Buffalo LakeOntario LEGEND Comparative Mortality Figure (CMF) 0.30 - 0.69 0.69 - 0.92 0.92 - 1.19 1.19 - 1.65 1.65 - 2.45 Missing Values 2 3 Kilometers N 1 0 1 Hatching indicates statisticallysignificant CMFs Strengths and Weaknesses • Often suggestive of relationships and hypotheses • Can educate public and officials • Open to abuse with cartographic tricks • Can be misinterpreted as causal Exploration • Searching for relationships with maps meeting certain conditions (usually defined with Boolean or set operations: AND, OR) • E.g., high mortality and high pollution Sulfate (SO4) Air Pollution Levels and Mortality Rates (All Cause) # Y # Seattle Y # Y # Y # Y # Y # Y # Y # Billings Y # Y # Y # Y # Y Salt Lake City # Y # Y ## Y Y # Y # Y # Y # Y San Francisco # Y # Y # Y # Y Denver Los Angeles # Y ## YY # Y # Phoenix Y # Y # Y # Y ## YY # Boston #Y # Y Y # # #Y # Y Y Y# Y # Y # Y # Y ## # YY Y # Y # # Y# Y # Y # # Y ## Y Y Y # Y # # Y # # Y Y Y Y # Y # Y Detroit # Y # # New York YY ## # YY # # Y Y ## Y # YY # Y Y ## # # YY # Y # Y # # Y Y YY Y Gary # # # ## # Y Y YY Y # Y # Y # Y # # Y Y # Y # Y # Y # Washington Y # Y ## YY # ## Y YY Kansas City # Y Steubenville ## # YY # Y # Y Y # Y # Y Charleston # Y # Y # Y Nashville # Y # Y # Y # Y # Y Oklahoma City Memphis # Y # # # Y Y Y # Y # Atlanta Y # Y # Y # Y # Y Minneapolis Johnstown # #Y Y # # # # Dallas # Y # Y # Y # Y # Y # Y # Y # Y # Y # Y Sulfate Cohort Loc'ns (151) High Sulfate Air Pollution Note: Low Mortality rate is not present in High Mortality areas with High Sulfate Air Pollution Medium Mortality Medium Sulfate Air Pollution High Mortality Medium Mortality Low Mortality Low Sulfate Air Pollution High Mortality Medium Mortality Low Mortality Houston # Y # Y # Y # Y New Orleans Tampa # Y # Y # Y # Y 500 0 500 1000 1500 Kilometers 500 0 500 1000 Miles Modelling • Usually tests for spatial dependence in the data or spatial association • Assesses against a random or control pattern • Five spatial processes underlie modelling Spatial Autocorrelation • Tobler’s Law: Everything is connected to everything else, but near things tend to be more connected than distant ones • Deals with correlation of the same variable at different spatial locations • Occurs when values at one location depend on values at nearby locations What Causes Spatial Autocorrelation? • Spatial interaction (e.g., pollution drifting from one jurisdiction to another) • Mis-sized units of analysis that don’t reflect the real world • Diffusion of lifestyles, diseases, etc. Autocorrelation Tests • Global Moran’s I statistic most common • Global tests measure overall tendency of high (or low) values to cluster together • Local tests measure smaller areas in relation to all data points (hot spots) (Local G and Moran most common) Figure 3.12. G Statistic for Sulfate Cohort All Cause Relative Risk Ratio ( Lag Distance = 600 km ) # Y # Y # Y # Y # Y # Y # Y # Y # Y # Y # Y # Y # Y # # #Y Y Y Y # ## ## YY # # Y YY # Y Y # Y # Y ## # YY Y # Y ## # Y Y # Y # Y # ## ## Y YY Y #Y Y # # Y Y Y # Y # Y ## YY # # ## # Y Y YY # # Y YY Y # # ## # #Y # YY Y Y # # # ## Y Y Y YY YY ## # Y #Y Y # Y # Y # # Y Y # Y # Y # Y ## # YY Y ## # YY Y # Y # Y # # # Y Y Y # Y # Y ## YY # Y # Y # Y # Y # Y # Y # # Y Y # Y # Y # Y # Y # Y # Y # Y # Y # Y # Y # Y # Y # Y # Y # Y # Y # Y # Y # Y # Y # Y # Y # Y # Y # Y # Y # Y # Y # Y # Y # Y ## YY # Y # Y # Y ## Y Y # Y # Y # Y # Y # Y # Y # Y # Y # Y # Y # Y # Y ACS Sulfate Cohort Loc'ns (151) Significant ACS Cases All Causes G Statistic < -1.96 -1.96 - -1.15 -1.15 - 1.15 1.15 - 1.96 > 1.96 # Y 500 500 0 0 500 500 1000 1500 Kilometers 1000 Miles Figure 3 Percentage of Low Income - Getis Statistic Industrial Core Niagara Escarpment z-scores -2.56 - -1.09 -1.09 - 1.1 1.1 - 3.28 Significance No Data N Hamilton Harbour HW 3 40 Y. MAIN ST. KING S T. Lake Ontario ST. Q.E. W . ER J AM ES UPP "LIN C" HW Y . 53 0 2 4 6 Kilometers Spatial Autocorrelation Implications • Forms basis of many geostatistical methods • Can suggest new hypotheses • Can render traditional statistical tests invalid because it violates the independent observation assumption Interpolation • Definition: estimating attribute values at unsampled sites within the area covered by existing (point-attribute) observations • Goal: to fit a plausible surface model to depict spatial variation Modelled (Kriged) Sulfate (SO4) Surface # Y # Seattle Y # Y # Y # Y # Y # Billings Y # Y # Y # Y # Y Johnstown Steubenville # Y # Y Salt Lake City # Y # Y ## Y Y # Y # Y # Y # Y San Francisco # Y # Y # Y # Y Denver Los Angeles # Y ## YY # Y # Phoenix Y # Y # Y # Y ## YY # Boston # Minneapolis Y #Y # Y Y # Y # # #Y # Y Y Y# Y # Y # Y # Y ## # YY Y # Y ## # YY Y # Y # Y # ## Y Y # Y # # Y # # Y Y Y Y # Y # Y Detroit # Y # # New York YY # ## # # Y YY Y Y ## # YY # Y Y ## # # YY # Y # Y #Y # Y YY Y Gary # # # ## # Y Y YY Y # Y # Y # Y # # Y Y # # Y Y # Y # Y Washington # ## Y YY # ## Y YY Kansas City # Y ## YY # # # Y Y Y # Y # Y Charleston # Y # Y # Y Nashville # Y # Y # Y # Y # Y Memphis Oklahoma City # Y # # Y Y # Y # Y # Atlanta Y # Y # Y # Y # # Sulfate Cohort Loc'ns (151) Y Y Dallas Sulfate (SO4 ) [ ugm-3 ] # Y # Y # Y # # # # N =151 # Y # Y # Y # Y # Y # Y 30 25 X Houston # Y # Y # Y # Y New Orleans Y Tampa # # Y # Y # Y 20 N 15 10 5 0 4.0 5.5 7.0 8.5 10.0 11.5 13.0 14.5 16.0 17.5 19.0 20.5 22.0 23.5 25.0 500 500 0 500 1000 1500 Kilometers 0 500 1000 Miles < 4.00 5.00 6.00 7.00 8.00 9.00 10.00 11.00 12.00 13.00 14.00 15.00 16.00 17.00 18.00 19.00 20.00 21.00 22.00 > 4 .00 - 4.99 - 5.99 - 6.99 - 7.99 - 8.99 - 9.99 - 10.99 - 11.99 - 12.99 - 13.99 - 14.99 - 15.99 - 16.99 - 17.99 - 18.99 - 19.99 - 20.99 - 21.99 - 22.99 23.00 SO4 [ugm ] -3 Standard (Estimation) Error Associated with SO4 Kriging # Y # Seattle Y # Y # Y # Y # Y # Billings Y # Y # Y # Y # Y # Y # Y Minneapolis # Y # Y # Y # Y # Y # Y # Y # Y # Y # Y # Y # Y Salt Lake City # Y # Y ## Y Y # Y # Y # Y # Y # Y Gary San Francisco # Y # Y # Y # Y Denver # Y Kansas City # Y # Y ## YY # Boston # #Y # Y Y #Y Y # # ## Y Y YY # Y # Y ## YY # Y # Y ## YY # Y # Y # ## # Y Y Y # # # Y Y Y Y # Y Detroit # # New York YY ## YY # # Y # Y ## Y YY # ## Y # YY # Y ## # Y YY Y # # ## # Y Y YY Y # Y # Y # Y # Y # Y # Y Washington # Y ## Charleston YY ## # YY Y # Y Steubenville # # Y Y # Y # Y # Y # Y # # # Pittsburgh Nashville # Y # Y # Y # Y # Y # Y # Y # Y Los Angeles # Y ## YY # Y # Phoenix Y # Y # Y Oklahoma City # Y # Y # Y # Y Memphis # Y # Atlanta Y # Y # Y Dallas # Y # Y # Y # Y # Y # Y # Y # Y # Y Houston # Y # Y # Y # Y New Orleans Tampa # Y # Y # Y # Y # Y Sulfate Cohort Loc'ns (151) SO4 Kriging Standard Error [ ugm-3 ] 0.04 - 0.10 0.11 - 0.20 0.21 - 0.30 0.31 - 0.40 0.41 - 0.50 0.51 - 0.60 500 500 0 500 1000 1500 Kilometers 0 500 1000 Miles Sulfate (SO4) 3-D LOESS Surface 8.0 2.0 -10.0 Sulfate (S O4) [μg m-3 ] Modeled (Kriged) All Relative Risk Surface Sulphate and OverallCause Mortality Surface Y # Y # Y # Y # Spokane Y # Y # Great Falls Y # Y # Billings Y # Y # Y # Y # Y # Reno Y # Y # Y # Y # Y # Y # Y # Fresno Y # YY ## Y # Y # Y # Y # Y # Y # Y # Y # YY ## YY ## YY ## Y YY # ## YY ## Y # Y # Y # Y # YY ## Y # Y # YY ## Y # YY ## Y # YY ## Y Y # Y# Y # Y # Y # Y # Y Waterloo# # # # YY ## YY ## Y # Y # Y # YY ## YY Y # Y YY Y # ## # NY, OH, PA, and WV YY ## Y # Y # Y Y YY Y # # ## # Y # # Y Localities in the 1.2 < Y # Y # Y # Y # Y # Y# # Y Relative Risk Cluster: Y # YY ## YY ## Y # Y # Buffalo YY ## Y # Y # Y # Y # Y # Erie YY ## Y # Youngstown Y # Y # Y # Cleveland Y # Y # Y # Akron Y # Y # Y # Y # Y # Y # Canton Y # Y # Sharon Y # Wichita Falls, TX Y # Y # Y # Steubenville Y # Pittsburgh Y # Y # YY ## Johnstown Y # Y # Y # Altoona Y # Y # Y # Y # Y # Y # Y # Y # Y # Sulfate Monitoring Loc'ns (151) Conterminous 48 States Relative Risk (All Causes) 0.924 - 1.000 1.001 - 1.100 1.101 - 1.200 1.201 - 1.283 State College Charleston Huntington 1500 500 400 0 Kilometers 0 Miles 500 400 1000 800 1200 LOG Mortality Relative Risk 3-D LOESS Surface 0.100 0.000 -0.300 LOG Mort ality Rela tive Risk Visualization Via Interpolation • Often useful to model surfaces to allow for visualization • Can be used for assigning exposures • Must always question “scale” of analysis Point Patterns Goals: 1. To assess point density against some random or control distribution, or 2. To assess whether points with like attributes are clustered together (e.g., infectious disease W. Nile Virus) • • Will not control for age Major shortcoming Figure 6: Kernel Estimate of Mortality in Hamilton, ON 1985-94 Figure 7: Female Asthma Symptoms and High TSP Pollution Have you had wheezing or whistling in your chest at any time in the last 12 months? An Average of Total Suspended Particulate (TSP) Levels in Hamilton from 1985 - 1994 N Hamilton Harbour Has your child ever had asthma? Females 13-14 Lake Ontario y. 40 3 ER J AM ES ST. H W M AI NS T. KIN GS T. QE W UPP "L IN HW Y. 5 3 C" Industrial Core Niagara Escarpment High Significance 0.9 - 0.95 0.95 - 1 Q6: rates for female age 13-14 70 - 130 131 - 170 171 - 210 211 - 250 251 - 360 2 0 2 4 Kilometers Spatial Association • May want to predict mortality from colocated variables such as pollution and income • Autocorrelation in the residuals inflates significance of predictor variables What Autocorrelated Residuals Mean • Systematic mis-measurement in the dependent variable • Significant variables may be missing Data Dependent variable (N = 151): • Relative risk ratios obtained from multilevel models, adjusted city level effects for individual risk factors (e.g., smoking) Independent variables (N = 151): • Mean SO4 variable • Demographic, socioeconomic covariates • Other environmental variables/co-pollutants Figure 2. Spatial Modeling Framework Individual Tobacco Variables Other Individual Variables & Obesity Ecologic Covariates Social Environment Health Determinants Physical Environment Health Care System Ecologic Covariates Income Income Disparity Poverty Employment Health Determinants Physical Environment Percent White Percent Black Education Social Environment Population Change Health Care System Ecologic Covariates Income Income Disparity Poverty Employment Health Determinants Physical Environment Temperature (variation) Humidity (variation) NO2 Altitude Water Hardness CO Health Care System O3 SO2 Percent White Percent Black Education Social Environment Population Change Ecologic Covariates Income Income Disparity Poverty Employment Health Determinants Physical Environment Temperature (variation) Humidity (variation) NO2 Altitude Water Hardness CO Health Care System Physicians Hospital Beds O3 SO2 Percent White Percent Black Education Social Environment Population Change LOG Mortality Relative Risk 3-D LOESS Surface 0.100 0.000 -0.300 LOG Mort ality Rela tive Risk Modelled (Kriged) Sulfate (SO4) Surface # Y # Seattle Y # Y # Y # Y # Y # Billings Y # Y # Y # Y # Y Johnstown Steubenville # Y # Y Salt Lake City # Y # Y ## Y Y # Y # Y # Y # Y San Francisco # Y # Y # Y # Y Denver Los Angeles # Y ## YY # Y # Phoenix Y # Y # Y # Y ## YY # Boston # Minneapolis Y #Y # Y Y # Y # # #Y # Y Y Y# Y # Y # Y # Y ## # YY Y # Y ## # YY Y # Y # Y # ## Y Y # Y # # Y # # Y Y Y Y # Y # Y Detroit # Y # # New York YY # ## # # Y YY Y Y ## # YY # Y Y ## # # YY # Y # Y #Y # Y YY Y Gary # # # ## # Y Y YY Y # Y # Y # Y # # Y Y # # Y Y # Y # Y Washington # ## Y YY # ## Y YY Kansas City # Y ## YY # # # Y Y Y # Y # Y Charleston # Y # Y # Y Nashville # Y # Y # Y # Y # Y Memphis Oklahoma City # Y # # Y Y # Y # Y # Atlanta Y # Y # Y # Y # # Sulfate Cohort Loc'ns (151) Y Y Dallas Sulfate (SO4 ) [ ugm-3 ] # Y # Y # Y # # # # N =151 # Y # Y # Y # Y # Y # Y 30 25 X Houston # Y # Y # Y # Y New Orleans Y Tampa # # Y # Y # Y 20 N 15 10 5 0 4.0 5.5 7.0 8.5 10.0 11.5 13.0 14.5 16.0 17.5 19.0 20.5 22.0 23.5 25.0 500 500 0 500 1000 1500 Kilometers 0 500 1000 Miles < 4.00 5.00 6.00 7.00 8.00 9.00 10.00 11.00 12.00 13.00 14.00 15.00 16.00 17.00 18.00 19.00 20.00 21.00 22.00 > 4 .00 - 4.99 - 5.99 - 6.99 - 7.99 - 8.99 - 9.99 - 10.99 - 11.99 - 12.99 - 13.99 - 14.99 - 15.99 - 16.99 - 17.99 - 18.99 - 19.99 - 20.99 - 21.99 - 22.99 23.00 SO4 [ugm ] -3 S u lfa te E ffe c t o n A ll C a u s e M o r ta lity 1 .5 1 .4 1 .3 Relative Risk 1 .2 1 .1 1 .0 In d . O b s . Cox 2 -S ta g e C lu s te r e d I.C . by C ity N .N . - 2 -S ta g e R .A . F .M . F .B .S . C o m b in e d S u lfa te a n d S u lfu r D io x id e E ffe c t 1 .5 1 .4 1 .3 Relative Risk 1 .2 1 .1 1 .0 0 .9 SO 4 Cox SO 2 C lu s te r e d I.C . by C ity - 2 -S ta g e R .A . N .N . F .M . F .B .S . Future Prospects • Increasing use as technology improves and health professionals discover that space and place matter • Software has improved, but still is difficult • Most advanced analysis must be performed with statistical software, rather than GIS • Data access and quality a major problem • Web-based systems show promise Query Tool • Allows queries on a specific area • Shows associated data attributes • Site programmer chooses what data is on page References • www.healtheffects.org • Downloadable report illustrating many of these techniques • THANK YOU FOR YOUR TIME TODAY! Acknowledgements • Toxic Substances Research Initiative, a collaborative program of Health Canada and Environment Canada • Health Effects Institute • Social Sciences and Humanities Research Council • National Centres for Excellence: Geomatics for Informed Decisions ...
View Full Document

This note was uploaded on 09/29/2009 for the course NUTRI SCI 10 taught by Professor Amy during the Spring '08 term at University of California, Berkeley.

Ask a homework question - tutors are online