Date of Award

5-2016

Document Type

Dissertation

Degree Name

Doctor of Philosophy (PhD)

Program

Health Outcomes and Policy Research

Track

Health Policy

Research Advisor

James E. Bailey, M.D.

Committee

Virginia Trotter-Betts, J.D. Erik L. Carlton, DrPH Cyril F. Chang, Ph.D. Dawn M. FitzGerald, M.S. J. Carolyn Graff, Ph.D. Esra Ozdenerol, Ph.D. George E. Relyea, M.S.

Abstract

This three-essay dissertation was focused on geographic variation of super-utilization, or the disproportionately high healthcare utilization and costs attributed to a small sub-set of the inpatient population. For purposes of this research, super-utilization was operationalized as high repeat utilization (HRU) and referred to inpatient utilization and inpatient readmission expenditures attributed to beneficiaries with four or more 30-day readmissions per year. The overall purpose of the research was to identify geographic areas at increased risk for HRU. These areas corresponded to where beneficiaries live and were aligned with the geographically-bound healthcare delivery systems. Each essay employed an observational study design using 100% Medicare Part A claims data on beneficiaries ages 65 and older residing in Tennessee hospital referral regions during the 2012 study period. The first essay focused on the impact of super-utilization on population-based rates of readmission across healthcare delivery systems. Specific aims of the first essay were: 1) to assess geographic variation in a population-based overall rate of readmissions across local healthcare delivery systems by conducting one sample means testing use the Z statistic to determine whether rates were lower, higher, or no different from the state average; 2) to contrast the number of beneficiaries, readmission events, and inpatient readmission expenditures attributed to beneficiaries with one compared to four or more readmissions per year between local delivery systems in the 10th and 90th percentiles of readmission rates using descriptive statistics; and 3) to assess the effect of the number of readmissions by beneficiaries with one, two, three, and four or more readmissions per year on overall readmission rates using a quasi-experimental approach to linear regression. The second essay focused on identifying clusters of super-utilization across healthcare delivery systems. Specific aims of the second essay were: 1) to detect statistically significant clusters of concentrated readmission events attributed specifically to HRU by using the SatScanTM method for spatial scan statistics; 2) to explore overlap of identified clusters with population-based rates of readmission using chloropleth mapping to visually depict the relationship; and 3) to assess differences in the geographic distribution of readmission events attributed to HRU between urban and rural locations within cross-border areas using the Mann-Whitney U test to determine statistical significance. The third essay focused on predicting risk of super-utilization across healthcare delivery systems using community demographic variables. Specific aims of the third essay were: 1) to assess the effect of rurality, income, and race on the presence of HRU using a logistic regression model; 2) to determine whether differences in model effects existed among regional healthcare delivery systems by including region as a class variable within the model; and 3) to evaluate whether differences in model effects existed at various concentrations of low-income households by performing decomposition analysis using contrasts based on percentiles of the distribution of low-income households. Findings on the whole suggest that local healthcare delivery systems with high population-based rates of overall readmissions are also more likely to have underlying issues related to super-utilization. In fact, half of all inpatient readmission expenditures attributed to super-utilization across the study area were concentrated in local healthcare delivery systems in the 90th percentile of readmission rates. Unsurprisingly, clusters of super-utilization overlapped these local healthcare delivery systems with high rates. However, clusters of super-utilization were identified in all regions including some local healthcare delivery systems with rates no different or lower than the statewide per capita rate of readmissions. Furthermore, the highest risk of super-utilization occurred in areas with the highest concentration of low-income households, regardless of rural-urban designation, household race, or region.

ORCID

http://orcid.org/0000-0003-0497-0736

DOI

10.21007/etd.cghs.2016.0403

Available for download on Friday, June 23, 2017

Share

COinS