LAUSR

Study Objectives: The spatial distribution of COVID-19 remains to be described, though there is growing evidence of an increased burden among already disadvantaged populations and neighborhoods Understanding the pattern of population risk is critically important for health systems and policy makers responding to the pandemic Our aims were: 1) to describe the association between neighborhood factors and incident cases of COVID-19;and 2) to examine the changes in cases over time We hypothesized that there would be an association between disadvantaged neighborhoods and case clusters Methods: We analyzed data from patients presenting to a large health care system in Boston, MA from 2/5/20 to 5/4/20 Patient mailing addresses were geocoded to census tracts within a 20-mile radius of Boston COVID-19 incidence per census tract was calculated using Empirical Bayes smoothed rates to adjust for small area estimation Clustering of cases at the census tract level were assessed using local Moran’s I, accounting for multiple comparisons Quantile local spatial autocorrelation was used to determine the spatial association between neighborhood demographic and disadvantage measures (from the American Community Survey) and census tracts with high incidence of COVID-19 Poisson regression models were used to assess the independent associations between neighborhood factors and COVID-19 Finally, we mapped the distribution of cases in the study area over time Results: As of May 4, 2020, there were 9,898 patients in the study area who had been treated in the health care system for COVID-19 The overall crude incidence was 31 8 cases per 10,000 population;adjusted incidence per census tract ranged from 2 3 to 405 1 per 10,000 population Two case clusters were identified in the Chelsea/Everett and Lynn areas (p=0 007) We found statistically significant co-location of the top quintile of cases with several neighborhood factors (all p<0 05): % of population Hispanic (n=72 census tracts), black (n=36), uninsured (n=33), receiving Supplemental Nutrition Assistance Program (SNAP) benefits (n=39), and living in poverty (n=23) In the adjusted model, factors associated with increased incidence of COVID-19 were a higher proportion of Hispanic population (aIRR 1 24, 95% CI 1 21-1 28) and households receiving SNAP benefits (aIRR 1 08, 95% CI 1 02-1 13) The distribution of cases varied over time, but with persistently high incidence in communities north of Boston Conclusion: We found a significant association between neighborhood disadvantage measures and high incidence rates of COVID-19 Limitations include case ascertainment challenges due to access to testing and possible selection bias from use of a single health care system These results suggest that policy makers should consider health inequities as they respond to the ongoing pandemic and plan for future health needs [Formula presented]