Penalised maximum likelihood estimation in multistate models for interval-censored data

TL;DR

This paper introduces an automatic penalised maximum likelihood approach for flexible multistate models with spline-based hazards, effectively handling interval-censored transition data.

Contribution

It proposes an efficient, automatic method for estimating multistate models with splines under interval-censoring, improving flexibility over traditional parametric approaches.

Findings

Method performs well in simulations

Applied to real data successfully

Outperforms existing approaches in flexibility

Abstract

Multistate models can be used to describe transitions over time across states. In the presence of interval-censored times for transitions, the likelihood is constructed using transition probabilities. Models are specified using proportional hazards model for the transitions. Time-dependency is usually defined by parametric models, which can be too restrictive. Nonparametric hazards specification with splines allow for flexible modelling of time-dependency without making strong model assumptions. Penalised maximum likelihood is used to estimate the models. Selecting the optimal amount of smoothing is challenging as the problem involves multiple penalties. We propose an automatic and efficient method to estimate multistate models with splines in the presence of interval-censoring. The method is illustrated with a data analysis and a simulation study.