Probabilistic cause-of-disease assignment using case-control diagnostic tests: A latent variable regression approach.

Area of Science:

• Epidemiology
• Biostatistics
• Public Health

Background:

• Effective disease prevention and treatment necessitate understanding the distribution of causes for multifactorial diseases like pneumonia.
• Cause-specific case fractions (CSCFs) are crucial for this understanding but can be influenced by various factors.
• Current methodologies inadequately address regression challenges in case-control studies with complex etiological data.

Purpose of the Study:

• To propose a novel, unified regression modeling framework for estimating covariate-dependent CSCF functions.
• To address the estimation of CSCFs using multivariate, non-gold-standard diagnostic data and covariate information within a case-control design.
• To provide a robust method for probabilistic cause assignment in disease etiology research.

Main Methods:

• Development of a regression modeling framework utilizing multivariate binary non-gold-standard diagnostic data.
• Leveraging control data for accurate probabilistic cause assignment to cases.
• Implementation of an efficient Markov chain Monte Carlo algorithm for flexible posterior inference.

Main Results:

• The proposed model yields less biased estimates and more valid inference of overall CSCFs compared to analyses omitting covariates.
• Simulations demonstrate the model's effectiveness in inferring CSCF functions.
• Analysis of pediatric pneumonia data shows CSCFs vary significantly with season, age, HIV status, and disease severity.

Conclusions:

• The developed regression framework offers a significant advancement for estimating covariate-dependent CSCFs in case-control studies.
• The findings highlight the importance of considering covariates for accurate CSCF estimation in disease etiology.
• The model provides a valuable tool for informing public health strategies and disease prevention efforts.