A Random-Effect Model on Correlated Health Care Costs with Zeroes

• Activity Number: 194
• Type: Contributed
• Date/Time: Monday, August 5, 2013 : 10:30 AM to 12:20 PM
• Sponsor: Biometrics Section
• Author(s): Y V Hui*+ and Andy H. Lee and Kelvin K W Yau
• Companies: City University of Hong Kong and Curtin University and City University of Hong Kong
• Keywords: fixed effect model ; correlated ; zeroes

Abstract

Extreme values (large medical expenses and zeroes) at both ends are often observed in a healthcare costs data set. For instance, healthy people may incur no medical expense while sick people generally consume more healthcare resources. These data are often correlated due to certain diseases and/or living environment. We propose a two-part random effect model to analyze the heterogeneous correlated data. A logistic mixed regression model and a gamma mixed regression model are chosen to model the binary part and the continuous part respectively. An efficient estimation algorithm is presented. We also discuss an illustrative example on a compensation claims data set that demonstrates the model application.