Walkability vs. Several Health Diagnoses for Klamath Falls, OR John Ritter, Ph.D. Geomatics Dept, Oregon Tech Stephanie Van Dyke, MD, MPH Medical Director, Sky Lakes Wellness Center Katherine Pope, RN, MPH Program Director, Sky Lakes Wellness Center
Purpose, Objectives, and Background Purpose: - Be a part of the movement to increase the health and wellness in Klamath County Objectives: - Map disease incidents for Klamath Falls, OR - Determine if disease incidents can be related to local walkability and/or demographics Klamath County / Klamath Falls, OR - Use spatial analyses to help direct resources for intervention efforts
Purpose, Objectives, and Background Background - The health ranking for each Oregon county was determined using composite scores in the following categories: - Length of life - Quality of life - Health behaviors - Clinical care - Social and economic factors - Physical environment - Using data for 2016, Klamath County ranked 35 th out of 36 ranked counties (http://www.countyhealthrankings.org)
Purpose, Objectives, and Background Background - Klamath awarded Blue Zones Project Could serve as model for the rest of Oregon (Herald & News, Mar. 26, 2015) - Klamath Falls has been chosen to lead the state in a new community health initiative. The city, and surrounding areas in the 97601 and 97603 area codes, have been chosen to participate in the state s first Blue Zones Project. (ibid) - The project is a well-being improvement initiative designed to make making healthy choices easier. It focuses on five categories of well-being: purpose, social, financial, community and physical. (ibid) - The initiative is based on studies of communities around the world with the highest number of residents who lived to be more than 100 years old. (ibid)
Purpose, Objectives, and Background Background - Sky Lakes Medical Center (SLMC) and Cambia Health Solutions have assumed a major role in advancing efforts towards the local Blue Zones initiative by creating the Sky Lakes Wellness Center (SLWC) (liveyoung.skylakes.org)
Methodology Walkability - An objective of this study was to determine if local walkability could explain a portion of the variance in observed disease incidence - A walkability model was constructed using the esri ModelBuilder utility based on several local parameters
Methodology Variables used to determine walkability score: Parameter Local slope (i.e. steepness of road) Zoning Service areas Population density Road density Housing density Speed limit on roadway Basis for a High Score Low slope (i.e. flat terrain) Medium density residential, neighborhood commercial, general commercial, or public facility zoning Close proximity to public points of interest High population density High road density High housing density Low speed limit
Methodology Disease Incidents - Working in tandem with personnel from SLMC and SLWC, approximately 60,000 (de-identified) patient records for 2012 were obtained for the following disease categories: - Diabetes (DM) - Hypertension (Htn) - Obesity (OB) - Cholesterol (HL) - Stroke (ST) - Heart Disease (HD) - Smoking (SM) - Data were also obtained on the total cost of care
Methodology Disease Incidence Mapping - Addresses from patient records were geocoded and aggregated to 2010 US Census Block Groups located within the study area - Incident density maps were created for each disease category - A web app was created to show a comparison of the spatial distribution of walkability vs. obesity, diabetes, high blood pressure, and cholesterol over the study area: Walkability vs. Disease Incident Density
Methodology Other Data Sources - 2015 esri Consumer Expenditure indices - Fresh vegetables - Cable & Satellite TV - Cigarettes - Beef - Seafood - Travel - Religious contributions - Many more - 2010 US Census Data - Population density - Race - Median age
Methodology Other Data Sources (cont d) - 2009-2013 American Community Survey Data - Language spoken at home - Income to poverty ratio - Per capita income - Health care insurance provider - Highest attained education for 75% of households - Average household income - Percent receiving public assistance/food stamps - Derived Data - Walkability
Methodology Regression Analysis - A regression analysis was performed using the esri Ordinary Least Squares (OLS) utility to determine if a properly specified model could be found to describe the variance in disease incidence data. y = β 0 + β 1 x 1 + β 2 x 2 + + ε (Observed value = Predicted value + Residual)
Methodology Regression Analysis - Six key factors for the formation of a properly specified model 1. The coefficient must support your hypothesis - Each explanatory variable must exhibit expected behavior 2. Each explanatory variable should be statistically significant (max coefficient p-value: 0.01) 3. The residual should not be clustered in space - Test for spatial autocorrelation using Moran s I - Min Spatial Autocorrelation p-value: 0.1
Methodology Regression Analysis - Six key factors for the formation of a properly specified model 4. Residuals must be normally distributed in value - Min Jarque Bera p-value: 0.1 5. Each variable should describe a different aspect of the variation (i.e., Principle Component Analysis) - Max Variance Inflation Factor (VIF): 7.5 6. Evaluate the model performance - Appropriate adjusted R-Squared values
Results Properly Specified Models (PSM) were found for - Cholesterol (HL) - Hypertension (Htn) - Diabetes (DM) Data from the Consumer Expenditure dataset were not incorporated into the analysis due to problems with multicollinearity between variables
Ordinary Least Squares (OLS)
Results Heart Disease (HD) - Variable significance - HL, Walk, AvgHHInc passed - Spatial Autocorrelation passed - Normally Distributed in value passed - VIF passed - Adjusted R-Sq: 0.43 - Map of locally varying R-sq
Results Cholesterol (HL) - Variable significance - Htn, HD, Hispanic Minority passed - Spatial Autocorrelation passed - Normally Distributed in value passed - VIF passed - Adjusted R-Sq: 0.60 - Map of locally varying R-sq
Results Diabetes (DM) - Variable significance - OB, Htn, HD passed - Spatial Autocorrelation passed - Normally Distributed in value passed - VIF passed - Adjusted R-Sq: 0.62 - Map of locally varying R-sq
Results Hypertension (Htn) - Variable significance - DM, HL passed - Spatial Autocorrelation passed - Normally Distributed in value passed - VIF passed - Adjusted R-Sq: 0.67 - Map of locally varying R-sq
Hot Spot Analysis Getis-Ord Gi*
Results Hot Spot Analysis - Cholesterol - Hypertension - Obesity - Diabetes - Heart Disease - A side-by-side comparison of hot spots for Cholesterol, Hypertension, Obesity, and Diabetes
Cluster & Outlier Analysis Anselin Local Morans I
Results Cluster & Outlier Analysis - Cholesterol - Hypertension - Obesity - Diabetes - Heart Disease - A side-by-side comparison of cluster & outliers for Cholesterol, Hypertension, Obesity, and Diabetes
Grouping Analysis
Results Grouping Analysis - A grouping analysis was done on the following five combinations of variables. The values for the pertinent variables for each census block group were placed into one of three groups such that values for features within a group are most similar to each other. - Diabetes, Obesity, Hypertension - Diabetes, Obesity, Cholesterol - Diabetes, Cholesterol, Hypertension - Heart Disease, Cholesterol, Hypertension * Heart Disease, Diabetes, Obesity
Results Heart Disease, Diabetes, and Obesity Green areas are highest in OB and lowest in HD Red areas are highest in DM and HD Blue areas are uniformly low in OB, DM, and HD
Conclusions Maps were produced that show the incidence of hypertension, diabetes, heart disease, cholesterol, and obesity in Klamath Falls, OR Properly specified regression models were obtained for: Dependent Variable Explanatory Variables R-Sq Hypertension Diabetes, Cholesterol 0.67 Cholesterol Hypertension, Heart Disease, 0.60 Pct Hispanic Minority Diabetes Hypertension, Cholesterol, 0.62 Smoking
Conclusions The results of this analysis show that remediation efforts should be specifically directed towards the extended downtown and south-central suburbs Grouping analysis shows OB, DM, Htn, and HL to be high in the extended downtown area and DM, Htn, OB, HL, and HD to be high in the south-central area Based on these results a protected bike lane has been proposed that would bisect the extended downtown area
Results Obesity (OB) - Variable significance - SM passed - Spatial Autocorrelation passed - Normally Distributed in value passed - VIF passed - Adjusted R-Sq: 0.32 - Map of locally varying R-sq
Results Diabetes, Obesity, and Hypertension Red areas have higher DM & Htn Blue area have lower values for OB, DM, and Htn
Results Diabetes, Obesity, and Cholesterol OB is highest in red areas and lowest in blue areas
Results Diabetes, Cholesterol, and Hypertension Green areas are highest in DM, Htn, and HL Red areas are uniformly lowest in DM, Htn, and HL
Results Heart Disease, Cholesterol, and Hypertension Red areas are high in Htn and HL Blue areas are low in Htn, HL, and HD
Conclusions Downtown Area - Hot spot and clusters of high values for OB - Generally a cold spot for HL, but contains a high outlier of HL - High outlier for Htn Extended Downtown Area - Cluster of high values for OB - Generally a cluster of low values for HL - The northern area has a high outlier for DM and a cool-cold spot for HD
Conclusions South-Central Suburbs - Clusters & hot spots of high values of HL, DM, HD, Htn Eastern Suburbs - High outlier for OB in the area near the intersection of Crater Lake Pkwy and Shasta Way - Clusters & hot spots of high values of HL