Multivariate disease progression modeling with longitudinal ordinal data.

Area of Science:

• Biostatistics
• Longitudinal Data Analysis
• Disease Modeling

Background:

• Traditional disease progression models often rely on continuous data like biomarkers.
• Categorical and ordinal data, such as questionnaire responses, offer valuable insights into disease progression.
• Existing models may not fully capture the complexity of disease heterogeneity and dynamics.

Purpose of the Study:

• To develop a novel disease progression model capable of analyzing ordinal and categorical data.
• To extend the principles of disease course mapping to incorporate item response theory.
• To provide a more granular understanding of disease progression and patient heterogeneity.

Main Methods:

• Developed a disease progression model based on disease course mapping principles.
• Integrated ordinal and categorical data analysis into the modeling framework.
• Applied the model to the Parkinson's Progression Markers Initiative (PPMI) cohort.

Main Results:

• The model provides a fine-grained description of Parkinson's disease progression at the item level, surpassing aggregated scores.
• Achieved improved predictions of future patient visits compared to conventional methods.
• Identified distinct disease heterogeneity patterns, including tremor-dominant and postural instability/gait difficulties subtypes.

Conclusions:

• The proposed model effectively analyzes ordinal and categorical data for disease progression modeling.
• This approach enhances the understanding of disease dynamics and patient-specific trajectories.
• The model offers valuable insights for personalized medicine and clinical trial design in neurodegenerative diseases.