Sensitive, accessible, and biosafe sampling methods for COVID-19 reverse-transcriptase polymerase chain reaction (RT-PCR) assays are needed for frequent and widespread testing. We systematically evaluated diagnostic yield across different sample collection and transport workflows, including the incorporation of a viral inactivation buffer. We prospectively collected nasal swabs, oral swabs, and saliva, from 52 COVID-19 RT-PCR-confirmed patients, and nasopharyngeal (NP) swabs from 37 patients. Nasal and oral swabs were placed in both viral transport media (VTM) and eNAT™, a sterilizing transport buffer, prior to testing with the Xpert Xpress SARS-CoV-2 (Xpert) test. The sensitivity of each sampling strategy was compared using a composite positive standard. Overall, swab specimens collected in eNAT showed superior sensitivity compared to swabs in VTM (70% vs 57%, P=0.0022). Direct saliva 90.5%, (95% CI: 82%, 95%), followed by NP swabs in VTM and saliva in eNAT, was significantly more sensitive than nasal swabs in VTM (50%, P<0.001) or eNAT (67.8%, P=0.0012) and oral swabs in VTM (50%, P<0.0001) or eNAT (56%, P<0.0001). Saliva and use of eNAT buffer each increased detection of SARS-CoV-2 with the Xpert test; however, no single sample matrix identified all positive cases.