Graduation Year

2022

Document Type

Dissertation

Degree

Ph.D.

Degree Name

Doctor of Philosophy (Ph.D.)

Degree Granting Department

Industrial and Management Systems Engineering

Major Professor

Devashish Das, Ph.D.

Co-Major Professor

Mingyang Li, Ph.D.

Committee Member

Tapas K. Das, Ph.D.

Committee Member

Yasin Yilmaz, Ph.D.

Committee Member

Lu Lu, Ph.D.

Keywords

Alcohol Use Disorder, Emergency Department, Healthcare Service Quality, Statistical Monitoring Methods, Stochastic Process Models

Abstract

In today’s healthcare industry, quality of care is a growing focus in the delivery of healthcare. To improve the quality of care in healthcare delivery, many studies focus on the longterm operational decision making to meet the expectations of healthcare providers and users, such as medical resource allocation, bed planning, staff scheduling, etc. These problems are typically parts of long-term operational decision making, however, time is essential in healthcare system. To ensure the adherence to a high quality of care and detect deterioration in real time, the quality of service should be measured over days or hours instead of just months or years. Therefore, it is critical to develop effective statistical monitoring methods for detecting the deterioration in the quality of healthcare services. In this dissertation, a series of statistical monitoring methods based on stochastic process models are developed for improving the service quality in healthcare including acute and chronic care services. First, a novel statistical monitoring method based on quadratic contrast estimation technique is proposed for detecting changes in the departure intensity function in emergency department. The proposed method is based on an approximate likelihood function that alleviates the issue of needing to numerically maximize a complex likelihood function for estimating the in-control parameters and obtaining test statistics. Second, likelihood-ratio based cumulative sum (CUSUM) control charts are proposed for monitoring the service rate of queueing network with time-inhomogenerous state dependent queues. The proposed approaches could overcome the limitation of the normality assumption of traditional multivariate control charts and do not need to know the potential change in service rate of the queueing nodes in a queueing network, and thus have important practical applications. Third, a continuous-time stochastic process model is proposed to monitor and measure the treatment process for patients with alcohol use disorder (AUD) based on the Cascade of care (COC) framework. The proposed work learns the ideal patterns in the initiation and duration of AUD treatment, from which benchmarks for COC can be developed and factors that are correlated to undesirable patient outcomes identified. Simulation studies and real case studies are considered to illustrate the proposed statistical monitoring methods and demonstrate their superior performance over traditional methods.

Download

Share

COinS