Adapting backcalculation methods to estimate the incidence and infectiousness distributions of tuberculosis

Boston University Medical Campus, Boston MA

Investigators

Abstract

Project Summary/Abstract Despite being a leading cause of death, the global tuberculosis (TB) burden is not well defined. Several methods exist to quantify TB incidence, but such methods are time and/or resource intensive and have at times been demonstrated to be inaccurate. Similar issues in estimating disease burden were faced in the beginning of the AIDS epidemic. A method referred to as backcalculation was developed to estimate HIV incidence with minimal assumptions. This method considers the distribution of observed AIDS cases to be a convolution of the incubation period distribution of AIDS and the incidence distribution of new HIV cases, and thus calculates the HIV incidence distribution via deconvolution. New estimates of key TB natural history parameters allow us to adapt backcalculation methods for TB by developing a Bayesian implementation that incorporates prior data from multiple sources to produce estimates of the TB disease incidence distribution and the TB infectiousness period (time from development of bacteriologically positive TB disease until notification) distribution. Accurate estimates of the TB disease burden are necessary for informing proper resource allocation to diagnose and ultimately eliminate TB. Understanding when and for how long people remain infectious is a relatively understudied area and will also aid strategies to reduce TB morbidity and mortality. My overall objective is to develop novel methods to add to our understanding of the TB burden and its natural history. Aim 1 is to modify the backcalculation framework by developing a Bayesian implementation to estimate TB disease incidence. Aim 2 is to develop a novel method using a Bayesian backcalculation framework to estimate the infectiousness period of TB. For each aim, we will assess the performance of our new statistical models using simulation studies and then apply our methods to reported TB disease notification data from multiple countries. Our methods will add to the currently limited toolbox for estimating key epidemiologic parameters for TB and may be adapted to estimate incidence and infectiousness duration distributions for other infectious diseases. Advancing statistical methods for developing estimates of the TB burden and its natural history are necessary to inform policy making decisions and identify intervention strategies for an epidemic that continues to have devastating effects on low- and middle- income countries. My mentoring team is committed to supporting me in my training and professional development and has outstanding experience in epidemiological research of TB, infectious disease modelling, and Bayesian methods. We have designed a training plan which includes study and workshops in TB epidemiology, infectious disease, and Bayesian modelling; career development; and grant writing. Through this fellowship, I will develop the skills to become an independent researcher with expertise in statistical methods for TB research.