Saliva has been demonstrated as feasible alternative to naso-oropharyngeal swab (NOS) for SARS-CoV-2 detection through reverse transcription quantitative/real-time polymerase chain reaction (RT-qPCR). This study compared the diagnostic agreement of conventional NOS, saliva with RNA extraction (SE) and saliva direct without RNA extraction (SalivaDirect) processing for RT-qPCR in identifying SARS-CoV-2. All techniques were also compared as separate index tests to a composite reference standard (CRS) where positive and negative defined as SARS-CoV-2 detection in either one and no sample, respectively. Of 518 paired samples, SARS-CoV-2 was detected in 150 (28.96%) NOS and 151 (29.15%) saliva specimens. The saliva-based tests were noted to have a sensitivity, specificity and accuracy (95% confidence interval) of 92.67% (87.26%, 96.28%), 97.55% (95.41%, 98.88%) and 96.14% (94.10%, 97.63%), respectively, for SE RT-qPCR and 90.67% (84.84%, 94.80%), 98.91% (97.24%, 99.70%) and 96.53% (94.56%, 97.93%), respectively, for SalivaDirect RT-qPCR compared to NOS RT-qPCR. Compared to CRS, all tests demonstrated diagnostic performance that is statistically similar to each other. These suggest that both conventional and streamlined saliva RT-qPCR are non-inferior to conventional NOS RT-qPCR in detecting SARS-CoV-2.