South Africa is considered to have the largest number of people infected with HIV and AIDS deaths . Government denial in the late 1990s, despite signs of the rapidly increasing epidemic fuelled by the social conditions created by apartheid and migrant labour, resulted in lost opportunities to avert huge loss of life [2,3] and has contributed to pervasive stigma associated with the disease.
Despite having a functional civil registration and vital statistics system, South Africa has been unable to monitor the number of AIDS deaths from routine cause-of-death statistics. Because of a backlog in reporting cause-of-death data at the time, the South African Medical Research Council used the age and sex details of deaths from the National Population Register to draw attention to the emerging increase in young adult mortality . It was concluded that by 2000, AIDS had become the leading single cause of death, with a demographic projection model suggesting that AIDS accounted for 25% of all deaths . After initial attempts by government to repudiate the report , the national statistical office was resourced to overcome the backlog in the production of the cause-of-death statistics . However, under registration, the high proportion of ill-defined natural causes and misattribution of HIV/AIDS as the underlying cause of death made it impossible to provide a reliable count of AIDS deaths in the country. Although cause-of-death coding in South Africa is undertaken according to the standards of the International Classification of Diseases, studies reveal that information regarding HIV status is frequently omitted by doctors from death notifications , and some of the rural deaths are reported by traditional headmen  with consequent underreporting of AIDS as a cause. Despite a rapid increase in the number of deaths  and clear epidemiological indications of an extensive AIDS pandemic (antenatal HIV prevalence of 30%  and population prevalence of 12% ), vital statistics continue to report only 2–3% of deaths being because of HIV/AIDS .
Groenewald et al. [13,14] noted the distinct HIV/AIDS-related age pattern of the mortality increase among nine diseases, including AIDS indicator conditions. Assuming the excess mortality for these conditions compared with rates in an earlier period, when there was little AIDS mortality, were misattributed AIDS deaths, they estimated that 61% of the HIV/AIDS deaths in 2000 were reported to other causes. Using a slightly different approach, Birnbaum et al.  estimated the misattribution of AIDS deaths by assuming that cause-specific mortality rates have a universal age pattern. They compared the ratio of death rates in each age group divided by rates for the 70–84 age group to similar ‘relative death rates’ from the pooled WHO mortality tables of several countries excluding South Africa. The excess between the South African relative death rates and the universal pattern was assumed to be misattributed AIDS deaths, and the authors estimated that that more than 90% of HIV/AIDS deaths during 1996–2006 were misattributed to other causes .
Fazito et al.  applied a similar approach to track AIDS mortality in Brazil. They used WHO estimates of death rates to adjust for completeness of registration. Although the Brazilian cause of death statistics allowed them to make use of pre-AIDS mortality rates dating back to 1985, they did not attempt to account for any changes in the background trend in the source causes. They found that at the beginning of the impact of the epidemic, high proportions of AIDS deaths (75%) were underreported, but that it declined steadily to 13% in 2009 .
Both of the South African studies [13,15] adjusted for overall underregistration of deaths by scaling the vital registration numbers to the estimated number of deaths from models (ASSA2000 and ASSA2003, respectively ), which are now known to overestimate the number of AIDS deaths . In addition, the Birnbaum et al. estimates ignored the fact that significant numbers of injury deaths fall into the broad ill-defined category, and thus they incorrectly counted some of these as misclassified AIDS deaths. Models on the other hand either are not calibrated to observed deaths at all [e.g. Spectrum, the model used by the Joint United Nations Programme on HIV/AIDS (UNAIDS)] or where some effort is made to calibrate to observed mortality [e.g. Global Burden of Disease Study (GBD)], the parameterization of the models is not country specific leading to estimates for a country being determined to some extent by the need to fit data from other countries.
Our study extends the approach developed by Groenewald et al.  and uses empirically based estimates of the completeness of death registration to avoid possible distortion of the estimates through calibrating estimates from an epidemiological/demographic projection model.
Statistics from South Africa have reported annual cause-of-death information from death notifications  using a standard process of coding to International Classification of Diseases-10 and applying the Automated Classification of Medical Entities software since 1997 . Individual-level record data were checked for consistency of age, sex, and underlying cause  and then aggregated into the South African National Burden of Disease Study (NBD) analysis codes . Kaposi's sarcoma (C46) was added to HIV/AIDS deaths (B20-B24) as an indicator condition of AIDS. Several codes have been identified as HIV pseudonyms (B33, B45, B59, and D84) as they are used to code terms such as ‘retroviral disease’ and ‘immune suppression’ that appear on death notifications.
Completeness of death registration was estimated  using death distribution methods for adults (15+ years), comparison with survey and census-based estimates of under 5 mortality for childhood mortality, and linear interpolation between these completeness estimates for intervening age groups. Overall, completeness of death registration increased from 75% in 1997 to 91% in 2010, with the level of registration of adult deaths being higher than that of children. Death rates were calculated using a set of mid-year population estimates consistent with the 2011 census age profile, corrected for age exaggeration at the oldest ages .
Estimation of misclassified AIDS
The trends in the mortality rates of each condition by age were reviewed and those that exhibited the characteristic HIV/AIDS-patterned increase in age-specific rates in young adult ages (the ‘AIDS signature’), excluding maternal causes, were considered to be potential source causes (i.e. causes under which AIDS deaths are misclassified) for misattributed AIDS deaths. The identified source causes are listed in Webappendix A, http://links.lww.com/QAD/A817. Following an approach proposed by Dorrington , a log-linear regression was fitted to the age-specific death rates for all the combined source causes for the period prior to extensive roll-out of antiretroviral therapy (1997–2003) on the lagged antenatal HIV prevalence by sex . The antenatal prevalence was lagged by 5 years for adults, 1 year for infants, and 2 years for children aged 1–4 years (past prevalence rates are given in Webappendix C, http://links.lww.com/QAD/A817). This regression was used to estimate the mortality rate (free of misclassified AIDS deaths) for the source causes in aggregate when the prevalence of HIV was last zero (see Webappendix B, http://links.lww.com/QAD/A817, for a more detailed description of method). To estimate the death rates for each source cause (free of misclassified AIDS) at this point, it was assumed that the proportion of deaths because of each source cause immediately prior to the start of the epidemic is the same as recorded for these causes in 1997.
The background trend in mortality rates of each source cause was estimated from the observed trend in the age group 75–84 years, which was assumed to be largely unaffected by HIV (see Fig. 1). Restricting the age range avoided data challenges related to age misspecification observed in the 85+ years age group [25,26]. A linear regression of the mortality rate on the log of time was fitted to estimate the rate of change in background mortality for conditions (the results of which appear in Webappendix D, http://links.lww.com/QAD/A817).
The number of misclassified AIDS deaths was estimated as the difference between the source cause-specific registered deaths, corrected for underregistration, and the estimated number of deaths from the source causes had there been no HIV. These were obtained from the regression model by setting the HIV prevalence to zero and allowing for the statistically significant background trends (P < 0.01) (Webappendix E, http://links.lww.com/QAD/A817). Statistically significant declines were observed for ill-defined natural causes, tuberculosis (TB), protein energy malnutrition, and other nutritional deficiencies and statistically significant increases for septicaemia, meningitis and encephalitis, and endocrine nutritional and blood disorders. Although protein energy malnutrition mortality rates decreased significantly in older ages, it was decided not to apply this trend to other age groups because of the specific cause of the condition in older age groups. None of the other source causes (excluding AIDS) showed a significant trend. As the study was based on secondary analysis of anonymized data, it did not require ethical review.
Estimates of the number of HIV/AIDS deaths were compared with recently published model estimates in four age groups (0, 1–4 years, 5–14 years, and 15+ years).
Sensitivity of estimated number of HIV/AIDS deaths
Sensitivity of the result to the following specific assumptions was investigated:
- The estimate of completeness of registration whether constant with age and the completeness of nonnatural deaths;
- The redistribution of the ill-defined natural deaths (after removing apparent excess because of HIV/AIDS) to all natural causes including HIV/AIDS;
- The years to which the regression is fitted and allowing for uncertainty about the fit of the regression;
- The choice of source causes;
- Allowing for the uncertainty in trend in non-AIDS source cause deaths over time; and
- The redistribution of those with unknown age or sex.
Registered deaths increased from about 317 000 deaths in 1997 to just over 613 000 in 2006 before dropping to nearly 544 000 in 2010. After adjusting for underregistration of deaths (but not misattribution), 8094 deaths in 1997 were attributed to HIV/AIDS, accounting for 1.9% of the estimated total deaths. A further 6120 deaths were attributed to causes considered pseudonyms for HIV/AIDS, and together, these accounted for 3.4% of the total deaths in 1997.
Nineteen source causes were identified as exhibiting the ‘AIDS signature’. Including misattributed AIDS deaths from the other source causes increased the total number to 60 265 or 14.5% of the total deaths in 1997. Table 1 shows the estimated AIDS deaths, which increased to peak at 283 428 or 42% of all deaths in 2006 (with women experiencing a higher proportion and peaking in 2005). By 2010, the estimated number of AIDS deaths had decreased to 207 685, accounting for 35% of the total deaths. This contrasts with 8.7% attributed to HIV/AIDS or pseudonyms in that year.
Over the study period, only 6.9% of the estimated AIDS deaths were correctly reported as HIV/AIDS. A further 13.3% of the estimated AIDS deaths were attributed to causes considered pseudonyms for HIV/AIDS, whereas 24.1% were attributed to ill-defined natural causes, 23.3% to tuberculosis, 15.9% to lower respiratory infections, 11.0% to diarrhoeal disease, and 1% or less to other conditions. Figure 2 shows the age distribution of the estimated AIDS death by main source causes. In 1997, it was estimated that 30.5% of TB deaths had HIV as the underlying cause. This increased to 76.5% in 2005 and then decreased to 70.8% in 2010 (further details appear in Webappendix F, http://links.lww.com/QAD/A817).
The trends in reported and estimated HIV/AIDS death rates are shown in Fig. 3, highlighting little change in the reported mortality rates across all age groups. Estimated rates show rapid increases in the HIV/AIDS mortality, which started to decline for children under 5 years in about 2004, and for adults aged 15–59 years in 2006 for women and 2007 for men. Mortality rates for women 15–29 years increased to levels of over 5/1000, more than double the rates for men in this age group. In the older age groups, the rates for men were higher than those for women and the mortality rates for people of 60+ years continued to increase over the period, possibly peaking in 2009.
Comparing the estimated numbers of AIDS deaths (indicated as NBD and Birnbaum ) with estimates from local models [Actuarial Society of South Africa (ASSA)  and University of Cape Town (UCT)] and global models (Spectrum  with the 2015 UNAIDS assumptions, GBD2010  and GBD2013 ) it can be seen from Fig. 4 that our empirically based estimates for infants are considerably higher than others from 2004 to 2008, whereas our estimates for 1–4 are similar to most others. It can also be seen that for adults of 15+ years, our estimates lie between ASSA2008 (lowest throughout) and the remaining estimates up to 2005, after which our estimates are very similar to ASSA2008. For children aged 5–14, there is much variability between estimates because the numbers infected are small and there are no (few) new infections in this age group (whereas there are in the other age groups) and limited data on which models can rely, but our estimates are very similar to most of the others up to 2004 after which they level off and are lower than all other estimates.
In terms of the sensitivity analysis, only the following were found to have a significant impact on the number of HIV/AIDS deaths (overall and in adults 15+ years):
- Completeness of death registration: An increase or decrease in the estimate of completeness of all deaths has a direct impact on the estimate of the numbers of HIV/AIDS deaths. Thus, if the estimate of completeness were 5% above the true value, we would underestimate the number of HIV/AIDS deaths by 5%. Overestimation of completeness is more likely than underestimation given that completeness is close to 100%.
- Redistribution of the ill-defined natural deaths: Distributing the remaining ill-defined natural deaths (after removing HIV/AIDS deaths as per other source causes) to all natural causes including HIV/AIDS would, if incorrect, affect the estimate differentially over time. It would exaggerate the deaths by 10% in the early years, dropping to 5% at the peak, and increasing to about 7.5% by 2010.
- Assumption of trend in the source causes excluding HIV/AIDS: Assuming the trend where this is significant when the truth is that there is no trend would exaggerate the number of HIV/AIDS deaths by 10% in the early years, dropping to 0% near the peak, and increasing to about 5% by 2010.
There was greater uncertainty in the case of children with the impact of (2) and (3) being about double than that of adults. Although there is a need to investigate the aggregate uncertainty more systematically using Bayesian modelling, given the counteracting influences observed in the sensitivity evaluation, we consider that it is likely that in aggregate, the true number of HIV/AIDS deaths lie within 5% of the estimates produced in the study.
Our estimates are empirically based and indicate that the number of AIDS deaths in South Africa increased steadily to a peak of more than 283 000 deaths/year in 2006, around 770 deaths/day. From 2007, the number of AIDS deaths started to decline and reached about 207 000 in 2010. These results are consistent with the extensive roll-out of antiretrovirals in South Africa from 2004 [32,33] and effective prevention of mother-to-child transmission from 2003 . However, at more than 560 deaths/day, AIDS was still the leading single cause of death in South Africa.
Early modelling of the demographic impact of the epidemic, without significant interventions, predicted that a cumulative total of between 4 and 7 million deaths might be expected by 2010 . We estimate that by the end of 2010, slightly below 3 million people had died of HIV/AIDS. Much of the difference can be attributed to overestimation because of limited knowledge of the epidemiology, behaviour change, and treatment impact. However, the ASSA2008 model suggests that interventions have probably averted over 1 million deaths in this period.
Compared with other empirical approaches, our estimate of 144 955 HIV/AIDS deaths for the year 2000 is slightly lower than the 153 257 previously estimated by Groenewald et al. . This is most likely explained by no longer relying on a demographic and AIDS model to provide the overall completeness of death registration. Our estimate of 283 428 HIV/AIDS deaths in 2006 is also lower than the estimate of 354 500 by Birnbaum et al. . Although the difference is exacerbated by their inclusion of undetermined injuries as misattributed AIDS deaths, being further into the epidemic increases the impact of their having relied on a model to provide the overall completeness of death registration.
Our estimates of HIV/AIDS mortality are consistent with trends reported by local in-depth health and demographic surveillance sites which have observed the same age pattern and reversals in HIV-related mortality following a rapid increase . The epidemic experienced at the Africa Centre site [36,37] in rural KwaZulu-Natal appears to have been more rapid than that experienced in the Agincourt site in rural Mpumalanga . The two span the national trend that we have estimated.
When compared with modelled estimates, our estimates are lower than GBD2010 (close to 305 000) and UNAIDS 2014 estimate (nearly 388 000) for 2006, the year when AIDS deaths appear to have peaked. The more recent GBD2013  estimates (252 500 in 2006) appear to be more reasonable, although they fail to decline after that so that they are the second highest by 2010. Our estimate of the number of AIDS deaths is likely to understate the true number. With the exception of infant deaths, our estimates tend to be lower than the estimates from models. Our higher number of infant deaths, previously unnoticed when assessed in the broader age group of deaths under 5 years , can in part be attributed to an underestimation of the recent number of births by models. Evaluation of the age pattern of the deaths considered not to include misattributed AIDS death, however, suggests that a small number of AIDS deaths could possibly have been attributed to causes not included in this study. Stroke and asthma mortality rates showed a slight increase in young adult ages that followed the same timing as the AIDS deaths.
Of concern regarding the GBD and UNAIDS estimates is the difference in the timing of the peak of the epidemic and particularly the continued increase in adult AIDS deaths after 2008 that they are projecting, despite the extensive provision of antiretrovirals. In all probability, the estimates produced by models could be improved if greater efforts were made to calibrate the output from the models to all-cause mortality estimates, corrected for underreporting .
Although our study has limitations, it is clear that HIV/AIDS is underreported as a cause-of-death in the vital statistics in South Africa, with 93% of AIDS deaths misattributed to other causes. Our estimates of AIDS deaths are dependent on the estimates of completeness of registration of deaths which is reasonably robust for adult deaths, less certain for children under 5 years, and even more so for adolescents. Thus some uncertainty, about both level and trend over time, remains for these age groups. In addition, the trends in non-AIDS mortality for causes from which misclassified AIDS deaths were removed was set to that observed in the 75–84 age group. Implied by this assumption is a downward trend in mortality from non-HIV-related TB, indicating that the heavy TB mortality toll does not necessarily reflect a failing TB control programme, but rather a consequence of the HIV pandemic. This assumption needs to be validated through other epidemiological data. However, the TB cure rate has increased from 53.7% in 2001  to 75.8% in 2010  pointing to the possibility of a strengthening of the TB programme despite increasing numbers of TB cases. In contrast, the assumption indicates increasing mortality from septicaemia and encephalitis which may point to the possibility of over-stretched hospital services as demonstrated by the increasing in-hospital mortality rates reported at a tertiary care facility .
Efforts to improve the quality of medical certification have been initiated by government through a training programme in the public sector. However, until the stigma of the disease is eliminated and HIV is reported reliably, the interpretation of cause-of-death data will remain challenging, and analytical approaches will be required to estimate the true numbers. Government's mid-term review in 2009  identified the need to reduce stigma and discrimination as well as a mechanism to monitor changes. Despite the new era in the fight against AIDS  including the largest antiretroviral treatment programme in the world, a national counselling and testing campaign, promotion of medical male circumcision, and the successful implementation of a prevention of mother-to-child transmission programme, stigma around HIV and AIDS prevails. Reluctance to acknowledge AIDS as the cause of death of a politician in 2012 was poignantly highlighted by a news-editor: ‘Rest in peace, Sir, you with the name we dare not say in the same breath as the name of the illness whose name none of us speak even after 20 deadly years. Stigma killed you as much as the long illness did’ . The sooner HIV and AIDS are normalized as chronic health conditions, the sooner South Africa will be able to eliminate the epidemic. Reliably counting deaths from AIDS, an essential mechanism to monitor progress, has to become a routine.
We thank the South African National Burden of Disease (SA NBD) team for reviewing the estimates. The team consists of the following members (in alphabetical order) Debbie Bradshaw, Rob Dorrington, Pam Groenewald, Janetta Joubert, Ria Laubscher, Richard Matzopoulos, William Msemburi, Nadine Nannan, Ian Neethling, Edward Nicol, Beatrice Nojilana, Victoria Pillay-Van Wyk (project coordinator), Megan Prinsloo, Anastasia Rossouw, Nomfuneko Sithole, Nontuthuzelo Somdyala, and Theo Vos.
All authors conceptualized the article. D.B. and V.P.W. led the study. R.D. and W.M. led the development of the methodology. R.D. led the demographic analysis, R.L. led the analysis of data from Stats SA, P.G. led the classification of causes, W.M. led the programming and modelling. All authors reviewed the estimates and reviewed the paper.
The study was funded by the South African Medical Research Council and we thank Statistics South Africa for providing cause-of-death data and Ms Elize De Kock for administrative support on the SA NBD study.
Conflicts of interest
There are no conflicts of interest.
2. Chigwedere P, Seage GR 3rd, Gruskin S, Lee TH, Essex M. Estimating the lost benefits of antiretroviral drug use in South Africa
. J Acquir Immune Defic Syndr
3. Abdool Karim SS, Churchyard GJ, Karim QA, Lawn SD. HIV infection and tuberculosis in South Africa: an urgent need to escalate the public health response
4. Dorrington R, Bourne D, Bradshaw D, Laubscher R, Timaeus IM. The Impact of HIV
on Adult Mortality in South Africa
. Technical Report. South Africa
: Burden of Disease Unit, Medical Research Council; 2001. http://www.mrc.ac.za/bod/reports.htm
. [Accessed 16 October 2013]
5. Robins S. Long live Zackie, long live’: AIDS activism, science and citizenship after Apartheid
. J South Afr Stud
6. Statistics South Africa
. Causes of death in South Africa 1997–2001: Advance release of recorded causes of death
. Pretoria: Statistics South Africa
7. Yudkin PL, Burger EH, Bradshaw D, Groenewald P, Ward AM, Volmink J. Deaths caused by HIV disease under-reported in South Africa
8. Burger EH, Groenewald P, Rossouw A, Bradshaw D. Medical certification of death in South Africa: moving forward
. S Afr Med J
9. Bradshaw D, Laubscher R, Dorrington R, Bourne DE, Timaeus IM. Unabated rise in number of adult deaths in South Africa
. S Afr Med J
10. Department of Health. The National HIV and Syphilis Antenatal Prevalence Survey, South Africa, 2011
. Pretoria: Department of Health, 2012.
11. Shisana O, Rehle T, Simbayi LC, Zuma K, Jooste S, Zungu N, et al. South African National HIV Prevalence, Incidence and Behaviour Survey, 2012
. Cape Town: HSRC Press; 2014.
12. Statistics South Africa
. Mortality and causes of death in South Africa, 2010: Findings from death notification
. Statistical Release P0309.3. Pretoria: Statistics South Africa
13. Groenewald P, Nannan N, Bourne D, Laubscher R, Bradshaw D. Identifying deaths from AIDS in South Africa
14. Groenewald P, Bradshaw D, Dorrington R, Bourne D, Laubscher R, Nannan N. Identifying deaths from AIDS in South Africa: an update
15. Birnbaum JK, Murray CJ, Lozano R. Exposing misclassified HIV/AIDS deaths in South Africa
. Bull World Health Organ
16. Fazito E, Cuchi P, Fat DM, Ghys PD, Pereira MG, Vasconcelos AM, Pascom AR. Identifying and quantifying misclassified and under-reported AIDS deaths in Brazil: a retrospective analysis from 1985 to 2009
. Sex Transm Infect
2012; 88 (suppl 2):i86–i94.
18. Dorrington RE, Bradshaw D, Johnson L, Budlender D. The Demographic Impact of HIV/AIDS in South Africa. National indicators for 2004
. Cape Town: Centre for Actuarial Research, South African Medical Research Council and Actuarial Society of South Africa
19. Statistics South Africa
. Mortality and causes of death in South Africa, 2010: findings from death notification
. Statistical Release PO309.3. Pretoria: Statistics South Africa
20. United States National Center for Health Statistics. Automated Classification of Medical Entities (ACME) software version 2005.02. www.cdc.gov/nchs
[Accessed 29 July 2012]
21. van Wyk VP, Laubscher R, Msemburi W, Dorrington RE, Groenewald P, Vos T. Second South African National Burden of Disease Study: data cleaning, validation and SA NBD list. Cape Town: South African Medical Research Council; 2014, http://www.mrc.ac.za/bod/SANBDReport.pdf
[Accessed 29 January 2015].
23. Dorrington RE. Ins and Outs: Estimation of the HIV incidence on all women from the change in antenatal prevalence and the number of AIDS deaths from the CoD vital registration data. University of Cape Town Actuarial and Demography seminar
, Cape Town. September 2010.
25. Bradshaw D, Pillay-van Wyk V, Laubscher R, Nojilana B, Groenewald P, Nannan N, Metcalf C. Cause of death statistics for South Africa
: challenges and possibilities. Cape Town: Burden of Disease Research Unit; 2010, www.mrc.ac.za/bod/cause_death_statsSA.pdf
[Accessed 29 July 2012]
26. Machemedze T, Dorrington RE. Levels of mortality of the South African aged population using the method of extinct generations
. J Afr Popul Stud
2011; 25 (suppl 1):63–76.
27. Johnson LF, Dorrington RE. Modelling the demographic impact of HIV/AIDS in South Africa and the likely impact of interventions
. Demographic Res
28. Johnson LF, Davies MA, Moultrie H, Sherman GG, Bland RM, Rehle TM, et al. The effect of early initiation of antiretroviral treatment in infants on pediatric AIDS mortality in South Africa: a model-based analysis
. Pediatr Infect Dis J
29. Stover J, Brown T, Marston M. Updates to the spectrum/estimation and projection package (EPP) model to estimate HIV trends for adults and children
. Sex Transm Infect
2012; 88 (suppl 2):i11–e16.
30. Lozano R, Naghavi M, Foreman K, Lim S, Shibuya K, Aboyans V, et al. Global and regional mortality from 235 causes of death for 20 age groups in 1990 and 2010: a systematic analysis for the Global Burden of Disease Study 2010
31. GBD 2013 Mortality and Causes of Death Collaborators. Global, regional, and national age-sex specific all-cause and cause-specific mortality for 240 causes of death, 1990–2013: a systematic analysis for the Global Burden of Disease Study 2013
32. Adam MA, Johnson LF. Estimation of adult antiretroviral treatment coverage in South Africa
. S Afr Med J
33. Carlson C, Bannerman M, Barron P, Baumann J, Chimbwete C, Conco D, et al. Sanelisiwe Tsela National strategic plan 2007–2011 mid-term review
. 2009; Johannesburg: Health Development Africa, http://http://www.irinnews.org
/pdf/Mid_Term_Review_of_the_NSP_(preliminary_report).pdf. [Accessed 12 August 2012].
34. Goga A, Dinh T-H, Jackson D, for the SAPMTCTE study group. Evaluation of the Effectiveness of the National Prevention of Mother-to-Child Transmission (PMTCT) Programme on Infant HIV measured at six weeks postpartum in South Africa, 2010
. Cape Town: South African Medical Research Council, National Department of Health of South Africa
and PEPFAR/US Centers for Disease Control and Prevention, 2012.
35. Streatfield PK, Khan WA, Bhuiya A, Hanifi SM, Alam N, Millogo O, et al. HIV/AIDS-related mortality in Africa and Asia: evidence from INDEPTH health and demographic surveillance system sites
. Glob Health Action
36. Herbst AJ, Mafojane T, Newell ML. Verbal autopsy-based cause-specific mortality trends in rural KwaZulu-Natal, South Africa, 2000–2009
. Popul Health Metr
37. Nabukalu D, Klipstein-Grobusch K, Herbst K, Newell ML. Mortality in women of reproductive age in rural South Africa
. Glob Health Action
38. Kabudula CW, Tollman S, Mee P, Ngobeni S, Silaule B, Gómez-Olivé FX, et al. Two decades of mortality change in rural northeast South Africa
. Glob Health Action
39. Murray CJ, Ortblad KF, Guinovart C, Lim SS, Wolock TM, Roberts DA, et al. Global, regional, and national incidence and mortality for HIV, tuberculosis, and malaria during 1990–2013: a systematic analysis for the Global Burden of Disease Study 2013
40. Kerber KJ, Lawn JE, Johnson LF, Mahy M, Dorrington RE, Phillips H, et al. South African child deaths 1990–2011: have HIV services reversed the trend enough to meet Millennium Development Goal 4?
41. Dorrington RE, Bradshaw D, Laubscher R, Nannan N. Rapid mortality surveillance report 2013
. 2014; Cape Town: South African Medical Research Council, http://http://www.mrc.ac.za
/bod/RapidMortalitySurveillanceReport2013.pdf. [Accessed 12 March 2015].
42. Barron P, Day C, Loveday M, Monticelli F. The District Health Barometer Year 1 January-December 2004
. 2005; Durban: Health Systems Trust, http://http://www.hst.org.za
/publications/district-health-barometer-year-1-january-december-2004. [Accessed 15 July 2015].
43. Massyn N, Day C, Barron P, Haynes R, English R, Padarath A. District Health Barometer 2011/12
. Durban: Health Systems Trust; 2013; http://http://www.hst.org.za
/publications/district-health-barometer-201112. [Accessed 15 July 2015].
44. Myer L, Smith E, Mayosi BM. Medical inpatient mortality at Groote Schuur Hospital, Cape Town, South Africa, 2002–2009
. S Afr Med J
45. Mayosi BM, Lawn JE, van Niekerk A, Bradshaw D, Abdool Karim SS, Coovadia HM. Lancet South Africa team. Health in South Africa: changes and challenges since 2009