Profiling Hospitals Based on Degree of Aggressiveness in Treating Patients with Advanced Cancer

• Activity Number: 184
• Type: Contributed
• Date/Time: Monday, August 1, 2016 : 10:30 AM to 12:20 PM
• Sponsor: Section on Bayesian Statistical Science
• Author(s): Tianyi Cai* and Sherri Rose and Deborah Schrag and Francesca Dominici
• Companies: Harvard and Harvard Medical School and Dana-Farber Cancer Institute and Harvard T.H. Chan School of Public Health
• Keywords: hospital profiling ; causal inference ; bayesian hierarchical modeling ; palliative care

Abstract

Hospital profiling is a very established topic in the literature. In this paper we are interested in: profiling hospitals based on outcomes related to aggressiveness of end-of-life (EOL) treatments (e.g. chemotherapy, re-admissions, ICU visits) for patients with advanced cancer; and identifying which hospital-specific characteristics explain the variation in these outcomes across hospitals. We are also interested in a hospital-level analysis to estimate the average causal effect of access to palliative care on the aggressiveness of EOL treatments. Towards these goals, we develop Bayesian hierarchical models for our cohort of 45,000 Medicare patients with advanced lung, pancreas, colorectal, or brain cancer. At the first stage of the model, we estimate hospital-level risk of EOL outcomes accounting for patient-level data. At the second stage, we formulate a potential outcome framework to estimate the average causal effect of access to palliative care on EOL outcomes. Characterizing between hospital variability and determining whether access to palliative care reduces treatment aggression is important due to the important health policy implications in the utilization of EOL care.