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

José L. Zayas-Castro, Ph.D.

Committee Member

Carla VandeWeerd, Ph.D.

Committee Member

Hongdao Meng, Ph.D.

Committee Member

Sriram Chellappan, Ph.D.

Keywords

Applied Statistics, Cognitive Trajectory Modeling, Data-driven Modeling, Length-of-Stay Prediction, Nursing Home Evacuation

Abstract

The United States (US) is experiencing rapid growth in its older adult population, who may suffer from multiple chronic diseases, injuries, and impairments. To meet with the excess demand without compromising the quality of care for older adults, the current aged care systems, such as nursing home systems, will face unprecedented challenges of healthcare resource shortage with rising costs. Accurate prediction of health outcomes of individual older adults will facilitate the aged care professionals to better prioritize healthcare resources for the most at-risk individuals with more focused care and provide more proactive and individualized treatment and care delivery. In addition, since older adults are highly vulnerable to natural disasters, accurate prediction of emergency response of aged care facilities, such as evacuation response of nursing homes during hurricane, will further facilitate the emergency operations agencies to provide more proactive and targeted support and coordination with adequate resources to ensure the health outcomes of older adults during hazard scenarios. In this dissertation, a series of healthcare data analytics models, algorithms, and tools are developed to improve modeling and prediction of individual older adults’ health outcomes as well as aged care systems’ emergency responses. First, a bi-level longitudinal data modeling approach is proposed to characterize the heterogeneous degradation of cognitive performance outcomes among community-dwelling older adults at both sub-population level and individual levels. The proposed model comprehensively investigates both the temporal heterogeneity at sub-population and individual levels with relaxed statistical assumptions and improved prediction accuracy. Second, a discharge outcomes prediction model is proposed to characterize the heterogeneous length of stays of post-acute care residents in the nursing home with multiple and competing discharge dispositions. The developed modeling approach allows accurate prediction of re/hospitalization risk and community discharge likelihood over time of individual post-acute care residents with various individual risk factors identified and quantified. Third, a GIS data integrated predictive analytics approach is proposed to improve the prediction of nursing home evacuation response under natural disaster scenarios of hurricane. The proposed work improves the prediction accuracy of evacuation response by integrating spatial and temporal rich storm characteristics information with varied facility characteristics and resident characteristics of nursing home facilities at different spatial locations. Real case studies are considered to illustrate the proposed modeling approaches and demonstrate their superior performances over existing benchmark models.

Download

Share

COinS