An Analysis of COVID -19 Social Vulnerability and Racial Disparity in the Mid -South: Model Prediction and Machine Learning

Overview

Increasing research shows that COVID -19 has disproportionately impacted racial minorities and people with low socioeconomic status (SES). By using statistical methods and machine learning data analysis, my study predicted how the interactions between community -level social vulnerability and individu al-level factors affected COVID-19 infection, hospitalization, and in -hospital mortality in the Mid -South’s tri-states (Arkansas, Mississippi, and Tennessee). Community -level Social Vulnerability Index (SVI) was derived from the Centers for Disease Control and Prevention (CDC)’s 2020 SVI dataset and matched with patients’ individual-level data by zip codes. Patients’ demographics, COVID -19 testing results, hospitalization, and mortality statuses were retrospectively extracted from an electronic medical reco rd system. Risk-adjusted multivariate logistic regression models were used to assess the associations between risk factors and COVID-19 outcomes. Bootstrapping machine learning methods were used to improve model predictions. My findings showed that people living in communities with vulnerable household composition and a high percentage of minority residents were more likely to get infected. Hispanics and Blacks living in communities with high SVIs also suffered the most from COVID -19 infection. No statistically significant differences were found from social vulnerability in predicting hospitalization, as age and race were the dominant predictors. Moreover, no statistically significant effects of race or social vulnerability on mortality were found. Bootstrapping-based machine learning offered similar results. My findings showed the racial disparity and community -level social determinants of COVID -19 infection. My study contributed to the existing literature by showing the interactive relationships between community-level social vulnerability and race affecting COVID-19 infection in the understudied Mid-South.