Maximum likelihood estimation for semiparametric regression models with interval-censored multistate data

Area of Science:

• Biostatistics
• Epidemiology
• Chronic Disease Research

Background:

• Chronic diseases often involve transitions between multiple health states.
• Observational data frequently feature interval censoring, where event times are known only within intervals.
• Analyzing such complex data requires advanced statistical methodologies.

Purpose of the Study:

• To develop a statistical framework for analyzing interval-censored multistate data in chronic disease studies.
• To model the influence of time-dependent covariates on disease progression.
• To provide robust estimation and inference procedures for these complex data structures.

Main Methods:

• Utilized semiparametric proportional intensity models with random effects.
• Employed nonparametric maximum likelihood estimation under general interval censoring.
• Developed a stable expectation-maximization algorithm for parameter estimation.

Main Results:

• Demonstrated consistency of parameter estimators.
• Established asymptotic normality for finite-dimensional components.
• Showcased that the covariance matrix achieves the semiparametric efficiency bound and is consistently estimable.
• Validated the methods through extensive simulations and a real-world cohort study.

Conclusions:

• The proposed methods offer a reliable approach for analyzing interval-censored multistate data in chronic disease epidemiology.
• The statistical procedures are computationally stable and provide efficient, asymptotically normal estimates.
• This work advances the statistical toolkit for understanding complex disease trajectories and covariate impacts.