Peterson Family Health Sciences Hall, #4115
April 17, 2018, 03:00 PM to 05:00 PM
This dissertation explores unseen and unexamined assumptions that affect our decision-making capabilities in the regulation and provision of healthcare. The first two chapters study prostate cancer prescription, while the third chapter examines the Hospital Readmissions Reduction Program.
The first chapter develops a methodology to test whether randomized controlled trials of treatment efficacy produce results that are relevant to individuals who are actually treated by physicians. Randomized controlled trials face power limitations and may not be able to account for all factors a physician may utilize when choosing to prescribe treatment. By using propensity score matching, I am able to identify whether clinicians assign treatment using more information than current randomized controlled trials account for. I apply this methodology to prescription of surgical treatments for prostate cancer, and find that current randomized controlled trials of prostate cancer efficacy are specified in a manner relevant to real world treatment prescription.
The second chapter studies the relative importance of patient medical and socioeconomic characteristics in prostate cancer treatment assignment. Using logistic regression and Shapley-Owen decomposition, I examine the extent to which patient treatment decisions are explained by patient characteristics that have no relationship to lifespan maximization. I find that socioeconomic characteristics have significantly less explanatory power than medical characteristics in predicting prostate cancer treatment assignment, but that neither medical nor socioeconomic characteristics explains the majority of treatment decisions.
The third chapter analyzes the impact of design choices in the implementation of the Hospital Readmissions Reduction Program (HRRP). The program aims to penalize hospitals that have high rates of patient readmission. However, design choices in the creation of the program inadvertently penalize hospitals that receive a large percentage of their total revenue from Medicare and hospitals which perform specific penalized conditions with more frequency, even after accounting for their rates of readmission. Other papers have demonstrated that structural characteristics of hospitals, such as socioeconomic status of their population, impact ability of reduce readmission rates, but this paper shows that the formula used in calculating HRRP penalty affects hospitals based on characteristics entirely unrelated to their readmission performance.095