Online Program

Return to main conference page

Friday, January 12
Fri, Jan 12, 8:30 AM - 10:15 AM
Crystal Ballroom F
Hierarchical Modeling

Measuring performance for end-of-life care. (304150)

*Sebastien Haneuse, Harvard University 
Kyu Ha Lee, The Forsyth Institute 

Keywords: Provider performance; Semi-competing risks; Hierarchical model

CMS currently uses hospital-specific risk-adjusted standard readmission rates to rank and profile hospitals. These rates are calculated for a range of acute health conditions for which prognosis is good, including pneumonia and heart failure. For advanced health conditions where prognosis is poor, such as pancreatic cancer, use of readmission rates alone to measure hospital performance is problematic because they ignore covariation in mortality rates. Building on the semi-competing risks framework, we propose hospital-specific risk adjusted standardized readmission and mortality cumulative rate functions as a novel means of measuring quality of end-of-life care. Estimation follows within the Bayesian paradigm, using a recently developed hierarchical modeling framework. In addition, using a decision-theoretic approach, we also develop estimators of a range of non-standard but important inferential targets including: identification of extreme performers; ranking of hospitals; and, estimation of the empirical distribution function. The ideas and methods are illustrated using data on all Medicare beneficiaries diagnosed with pancreatic cancer between 2000-2013.