In Reply We thank Kim and colleagues for their letter on the need for cautious interpretation of negative GeneXpert MTB/RIF (Xpert; Cepheid) results in patients undergoing evaluation for possible tuberculosis (TB). They correctly note that although Xpert accurately identified all cases of culture-confirmed TB in our study1—including 1 sputum smear–negative case—its sensitivity among patients with negative smear results has been shown to be only moderate in some settings.2 However, to better understand the absolute risk of false-negative results, it would be useful to know the total number of patients tested in the study by Kim and colleagues in order to calculate the TB prevalence and the negative predictive value of Xpert in that setting. Importantly, the literature suggests that false-negative results are very rare with single Xpert algorithms, and they have not occurred in previous studies with algorithms incorporating 2 Xperts.3-6 Nevertheless, despite the important incremental sensitivity of Xpert over smear, we agree that providers should exercise care when making clinical decisions based on negative Xpert results, particularly in intermediate and high TB-burden settings where the negative predictive value of Xpert may be less optimal. In our pragmatic implementation study,1 clinicians were free to decide whether to use Xpert results to guide isolation discontinuation. Furthermore, our algorithm recommended 2 negative results by the molecular assay and/or smear prior to isolation discontinuation in order to reduce the risk of false-negative results. We recommend that hospital policymakers in other settings allow clinicians similar flexibility and consider the local burden of paucibacillary disease when developing Xpert-based algorithms to guide care of patients with possible TB.