Infectious Disease
Does county tell the whole story? Assessing within-county heterogeneity in COVID-19 burden across levels of urbanicity and rurality. John T. Kubale* John Kubale Robert Melendez Andrew Hoover Paul Schulz Philippa Clarke Sonia T. Hegde Fan Bu Brady T. West Grace Noppert
Introduction: Within-county heterogeneity in infectious disease burden is well-documented in metropolitan areas (e.g., neighborhood hotspots during outbreaks). However, in micropolitan and rural areas, infectious disease burden is often assumed to be homogenous below the county level, and this may not be accurate. Assessing the level of within-county heterogeneity in infectious disease burden in non-metropolitan areas has key implications for forecasting and ensuring that appropriate policy measures are implemented.
Methods: Using census tract-level data for COVID-19 monthly case counts in Wisconsin and New Mexico (2020-2022), we fit two generalized additive mixed models with a generalized Poisson distribution. In model 1, a single random effect for county was included, while in model 2, an additional random effect for census tract was also included. Calendar month and number of cases in the prior month were included as predictors in all models using cubic splines. Using Rural Urban Commuting Area (RUCA) codes, communities were classified as metropolitan (RUCA 1-3), micropolitan (4-6), or rural (7-10), with models fit to each class. The importance of within-county heterogeneity in COVID-19 burden across levels of rurality was examined by comparing model 1 to model 2 using likelihood ratio tests.
Results: In Wisconsin, we observed significant within-county heterogeneity in disease burden within metropolitan areas (p<0.0001), but not in micropolitan or rural areas. In New Mexico, we observed significant within-county heterogeneity in both metropolitan and rural areas (p<0.0001), but not micropolitan areas.
Conclusions: Within-county heterogeneity does exist in non-metropolitan areas, but it is not consistent across states. Better understanding of the drivers of this heterogeneity is needed to recognize the importance of more spatially granular data for public health decision-making.