Implementing competing risks in discrete event simulation: the event-specific probabilities and distributions...

Area of Science:

• Biostatistics
• Computational Biology
• Health Informatics

Background:

• Discrete event simulation (DES) lacks robust methods for competing events with censored data.
• Existing strategies do not adequately address event-specific probabilities and distributions (ESPD) in censored scenarios.

Purpose of the Study:

• To define and illustrate the ESPD approach for modeling competing events when dealing with censored data.
• To evaluate the performance and applicability of the ESPD strategy in simulation and real-world case studies.

Main Methods:

• The ESPD approach models events via a two-step process: event type selection and time-to-event sampling, both potentially covariate-dependent.
• Performance was assessed using simulation studies varying sample size and censoring levels.
• An oncology case study demonstrated implementation in R using frequentist and Bayesian frameworks.

Main Results:

• The ESPD approach demonstrated good performance in simulation studies.
• Accuracy decreased with smaller sample sizes and higher censoring levels.
• The oncology case study yielded realistic results, confirming the approach's practical utility.

Conclusions:

• The ESPD approach is a viable method for modeling competing events in DES with censored data.
• Further research is needed to compare ESPD with other DES modeling techniques and assess its utility for cumulative event incidence estimation.