Graduation Year

2017

Document Type

Dissertation

Degree

Ph.D.

Degree Name

Doctor of Philosophy (Ph.D.)

Degree Granting Department

Industrial and Management Systems Engineering

Major Professor

Jose L. Zayas-Castro, Ph.D.

Committee Member

Jay Wolfson, Dr.P.H., J.D.

Committee Member

Peter Fabri, M.D.

Committee Member

Alex Savachkin, Ph.D.

Committee Member

Stephanie L. Carey, Ph.D.

Keywords

Interventions, machine learning, Patient outcomes, policy analysis, predictive model

Abstract

The high expenditure of healthcare in the United States (U.S.) does not translate into better quality of care. Indeed, the U.S. healthcare system is recognized by its lack of efficiency and waste (which represents about 20% of the country’s healthcare expenses). Lack of coordination is one of the most referenced causes of waste in the U.S. healthcare system, and preventable hospital readmissions have been acknowledged to be evidence of poor coordination of care. In fiscal year 2013, the Centers for Medicare and Medicaid Services (CMS) established financial penalties for inpatient care reimbursements in hospitals with excessive readmissions. All the same, the preliminary results of this effort have yet to result in a consistent reduction of readmission rates. Research in healthcare policy is usually reported through case studies, which makes it difficult to apply that research to different spatiotemporal contexts. Additionally, relevant research can remain overlooked due to the challenge of translating it from other fields. Therefore, in order to create effective healthcare policies, a system that can provide the most accurate information to stakeholders about their decisions and the future impact of those decisions should be developed. This dissertation proposes a decision-based support system that could aid hospital administrators in the design of disease-specific interventions that target specific groups of patients who are at risk for readmission. First, the use of disease-specific interventions that were designed to reduce readmissions will be explored. Second, a variety of predictive tools for readmissions will be developed and compared to complete the search for the best tool. Finally, an optimization model bringing together the two ideas will be formulated so that hospitals can use it to design interventions. This model will target specific patients depending on their risk for readmission and minimize the cost of intervention while ensuring quality hospital performance. In sum, this work will help hospital administrators to better plan in the reduction of readmissions and in the implementation of interventions. In addition, it will deepen knowledge about the impacts of economic penalties on hospitals and facilitate the construction of stronger arguments for decisions about healthcare policy.

Share

COinS