Operational modelling: the mechanisms influencing TB diagnostic yield in an Xpert super( registered ) MTB/RIF-based algorithm

SETTING: Cape Town, South Africa. OBJECTIVE: To compare the diagnostic yield for smear/culture and Xpert super( registered ) MTB/RIF algorithms and to investigate the mechanisms influencing tuberculosis (TB) yield. METHOD: We developed and validated an operational model of the TB diagnostic process, first with the smear/culture algorithm and then with the Xpert algorithm. We modelled scenarios by varying TB prevalence, adherence to diagnostic algorithms and human immunodeficiency virus (HIV) status. This enabled direct comparisons of diagnostic yield in the two algorithms to be made. RESULTS: Routine data showed that diagnostic yield had decreased over the period of the Xpert algorithm roll-out compared to the yield when the smear/culture algorithm was in place. However, modelling yield under identical conditions indicated a 13.3% increase in diagnostic yield from the Xpert algorithm compared to smear/culture. The model demonstrated that the extensive use of culture in the smear/culture algorithm and the decline in TB prevalence are the main factors contributing to not finding an increase in diagnostic yield in the routine data. CONCLUSION: We demonstrate the benefits of an operational model to determine the effect of scale-up of a new diagnostic algorithm, and recommend that policy makers use operational modelling to make appropriate decisions before new diagnostic algorithms are scaled up.