Developing a Model: Maximizing Doctor Utility Given Patient Severity and Invasiveness of Treatment

Open Access
Tamburo, Kelly
Area of Honors:
Bachelor of Science
Document Type:
Thesis Supervisors:
  • Edward Green, Thesis Supervisor
  • Russell Chuderewicz, Honors Advisor
  • doctor utility
  • health economics
  • severity index
  • regional disparity
One of the challenges in a doctor-patient-insurer relationship is finding the right balance between costly, invasive treatment, and cheap, noninvasive treatment for a given patient’s severity level. This leads to two key questions: (1) From the perspective of the doctor and patient, what is the right severity threshold at which to prescribe the most invasive treatment; and (2) From the insurer perspective, can we estimate from the decisions made by a doctor over time, what is the threshold used by that doctor to prescribe the most invasive treatment? This paper will utilize the Healthcare Cost and Utilization Project, Agency for Healthcare Research and Quality National Inpatient Sample from 2014. A two-part model that optimizes patient life outcomes is used in this paper and consists of an ordered logit regression model and conditional logit discrete choice model. This model determines a doctor’s utility from medical versus surgical treatment based on a patient’s severity level and life outcome. In the model findings, medical treatment stochastically dominates outcomes. Additionally, in the model, doctor utility preferences result in a ranking of outcomes as semi-cured (lowest), death, and cured (highest). This discrepancy may be driven by two explanations. The assumption that doctors maximize a utility function that depends only on, and that is increasing in, the patient’s outcome is violated, or the assumption that severity, as measured in the data, is a sufficient statistic for the doctor’s best judgement about the appropriate treatment, given all the evidence that the doctor sees does not hold. There is evidence that suggests that the first model assumption could be violated through regional disparities and further research would be needed to determine if the second model assumption is violated.