Prediction of the tuberculosis reinfection proportion from the local incidence

Background. Reinfection is a major contributor to tuberculosis (TB). It seems that the higher the local incidence, the higher the proportion of reinfection. Methods. Based on a systematic review of the literature, we established a regression model to predict the reinfection proportion from the local incidence. We then used our local data to verify the algorithm. Results. Of the 23 studies addressing reinfection in recurrent TB, 6 were population based. The reinfection proportion was correlated with the local incidence (reinfection proportion = -29.7 + 36.8 X log^Incidence) (95% confidence interval [CI] for coefficient, 15.3-58.3; R² = 0.849). The reinfection proportion in Taiwan (incidence, 62.4/ 100,000 people) was estimated to be 36% (95% CI, 3%-69%). Of our 49 recurrent patients, 51% had reinfection. Patients with reactivation seemed more likely to have underlying diseases and less likely to be smear positive. The relapse isolates seemed more resistant than the initial isolates. Conclusions. The regression model could possibly predict the TB reinfection proportion from the local incidence. This algorithm is probably helpful in policy making for TB control programs. In areas where TB is endemic, reinfection might be responsible for >50% of TB cases, and aggressive surveillance to detect asymptomatic carriers could be an important strategy for controlling the disease.