Selected sensitivity analyses.

Sensitivity of the model for the prevalence of tuberculosis, for the prevalence of multidrug-resistant tuberculosis, and for the accuracy of clinical diagnosis. Patterns of ICERs in 2010 US$ for varying the proportion of individuals with suspected TB in the cohort who have smear-positive TB (A, D, G); for varying the proportion of new patients with TB who have multidrug-resistant TB (MDR-TB, B, E, H); and for varying the specificity of the clinical diagnosis of TB in the base case (C, F, I). (A, B, and C), South Africa; (E,D, and F), India; (G, H, and I), Uganda. Black lines, Xpert assay in addition to sputum smear examination; grey lines, Xpert assay as replacement of sputum smear examination. The proportion of individuals with suspected TB in the cohort who have smear-negative TB varies along with the proportion of individuals with suspected TB in the cohort who have smear-positive TB in a linear manner, depending on the HIV-infection prevalence (A, D, G; see Table 1 and Text S1). Similarly, the proportion of previously treated patients with TB who have MDR-TB varies linearly with the proportion of new patients with TB who have MDR-TB (B, E, H; see Table 1 and Text S1). The sensitivity of clinical diagnosis in the base case varies inversely with the specificity (range, 8%–80%; C, F, I).