We are upgrading the repository! A content freeze is in effect until December 6th, 2024 - no new submissions will be accepted; however, all content already published will remain publicly available. Please reach out to repository@u.library.arizona.edu with your questions, or if you are a UA affiliate who needs to make content available soon. Note that any new user accounts created after September 22, 2024 will need to be recreated by the user in November after our migration is completed.
Mental Health Readmissions Among Veterans: An Exploratory Endeavor Using Data Mining
Author
Price, Lauren EmilieIssue Date
2015Advisor
Shea, Kimberly D.
Metadata
Show full item recordPublisher
The University of Arizona.Rights
Copyright © is held by the author. Digital access to this material is made possible by the University Libraries, University of Arizona. Further transmission, reproduction or presentation (such as public display or performance) of protected items is prohibited except with permission of the author.Abstract
The purpose of this research is to inform the understanding of mental health readmissions by identifying associations between individual and environmental attributes and readmissions, with consideration of the impact of time-to-readmission within the Veterans Health Administration (VHA). Mental illness affects one in five adults in the United States (US). Mental health disorders are among the highest all-cause readmission diagnoses. The VHA is one of the largest national service providers of specialty mental health care. VHA's clinical practices and patient outcomes can be traced to US policy, and may be used to forecast national outcomes should these same policies be implemented nationwide. In this research, we applied three different data mining techniques to clinical data from over 200,000 patients across the VHA. Patients in this cohort consisted of adults receiving VHA inpatient mental health care between 2008 and 2013. The data mining techniques employed included k-means cluster analysis, association-rule mining, and decision tree analysis. K-means was used during cluster analysis to identify four statistically distinct clusters based on the combination of admission count, comorbidities, prescription (RX) count, age, casualty status, travel distance, and outpatient encounters. The association-rule mining analysis yielded multiple frequently occurring attribute values and sets consisting of service connection type, diagnoses/problems, and pharmaceuticals. Using the CHAID algorithm, the best decision tree model achieved 80% predictive accuracy when no readmissions were compared to 30-day readmissions. The strongest predictors of readmissions based on this algorithm were outpatient encounters, prescription count, VA Integrated Service Network (VISN), number of comorbidities, region, service connection, and period of service. Based on evidence from all three techniques, individuals with higher rates of system-wide utilization, more comorbidities, and longer medication lists are the most likely to have a 30-day readmission. These individuals represented 25% of this cohort, are sicker in general and may benefit from enrollment in a comprehensive nursing case management program.Type
textElectronic Dissertation
Degree Name
Ph.D.Degree Level
doctoralDegree Program
Graduate CollegeNursing