The objectives of the course are to disseminate the concepts, methods, and the recent statistical tools for risk prediction with data involving competing risks, and to enrich a network of researchers who use data science and analytic methods for medical decision making.

Specifically, the course will introduce two types of competing risks, describe statistical methods associated with each type, software packages that can be used for each method, and methods for assessing and evaluating the risk prediction models. At the end of the course, attendees will understand why standard statistical regression models are not appropriate for analyzing data of competing risks, and will be able to identify two different types of competing risks in practice, and understand the importance of diagnostics in risk prediction modeling. Outline of the class is summarized below.

1. Risk prediction
2. Time-to-event data and censoring.
3. Cox proportional hazards regression model.
4. Competing risks and two types of competing risks (real-world and counterfactual).
5. Risk prediction under real-world setting
   1. Examples
   2. Non-regression method (cumulative incidence function)
   3. Regression method (Fine-Gray subdistribution hazards model)
   4. Other advanced methods (random forest, neural network, etc.)
   5. Software package
6. Risk prediction under counterfactual setting
   1. Examples
   2. Non-identifiability issue
   3. Normalized subdistribution function and marginal failure function
   4. Regression methods (copula-based, joint modeling, model with random signs censoring)
   5. Software package
7. Model diagnostics and evaluations
   1. Discrimination (c-statistic)
   2. Calibration (Brier score, O/E, calibration plot)
   3. Accuracy (mean squared errors)
   4. Software package
8. Summary