ABSTRACT
The situation of competing risks occurs in time to event analysis when the person (or subject) will encounter p mutually exclusive causes of death (failure). Different methods have been implemented in the last few decades to model time-to-event data in the face of competing risks, where some of the methodologies have not yet received much attention. In this study, we are briefly discuss some important aspects and impacts of various classical approaches towards the competing risk framework in survival analysis. Competing risks initially modeled in terms of latent failure time approach and could not attracted due to awkward interpretation in real life. Later, cause-specific quantities such as cause specific hazard function, cumulative incidence function and subdistribution hazard function gets considerable attention in survival analysis. In addition to this, we focused on some alternative techniques viz., mixture model, direct parameterization and machine learning modeling pertaining to the competing risks application for estimating cumulative incidence function. We analysed breast cancer data with help of inbuilt function in R software for illustration of competing risks application. Finally, we have given discussion and concluding remarks on the advantages and disadvantages of these methods.
