|
Abstract:
|
Accurate prediction of health care costs of patients is important for risk-adjusted physician payments and disease management of high cost patients. However, building a good prediction model is hindered by some special distributional features of health care costs, including a highly skewed distribution, excess zero values, and heteroscedasticity. Much research has been done to develop new statistical methods for addressing these special distributional features in predicting health care costs of patients. In this roundtable, I will discuss some of newly developed predictive models for health care costs of patients and the computer programs for implementing them in practice. I will also illustrate their applications in some real examples.
|
