Graduation Year

2024

Document Type

Dissertation

Degree

Ph.D.

Degree Name

Doctor of Philosophy (Ph.D.)

Degree Granting Department

Marketing

Major Professor

Sajeev Varki, Ph.D.

Committee Member

Brianna Paulich, Ph.D.

Committee Member

Mark Bender, Ph.D.

Committee Member

Eunsook Kim, Ph.D.

Keywords

Healthcare, Pandemic, Item Response Theory, Measurement, First-order Autoregressive Model, Structural Equation Modeling, Latent Growth Model

Abstract

The hospital sector is the largest revenue-generating industry in the United States, with a revenue of $1.4 trillion in 2022, accounting for 5.8% of the U.S. GDP. Its service quality profoundly impacts societal welfare, especially given the aging population reached 55.8 million in 2020 and is expected to reach 82 million by 2050. Therefore, multiple policies aim to incentivize sustained healthcare service quality improvements. As an integral component of the Hospital Value-Based Purchasing (HVBP) Program, the Hospital Consumer Assessment of Healthcare Providers and Systems (HCAHPS) survey, administrated by the Centers for Medicare and Medicaid Services (CMS), serves as a cornerstone in fostering patient-centered care, enhancing informed decision-making, and optimizing healthcare resource allocation.

Conventional empirical studies on HCAHPS typically use linear or multilevel linear regression models and Maximum Likelihood Estimation (MLE) to examine cross-sectional data of the rating scores. However, this approach is deficient in addressing measurement errors and identifying changes in the ratings over time. In contrast, this research employs a multilevel Item Response Theory (MLIRT) model and Bayesian estimation to investigate the hierarchical structure of HCAHPS data and capture the longitudinal changes in latent hospital service quality. Two publicly available sources, HCAHPS ratings from CMS and hospital characteristics from the American Hospital Directory (AHD), are integrated to uncover (a) which service attributes best reflect hospitals’ latent service quality, (b) how this service quality changes over time, particularly pre- and post-COVID, and (c) how hospital heterogeneity influence the longitudinal change of latent service quality.

By isolating latent hospital service quality and item parameters of discrimination and threshold, the MLIRT model effectively addresses measurement errors, facilitating more accurate estimation. Bayesian estimation excels over MLE due to its superior ability to incorporate prior knowledge or beliefs about these parameters and handle heterogeneity issues. Compared to the conventional approach, the proposed model offers more in-depth insights into the change of latent service quality over time and the moderating effects of hospital characteristics on this change.

This research helps hospitals develop effective strategies and optimize resource allocation to improve service quality by identifying the foremost service attributes reflecting latent service quality. The observed 14% discrepancy between raw HCAHP scores and latent service quality underscores the need for policymakers to review and potentially refine existing evaluation mechanisms. By examining the changes in service quality before and after the pandemic, the research offers hospitals and policymakers a more comprehensive understanding of the pandemic's impact on the healthcare industry. This understanding will facilitate the development of appropriate strategies for similar crises in the future. Methodologically, the proposed model can be used with any panel data involving latent variables, which can help researchers minimize measurement errors and reveal time effects on the focal variables.

Download

Share

COinS