Data-Driven Development of a Small-Area COVID-19 Vulnerability Index for the United States

As the COVID-19 pandemic continues to surge in the United States, it has become clear that infection risk is higher in certain populations, particularly socially and economically marginalized groups. Social risk factors, together with other demographic and community characteristics, may reveal local variations and inequities in COVID risk that could be useful for targeting testing and interventions. Yet to date, rates of infection and estimations of COVID risk are typically reported at the county and state level. In this study we develop a small area vulnerability index based on publicly-available sociodemographic data and 668,428 COVID diagnoses reported in 4,803 ZIP codes in the United States (15% of all ZIP codes). The outcome was COVID-19 diagnosis rates per 100,000 people by ZIP code. Explanatory variables included sociodemographic characteristics obtained from the 2018 American Community Survey 5-year estimates. Bayesian multivariable techniques were used to capture complexities of spatial data and spatial autocorrelation and identify individual risk factors and derive their respective weights in the index. COVID-19 diagnosis rates varied from zero to 29,508 per 100,000 people. The final vulnerability index showed that higher population density, higher percentage of noninsured, nonwhite race and Hispanic ethnicity were positively associated with COVID-19 diagnosis rates. Our findings indicate disproportionate risk of COVID-19 infection among some populations and validate and expand understanding of these inequities, integrating several risk factors into a summary index reflecting composite vulnerability to infection. This index can provide local public health and other agencies with evidence-based metrics of COVID risk at a geographical scale that has not been previously available to most US communities.