Essays on incentive design and healthcare delivery reform in health economics
OA Version
Citation
Abstract
This dissertation considers mechanisms to improve healthcare delivery andreduce healthcare costs, both through the incentives created by risk adjustment methodology and the study of an intervention to improve care for children with medical complexity.
The first chapter, joint with Randall Ellis, Jeffrey Siracuse, Allan Walkey, Karen Lasser, Brian Jacobson, Alex Hoagland, Ying Liu, Chenlu Song, Tzu-Chun Kuo, and Arlene Ash, characterizes a novel diagnosis classification system, the Diagnostic Items Classification (DXI) System. The system leverages the detail embedded in the International Classification of Diseases, Tenth Revision, Clinical Modification. The system performs better than benchmark risk adjustment models across a range of measures of model fit, especially for individuals with rare diseases. Addressing systemic underpayment for individuals with rare diseases has potential implications for the quality of care they receive.
The second chapter, joint with Randall Ellis, Jeffrey Siracuse, Alexander Hoagland, Heather Hsu, Allan Walkey, Karen Lasser, Tzu-Chun Kuo, and Arlene Ash proposes a regression algorithm for variable selection in risk adjustment models. Risk adjustment systems are vulnerable to gameability, particularly in the form of upv coding. This work develops a computationally feasible, transparent, and clinically informed approach to undermine gaming incentives while maintaining predictive power.
The third chapter evaluates a project to improve care for children with medical complexity (CMC): The Collaborative Improvement and Innovation Network to Advance Care for Children with Medical Complexity (CoIIN). While CMC represent a small share of the population of children, they are associated with a disproportionately large share of healthcare spending, and their families face significant burdens. I conduct a claims-based evaluation of the effect of the programs at two sites on a range of utilization outcomes, including inpatient (IP) admissions and length of stay (LOS), emergency department (ED) visits, outpatient (OP) provider encounters, and days with OP provider encounters. I do not find robust statistically significant changes in utilization from the interventions. This study lays the foundation for future work, and provides informative descriptive statistics to inform future study design.
Description
License
Attribution 4.0 International