Objectives Frequently used rapid rifampicin drug susceptibility tests (RMP-DST) miss certain rifampicin resistance (RR)-conferring mutations, leaving RR-tuberculosis undetected. Unknown for RR-TB is the therapeutic threshold, the probability of disease at which there is equipoise between treating and not treating. In Mozambique, in a patient not responding to first-line treatment, clinicians decided to start RR-TB treatment without bacteriological proof of RR-TB. We determined the probability of RR-TB in this patient. Methods We converted probabilities and odds ratios of clinical arguments for RR-TB from literature to likelihood ratios. We then combined the associated confirming and excluding power of those arguments to estimate the probability of RR-TB when the patient was started on RR-TB treatment, and simulated its variation. We used a log-odds scale to illustrate the effect of confirming and excluding arguments. Results The starting point was the prevalence of RR-TB in Mozambique. Positive HIV-status, treatment failure after a first treatment and after retreatment were included as confirming arguments, and RMP-DST showing rifampicin susceptibility as excluding argument for RR-TB. In this patient, the probability of RR-TB was 46.6% (95% uncertainty interval: 25.0%-72.0%) when RR-TB treatment was started. Treatment failure and retreatment failure provided strong confirming arguments, and the RMP-DST result a strong excluding argument for RR-TB. Conclusions The therapeutic threshold to start RR-TB-treatment is unknown but probably lower than 47%. The uncertainty in our estimation reflects the clinical uncertainty in low-resource settings. Determining the RR-TB therapeutic threshold is needed to guide clinical decisions.