A joint modelling approach for multistate processes subject to resolution and under intermittent observations

Area of Science:

• Biostatistics
• Health Data Analysis
• Statistical Modeling

Background:

• Multistate processes are useful for transition intensity analysis.
• Unobserved events that stop future transitions complicate inference.
• Patient resolution in health studies is a key example of such an event.

Purpose of the Study:

• To propose a novel statistical model for multistate processes with unobserved, state-stopping events.
• To address challenges in health studies where resolution status is unknown.
• To improve the characterization of transition intensities in the presence of latent events.

Main Methods:

• A modified multistate model is proposed, expanding the state space.
• The state for absent symptoms is partitioned into a latent absorbing resolved state and a temporary transient state.
• The expanded state space explicitly models uncertainty about resolution occurrence.

Main Results:

• The methodology allows for explicit distinction between temporary and resolved states of absent symptoms.
• The likelihood construction effectively captures the uncertainty of resolution.
• The model was successfully illustrated on a psoriatic arthritis dataset with disability scores.

Conclusions:

• The proposed model enhances the analysis of multistate processes with unobserved resolution events.
• It provides a robust framework for health studies where symptom resolution is intermittent or uncertain.
• Estimated probabilities of resolving can be obtained, offering valuable clinical insights.