Electronic Theses and Dissertations

Identifier

4840

Date

2016

Date of Award

12-3-2016

Document Type

Dissertation (Access Restricted)

Degree Name

Doctor of Philosophy

Major

Epidemiology

Committee Chair

Fawaz Mzayek

Committee Member

Vikki Nolan

Committee Member

Xinhua Yu

Committee Member

George Relyea

Abstract

Patients readmitted to the same or different healthcare facility within 30 days of initial inpatient discharge are of great concern to hospitals due to quality and financial implications. Prevention of unplanned hospital readmissions is one of the primary quality metrics mandated by the Centers for Medicare and Medicaid Services (CMS) under the Affordable Care Act (ACA). A retrospective longitudinal study was performed using medical records of the patients discharged from Methodist Le Bonheur Healthcare (MLH) between: January 1, 2007 and December 31, 2013. Textual data mining was performed on the unstructured data using SAS Enterprise Miner version 12.1® (SAS Institute, Cary, NC). Unstructured data elements were then created by using the keywords within the SCAN function in SAS version 9.4® (SAS Institute, Cary, NC). Descriptive statistics, bivariate and multivariable logistic regressions were used to determine the factors associated with unplanned readmission and multiple unplanned readmissions. Eight new variables created using text mining of unstructured data are: homeless, medication assistance, food stamps, travel vouchers/assistance, meal vouchers/assistance, morbid obesity, family support, and non-compliance to medications. We did not notice any improvement in the discrimination power or c-statistic for the enhanced multivariable logistic regression model (combination of structured and unstructured data elements) compared to the general model (only structured data elements). The c-statistic for general and enhanced models are: 0.74 (SA2: unplanned readmissions) and 0.62 (SA3: multiple unplanned readmissions). Even though “homelessness” and “family support” showed significance in bivariate results; none of these variables derived from the unstructured data stayed in the final enhanced model, highlighting that solely depending on unstructured data may not be a good research strategy.This study demonstrated the process of developing variables from unstructured data and how they can be added to the structured data to explore a complicated topic like unplanned and multiple (two or more) unplanned readmissions.

Comments

Data is provided by the student.

Library Comment

Dissertation or thesis originally submitted to the local University of Memphis Electronic Theses & dissertation (ETD) Repository.

Share

COinS