Quantifying Neighborhood-Level Social Determinants of Health in the Continental United States: A Multivariate Small-Area Analysis

Authors: Marynia Kolak*, University of Chicago
Topics: Geography and Urban Health, Quantitative Methods, Spatial Analysis & Modeling
Keywords: social determinants of health, SDOH, public health, health disparities, healthy inequalities
Session Type: Paper
Day: 4/4/2019
Start / End Time: 3:05 PM / 4:45 PM
Room: Forum Room, Omni, West
Presentation File: No File Uploaded


The impacts of social determinants of health (SDOH) have increasingly dominated public health discussions in the United States as population health outcomes have not kept pace with other developed nations, despite spending more per person in medical service expenses (Bravemen et al 2011, Murray et al 2013, McGinnis et al 2002). Despite a complex presentation of SDOH, they are often modeled using singular variables, or as one-dimensional indices. We develop a multidimensional SDOH typology of the continental US at tract level using a principal component analysis and spatially-sensitive K-means clustering. Then, we predict neighborhood-level mortality rates using SDOH principal components in the City of Chicago in 2014, and controlling for violent crime. Four identified SDOH components explain over 75.6% of the variance for all census tracts in the continental United States for the time period, and are characterized as the following SDOH indices: socioeconomic disadvantage, mobility-related isolation, urban housing and transportation, and family social cohesion. These can be represented as 7 multidimensional SDOH typologies across geographic space. An especially vulnerable typology of “extreme poverty” makes up 9.4% of all continental U.S. tracts and includes areas of known public health crises, including Flint, Michigan; Butler County, Alabama; and South and West Sides of Chicago. There is an extremely strong association between SDOH indices and age-adjusted mortality in Chicago, even after accounting for spatial autocorrelation and violent crime rates (R2=0.63, p<0.001).

Abstract Information

This abstract is already part of a session. View the session here.

To access contact information login