Abstract
Tuberculosis (TB) transmission is a key factor for disease-control policy, but the timing and distribution of transmission and the role of social contacts remain obscure.
We develop an agent-based epidemic simulation of a TB in a single population, and consider a hierarchically structured contact network in three levels, typical of airborne diseases. The parameters are adopted from the literature, and the epidemic simulation model is calibrated to a setting of high TB incidence.
We model the dynamics of disease transmission at the individual level, and study the timing of secondary infections from a single source throughout the duration of the disease. Using simulation, we compare the patterns of disease transmission among different networks and discuss implications. Sensitivity analysis of outputs indicates the robustness of the results to variations in the parameter values.
Epidemiology Simulation with Agent-Based Modeling Software
TB is an airborne disease transmitted through infectious contact with an active case. TB transmission is one of the key determinants of epidemic severity, and has important implications for design, implementation, and scaling-up of control interventions (e.g., improved diagnosis, active case finding) that aim to reduce the rate of disease transmission.
Unlike other airborne diseases such as influenza, however, TB transmission is not directly measurable, i.e., available diagnostic techniques cannot estimate the original timing of infection in diagnosed cases. TB also has a predilection for establishment of a latent state that is non-infectious and asymptomatic, but may progress to active, infectious disease at any time.
As a result, one cannot reliably differentiate between primary infection with rapid progression to active disease, re-infection following a remote initial infection, or reactivation of a previous latent infection. Such limitations pose several challenges to the study of disease transmission dynamics across populations, including the lack of informative data to trace the chain of transmission across various contact networks in retrospective studies.
The prospective following a cohort population, on the other hand, are prohibitively expensive and restricted by the time and budget constraints. In such settings, the relationships between the duration of disease, symptom burden, contact networks, diagnosis/treatment, and patients’ infectiousness remain obscure.
We propose an agent-based epidemic simulation model for the study of disease (TB) transmission dynamics and to discover the role of various contact networks. Our model simulates the TB epidemic course across a single population and uses a hierarchical network of contacts at three levels, typical to the transmission of airborne diseases.
Parameters are chosen from the literature, and the model is calibrated to a setting of high TB incidence. We use our epidemic simulation model to study disease transmission dynamics at an individual level, with regard to the timing and distribution of secondary infections from a single source.
The average time for disease transmission to reach 50% of infections at an individual level is estimated, and the timing patterns are compared among different networks. We perform sensitivity analysis of results with regard to multiple parameter values, and discuss the implications for TB control policy.
Modeling TB dynamics has a long history, including mathematical models and analytical techniques to describe and predict disease prevalence at the population level. Analytical studies, however, are usually restricted by their simplifying assumptions regarding the population heterogeneity, network structure, and parameter uncertainty, and do not provide a realistic representation of disease transmission dynamics unlike epidemic simulation models.
Simulation modeling of TB epidemics in human populations, on the other hand, has a shorter history. One group of studies uses system dynamics to model disease prevalence at the population level.
Following a top-down approach, these studies divide the population into different health states, and use transition rates to describe the disease’s natural history. A system of differential equations is used to model disease prevalence through time. In comparison to the analytical approach, such studies apply a semi-Markov system in which transition rates can change with time, and are able to capture output uncertainty.
Other researchers have developed discrete-event simulation (DES) models to evaluate the impact of new diagnostic tools, or to study more complex structure as in the coinfection of HIV/TB. The DES studies use a schedule of events that are executed in chronological order, and model disease transmission using random generation of Poisson distributions in each mixing group.
The aggregate (top-bottom) modeling nature of DES, however, offers low flexibility for direct modeling of contacts (and disease transmission events) at the individual level, and restricts application of such models to complex social networks.
In this article we consider a simulation model of disease transmission involving three contact networks that represent the main social relationships in the transmission of an airborne disease. Our epidemic simulation model simulates the stochastic contact events at an individual level and enables us to study patterns of disease transmission across different networks.
