A joint model for interval-censored functional decline trajectories under informative observation

Area of Science:

• Biostatistics
• Epidemiology
• Health Services Research

Background:

• Multi-state models are crucial for tracking disease progression through discrete health states.
• Disease observation often occurs at clinical visits, with visit schedules potentially influenced by patient health status.
• Ignoring this 'informative observation' can lead to biased inference in disease progression models.

Purpose of the Study:

• To develop a joint model that simultaneously analyzes disease progression and the observation process.
• To address potential bias arising from informative, status-dependent clinical visit schedules.
• To ensure valid statistical inference in disease modeling with interval-censored data.

Main Methods:

• A Markov process models disease transitions, incorporating bivariate subject-specific random effects.
• These random effects link the disease model with the observation process model.
• Inference is conducted within a Bayesian statistical framework.

Main Results:

• The developed joint model provides a statistically valid method for analyzing disease progression.
• It accounts for the informative nature of observation times, reducing potential bias.
• The model was successfully applied to a large study of palliative care patients.

Conclusions:

• Joint modeling of disease and observation processes is essential when visit schedules are informative.
• This approach yields more accurate insights into disease progression trajectories.
• The methodology is particularly relevant for studies with interval-censored data and status-dependent follow-up.