Trends in Untreated Tuberculosis in Large Municipalities, Brazil, 2008–2017

M any countries that have considerable subnational variation in tuberculosis (TB) burden also have decentralized the management and implementation of control policies. In this context, local estimates of TB burden can convey actionable insights for these TB control decisions. Reported cases are commonly used as a proxy for TB burden; however, reported cases may not reflect the true burden because areas of apparently low burden may instead represent areas of inadequate case detection. Modeling approaches have been proposed to adjust for this bias and enable valid inference of TB incidence, but these approaches typically require primary data collection (1,2). Alternative methods make use of routinely collected data (3-5). We applied a recently developed Bayesian method to report unbiased estimates of TB incidence and the completeness of case detection in Brazil's state capitals and 100 most populous municipalities during 2008-2017 (Appendix, https://wwwnc.cdc.gov/EID/article/27/3/20-4094-App1.pdf). The Office of Human Research Administration at Harvard T.H. Chan School of Public Health reviewed the initial study submission (protocol no. IRB18-0759) and determined that it met the criteria for exemption from ethics board review.