Cost-effectiveness analysis of a prediction model for community-based screening of active tuberculosis.

Abstract

Active case finding could effectively detect tuberculosis (TB) patients, but it can be costly. Therefore, a feasible, cost-saving, and efficient algorithm for community-based TB screening is needed. The study population was based on a previous TB prevalence survey conducted in the Zambia/South Africa Tuberculosis and HIV/AIDS Reduction trial. We developed predictive scoring models for HIV-positive and HIV-negative/unknown populations for practical purposes. We compared the cost-effectiveness of our models with that of WHO tools through average cost-effectiveness ratio (ACER) and incremental cost-effectiveness ratio. The prediction model for HIV-positive population presented higher area under the curve (AUC) = 0.652, 95% confidence interval (CI) = 0.602-0.701) than WHO-recommended four-symptom screen (AUC = 0.568, 95% CI = 0.524-0.612) among South African participants. The AUC of the model for HIV-negative/unknown population was 0.673 (95% CI = 0.648-0.697), which was higher than that of the WHO tools as well. The ACER of our model can range from 246 to 1670 USD per TB case detected in the South African communities. The scoring system for active TB case finding presented a better performance and showed cost-effectiveness, which can provide new strategies for active TB case finding with multiple options under budget considerations.