A longitudinal analysis of the completeness of maternal HIV testing, including repeat testing in Cape Town, South Africa

Abstract Introduction The virtual elimination of mother‐to‐child transmission of HIV cannot be achieved without complete maternal HIV testing. The World Health Organization recommends that women in high HIV prevalent settings repeat HIV testing in the third trimester, and at delivery or directly thereafter. The Western Cape Province (South Africa) prevention of mother‐to‐child transmission (PMTCT) guidelines recommend a repeat maternal HIV test between 32 and 34 weeks gestation and at delivery in addition to testing at the first antenatal visit (ideally <20 weeks gestation). There are few published longitudinal studies on the uptake of initial and repeated maternal HIV testing programmes in sub‐Saharan Africa. We aimed to investigate the implementation of initial and repeat maternal HIV testing guidelines in Cape Town, South Africa. Methods Between 2013 and 2016 we established an electronic PMTCT register that consolidated routine data from a primary healthcare facility and its secondary and tertiary referral sites in Cape Town. This provided a longitudinal record for each participant, from first antenatal visit to delivery. Utilizing these data, we conducted a retrospective analysis investigating the completeness of maternal HIV testing according to the PMTCT HIV testing guidelines in Cape Town, and predictors of complete testing, from 2014 to 2016. Results Among 8558 enrolled pregnant women, 7213 (84%) were not known to be HIV positive at their first visit and thus eligible for HIV testing; 91% of them received ≥1 HIV test during pregnancy/delivery. Testing at the first visit was 98% among the 85% of women who attended antenatal care. Among women eligible to receive all three recommended HIV tests, only 11% achieved all three tests. Delivery HIV testing completion among all women without an HIV‐positive diagnosis was 23%. HIV prevalence at delivery was 21% and HIV incidence between first visit and delivery in those with ≥2 HIV tests was 0.2%. Women who enrolled after 2014 were more likely to receive the three recommended tests (aOR: 1.41; 95% CI: 1.10 to 1.81) and retest at delivery (aOR: 1.20; 95% CI: 1.05 to 1.39). Conclusions Implementation of maternal HIV testing in Cape Town improved between 2014 and 2016 but major gaps remain, particularly at delivery.


| INTRODUCTION
The implementation of policies recommending immediate lifelong triple antiretroviral therapy (ART) initiation for all HIVpositive pregnant women (Option B+) has been shown to be effective in reducing mother-to-child transmission (MTCT) of HIV worldwide [1,2]. Women's access to these and other HIVrelated services in the prevention of MTCT (PMTCT) continuum relies on complete universal antenatal care (ANC) HIV testing, which is essential for the virtual elimination of MTCT of HIV [3,4]. It is estimated that failure to detect maternal HIV infection during ANC is responsible for up to 54% of MTCT in resource-limited settings [5].
The World Health Organization (WHO) recommends that provider-initiated HIV testing and counselling (PITC), forming part of ANC services, be offered to pregnant women at their first ANC visit (recommended at ≤12 weeks gestation) [6,7]. Not only will this enable the timely diagnosis of undiagnosed pregnant women living with HIV, it will also provide HIV-negative women with guidance counselling and services to assist with the prevention of HIV acquisition and the opportunity to retest later during the pregnancy [8]. In 2010, WHO recommended that women in areas with generalized HIV epidemics be re-tested for HIV in either the third trimester, at delivery, or directly thereafter [9]. In 2014, in the Western Cape (WC) Province, South Africa (SA), the Western Cape Prevention of Mother-to-Child Transmission of HIV (PMTCT) Clinical Guidelines recommended a repeat ANC HIV test between 32 and 34 weeks gestation and again at delivery, subsequent to testing at first ANC visit (recommended at <20 weeks gestation) [10]. Retesting is important as pregnant women are more likely to acquire HIV infection compared to non-pregnant women, and, if acutely infected, MTCT risk is increased in comparison with women who acquire HIV infection prior to pregnancy [11][12][13][14][15][16]. Incident maternal HIV infection was predicted to account for up to 34% of vertical infant infections in 2014 [17]. In addition, the rapid antibody assays may not accurately diagnose HIV during the window period (up to three months post acquisition) [13]. In settings such as SA where >80% of deliveries occur in facilities [18], HIV testing at delivery provides a particular opportunity to test women with an unknown HIV status who do not attend ANC [19,20].
Across sub-Saharan Africa (SSA) initial antenatal HIV testing proportions range from 35% in Ethiopia to 99% in Malawi, but are mostly below the minimum 90% required to enable 90% of HIV-positive pregnant women access to ART [4,[21][22][23][24][25]. However, most studies were conducted before 2015 with very few reporting on post-Option B+ policy implementation testing completion. Furthermore, these studies often use self-report to determine HIV testing completion which could skew results due to reporting bias. In order to better understand the current (i.e. post-Option B+) implementation and uptake of antenatal HIV testing in SSA studies utilizing recent routine data from ANC services are required. Additionally, since repeat HIV testing is provided sequentially throughout pregnancy, potentially involving multiple service providers, longitudinal individual maternal data are needed to accurately estimate repeat testing coverage. However, very few studies have collected such data and thus reports on coverage of repeated HIV testing in SSA are lacking [16,[26][27][28].
Utilizing prospectively collected, individual-level, longitudinal patient data from a primary healthcare facility and its referral sites (secondary and tertiary) we aimed to investigate the implementation of and adherence to initial and repeat maternal HIV testing PMTCT guidelines in Cape Town, SA. We assessed: the coverage and timing of initial HIV testing during pregnancy, repeat HIV testing in the third trimester and at delivery, HIV prevalence and incidence among those tested, and the predictors of maternal HIV testing completion.

| Study design
This study was a retrospective analysis conducted on antenatal maternal HIV testing data that were prospectively collected for the Closing the Gaps (CTG) study which has been described previously [29]. Briefly, the CTG study established an integrated electronic PMTCT register (e-register) that consolidated routine data from a primary care obstetric facility, Mitchells Plain Midwife Obstetric Unit (MPMOU), and its referral sites: Mitchells Plain District Hospital (MPDH), Mowbray Maternity Hospital (MMH) and Groote Schuur Hospital (GSH), within the urban Klipfontein/Mitchells Plain Sub-Structure of the WC, SA.

| Study setting
In the Western Cape, women with uncomplicated pregnancies can access antenatal care at Midwife Obstetric Units (MOUs) which can manage normal deliveries, or basic antenatal care (BANC) facilities which provide outpatient services only (these women will present to the MOU for delivery). MPMOU is within the Mitchells Plain Community Health Centre which is accessed by approximately 1.2 million people. Among community members, an estimated 24% are unemployed and 16% have lower grade education. At the time of the study MPMOU (primary care) provided ANC, including HIV testing, counselling and treatment, as well as midwife-assisted uncomplicated vaginal delivery services [30]. High-risk pregnancies or those requiring advanced care were referred from MPMOU to a secondary or tertiary care facility during the antenatal or peri-partum period. MPDH and MMH (secondary level) are equipped with operating theatres and neonatal intensive care unit (ICU) facilities respectively while GSH (tertiary level) has both adult and neonatal ICU facilities.
The 2015 WC antenatal HIV prevalence was reported to be 19% [31].

| Study participants
In the CTG study, pregnant women, irrespective of HIV status, were passively enrolled at their first ANC visit at MPMOU, or alternatively, upon presentation for delivery if they had not received ANC. HIV tests (rapid antibody assay) were performed by community care workers as part of routine ANC services and by nurses in the labour ward of MPMOU and the labour/postnatal ward of the hospitals.
We included data from the CTG e-register for all enrolled women who delivered at MPMOU or one of its referral sites, between 1 July 2014 and 31 December 2016, with a live or still birth pregnancy outcome. We excluded data from women who (a) presented for delivery having attended ANC at a BANC since we did not have access to complete longitudinal HIV testing data for these individuals or (b) for whom no pregnancy outcome could be found as we could not exclude pregnancy loss, in which case PMTCT testing guidelines would no longer be applicable.

| Procedures and measurements (Box 1)
A longitudinal record, including HIV testing history from first antenatal visit (if attended) through to delivery, was extracted from the CTG e-register for each woman. We categorized women as having attended ANC if they presented to a health facility >5 days before delivery. We defined early ANC attendance as ≤22 weeks gestation. We deemed all women eligible for an initial test if on their first visit, whether antenatally or at delivery, they were not already known to be HIV positive; and eligible for repeat testing if they had not had a prior positive antenatal HIV test. Tests were considered to be retests if women underwent an HIV test within the third trimester and/ or at delivery following a prior negative HIV test. We considered women eligible for retesting in the third trimester if they delivered >35 weeks gestation. According to the guidelines there was no minimum time period between tests. The primary outcome was receipt of an HIV test (yes/no): (1) at first visit (ideally ≤22 weeks, but up to delivery); (2) in the third trimester as a repeat test; and (3) at delivery as a repeat test. The secondary outcome was HIV-positive diagnoses (yes/no) at each test.
We calculated testing completion at: • First ANC visitin all eligible women who attended ANC.
• <28 weeks gestation (first test only)in eligible women who attended their first ANC visit <28 weeks gestation. • Third trimester (first and repeat testing assessed)in eligible women who attended their first ANC visit before/during the third trimester (i.e. ≥28 weeks gestation); were not previously diagnosed HIV positive and delivered >35 weeks gestation. • Delivery (AE5 days) (first and repeat tests)in eligible women irrespective of gestational age at delivery, ANC visit attendance or testing history.
Women who tested at first ANC visit were also categorized as having tested <28 weeks gestation, in the third trimester and/or at delivery. Testing completion at each recommended window was calculated only among women eligible to be tested at that respective window ( Figure 1; Table 2 column 1). Data abstracted from the e-register did not record ANC visits between the first visit and delivery unless a test was recorded. Eligibility for repeat testing in the third trimester was hence based on prior ANC attendance.
In addition, we calculated the testing completion restricted to the windows defined in the PMTCT guidelines (best practice) as follows: • Eligible women must have attended ANC early (≤22 weeks gestation) and delivered >35 weeks gestation with a prior HIV-negative status. • Completion was defined as testing at three timepoints: ≤22 weeks gestation, in the third trimester, and at delivery.
Individuals diagnosed HIV positive during the study period were coded as "diagnosed at enrolment" if they tested HIV positive at their initial ANC test, or as "seroconverts" if they  The PMTCT guidelines assume best practice. Women do not attend ANC at recommended gestational ages and guidelines may be loosely implemented, for example, women may attended their first ANC visit and retest within the third trimester.
We assessed HIV testing completeness as follows: 1. Testing completion at first visit at any timepoint (i.e. < or ≥28 weeks gestation, or at delivery).
▪ Longitudinal testing was assessed with both relaxed ( Table 2) and strict (Table 3) eligibility criteria for retesting: O Relaxedwomen could retest in third trimester and/or at delivery regardless of gestational age at first test. O Strictwomen could only retest in third trimester and at delivery if they tested ≤22 weeks gestation and retested in the third trimester respectively.
tested HIV-positive after a previous negative antenatal HIV test. We categorized women as having an "uncertain HIV status at delivery" if they tested HIV negative at least once during ANC but had not received a test within three months prior to or at delivery; and having an "unknown HIV status at delivery" if they never tested during the current pregnancy. For women with two pregnancies within the study period, we included the pregnancy for which most data were available. Premature delivery was defined as delivery <37 weeks gestation.

| Data analysis
Data analysis was carried out using Stata version 13.0 (Stata Corporation, College Station, Texas, USA). Differences between the characteristics of all participants, stratified by HIV status post delivery, were assessed using the two-sample t-test (normal distribution) or the Wilcoxon rank-sum test (non-normal distribution) for numerical data and the v 2 or Fishers Exact test for categorical data. A descriptive analysis (proportions) was used to assess HIV testing completion during pregnancy among participants for whom gestational age data was available. We used logistic regression to assess predictors of maternal HIV testing completeness. Multivariable models included several variables selected a priori as being possible risk factors (age at enrolment, gravidity, and year of enrolment) for the outcome. Additional variables were included by sequentially adding them based on univariable analysis of significance. Those that were either significantly associated with the outcome of interest after adjustment for other variables (p < 0.05) or that altered the odds ratios (ORs) for other variables in the model by ≥10% were retained [32].

| Ethics
The University of Cape Town Human Research Ethics Committee and the Provincial Government of the Western Cape Department of Health Research Committee approved the study. The CTG study was granted a waiver of informed consent for the e-register as all data had already been collected routinely by health services. No participant recruitment was required.

| Participant characteristics
Between July 2014 and December 2016, 8558 women delivered at MPMOU or one of its referral sites with a live or still birth pregnancy outcome (Table 1). Within the study period, 153 women had two pregnancies of which only one pregnancy was included.

<28
weeks gestation, in the third trimester or at delivery; d % positive was calculated as a proportion of woman who received a repeat test in the third trimester or at delivery. In the third trimester the denominator included individuals who had their first ANC visit and had >1 test within the window; e Completion of retesting in the third trimester does not distinguish between visit coverage and re-testing coverage as we did not have a record of visit data separate from testing data.

| DISCUSSION
This is, to the best of our knowledge, one of the first longitudinal studies of the implementation of maternal HIV testing guidelines in a routine setting in SA. The majority of women eligible for HIV testing tested at least once (91%) during pregnancy or at the time of delivery. Although testing completion at first ANC visit was high (98%), women attended ANC late with only 39% doing so <20 weeks gestation. While HIV testing implementation improved over time, with increased completion in the second year after guideline change, there were substantial missed testing opportunities with a third of women not testing within three months of or at delivery and only 11% of eligible women being tested in all three recommended windows.
The high proportion of women receiving ≥1 HIV test in our study is encouraging; however, it falls short of a) the global PMTCT target of having over 95% of pregnant women aware of their HIV status and b) the >95% "uptake of antenatal HIV testing" rate reported for SA between 2010 to 2012 [15]. This observed difference may be due to our use of operational data from a single referral chain of health facilities in one area of Cape Town, including testing completeness in women presenting at delivery with no prior ANC, versus the use of nationwide aggregate antenatal survey data. This is underscored by earlier gestational age at first ANC visit (per additional week gestation) being a significant predictor of initial HIV testing in pregnancy in our analysis, concurring with results from Zimbabwe [33]. These results suggest a need for additional interventions to improve the implementation of initial HIV testing among women who present to ANC late in pregnancy or not at all.
In SA attending ANC before 20 weeks gestational age, regardless of HIV status, is a PMTCT-related indicator [15]. It is also recommended in national and WC guidelines that women are HIV tested at their first visit [8,10]. Reassuringly, in our study the implementation of HIV testing at first ANC visit was excellent, but unfortunately many women attended ANC late i.e. after 20 weeks gestation. Furthermore, a high proportion of women presented at delivery with unknown HIV status and having not attended ANC. Of the women who did not attend ANC and tested at delivery, 1.5% tested HIV positive which may be an underestimate as less than half of women who did not attend ANC received an HIV test at delivery. These women and their infants left the facility with unknown HIV and HIV exposure status respectively. Untested women may have had a different HIV risk profile to tested women.
Reassuringly, implementation of maternal HIV testing in our setting improved over time following the adoption of the updated PMTCT guidelines in 2014, similar to results from Kenya [26]. Women were more likely to complete testing if they enrolled in 2015/2016 as opposed to in 2014. Notwithstanding, we observed several missed repeat testing opportunities; a third of women had an unconfirmed HIV status after delivery as they had either never tested or last tested ≥3 months before delivery. This is lower than that observed in Kenya [26]. Only a small proportion of eligible women received all three recommended HIV tests. The data suggest that poor implementation of delivery testing was responsible for the bulk of missed opportunities.
At delivery, women were more likely to HIV retest at either of the three referral hospitals (MPDH/MMH/GSH) versus MPMOU. At these facilities there were dedicated PMTCT nurses designated to do HIV testing in contrast to the MOU where HIV testing fell within the remit of all midwives. Women were often tested in the labour ward upon arrival, but otherwise had other opportunities to test in the postnatal ward during an admission of a few days. In primary care, women are often discharged within six to eight hours of an uncomplicated birth.
HIV prevalence among women with a known HIV status at delivery was similar to that reported for the WC [31]. Of the HIV-positive women the vast majority were known positive at   enrolment indicating the maturity of the HIV programme with high overall testing coverage. It is also reassuring that the first of the 90-90-90 targets [15], that 90% of HIV-positive people know their status, has been met. It should however be noted that this proportion may be slightly overestimated due to missed positive diagnoses among women that never tested. Furthermore, awareness of HIV status does not guarantee that women have initiated and adhered to ART. Further studies are required to assess viral load suppression among known positive women. The overall HIV incidence among women that initially tested HIV negative was estimated to be substantially lower than that previously reported in SA and SSA (3 to 4%), but similar to that of non-African countries (0.3%) [12,34]. While our study may substantially underestimate incidence due to the low completeness of repeat HIV testing, assessment of the most cost-effective number and timing of maternal HIV tests in different incidence settings, and feasibility of implementation, should be considered for future studies.

| Strengths and limitations
The use of individual-level longitudinal data allowed for a more accurate assessment of testing completion, based on each woman's progress through the PMTCT continuum as opposed to using aggregate data. The use of longitudinal data also enabled the description of (a) the timing of HIV-positive diagnoses and (b) HIV incidence estimates over the course of pregnancy. Point-of-care HIV testing data are not routinely digitized in SA so to our knowledge these data are rare. We had a large sample size which included antenatal and delivery HIV testing data from primary through tertiary care facilities. Positively, our study demonstrated the real-world implementation of PMTCT testing guidelines; however, we were dependent on the quality of routine operational data and could not account for missing data. Furthermore, qualitative data were not routinely collected, and we could not explore qualitative factors associated with testing completion. At visits other than first ANC visit and delivery, we did not have a record of visit data separate from testing data and therefore could not distinguish between visit coverage and re-testing coverage in the third trimester. It is therefore possible that retesting coverage in the third trimester is reflective of poor ANC attendance as opposed to poor implementation of PMTCT guidelines. Although we excluded women who attended any documented ANC visits at facilities outside MPMOU, we were unable to determine whether women had received point-of-care HIV testing at facilities not included in the study, or outside the WC province. The subset of women without recorded pregnancy outcomes may have had systematically different testing outcomes to those for whom outcomes were recorded and their exclusion from the dataset may therefore have introduced selection bias into the study results, thereby slightly overestimating testing completeness. Although the results of this study may be externally valid to the WC, HIV testing is context-specific, and results should be treated with caution when generalizing to other settings.

| CONCLUSIONS
The results of this study illustrate that although there has been maturation of maternal HIV testing within the PMTCT programme over time, gaps remain in late pregnancy followup testing, particularly at delivery. Interventions are required at facility level to improve delivery HIV testing among women starting ANC late/not attending ANC and those that have high-risk pregnancies in order to limit the transition of undiagnosed HIV-positive women to the postnatal period without access to lifelong ART, feeding support and infant post-exposure prophylaxis. Additional research to assess (a) the viral suppression among known HIV-positive women and (b) the cost-effectiveness of PMTCT testing guideline implementation is required.

SUPPORTING INFORMATION
Additional information may be found under the Supporting Information tab for this article. Figure S1. Delivery HIV testing completion at respective delivery facilities. Figure S2. Diagnoses among HIV-positive women a (n = 1344).