Routine Electronic Mother-Infant Data (REMInD): a proof-of-concept Data to Care study to support retention in maternal HIV treatment and infant HIV testing in Cape Town, South Africa.

Data to Care (D2C) strategies - using routine data to facilitate identification and linkage back to care of people living with HIV who are not in care - have shown promise in high-income settings but received little attention in lower resourced or vertical HIV transmission prevention (VTP) contexts. In this proof-of-concept study, we monitored existing linked electronic medical records in near real-time to identify key gaps in postpartum VTP steps among 336 mothers living with HIV and their infants in Cape Town, South Africa (recruited March 2021 - April 2022). We attempted to confirm observed gaps through source data systems and telephonic tracing, and facilitated re-engagement in care where needed. There were 302 gaps observed in the routine data; 123 (41%) were false gaps and 179 (59%) were considered probable gaps (133 mother-infant pairs). Overall, 54 mothers (16%) did not link to HIV care within 12 weeks of delivery, 43 mothers (13%) linked to care but had a gap in ART dispensing by nine months postpartum, 25 infants (10%) did not have an HIV test around 10 weeks and 57 (17%) had no HIV test around 6 months of age. Only 100 of the probable gaps (56%) could be confirmed through telephonic tracing and, of those, only 47 were successfully re-linked to care. Mobility and clinic transfer, fear of stigma and employment-related challenges were commonly reported reasons for gaps in VTP steps. This study highlights that linked routine data sources linking mother-infant pairs across health facilities has the potential to streamline tracing efforts; however, implementation is challenging and, even when gaps are identified, re-engagement in care may be difficult. Further research is needed to combine D2C strategies with interventions addressing broader social and structural determinants of health, and to tailor D2C strategies to fit available resources and data sources in low-resource settings.