Regression of Years of Potential Life Lost using Computed Mortality-type Percentages and OLS with a Spatial Lag Model

Dublin Core

Title

Regression of Years of Potential Life Lost using Computed Mortality-type Percentages and OLS with a Spatial Lag Model

Subject

Spatial regression analysis of the top five causes of death from 2013-2017 and their effect on YPLL in the Four Corners Region

Description

In order to better understand what covariates impact YPLL the most in the Four Corners region, a 5-year age-adjusted summary of deaths for all counties in the Four Corners Region was downloaded from the CDC's WONDER database. The top 5 causes of death with > 90% Counties reporting values were selected for OLS analysis. Ratios were computed for the number of deaths per category divided by the total number of deaths reported from 2013 - 2017 for all counties.
OLS was performed using ArcGIS Pro 2.6.1 with YPLL as the response variable and the %'s for Deaths due to Neoplasms, Heart Disease, Respiratory Disease, Digestive Tract Disease, and External/Accidental Causes as the chosen explanatory variables. All other causes of death were excluded due to their relative rarity. All counties with suppressed data were excluded from the analysis.
OLS diagnostics revealed a high degree of spatial stationarity. To remedy this, the shapefile was exported to GeoDa where OLS with a spatial lag was performed using a 2nd Order Queen Neighbors weight matrix. The residuals were added to the shapefile and the shapefile was reimported into ArcPro for final visualization and spatial autocorrelation analysis.
Moran's I was run and a value of 0.12 was found to be significant at p < 0.005. This indicates that there is slight clustering in how the residuals relate to each other over space. A computed R-squared value of 0.54 was also found after running the second analysis. Deaths due to neoplasms and digestive tract diseases were significant at p < 0.000. Deaths due to respiratory-related diseases were found to be significant at p < 0.01. These three statistically significant factors were then visualized in ArcPro.
It was expected that heart disease would be a major contributor to YPLL, but this analysis suggests otherwise and should be considered for further research. It is hypothesized that cirrhosis of the liver is responsible for the vast majority of digestive related diseases, but this is impossible to tell with the scheduled mortality data in its aggregated form.
Taken together, these findings suggest that premature mortality is affected in the Four Corner States primarily by cancers and diseases that affect the digestive system and the respiratory system. It is unknown whether lifestyle choices such as heavy drinking and smoking are acting as cofounder variables for this analysis of YPLL. These questions could provide stepping points for future additions to the Four Corners Health Atlas

Creator

David Gunther

Source

CDC WONDER – Underlying Cause of Death Data from 2013 - 2017

Country Shapefiles - ESRI Living Atlas
U.S. State Shapefiles - ESRI Living Atlas
U.S. County Shapefiles - ESRI Living Atlas
U.S. State Capitals - ESRI Living Atlas

Publisher

Website maintained by Chantel Sloan, Associate Professor in the Public Health Department at Brigham Young University.

Date

Underlying Cause of Death Data taken from a 2013 – 2017 analysis of CDC WONDER data.

Contributor

David Gunther

Rights

All shapefiles are ESRI proprietary data

Format

[PDF], [PNG]

Language

[English]

Type

Map file generated using ArcGIS Pro 2.6.1; spatial autocorrelation report generated with ArcGIS Pro 2.6.1; Text output of GeoDa results

Coverage

This is a series of 1:6,500,00 maps of the Four Corners region centered on Arizona, Colorado, New Mexico, and Utah.

Files

OLS_CauseOfDeath_Map.pdf
CoD_Spatial Autocorrelation_Report.pdf
GeoDa_CoD_Output.PNG

Reference

Regression of Years of Potential Life Lost using Computed Mortality-type Percentages and OLS with a Spatial Lag Model, David Gunther, Website maintained by Chantel Sloan, Associate Professor in the Public Health Department at Brigham Young University., Underlying Cause of Death Data taken from a 2013 – 2017 analysis of CDC WONDER data.