A new framework for semi-Markovian parametric multi-state models with interval censoring.

Area of Science:

• Biostatistics and Computational Statistics
• Survival Analysis and Longitudinal Data Modeling

Background:

• Limited availability of computational and methodological tools for general multi-state models with interval censoring.
• Need for flexible frameworks to incorporate various parametric models and covariates in multi-state data analysis.

Purpose of the Study:

• To propose a general parametric inference framework for interval-censored multi-state data.
• To develop an efficient R package (smms) for implementing the proposed framework and computing high-dimensional integrals.

Main Methods:

• Development of a general likelihood construction method for parametric multi-state models with interval censoring.
• Implementation of the framework in the R package 'smms', available on GitHub.
• Efficient computation of high-dimensional integrals within the R package.

Main Results:

• The proposed framework accommodates arbitrary parametric models for transition times and various covariate inclusion methods.
• The 'smms' R package provides a ready-to-use tool for analyzing interval-censored multi-state data.
• Demonstrated application of the framework using heart transplant patient data and simulation studies on model misspecification.

Conclusions:

• The developed framework offers a novel and flexible approach to parametric inference for interval-censored multi-state data.
• The 'smms' package enhances the practical analysis of such data, addressing computational challenges.
• The study contributes to the field of survival analysis by extending multi-state modeling capabilities.