Extensions to Bayesian generalized linear mixed effects models for household tuberculosis transmission.

Area of Science:

• Epidemiology
• Infectious Disease Modeling

Background:

• Household contact studies are crucial for understanding tuberculosis transmission.
• These studies often incorrectly assume all infections are acquired within the household.
• Strain genotyping challenges the assumption of solely household-acquired tuberculosis infections.

Purpose of the Study:

• To develop and validate a model estimating the probability of community-acquired tuberculosis infection in household contacts.
• To simultaneously estimate predictors of tuberculosis transmission within households.
• To address biases in transmission predictor estimates caused by misattributing infection sources.

Main Methods:

• Development of a novel household-community transmission model.
• Simulation studies to assess the model's accuracy in predicting community infection probability.
• Application of the model to real-world tuberculosis contact data from Vitória, Brazil.

Main Results:

• The model accurately predicts community infection probability across various scenarios.
• Failure to account for community-acquired infections can significantly bias estimates of household transmission risk factors.
• Analysis of Brazilian data revealed different risk factor estimates for sleeping proximity and disease severity compared to standard methods.

Conclusions:

• Estimating both community infection probability and household transmission predictors is feasible and essential.
• Standard tuberculosis transmission models may underestimate the risk associated with key transmission predictors.
• Accurate modeling of infection sources is vital for effective tuberculosis control strategies.