In this combined prospective and retrospective observational cohort study, clinic records for HIV-infected pregnant women indicated multiple delays and interruptions as well as loss to follow-up in the initiation and continuation of AZT and lifelong ART. Maternal age, gravidity, parity, and baseline CD4 count of ART-eligible women diagnosed with HIV at first ANC were similar to previously reported studies in South Africa; median gestational age was slightly younger [3-5]. In this study, ART-eligible women presented at a median gestation age of 24 weeks. The proportion of ART-eligible women who initiated treatment within 60 days was surprisingly low (21%). While women who did not initiate ART within the 60 days follow-up may still have initiated ART prior to delivery, duration on ART, and therefore the potential effectiveness of ART [6,7], would have been limited. The long delays in initiation, treatment interruptions, and high loss to treatment found in this study occurred despite large-scale implementation of interventions in PMTCT programs in South Africa, including fast-tracking of pregnant women with reduced adherence counseling and ART preparation, nurse-initiation of ART, and integration of ANC and ART services at the same clinic.

Initiation of AZT was also below expectation. According to clinic files, only 69% of women who were documented as being eligible for initiation of AZT received AZT at their first visit. 11% of women were missing Hb test results, which has clinical implications for the management of anemia as well as AZT eligibility. Problems which limited the same day initiation of AZT in this study are likely to carry over to the new Option B+ guidelines implemented in South Africa, which call for all HIV-infected women to receive a fixed-dose, triple therapy combination at first ANC visit.

The current study was not designed to determine the cause of the problems with initiation and adherence. Clinic managers and nursing staff at the sites indicated that AZT, when available, was always dispensed. However, multiple health systems issues that have been identified elsewhere as barriers to adherence to PMTCT regimens [8] were encountered during the study period. One clinic experienced a stock-out of HIV rapid test kits and enrollment was interrupted as women had to return to the clinic for laboratory-generated HIV results before initiating CD4 test and receiving AZT. The other clinic experienced repeated stock-outs of AZT, including one period of more than 6 weeks (just prior to the study period) when AZT was not available. Health care worker shortages caused difficulties for patient care as well, especially during holiday periods as staff members on leave for studying, vacation, or illness were not replaced. Thus, women may have presented for care but been turned away because of staff, drugs, or consumable shortages.

The two clinics were selected for convenience reasons and may not be generalizable to other clinics within South Africa. However, both clinics are located within the city of Johannesburg and thus, on average, are more likely to have access to staffing, consumables, and supervision than facilities in more rural and remote areas.

The study was designed to evaluate the early steps of the PMTCT cascade, and in particular whether interventions such as same day initiation of AZT for all women and fast tracking of ART initiation for eligible women were occurring; therefore, a short period of follow-up was chosen. This study also did not match mother-infant pairs to determine if the delays, losses and interruptions documented here affected transmission; it may be that treatment is robust despite these challenges. While women may have initiated ART, continued AZT beyond the 60 days, or delivered a healthy, HIV-free baby, earlier initiation maximizes the opportunity for prevention [6,7].

A limitation of this study is that it was a review of clinic records; it may be that clinical practice differed from what was recorded in clinic records. However, for HIV-infected persons initiating ART, patient medical information is written in individual patient files, kept by the facility and stored within the ART clinic; one of the clinics also captured patient files into an electronic medical record system for HIV management. For this set of pregnant women, who are both ANC and ART patients, their medical records from first presentation at the ART clinic within the sites are as good as for any other ART patients (and only 21% known eligible initiated treatment with 60 days of HIV testing).

However, within the ANC clinic itself, no individual clinic-maintained patient files or clinic cards were in use at the time of the study. ANC patient cards are to be completed at each visit and women are strongly encouraged to bring these patient cards to each visit and at delivery. Nursing staff are asked to duplicate, by hand, information recorded in the patient cards in the clinic ANC register as well. It may be that some of the missing visits identified in this analysis were completed by the patients and recorded in the patient cards but never recorded in the ANC clinic register.

Each ANC clinic had one ANC register, which is a very large format hardcover, pre-printed book with follow-up visits marked at 6-weekly intervals. Because each clinic had one large register, it was kept in a central location. With more than one nurse providing antenatal care, so that the single ANC register could not always be in front of the nurse for initial or follow-up visits, nurses often could not enter patient information into the clinic’s ANC register during the consultation. Nurses solved this constraint by writing patient-visit information onto separate sheets of paper during a patient interaction, with the intent of later transcribing such information back into the ANC register. This may have been a reason for poor capturing into the register of Hb, gestational age, and WHO staging. During the study period, however, study staff remained at the sites to complete this transcribing for the sites so that information extracted from the ANC register during the study period was as complete as information obtained at the sites. Study staff also used a pregnancy due date wheel tool from the clinic to fill in missing information for last menstrual period, weeks gestation at time of clinic visit, or expected delivery date. Beyond the study period, such after-the-fact transcribing into the large-format ANC register was no longer completed by study staff and remains a barrier to information management at the sites.

Printed registers and patient cards with follow-up visits marked at 6-weekly intervals for ANC care of HIV-negative women may also have contributed to poor adherence by both the facilities and the patients to the increased number of visits required for ANC care if HIV-infected. Patients were told during their first visit to return in a week for CD4 results, dispensed a 4-week supply of AZT, and received an ANC patient card indicating 6-week visit. For those who returned for a second for a second visit, the median time from first visit to CD4 results was 28 days, coinciding with a need to return to the clinic for a 4-week AZT refill.

Additional registers (e.g. recording return visits for CD4 results, referral to ART initiation, and counseling for ART initiation) were in use throughout the two clinics, mostly informal hand-drawn columns in notebooks. These registers were organized according to the particular use of that section of the clinic (e.g. date of first visit, date of current visit, patient name, patient date of birth, or ANC register number), which caused difficulties in tracing patients through the services. Neither clinic site had dedicated staff for assisting in record keeping or data management for these patients; therefore, study staff assisted the clinics in collecting information from the different sources and completing registers during the period. Again, beyond the study period, the time required for this after-the-fact tracing of patients across the multiple registers is a barrier to information and patient management at the sites.