This study aimed to evaluate the effectiveness of the algorithm used in HIV diagnosis and to propose an effective new algorithm for rapid diagnosis. In accordance with CDC algorithm, our laboratory uses Architect HIVAg/Ab for screening and Geenius HIV1/2 and Artus HIVirus-1 QS-RGQ for confirmation. The Geenius test was used as a reflex and the HIV-1-RNA required clinician order. The HIVAg/Ab test was performed in 82,882 sera and found to be reactive in 262 (0.3%). HIV-antibody confirmatory testing was performed on 79% of samples with a reactive screening test, and the presence of HIV-1 antibodies was confirmed in 51% (105/206). Half of the samples with positive-screening but negative-antibody confirmatory results were tested for HIV1-RNA, and viremia was detected in 5, confirming acute HIV1 infection. HIV1-RNA was not ordered for 49 samples with positive-screening and negative antibody-confirmation tests, and 16 of these were considered false-reactive by the clinician. The Geenius assay result was indeterminate in 1.45% (3/206) of the samples. In the algorithm, the number of Geenius tests would have been reduced by 25% if HIV-1-RNA had been applied as a reflex test to HIV-Ag/Ab positive samples and Geenius testing had been performed on RNA negative samples. A retrospective analysis showed that the HIV diagnostic algorithm was not fully implemented. An important factor was that clinicians did not order HIV-1-RNA-PCR from ELISA reactive and Geenius test negative patients. Requesting HIV-1 RNA PCR as a reflex test is thought to prevent patient losses and shorten the turnaround time of the HIV diagnosis.