Data Analysis with Competing Risks and Intermediate States

Data Analysis with Competing Risks and Intermediate States explains when and how to use models and techniques for the analysis of competing risks and intermediate states. It covers the most recent insights on estimation techniques and discusses in detail how to interpret the obtained results. After introducing example studies from the biomedical and epidemiological fields, the book formally defines the concepts that play a role in analyses with competing risks and intermediate states. It addresses nonparametric estimation of the relevant quantities. The book then shows how to use a stacked data set that offers great flexibility in the modeling of covariable effects on the transition rates between states. It also describes three ways to quantify effects on the cumulative scale. Each chapter includes standard exercises that reflect on the concepts presented, a section on software that explains options in SAS and Stata and the functionality in the R program, and computer practicals that allow readers to practice with the techniques using an existing data set of bone marrow transplant patients. The book’s website provides the R code for the computer practicals along with other material. For researchers with some experience in the analysis of standard time-to-event data, this practical and thorough treatment extends their knowledge and skills to the competing risks and multi-state settings. Researchers from other fields can also easily translate individuals and diseases to units and phenomena from their own areas.

Basic Concepts Introduction Examples Data structure On rates and risks Non-informative observation schemes? The examples revisited Notation Basic techniques from survival analysis Summary and preview Exercises R code for classical survival analysis Computer practicals Competing Risks; Nonparametric Estimation Introduction Theoretical relations Estimation based on cause-specific hazard Estimation; the subdistribution approach Standard errors and confidence intervals Log-rank tests and other subgroup comparisons Summary; three principles of interpretability Exercises Software Computer practicals Intermediate Events; Nonparametric Estimation Introduction; multi-state models Main concepts and theoretical relations Estimation Example: HIV, SI, AIDS and death Summary; some alternative approaches Exercises Software Computer practicals Regression; Cause-Specific/Transition Hazard Introduction Regression on cause-specific hazard; basic structure Combined analysis and type-specific covariables Why does the stacked approach work? Multi-state regression models for transition hazards Example: causes of death in HIV infected individuals Summary Exercises Software Computer practicals Regression; Translation to Cumulative Scale Introduction From cause-specific/transition hazard to probability Regression on subdistribution hazard Multinomial regression Summary Exercises Software Computer practicals Epilogue Which type of quantity to choose? Exercises Bibliography Appendix: Answers to Exercises Index

Ronald B. Geskus is an associate professor at the Academic Medical Center in Amsterdam. He received a Ph.D. in mathematics from the Delft Technical University. His main research interests include competing risks and multi-state models, prediction of events based on time-updated marker values, and causal inference.