Efficient algorithms for survival data with multiple outcomes using the frailty model

Area of Science:

• Biostatistics
• Computational Biology
• Neuroscience

Background:

• Survival data with multiple outcomes are common in biomedical research, such as tracking progression in Alzheimer's Disease Neuroimaging Initiative (ADNI) studies.
• Modeling correlated transitions (e.g., normal cognition to mild cognitive impairment, MCI to Alzheimer's disease dementia) often involves frailty models.
• Frailty models present computational challenges due to multiple integrations, especially with high-dimensional covariates.

Purpose of the Study:

• To propose efficient minorization-maximization algorithms for frailty models with multiple survival outcomes.
• To incorporate simultaneous variable selection and homogeneity pursuit using regularization and fusion.
• To address the computational intractability of existing estimation methods.

Main Methods:

• Development of novel minorization-maximization algorithms tailored for multi-outcome survival data within a frailty framework.
• Integration of the alternating direction method of multipliers (ADMM) for regularization and fusion.
• Application of these algorithms to analyze Alzheimer's Disease Neuroimaging Initiative (ADNI) data.

Main Results:

• The proposed algorithms demonstrate efficiency in handling complex survival data with multiple correlated outcomes.
• Simulations confirm the performance of the developed computational methods.
• The approach effectively facilitates variable selection and homogeneity pursuit in frailty models.

Conclusions:

• The novel algorithms provide an efficient computational solution for analyzing multi-outcome survival data in frailty models.
• These methods are applicable to complex biomedical data, including neurodegenerative disease progression.
• The study enhances the practical utility of frailty models in analyzing correlated survival data.