Jointly Modeling Healthcare Costs
*Joanne K. Daggy, Indiana University School of Medicine 

Keywords: zero-inflated, healthcare costs, correlation, multivariate analysis

Development of appropriate healthcare cost models provide decision-makers the tools necessary to study the impact of policy decisions and/or allocation of resources. To date, modeling research has primarily focused on marginal costs and addressing the issues of right-skewness, heteroskedasticity, and excess zeros. Because data are often collected longitudinally, clustered in terms of physician or treatment facility, or subdivided into cost categories, there is an interest in jointly modeling these costs. In this talk, I will describe a version of this joint modeling that marginally models each cost using a two-part generalized gamma distribution and correlates them using copulas. We argue this approach provides great flexibility, is easy to implement, and computationally less intensive then alternative methods such as random effects modeling.