APPLYING MACHINE LEARNING TO LABORATORY DATA: PREDICTING SUPPRESSION OF NEXT HIV VIRAL LOAD IN SOUTH AFRICA

By  Dr Mhairi Maskew  Crompton T, Sharpey-Schafer K, De Voux L  Jacob Bor  Rennick M, Pisa P  Jacqui Miot  |  | 

During 2018, South Africa was estimated to have more than seven million people living with HIV1, representing the largest single country epidemic2 and treatment program. 3 In September 2016, the National Department of Health revised its treatment guidelines to extend the availability of ART to all people living with HIV, irrespective of CD4 cell count and stage of disease3. This policy, widely referred to as “treat all” or “universal test and treat” (UTT) holds promise to offer substantial advancements not only in the health of those living with HIV 4-5, but also in the country’s efforts to meet 95-95-95 targets. However, implementation of a policy like UTT requires a rapid scale-up and expansion of the ART program on a country-wide level; a shift that is often challenging in these settings. By expanding eligibility, UTT has eliminated “pre-ART” care for most patients.6 In many cases, this translates into more rapid initiation of ART with fewer required clinic visits prior to dispensing drugs. In many cases, the pre-ART care cascade is compressed into a single visit with same day initiation (SDI) of treatment. Though this increases uptake of ART, retention of new patients may have suffered. The problem of patient loss to follow-up (LTFU) within the public sector in South Africa has been well documented.7 The potential for improved health through expanded ART availability will only be realized if individuals sustain engagement in HIV care.

Publication details

PDF