Graduation Year

2021

Document Type

Dissertation

Degree

Ph.D.

Degree Name

Doctor of Philosophy (Ph.D.)

Degree Granting Department

Industrial and Management Systems Engineering

Major Professor

Mingyang Li, Ph.D.

Committee Member

Tapas K. Das, Ph.D.

Committee Member

Susana K. Lai-Yuen, Ph.D.

Committee Member

Hongdao Meng, Ph.D.

Committee Member

Wenjun Cai, Ph.D.

Keywords

Performance Outputs, Latent Heterogeneity, Prediction Accuracy, Bayesian Statistics, Estimation Algorithm

Abstract

In both health systems engineering and reliability engineering, individual units, such as patients and product units, often exhibit highly heterogeneous performance due to the influences of various observed individual characteristics and unobserved/unknown factors. Successful modeling of the heterogeneous performance of individual units is of great importance. It will not only facilitate the identification and quantification of influencing factors for improving performance of individual units, but also improve prediction accuracy of their future performance. This will further facilitate better decisions, such as cost-effective and adaptive healthcare resource planning decision, and proactive maintenance policy at reduced cost. However, due to the highly complex data structure of the heterogeneous performance of individual units, heterogeneous performance modeling is challenging. Most of existing models often have restrictive modeling assumptions with limited modeling flexibility. In this dissertation, a generic modeling framework is established with a series of statistical models to characterize three major types of heterogeneous performance data in health systems engineering and reliability engineering, namely the heterogeneous time-to-event data, the heterogeneous trajectory data and the heterogeneous frequency response data. First, a modeling approach of heterogeneous time-to-event data is proposed to characterize the time-to-discharge and time-to-readmission observations of older adults. The proposed model improves service utilization modeling of individual older adults by considering individual latent heterogeneity as well as multiple types of healthcare settings. Second, a modeling approach of heterogeneous trajectory data with latent heterogeneity is proposed to characterize the heterogeneous service demand of nursing home residents. The proposed approach improves prediction accuracy of service demand of individual resident and further improves resource planning decisions via integration of the predictive models with computer simulation and stochastic optimization. Third, a modeling approach of heterogeneous trajectory data with covariates is proposed to characterize the heterogeneous tribological degradation performance data of deteriorating test units of cooper alloys. The proposed model improves prediction accuracy of degradation performance by considering mixed-type covariates as well as latent heterogeneity within each test unit. Last, a modeling approach of heterogeneous frequency response data is proposed to characterize the heterogeneous corrosion performance assessment data of corroding aluminum alloys in frequency domain. Both the individual latent heterogeneity and nonlinear fractional order system dynamics are considered to improve reliability assessment and prediction performance of individual test units. Real case studies in both health systems engineering and reliability engineering are considered to illustrate the effectiveness of the proposed modeling framework as well as the developed approaches, and further demonstrate their superior performances as well as benefits to stakeholders in both health systems engineering and reliability engineering.

Share

COinS