1513. Evaluating the Incremental Accuracy of HIV Self-Test Together with an App-Based Solution: A Secondary Data Analysis of Trial Data

BACKGROUND: According to the United Nations Programme on HIV/AIDS (UNAIDS), in 2019, two out of every seven new human immunodeficiency virus (HIV) infections, globally, were among young people aged 15 to 24 years old. HIV self-testing (HIVST) is a convenient strategy that helps increase knowledge of HIV serostatus in the young. In 2012, the Food and Drug Administration (FDA) approved the OraQuick In-Home HIV Test with a reported sensitivity of 92%. Digital supports such as applications (Apps) and websites, in conjunction with oral self-tests, have demonstrated a high feasibility, acceptability, and impact, yet data on accuracy with digital supports remain largely unexplored. METHODS: We performed a secondary data analysis of a quasi-randomized trial of oral HIVST with HIVSmart! conducted in South African township populations (2017-2019). We hypothesized that HIVSmart! guided interpretation increased test accuracy. We evaluated the diagnostic accuracy of the HIVSmart! guided interpretation of an oral self-test result against the reference standard (dual Elisa and HIV RNA). Stored picture of a self-test result was uploaded by the participants. RESULTS: Accuracy data from (n=1489) HIVST participants versus reference standard (2 rapid tests and HIV RNA) demonstrated: Sensitivity: 95.52% (95% CI: 94.48%-96.56%) Specificity: 99.93% (95% CI: 99.79%-100.00%) Positive Predictive Value: 99.22% (95% CI: 98.78%-99.67%) Negative Predictive Value: 99.57% (95% CI: 99.24%-99.90%) CONCLUSION: With the App, we noticed an improved sensitivity to 95.5% (from 92% with self-tests without the use of the App), specificity high at 99%, together with high positive and negative predictive values. Findings demonstrate that Smart-App based digital interpretation removed subjectivity and increased accuracy of test result interpretation, together with recording and storage of data for monitoring purposes. Findings suggest that Smart-App based readers could be useful adjuncts to improve the accuracy estimations of self-tests, translating to increased trust and confidence in self-tests. DISCLOSURES: All Authors: No reported disclosures