Flexible multistate models for interval-censored data: Specification, estimation, and an application to ageing...

Area of Science:

• Biostatistics
• Survival Analysis
• Health Economics

Background:

• Health-related processes are often dynamic, involving transitions between multiple states over time.
• Analyzing these transitions is complex, especially with interval-censored data where exact transition times are unknown.
• Existing survival models may lack flexibility in capturing complex hazard shapes.

Purpose of the Study:

• To present a novel method for estimating flexible continuous-time multistate survival models.
• To accommodate both parametric and semiparametric hazard specifications within a unified framework.
• To handle interval-censored transition times effectively.

Main Methods:

• Utilizes continuous-time multistate survival models.
• Constructs the likelihood using transition probabilities for interval-censored data.
• Employs P-splines and penalized maximum likelihood estimation for semiparametric hazards.
• Implements a scoring algorithm for model estimation.
• Applies the method to data from the English Longitudinal Study of Ageing.

Main Results:

• The proposed method allows for flexible specification of hazard functions (parametric and semiparametric).
• The scoring algorithm provides an efficient estimation procedure.
• Demonstrates successful application to real-world health-related data, specifically aging processes.

Conclusions:

• The developed method offers a flexible and robust approach to analyzing multistate health processes with interval-censored data.
• This enhances the ability to model complex health trajectories and transitions accurately.
• The approach is valuable for epidemiological and health services research.