Increased access to antiretroviral therapy is associated with reduced maternal mortality in Johannesburg, South Africa: An Audit from 2003-2012

Objective: To assess the impact of expanded access to antiretroviral treatment (ART) on maternal mortality in Johannesburg, South Africa between 2003 and 2012. Methods: Audit of patient files, birth registers and death certificates at a tertiary level referral hospital. Cause of death was assigned independently, by two reviewers. We compared causes of deaths and the maternal mortality ratios (MMR, deaths/100,000 live births) over three periods corresponding to changes in government policy on ART provision: period one, 2003-2004 (pre-ART); period two, 2005-2009 (ART eligibility with CD4 count <200cells/μL or WHO stage 4 disease); and period three, 2010-2012 (eligibility with CD4 count <350 cells/μL). Results: There were 232 deaths and 80,376 deliveries in the three periods. The proportion of pregnant women tested for HIV rose from 43.4% in 2003 to 94.6% in 2012. MMR was 301, 327 and 232 in the three periods, (p = 0.10). The third period MMR was lower than the first and second combined (p = 0.03). Among HIV-positive women, the MMR fell from 836 in the first time period to 431 in the third (p = 0.008) but among HIV negative women it remained unchanged over the three periods, averaging 148. Even in the third period, however, the MMR among HIV-infected women was 3-fold higher than in other women. Mortality from direct obstetric causes such as hemorrhage did not decline over time, but deaths from tuberculosis and HIV-associated malignancy did. In 38.3% of deaths, women had not attended antenatal care. Conclusion: Higher coverage of HIV testing and ART has substantially reduced MMR in this hospital setting. Though the gap in MMR between women with and without HIV narrowed, a third of deaths still remain attributable to HIV. Lowering overall MMR will require further strengthening of HIV services, increased antenatal care coverage, and improved care for obstetric emergencies at all levels of care. Charlotte Maxeke Johannesburg Academic Hospital (CMJAH) provides tertiary-level services for patients in a densely populated sub-district of greater Johannesburg. High-risk women from the surrounding seventeen primary and one secondary facilities are referred to the hospital during pregnancy or childbirth. However, deliveries are not confined to high risk women as only two other smaller facilities provide labour and delivery services and the majority of women in the district have their deliveries at CMJAH. About 95% of pregnant women in the province attend antenatal care at least once, and even more have a trained nurse, midwife, or doctor present at birth.[10] At labor onset, women present for assessment and admission to CMJAH’s 100-bed maternity unit. Provided a woman is well, she is usually discharged within 24 hours of childbirth, or after three days following complicated deliveries or Cesarean sections. Women are asked to attend their local clinic for postpartum assessment within a week of delivery and to continue HIV care if they are HIV positive, but uptake of these services is low.[8] HIV testing in South Africa was initially based on voluntary counselling and testing, but this changed to provider-initiated HIV testing and counselling (PITC) in 2008. For pregnant women, HIV testing occurred at the first visit for antenatal care and was repeated at the 32nd week of pregnancy in women who initially tested negative. If HIV testing was declined by a pregnant woman, it was offered again at subsequent visits and during labor. As indicated above, ART provision is characterized by three distinct time periods. In the first, 2003 to March 2004, ART was not available in the state sector, and HIV-infected women only received single-dose nevirapine for PMTCT prophylaxis, which was the prevailing PMTCT regimen until 2008, when zidovudine maternal and infant prophylaxis were introduced. In the second period, between April 2004 and March 2010, HIV-infected people (including pregnant women) with a CD4 count ≤200 cells/μL or WHO stage IV clinical disease, qualified for ART and were usually initiated on stavudine, lamivudine and nevirapine. Efavirenz was included as an alternative to nevirapine after the first trimester of pregnancy. Due to delays in implementing the ART guidelines for pregnant women at CMJAH, very few pregnant women were initiated on ART in 2004 “Fig 1”.[11] For this audit, therefore, the months April to December 2004 are considered part of period one. The third period begins from April 2010, when the CD4 cell count threshold for ART initiation was raised. Pregnant women with a CD4 count ≤350 cells/μL were offered ART, consisting of nevirapine, lamivudine and tenofovir. Because of concerns about teratogenicity at the time, efavirenz was no longer offered to pregnant women, barring clinical conditions that contraindicated the use of both nevirapine and lopinovir/ritonovir. However, following reports of maternal deaths related to nevirapine toxicity, the guidelines were reversed in mid-2012, with efavirenz replacing nevirapine in ART regimens.[12] As data on number of births were only available for 2010 and not for individual months, the months January to March 2010 was included in the third period, even though women with a CD4 count between 200 cells/μL and 350 cells/μL did not initiate ART during those three months. Cotrimoxazole prophylaxis was provided to HIV-infected pregnant women throughout the study period, as was isoniazid prophylaxis for women with no suspicion of tuberculosis. A Number of pregnant women initiated on ART in a sub-district of Gauteng Province, South Africa from 2004 to 2012, and key events. Key: vertical stripes = ART initiation at primary health care antenatal clinic (ANC); grey shading = ART initiation at district hospital ANC: white = ART initiation at Charlotte Maxeke Johannesburg Academic Hospital ANC; black = women who had initiated ART prior to pregnancy. Key events: 1 = ART initiation for people with a CD4 count < 200cells/μL or WHO stage 4, 2 = Labour dispute in health sector, 3 = Decentralisation of ART provision to primary care facilities, 4 = ART initiation for pregnant women if their CD4 count < 350 cells/μL. B Maternal mortality rate among women from 2004 to 2012 at Charlotte Maxeke Johannesburg Academic Hospital. Circle = HIV-positive women MMR; diamond = HIV-negative women. Weighted least-squares fit of linear association between year and number of deaths. The slope of the line for HIV-positive women is -54.4/year, 95%CI = -90.4 to -18.4; p = 0.007. The slope of the line for HIV-negative women is 0.08/year, 95%CI = -23.1–23.3; p = 0.994 Some aspects of service delivery warrant description. Firstly, in the second and third periods, women testing HIV positive had a CD4 cell count test and results were available two weeks later, as point of care tests were not available at the time.[11] Secondly, HIV testing and other services were affected for extended periods in 2007 and 2010 by a series of labor disputes, involving nurses, and by interruptions in payments for lay counselors.[13–15] Finally, in early 2008, ART was provided only at CMJAH and two large referral facilities, but by the end of 2012 the services had expanded progressively to cover all 17 facilities in the sub-district. Prior to 2008, women attending facilities where ART was not available were referred to CMJAH antenatal clinic if eligible for treatment. CMJAH patient and laboratory records were audited for all facility-based maternal deaths between 2003 and 2012, and data extracted on the characteristics of women who had died, their HIV status, obstetric history, clinical condition and timing of death in relation to pregnancy. Maternal deaths were defined as the death of a woman at the facility during pregnancy or within 42 days of childbirth.[16] Information was not available about maternal deaths that occurred at home or at other facilities. The cause of death in each case was assigned following discussions at mortality meetings of the CMJAH Department of Obstetrics and Gynaecology. An internal medicine specialist (AB) and an infectious diseases specialist (VB), using all available evidence including clinical records, prescription charts, nursing reports, laboratory results and mortality audit reports, then independently verified cause of death and its classification. Any discrepancies between reviewers were resolved through discussion. Women who had died following a spontaneous or induced abortion (termination of pregnancy before 20 weeks gestation)[17] were considered as having died during pregnancy. Deaths were classified as direct obstetric deaths (due to obstetric complications including pregnancy hypertensive disorders, hemorrhage, pulmonary embolism, pregnancy related infections and iatrogenic factors), indirect obstetric deaths (resulting from a pre-existing disease that was aggravated by the physiological effects of pregnancy, such as cardiac disease, end organ disease and non-pregnancy related infections), or unknown.[18] Deaths that were considered accidental or incidental were excluded. The total number of deliveries and HIV status of women was determined through a review of hospital birth registers. The District Health Information System (DHIS) provided information on the total number of women in the sub-district who had initiated ART or were already receiving treatment at the time of pregnancy (over two thirds of the women in the sub-district give birth at CMJAH). The annual maternal mortality ratio (MMR = number of maternal deaths at CMJAH per 100,000 live births at CMJAH) was calculated for the facility, and for women with or without HIV infection, or of unknown HIV status. The proportional reduction in MMR that would occur if no women had HIV was estimated by the population attributable fraction (PAF). This was calculated as PAF = [p(r1-r0)]/r where r1 is MMR in HIV positive women, r0 is MMR in HIV negative women, r is the overall MMR and p is prevalence of HIV. Data were analyzed using Intercooled Stata version 12.0 (Stata-Corp, LP, College Station, TX). Patterns in maternal deaths and in the performance of the HIV programme were compared between the three time periods. Chi-square tests were used to assess differences between categorical variables. A Chi-square test for trend determined whether there was a trend over the three periods in the proportions within each exposure category, together with a chi-squared test of homogeneity of odds was also done.[19] Approval for the study was given by the Human Research Ethics Committee of the University of the Witwatersrand (M101150).