To perform a multicentre study evaluating the performance of the microscopic observation drug susceptibility (MODS) assay for the detection of MDR-TB and XDR-TB in high-burden resource-limited settings. We performed a prospective diagnostic accuracy study of drug-resistant TB suspects from outpatient and inpatient settings in five laboratories in China. Sputum was tested by smear microscopy, liquid [mycobacterial growth indicator tube (MGIT)] culture and the MODS assay at each site. Drug susceptibility testing (DST) was by MODS and an indirect 1% proportion method. The reference standard for Mycobacterium tuberculosis detection was growth on MGIT culture; the 1% proportion method was the reference standard for rifampicin, isoniazid, ofloxacin, kanamycin and capreomycin DST. M. tuberculosis was identified by reference standard culture among 213/532 (40.0%) drug-resistant TB suspects. Overall MODS sensitivity for M. tuberculosis detection was 87.8%-94.3% and specificity was 96.8%-100%. For drug-resistant TB diagnosis, excellent agreement was obtained for all drugs tested at the majority of sites. The accuracy was 87.1%-96.7% for rifampicin, 87.1%-93.3% for isoniazid, 92.7%-100% for ofloxacin, 90.9%-100% for kanamycin and 90.2%-100% for capreomycin. The median time to culture positivity was significantly shorter for MODS than for the MGIT liquid culture (8 days versus 11 days, P<0.001). The contamination rate ranged between 2.1% and 5.3%. In the study settings, MODS provided high sensitivity and specificity for rapid diagnosis of TB and drug-resistant TB. We consider it to have a strong potential for timely detection of MDR-TB and XDR-TB in high-burden resource-limited settings. © The Author 2014. Published by Oxford University Press on behalf of the British Society for Antimicrobial Chemotherapy. All rights reserved. For Permissions, please e-mail: journals.permissions@oup.com.