ABSTRACT

Absolute risk estimates based on cause-specific models rely on survival analysis methods that were first widely applied for estimating pure risks following disease diagnosis. This chapter focuses on three important issues for modeling absolute risk: covariate selection, missing covariate data, and updating previously well-established risk models by adding new covariates. An important aspect of regression modeling is selecting covariates for the model. The criteria for and approach to covariate selection can depend on the intended use of the model and the scientific information available on potential risk factors from previous studies. Selecting a regression model by finding a subset of covariates that appear to perform well in the training data also has implications for prediction, and, in particular, for predicting the risk of disease for a new individual based on his or her covariates and the selected risk prediction model.