Bayesian feature selection in joint models with application to a cardiovascular disease cohort study

Area of Science:

• Biostatistics
• Epidemiology
• Cardiovascular Research

Background:

• Cardiovascular disease (CVD) cohorts track risk factors over time to understand event associations.
• Identifying specific features of risk factor trajectories linked to CVD is crucial for prevention.

Purpose of the Study:

• To develop and implement novel feature selection methods for joint models in CVD research.
• To adapt Bayesian sparse group selection (BSGS) for bi-level variable selection within longitudinal risk factors.

Main Methods:

• Developed a modified Bayesian sparse group selection (BSGS) prior for joint models.
• The new prior accounts for feature correlations within risk factors, improving selection efficiency.
• Applied the method to the Atherosclerosis Risk in Communities (ARIC) study data.

Main Results:

• Identified key cardiovascular risk factors: systolic and diastolic blood pressure, glucose, and total cholesterol.
• Determined specific trajectory features within these risk factors associated with CVD death.
• The modified BSGS prior demonstrated efficient exclusion of unimportant features.

Conclusions:

• The novel feature selection method enhances understanding of CVD risk factor trajectories.
• Systolic/diastolic blood pressure, glucose, and cholesterol are significant risk factors for CVD death.
• The approach effectively identifies important risk factors and their trajectory features in population studies.