Date of Award

Fall 2019

Document Type

Open Access Dissertation

Department

Epidemiology and Biostatistics

First Advisor

Anwar T. Merchant

Abstract

AIM: The primary aim of this study was to examine individual-level and community- level characteristics associated with ambulatory or primary care utilization, emergency department (ED) utilization, and ED charges among a sample of ED patients.

DESIGN AND SAMPLE: Data for this cross-sectional study were obtained from three distinct sources: (i) electronic medical records (EMR); (ii) billing records; and (iii) the 2013-2017 American Community Survey (ACS). The individual-level EMR and billing sample included all adults residing in Mecklenburg County, North Carolina who visited an Atrium Health ED in 2017. The ACS sample included population and demographic estimates from Mecklenburg County’s 27 ZIP code tabulation areas (ZCTAs).

METHODS: The total number of billed ED visits and associated ED charges were primary outcomes in the study. The total number of billed visits to ambulatory or primary care (APC) was both an outcome and a covariate. Other individual-level covariates were: insurance coverage type, race, ethnicity, age, and gender. ZCTA-level covariates were: residential segregation, measured using the dissimilarity index, and living in a public health priority area (PHPA), defined as areas with disproportionately low educational attainment and high poverty. Mean regression (i.e. negative binomial, and linear regression) models were used to assess associations between healthcare utilization and residential segregation on average. Quantile regression models were used to assess the relationship between covariates and ED utilization (avoidable utilization, ED visit frequency, and ED Charges) at the 25th, 50th, 75th, 95th, and 99th percentiles of the distributions.

RESULTS: Residential segregation was not associated with the average number of ED visits and was associated with the average number of APC visits during the study period. The relationships between residential segregation and not having any visits to APC in the past year, and average ED charges varied based on the race of the individual. There was heterogeneity in the association between APC utilization and avoidable ED scores by insurance type. Having Medicaid or Medicare insurance was positively associated with ED visits compared to those that were uninsured, at the 50th and 75th percentiles of the distribution. Medicaid and Medicare were positively associated with ED charges and having Private insurance was negatively associated with ED charges across all percentiles of the distribution. Visits to APC was positively and negatively associated with ED visit frequency, and living in a PHPA was positively and negatively associated with ED charges.

CONCLUSIONS: Residential segregation was associated with APC utilization and ED charges, but not with ED visits. The associations between APC utilization and avoidable ED utilization varied based on segments of the distribution and was significantly different among insurance stratum. The associations between APC visits and PHPA status with the outcomes of ED visits and ED charges varied by percentile of the distribution, and included relationships that were in qualitatively opposite directions. Modeling ED utilization outcomes using internal, distribution-based cut points described their relationships with independent variables more accurately than conventional methods that dichotomize the outcome or evaluate the average of the entire distribution.

Rights

© 2019, Carlene A. Mayfield

Included in

Epidemiology Commons

Share

COinS