Combining Particle Filtering and Transmission Modeling for TB Control

One third of the world's population is affected by tuberculosis (TB). This pandemic presents several challenges to the health professionals. Hence, forecasting and controlling TB epidemics play an important role to deal with this disease. Transmission models are valuable tools for projecting and evaluating control strategies against TB, but lack capability to easily integrate noisy information from epidemiological data into projections and policy analysis. To overcome this shortcoming, a dynamical system for TB based on particle filtering algorithm was developed. We evaluated the effectiveness of different levels of particle filtering by running the particle filter to year 2000, then disabling it and judging the discrepancy of model predictions vs. observed data for the period 2000-2007. The successive results revealed similar patterns including a rise in force of infection during the 1990s, but considering successively larger sets of observations yielded smaller discrepancies between particle filtered projections and observed data.