Bayesian estimation of variance partition coefficients adjusted for imperfect test sensitivity and specificity

Area of Science:

• Biostatistics
• Epidemiology
• Statistical Modeling

Background:

• The variance partition coefficient (VPC) is crucial for understanding disease clustering in specific populations.
• Accurate VPC estimation is vital for effective intervention strategies and study design.
• Imperfect diagnostic tests can introduce bias in VPC calculations.

Purpose of the Study:

• To develop a Bayesian discrete mixed model for estimating covariate-pattern-specific VPCs.
• To address bias introduced by imperfect diagnostic tests in VPC estimation.
• To demonstrate the model's utility in real-world applications and covariate analysis.

Main Methods:

• A Bayesian discrete mixed model was employed for VPC estimation.
• The model incorporated parameters for test sensitivity and specificity using beta distributions.
• The methodology was validated through three distinct case applications.

Main Results:

• Imperfect tests were found to bias VPC estimates towards the null.
• Modeling test sensitivity and specificity corrected for this bias, yielding more accurate VPCs.
• Comparison of adjusted VPCs revealed the extent to which covariates explain data heterogeneity.

Conclusions:

• The proposed Bayesian model accurately estimates covariate-pattern-specific VPCs even with imperfect diagnostic tests.
• Accounting for test characteristics is essential for unbiased VPC estimation and robust data interpretation.
• The model facilitates a better understanding of heterogeneity and informs targeted public health interventions.