Estimation in a three-state model with interval-censored data [comunicação oral]

Abstract

In many fields, including medical research, engineering, and the social sciences, analyzing time-to-event data is essential for uncovering underlying processes and facilitating decision-making. A common challenge in this analysis arises when events are confirmed to have taken place within specific time intervals, yet the exact timing within those intervals remains unknown, a phenomenon known as interval censoring. The focus is on a three-state progressive multi-state survival model where the intermediate state and/or the final state may be interval-censored. The primary aim is to estimate state occupation probabilities, which are crucial for understanding the dynamics of state transitions over time. Additionally, the estimation of the bivariate distribution of the gap times is considered. New estimation methods are introduced based on the Turnbull estimator of survival to address the challenges posed by interval-censored events. Imputation-based methods are also explored for estimating event times within the interval, such as using the midpoint, left-point, and right-point of the interval. Findings contribute to filling the gap in the literature regarding interval-censored multi-state models, providing valuable insights for researchers and practitioners dealing with such data.