Impacts of Broadband and the Social Determinants of Health on COVID-19 Mortality and Infant Death Rates in EPA-Designated Technical Assistance Communities Open Access

Ravichandran, Vivek (Spring 2021)

Permanent URL:


Purpose: This study aims to externally validate the findings from a previous broadband project that assessed the lack of access to broadband/telemedicine and its effect on health disparities throughout rural Georgia (Ravichandran, 2020). The Georgia Broadband Project found poor broadband coverage linked, at a statistically significant level, to increased COVID-19 Deaths Per 100K Residents (COVID DR), as well as higher Infant Deaths Per 1K Live Births (IMR) (Ravichandran, 2020).

Methods: A total of 41 counties were chosen from the EPA’s Office of Community Revitalization Technical Assistance (OCR TA) website, specifically from the Healthy Places for Healthy People and Cool & Connected programs (EPA, 2021). Due to the lack of granular reporting on COVID DR, COVID-19 Case Fatality Rate (CFR), and IMR, analyses were done at the county level, which presented a layer of unavoidable ecological bias. Each county was matched 1:2 with other counties in the same state with similar populations, to account for population density and mobility. This resulted in a total N = 123 counties to enter the analysis. Due to the non-normal nature of the aforementioned health outcomes of interest, spearman correlation assessments, two-sample-t-test hypothesis testing, and Poisson regression modeling were performed via Statistical Analysis Software (SAS) programming.

Results: The spearman correlations between broadband and the health outcomes were not strong nor significant. Based on the two-sample-t-tests, there was no statistical significance observed between OCR TA and non-OCR TA communities for the COVID DR and CFR (p-value = 0.9474 and 0.6870, respectively), though the former appears to possess significantly higher IMR (p-value = 0.0142). However, our secondary hypothesis of broadband being protective against the COVID health outcomes holds true based on the combination of the Two-Sample-T-Test (p-value = 0.0114) and the Simple Linear Regression Model (negative beta coefficient) with broadband as the predictor; however, it should be noted that the Two-Sample-T-Test did not portray significance for CFR or IMR (p-value = 0.3228 and 0.3550, respectively), though the point estimate displayed protection. Simple Poisson models for COVID DR exhibited a synergistic, protective relationship with broadband access, median household income, numeracy score, and literacy scores. An antagonist relationship was seen with the other predictors. All but EPA Designation (p-value = 0.6336) did not contain the null value for their respective beta coefficients. In terms of Simple Poisson models for COVID CFR, all of the beta coefficients contained the null (p-values ranged from 0.6875 to 0.9713) so judgements on synergistic vs. antagonistic qualities are not significant. Lastly, Simple Poisson models for IMR, revealed that median household income, urbanity, numeracy score, literacy score, and % foreign born were synergistic and protective, while poverty rate, minority presence, pandemic vulnerability score, EPA designation, SNAP, and % unemployed were all antagonistic.

Conclusions: The rise of COVID DR in broadband “deserts” compared to highly served areas further emphasizes the exigence for localized as well as federal broadband expansion initiatives. Due to rapid advances in technology, increased demand, and frequent adjustments to the Federal Communication Commission’s (FCC) definition of broadband, the latter should be analogous to the United States Rural Electrification Act (REA) of 1936, which expanded power to half of rural farmland. Because high speed and low latency networks are no longer considered luxuries but rather essential in the modern digital era, higher speed thresholds (> 100 mbps, 1gbps) should be utilized to prevent misclassification or underreporting bias in future broadband-related projects. Furthermore, the extensive effects of the social determinants of health as covariates presents a need to contextualize them within the broadband-health outcome pathway moving forward. Qualitative approaches, such as gathering first-hand residential insights via electronic data capturization tools and PhotoVoice, would corroborate the statistical analysis and aid in policymaking/agenda setting. These efforts can prompt effective interventions, as presented in Frieden’s 5-Tiered Pyramid Framework for Public Health Action (Frieden, 2010)

Table of Contents

1. Introduction. 1

1.1 Broadband and Exigence for this Study. 1

1.2 What is COVID-19?. 1

1.3 COVID-19 and its Impacts on Medical Treatment 2

1.4 Future of Telehealth. 3

1.5 Importance of the Social Determinants of Health. 3

1.6 EPA’s Technical Assistance Program.. 4

2. Methods. 6

2.1 Hypothesis. 6

2.2 Data Collection. 7

2.2.1 Predictors. 7

2.2.2 Outcomes. 9

2.3 Overview of Statistical Analysis. 10

2.3.1 Test for Normality. 11

2.3.2 Two-Sample-T-Tests to Assess Differences in Health Outcomes by Dichotomized Predictors. 11

2.3.3 Assessing for Multicollinearity. 12

2.3.4 Identifying Confounders. 12

2.3.5 Identifying Effect Modifiers. 13

2.4 Choosing Best-Fit Parsimonious Model to Quantify Broadband, SDOH, and COVID-19 Outcomes. 13

3. Results. 13

3.1 Demographic Table. 13

3.2 Results of Tests for Normality. 14

3.3 Spearman Correlation Matrices. 17

3.4 Two-Sample T-Tests. 18

3.5 Simple Poisson Regression Models. 20

3.6 Best-Fit “Parsimonious” Poisson Models for Health Outcomes. 21

4. Discussion. 22

4.1 Spearman Correlation Interpretations. 22

4.2 Two-Sample-T-Test Interpretations. 23

4.3 Confounding and Effect Modification Findings and Their Potential Impacts on Policymaking. 24

4.4 Synergistic vs. Antagonistic Determinants of Health Based on Individual Poisson Models. 25

5. Conclusions. 26

5.1 Recommendations. 26

5.1.1 Reduce Broadband Costs via Healthy Market Competition Initiatives. 26

5.1.2 Promote Smart Growth in Communities Through Broadband Designation Programs. 26

5.2 Next Steps. 27

5.2.1 Qualitative Aspects of Public Health Practice. 27

5.2.2 Elevate Broadband Coverage Threshold for Future Quantitative Analyses. 28

5.3 Final Takeaways. 29

6. References. 30

7. Supplemental SAS Programming Material 34

About this Master's Thesis

Rights statement
  • Permission granted by the author to include this thesis or dissertation in this repository. All rights reserved by the author. Please contact the author for information regarding the reproduction and use of this thesis or dissertation.
  • English
Research Field
Committee Chair / Thesis Advisor
Committee Members
Last modified

Primary PDF

Supplemental Files