Capture-recapture to estimate completeness of tuberculosis surveillance in two communities in South Africa

Reliable surveillance is essential for any tuberculosis (TB) control programme; however, under-registration of TB cases due to under-notification of patients on treatment or failure to initiate treatment has been well-documented internationally. To determine the contribution of capture-recapture methods in estimating the completeness of bacteriologically confirmed pulmonary TB registration in two high-incident communities in South Africa. Record linkage between the TB treatment register and two laboratory sputum TB result registers and three-source log-linear capture-recapture analysis. The number of bacteriologically confirmed pulmonary TB cases in the TB treatment register was 243, with an additional 63 cases identified in the two laboratory databases, resulting in 306 TB cases. The observed completeness of the TB treatment register was 79%. The log-linear model estimated 326 (95%CI 314-355) TB cases, resulting in an estimated completeness of registration of 75% (95%CI 68-77). Capture-recapture can be useful in evaluating the completeness of TB control surveillance and registration, including in resource-limited settings; however, methodology and results should be carefully assessed. Interventions are needed to increase the completeness of registration and to reduce the number of initial defaulters