African countries are implementing Option B+, but high loss to follow-up (LTF) among pregnant women on antiretroviral therapy (ART) threatens program success and mother/infant lives. Due to the lack of linked data, LTF estimates do not account for unreported transfers. We hypothesize that “clinic shopping” and rural-urban travel after delivery may inflate LTF estimates. To test this, we traced lost patients using a national lab database in South Africa to assess continuity of care and update LTF estimates.
We included all HIV+ women initiating ART during pregnancy at seven clinics in Gauteng Province from 1 Jan 2012 to 31 July 2013 and considered LTF (no visit >3 mo; n=312). Using name and date of birth, we manually searched the National Health Laboratory Service database. Records were categorized as from the initiation site or a new facility. Continued HIV care was defined as accessing care after ART initiation at a new facility shown by ≥1 CD4 or viral load test on record, or any record from a new ART clinic. “Clinic shoppers” were defined as seeking care at a new ART facility within Gauteng.
At ART initiation, median age was 29 years (IQR:25-33) and CD4+ cell value was 268 cells/µL (200-340). Median time between initiation and last clinic visit was 112 days (29-268). Records were missing—including from the initiation site—for 115 (36.9%) women. Of the 197 located, 97 (49.2%) continued HIV care at a new facility. Most (71.1%) were clinic shoppers; 28.9% sought care in other provinces. Overall median time out of care was 406 days (238-734). Compared to women accessing care in other provinces, clinic shoppers stayed out of care longer (median 530 days, IQR:332-808 vs. 269, IQR:72-409, p<0.01) and median CD4 upon care reentry trended lower (317 cells/µL, IQR:159-610 vs. 499, IQR:213-571). Considering all 97 women as engaged in care, cohort LTF drops from 38.1% to 26.3%.
We found substantial continued care among women considered LTF after initiating ART during pregnancy, both within the same city and in other provinces. This highlights the difficulty of producing accurate estimates of retention in care and underscores the need for a unique identifier and a national, linked health database. We also found that women are suspending care for extended periods of time with consequent immunosuppression. More must be learned as to how women choose HIV facilities, access care, and travel around the time of delivery; continuum of care estimates may be overly pessimistic.