Contribution of Reinfection to Annual Rate of Tuberculosis Infection (ARI) and Incidence of Tuberculosis Disease

Abstract

Abstract Background Modeling studies have concluded that 60–80% of tuberculosis (TB) infections result from reinfection of previously infected persons. The annual rate of infection (ARI), a standard measure of the risk of TB infection in a community, may not accurately reflect the true risk of infection among previously infected persons. We constructed a model of infection and reinfection with Mycobacterium tuberculosis to explore the predictive accuracy of ARI and its effect on disease incidence. Methods We created a deterministic simulation of the progression from TB infection to disease and simulated the prevalence of TB infection at the beginning and end of a theoretical year of infection. We considered 10 disease prevalence scenarios ranging from 100/100 000 to 1000/100 000 in simulations where TB exposure probability was homogeneous across the whole simulated population or heterogeneously stratified into high-risk and low-risk groups. ARI values, rates of progression from infection to disease, and the effect of multiple reinfections were obtained from published studies. Results With homogeneous exposure risk, observed ARI values produced expected numbers of infections. However, when heterogeneous risk was introduced, observed ARI was seen to underestimate true ARI by 25–58%. Of the cases of TB disease that occurred, 36% were among previously infected persons when prevalence was 100/100 000, increasing to 79% of cases when prevalence was 1000/100 000. Conclusions Measured ARI underestimates true ARI as a result of heterogeneous population mixing. The true force of infection in a community may be greater than previously appreciated. Hyperendemic communities likely contribute disproportionally to the global TB disease burden.