Exploring determinants of vaccine uptake in individuals aged 65+ using Multi-scale Geographically Weighted Regression (MGWR)

Authors

  • Subaita Mahmud Department of Geography and Geoinformation Science, George Mason University, Fairfax, VA
  • Taylor Anderson Department of Geography and Geoinformation Science, George Mason University, Fairfax, VA

Abstract

Traditional regression methods often assume relationships between variables remain constant across geographic space, but this is not always the case. Understanding determinants of vaccine uptake using spatial data requires advanced methodologies to capture spatial variations in these relationships. Therefore, this study investigates using spatial regression approaches such as Multi-scale Geographically Weighted Regression (MGWR). We compare global linear regression models and MGWR for identifying determinants of COVID-19 vaccine uptake in populations ages 65+ in 2021-2022. The linear regression shows that counties with more Democratic voters, primary care physicians, overcrowding, and households with more broadband access have more vaccine uptake, explaining 27-35% of the variation in the data. The residuals are spatially clustered meaning that the global linear regression is unable to predict vaccine uptake in certain locations, motivating the use of MGWR. While the linear regression assumes that these variables are significant across the US, the MGWR identified specific counties where these variables have impact and where they are not significant for vaccine uptake. The MGWR model explains 61% of the variation in the vaccine uptake, and the model is the best fit for counties in the West. This spatial regression approach supports geographically-targeted interventions aimed at increasing age 65+ vaccination.

Published

2024-10-13

Issue

Section

College of Science: Department of Geography and Geoinformation Science