Application of provincial data in mathematical modelling to inform sub-national tuberculosis program decision-making in South Africa.

• Main Authors: Piotr Hippner, Tom Sumner, Rein Mgj Houben, Vicky Cardenas, Anna Vassall, Fiammetta Bozzani, Don Mudzengi, Lindiwe Mvusi, Gavin Churchyard, Richard G White
• Format: Article
• Language: English
• Published: Public Library of Science (PLoS) 2019-01-01
• Series: PLoS ONE
• Online Access: https://doi.org/10.1371/journal.pone.0209320

South Africa has the highest tuberculosis (TB) disease incidence rate in the world, and TB is the leading infectious cause of death. Decisions on, and funding for, TB prevention and care policies are decentralised to the provincial governments and therefore, tools to inform policy need to operate at this level. We describe the use of a mathematical model planning tool at provincial level in a high HIV and TB burden country, to estimate the impact on TB burden of achieving the 90-(90)-90 targets of the Stop TB Partnership Global Plan to End TB. "TIME Impact" is a freely available, user-friendly TB modelling tool. In collaboration with provincial TB programme staff, and the South African National TB Programme, models for three (of nine) provinces were calibrated to TB notifications, incidence, and screening data. Reported levels of TB programme activities were used as baseline inputs into the models, which were used to estimate the impact of scale-up of interventions focusing on screening, linkage to care and treatment success. All baseline models predicted a trend of decreasing TB incidence and mortality, consistent with recent data from South Africa. The projected impacts of the interventions differed by province and were greatly influenced by assumed current coverage levels. The absence of provincial TB burden estimates and uncertainty in current activity coverage levels were key data gaps. A user-friendly modelling tool allows TB burden and intervention impact projection at the sub-national level. Key sub-national data gaps should be addressed to improve the quality of sub-national model predictions.