CURF Spring 2021 Research Fair

Associations Between Neighborhood-Level Characteristics and COVID-19 Incidence and Mortality in Southeastern Pennsylvania

The COVID-19 pandemic has disproportionately impacted racial/ethnic minority and socioeconomically disadvantaged groups in the United States including at-risk populations within Southeastern Pennsylvania. We sought to determine whether neighborhood-level health, demographic, and socioeconomic characteristics were associated with COVID-19 incidence and mortality at the zip code and municipality level, thereby establishing whether neighborhood-level disparities mirror individual-level ones. 

We collected data from several sources, aggregating by county subdivision geographical units. For Philadelphia those were zip codes, and for Bucks, Chester, Delaware, and Montgomery Counties those were municipalities. Cumulative data on COVID-19 cases and deaths were obtained from the public health departments of those counties, and data for individual long-term care facilities (LTCFs) were obtained from the Pennsylvania Department of Health. For corresponding geographic areas, demographic and socioeconomic status variables were obtained from the 2015-2019 American Community Survey 5-year estimates, and data on the health status and behaviors of local residents were obtained from the Southeast Pennsylvania Household Health Survey, conducted in 2012, 2015, and 2018 by the Public Health Management Corporation. Derived from the ACS 5-year estimates were income disparity and Area Deprivation Index, a validated score of socioeconomic deprivation.

For feature selection, we created univariable quasi-Poisson models with COVID-19 case and death counts with offsets for population counts to identify individual factors that were significantly associated having a p-value less than 0.01, then selected single variables among those with highly collinear terms in each model with collinearity established based on having variance inflation factors greater than 3. Including only the individual predictors that passed feature selection, we created two multivariable quasi-Poisson regression models with offsets for population counts to determine whether neighborhood-level variables were each associated with COVID-19 incidence and mortality. 

Among 276 zip codes and municipalities with complete data, the COVID-19 cumulative incidence through March 12, 2021 ranged from 112.7 to 1206.4 per 10,000 residents, and the COVID-19 mortality rate ranged from 0 to 7.6 per 10,000 residents. Cumulative incidence aggregated by county ranged from 5.25% for Bucks to 7.59% for Philadelphia, and cumulative mortality ranged from 0.14% for Chester to 0.23% for Delaware. 

Of 23 variables included in the final model for COVID-19 incidence, we found that the proportion of those aged 65 or older increased risk, while median gross rent and the proportion of individuals eating 3 or more servings of fruits or vegetables daily decreased risk. Additionally, the proportion of those aged 65 or older was also associated with increased risk of COVID-19 mortality out of 16 neighborhood-level variables included in that final model. 

In conclusion, neighborhood-level data can complement individual-level data, which is not always readily available, to help identify specific needs of vulnerable populations and inform policies to address health disparities related to COVID-19.

PRESENTED BY
Other
College of Arts & Sciences 2023
Advised By
Blanca E. Himes, PhD
Associate Professor of Informatics
Join Alexandra for a virtual discussion
PRESENTED BY
Other
NIEHS R25 ES021649
College of Arts & Sciences 2023
Advised By
Blanca E. Himes, PhD
Associate Professor of Informatics

Comments