K-Medoids-Surv: A Patients Risk Stratification Algorithm Considering Censored Data

Traditional survival analysis estimates the instantaneous failure rate of an event and predicts survival probabilities distributions. In fact, in a set of censored data there may exist several sub-populations with various risk profiles or survival distributions, for which regular survival analysis approaches do not take into consideration. Consequently, there is a need for discovering such sub-populations with unambiguous risk profiles and survival distributions. In this work, we propose a modified version of the K-Medoids algorithm which can be used to efficiently cluster censored data and identify diverse groups with distinct lifetime distributions.

Automated summary

Traditional survival analysis estimates the instantaneous failure rate of an event and predicts survival probabilities distributions. However, in a set of censored data, there may exist several sub-populations with various risk profiles or survival distributions that are ignored by regular survival analysis approaches. Therefore, it is essential to discover such sub-populations with unambiguous risk profiles and survival distributions. In this study, a modified version of the K-Medoids algorithm is proposed to efficiently cluster censored data and identify diverse groups with distinct lifetime distributions.