Date of Award

12-2023

Document Type

Dissertation

Degree Name

Doctor of Philosophy (PhD)

Program

Health Outcomes and Policy Research

Track

Health Services Research

Research Advisor

Charisse Madlock-Brown, PhD

Committee

Simonne S. Nouer, PhD; Rebecca B. Reynolds, Ed.D.; Satya Surbhi, PhD; Kristi Wick, DNP

Keywords

COVID-19;emergency department;machine learning;social vulnerability

Abstract

Background: COVID-19 is a contagious respiratory illness caused by the severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2). As of March 10, 2023, the virus infected nearly 104 million Americans. Many individuals were wary of seeking medical attention for issues unrelated to COVID-19 for fear of infection inside healthcare facilities. This hesitation caused a drop in emergency department (ED) visits for non-COVID-19 conditions. It is essential to ensure that individuals with critical healthcare needs receive the care they require and to mitigate the potential long-term consequences of delayed or deferred care. Objectives: The primary aim of this study is to examine how patterns of ED utilization shifted and may continue to shift for socially vulnerable populations during and beyond the COVID-19 pandemic. Accomplishing this entails understanding the broader baseline changes in ED utilization and then identifying patterns for vulnerable individuals. Insights from this step will be used to predict future ED utilization changes and assess downstream impacts on the healthcare system. Methods: An observational population-based cohort study design was chosen for this study. The onset of COVID-19 created a natural inflection point in time whereby changes in captured data could be examined. Electronic medical records from three large health systems in Tennessee was extracted from the Research Enterprise Data Warehouse (rEDW) from University of Tennessee Health Sciences Center (UTHSC), representing data from three Tennessee health systems: Methodist Le Bonheur (Memphis), University of Tennessee Medical Center (Knoxville), and St. Thomas Health (Nashville). These data included medical history, demographics, hospital procedures, diagnostic data, laboratory orders and results, and prescription medication information, and were adequate and equivalent pre- and post-COVID-19 cohorts for analyses. The data extract consisted of ED visits between March 2018 and March 2022. Results: This research revealed a significant overall decrease in ED utilization in the post-COVID-19 era. Vulnerable populations showed a significant reduction in mental health ED visits but an increase in substance-related concerns. Machine learning analysis highlighted a key predictor for ED volume in a census tract: the percent of single-parent households, indicating accessibility challenges. While the XGBoost model performed well overall (R2 = 0.96), it had limited applicability in Memphis MSA due to census tract heterogeneity. This study demonstrates the potential for accurate ED volume estimation using simplified models and census-level data. Conclusions: Emergency departments are critical to the US healthcare system and can be essential when dealing with life-threatening events and conditions. However, ED services can often be utilized for conditions that could be treated in an outpatient setting. This could impact cost and outcome issues in general; however, it becomes even more concerning amid a global pandemic. This study provides future researchers with insights, suggestions, and protocols for leveraging data and developing predictive tools for understanding emergency department utilization.

Declaration of Authorship

Declaration of Authorship is included in the supplemental files.

ORCID

0009-0006-2827-5876

DOI

10.21007/etd.cghs.2023.0650

Additional Files
2023-039-McCormack-DOA.pdf (168 kB)
Download

Share

COinS