Tuberculosis is a global public health threat as an infectious disease, and effective blocking of transmission relies on timely diagnosis. Although a number of laboratory tests are available in diagnosing Mycobacterium tuberculosis (MTB) and nontuberculous mycobacteria (NTM), it is still of great need to evaluate their diagnostic value in clinical samples. In this study, we evaluated five MTB diagnostic methods (including conventional sputum smear microscopy, sandwich cup sputum smear microscopy, sputum culture, Xpert-MTB/RIF, and CapitalBio TB/NTM kit detection test) in 3012 sputum specimens and compared their diagnostic performance of the single and combined tests. In the diagnosis of MTB, the Xpert-MTB/RIF had the highest sensitivity, 79.6% (0.770-0.819), among all the single diagnostic methods, and the combination of CapitalBio TB/NTM kit and culture approach significantly increased sensitivity to 88.4% (p <
0.05). In the diagnosis of NTM, the culture method has higher sensitivity (85.7%) compared with the Capital Bio TB/NTM kit method (45.7%). In the diagnosis of mycobacteria, the CapitalBio TB/NTM kit detection test has the highest sensitivity (77.1%) and combined with conventional sputum smear and culture significantly increased the sensitivity further to 84.2%. In conclusion, Xpert-MTB/RIF is a sensitive, rapid, and reliable method for TB detection in sputum samples, and other diagnostic methods including culture are still of great clinical values for improving the sensitivity of MTB diagnosis. The sensitivity of CapitalBio TB/NTM kit in diagnosing NTM is still insufficient in clinical practice.