Case Study on Fitting a Risk Prediction Model for Competing Risks

• Activity Number: 193
• Type: Contributed
• Date/Time: Monday, August 10, 2015 : 10:30 AM to 12:20 PM
• Sponsor: Biometrics Section
• Author(s): Haley Hedlin* and John Robbins
• Companies: Stanford University and UC Davis
• Keywords: risk ; prediction ; competing risks

Abstract

Many risk calculators are available for patients and physicians to estimate risk of a particular outcome. However, patients and physicians will be able to make better decisions with a calculator that considers risk of several outcomes simultaneously. Our goal is to create an individualized risk prediction model capable of providing risk estimates for several outcomes such as death, heart attack, stroke, fracture, and lung, breast, and colorectal cancer. To this end, we fit our model in a competing risk framework using previously developed methods. Special consideration is given to modeling choices, handling of missing covariate data, and model validation in this case study.