Adapting Back-calculation Methods to Estimate the Incidence of Tuberculosis.

Area of Science:

• Epidemiology
• Biostatistics
• Infectious Disease Modeling

Background:

• Global tuberculosis (TB) burden remains poorly defined due to inaccurate and resource-intensive incidence estimation methods.
• Back-calculation, a method used for HIV incidence, can be adapted for TB by considering disease duration and new case incidence.
• Advancements in TB natural history parameters enable the development of Bayesian back-calculation for TB.

Purpose of the Study:

• To develop and apply Bayesian back-calculation methods for estimating tuberculosis incidence.
• To assign case notification data to the precise time point of disease onset.
• To provide more temporally accurate TB incidence estimates than existing methods.

Main Methods:

• Utilized Bayesian back-calculation incorporating TB natural history parameters.
• Included a multiplier derived from prevalence-to-notification ratios to adjust for underreporting.
• Employed Markov chain Monte Carlo methods for sex-stratified TB incidence estimation in Vietnam, Cambodia, and The Philippines, assuming Poisson distributions and using penalized-likelihood priors.

Main Results:

• Annual estimated TB incidence was, on average, 19% higher than recorded notifications.
• Sex-stratified analysis showed higher TB incidence in males compared to females: 3.8% in Vietnam, 1.3% in Cambodia, and 2.5% in The Philippines.

Conclusions:

• The developed Bayesian back-calculation method provides more temporally accurate estimates of TB incidence.
• These estimates account for the time lag between subclinical disease and notification.
• The findings highlight the underestimation of TB burden by current reporting methods.