Estimating the survival functions for right-censored and interval-censored data with piecewise constant hazard functions

The exponential distribution is frequently used to model the survival time of a patient population, which assumes the hazard rate to be a constant over time. This assumption is often violated as the hazard function may vary over time and exhibit one or more change points in its values. Several methods exist in the literature for detecting a single change point in a piecewise constant hazard function for right-censored data. A sequential testing approach to detecting multiple change points in the hazard function using likelihood ratio statistics and resampling is proposed, which is applicable to both right-censored and interval-censored data. Numerical results based on simulated survival data and a real example show that the proposed approach can accurately detect the change points in the hazard function for both right-censored and interval-censored data.