SEARCH

SEARCH BY CITATION

Keywords:

  • HIV ;
  • maternal mortality;
  • review;
  • systematic;
  • sub-Saharan Africa;
  • pregnancy-related mortality

Abstract

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Conclusions
  7. Acknowledgements
  8. References
  9. Appendix

Objectives

To estimate the proportion of pregnancy-related deaths attributed to HIV in population-based studies in sub-Saharan Africa, and to document the methods used to make such attribution.

Methods

Four databases were searched for studies on causes of maternal and pregnancy-related mortality published from 2003 to June 2013. Data were extracted, and meta-analysis of proportions with random effects was used to obtain summary estimates.

Results

In the 19 studies found, the proportion of deaths attributed to HIV ranged from 0.0% to 27.0%. The summary proportion was 3.4% (95% confidence interval: 1.8–6.3), with high heterogeneity. Subregionally, the summary proportions were 1.1% (0.4–3.3%) in West Africa, 4.5% (1.7–11.2%) in East Africa and 26.1% (21.9–30.7%) in Southern Africa. Criteria for assigning HIV as a cause of maternal death were rarely reported, and overall, methods were poor.

Conclusions

The proportion of pregnancy-related/maternal deaths attributed to HIV is substantially lower than modelled estimates, but comparisons are hampered by the absence of standard approaches. Clear guidelines on how to classify pregnancy-related deaths as attributable to HIV are urgently needed, so that the effect of the HIV epidemic on pregnancy-related mortality can be monitored and action taken accordingly.

Objectifs

Estimer la proportion des décès liés à la grossesse, attribuables au VIH dans les études basées sur la population en Afrique subsaharienne et documenter les méthodes utilisées pour effectuer une telle attribution.

Méthodes

Quatre bases de données ont été recherchées pour les études sur les causes de mortalité maternelle liée à la grossesse et publiées de 2003 à juin 2013. Les données ont été extraites et la méta-analyse des proportions à effets aléatoires a été utilisée pour obtenir des estimations de synthèse.

Résultats

Dans les 19 études trouvées, la proportion des décès attribués au VIH allaient de 0.0% à 27.0%. La proportion de synthèse était de 3.4% (intervalle de confiance à 95%: 1.8–6.3%), avec une forte hétérogénéité. A l’échelle sous-régionale, les proportions de synthèse était de 1.1% (0.4–3.3%) en Afrique de l'Ouest, 4.5% (1.7–11.2%) en Afrique de l'Est et 26.1% (21.9–30.7%) en Afrique australe. Les critères pour l'attribution du VIH comme cause de décès maternels étaient rarement signalés et dans l'ensemble, les méthodes n’étaient pas pertinentes.

Conclusions

La proportion des décès liés à la grossesse/maternels attribués au VIH est substantiellement inférieure aux estimations modélisées, mais les comparaisons sont entravées par l'absence de méthodes standard. Des directives claires sur la façon de classer les décès liés à la grossesse comme imputables au VIH sont urgemment nécessaires, afin que l'effet de l’épidémie du VIH sur la mortalité liée à la grossesse puisse être surveillé et que des mesures soient prises en conséquence.

Objetivos

Calcular la proporción de muertes maternas atribuidas al VIH en estudios poblacionales en África subsahariana y documentar los métodos utilizados para atribuirlas.

Métodos

Se realizaron búsquedas, en cuatro bases de datos, de estudios sobre causas de mortalidad materna y relacionadas con el embarazo, publicados entre el 2003 y Junio del 2013. Se extrajeron datos y se utilizó un metaanálisis de las proporciones con efectos aleatorios para obtener los cálculos globales. Resultados En los 19 estudios encontrados, la proporción de muertes atribuidas al VIH estaban entre un 0.0% y un 27.0%. La proporción global era del 3.4% (intervalo de confianza del 95%: 1.8–6.3%) con una alta heterogeneidad. A nivel subregional, las proporciones globales eran del 1.1% (0.4–3.3%) en África Occidental, 4.5% (1.7–11.2%) en África Oriental y 26.1% (21.9–30.7%) en África del Sur. Los criterios para asignar el VIH como la causa de muerte materna eran rara vez reportados y en general, la metodología era pobre.

Conclusiones

La proporción de muertes maternas/relacionadas con el embarazo atribuidas al VIH es sustancialmente menor que en los cálculos modelados, pero las comparaciones se ven limitadas por la ausencia de un enfoque estandarizado. Se requiere urgentemente de guías claras sobre como clasificar las muertes relacionadas con el embarazo atribuibles al VIH, de forma que el efecto de la epidemia del VIH sobre la mortalidad relacionada con el embarazo pueda ser monitorizada y se puedan tomar las acciones requeridas.


Introduction

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Conclusions
  7. Acknowledgements
  8. References
  9. Appendix

Empirical estimates of the proportion of maternal or pregnancy-related deaths that are due to HIV/AIDS at the population level vary substantially. A systematic review of causes of maternal death using data published between 1997 and 2002 found that 6.2% of maternal mortality in Africa was due to ‘HIV/AIDS’ (Khan et al. 2006). This review included only seven studies from sub-Saharan Africa, and the criteria for assigning a death to HIV were not specified. A recent systematic review of studies comparing mortality during pregnancy and the post-partum in HIV-positive and HIV-negative women found that between 12% and 50% of pregnancy-related deaths were attributable to HIV in settings with an HIV prevalence of 2% and 15%, respectively (Calvert & Ronsmans 2013). Of 17 studies from sub-Saharan Africa, however, only two were population based. An analysis of data from six population-based studies in East and Southern Africa found the proportion of pregnancy-related deaths attributable to HIV to be 45% and estimated that across sub-Saharan Africa as a whole this proportion was 24% (Zaba et al. 2013).

Most global estimates of the proportion of maternal or pregnancy-related deaths attributable to HIV are based on mathematical models. Hogan et al., using data from national and subnational studies estimated 32% of maternal mortality in sub-Saharan Africa in 2008 to be due to HIV (Hogan et al. 2010; Rosen et al. 2012). The Maternal Mortality Estimation Interagency Group (MMEIG), using data from nationally representative studies, found 9% of maternal deaths in sub-Saharan Africa in 2008 to be due to HIV (WHO, 2010). The methods were very different, and the validity of either model is difficult to assess; supportive empirical data are ultimately needed.

The difficulty in estimating mortality attributable to HIV arises in part from the uncertainty over whether and when HIV should be considered an indirect maternal or a coincidental cause of death in pregnant/post-partum women (McIntyre 2005). Evidence for an effect of pregnancy on HIV progression or of HIV on direct obstetric causes of death is weak (French & Brocklehurst 1998; Graham & Hussein 2003; Van der Paal et al. 2007; Calvert & Ronsmans; Kumar et al. 1997). Models and guidelines differ in how they categorise deaths of HIV-positive pregnant women: the MMEIG model (WHO, 2010) assumed 50% of such deaths to be indirect maternal deaths; and the Hogan model (Hogan et al. 2010) took this proportion to be 100%. The WHO Human Reproduction Programme guidelines on applying ICD-10 to pregnancy-related deaths (ICD-MM guidelines) (WHO 2012b), on the other hand, state that deaths due to ‘the aggravating effect of pregnancy on HIV’ are indirect maternal deaths, while deaths meeting the criteria for AIDS and without direct obstetric complications should be considered coincidental to pregnancy. The diseases that may aggravate the progression of HIV in pregnant women are not specified, and it is uncertain which deaths in HIV-positive women qualify as indirect.

This study aims to support a better understanding of the proportion of pregnancy-related deaths attributed to HIV/AIDS in sub-Saharan Africa, in the general population. We systematically review and summarise population-based studies reporting the causes of pregnancy-related or maternal deaths published from 2003 to June 2013, stratifying the estimates by subregion. We also examine the approaches taken in assigning pregnancy-related or maternal deaths to HIV, in particular whether HIV-related deaths are considered to be indirect maternal deaths, and the criteria used to attribute deaths to HIV.

Methods

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Conclusions
  7. Acknowledgements
  8. References
  9. Appendix

Population and outcome of interest

The population of interest was any set of pregnancy-related/maternal deaths ascertained at the population level. The main outcome of interest was the proportion of deaths attributed to HIV, and the methods used to assign HIV as the cause of death.

Search strategy

On 8 July 2012, the databases Medline, Embase, Popline and African Index Medicus were searched for studies published from 2003 onwards. The review was updated on 20 June 2013. Extensive search terms followed the logic (‘maternal’ AND ‘mortality’) AND (‘cause of death’ OR ‘HIV’) AND (sub-Saharan Africa). The full search strategy is shown in Appendix 1.

Searches were conducted in English, with no limit on language. The reference lists of all retained studies were searched by hand. Demographic and Health Survey reports were checked, and advice was sought from experts at the WHO Department for Reproductive Health.

Study selection

Titles and abstracts were screened by one author (CG), and full texts were sought for potentially relevant publications. A random 50% sample of the titles/abstracts was double screened by a separate researcher. Discordant articles were re-evaluated by CG: no discordant articles met the inclusion criteria.

Eligibility criteria

Studies were included if they reported cause-specific pregnancy-related or maternal mortality in sub-Saharan Africa. Studies had to be population based, designed to capture all deaths in an area. This could include ascertainment of causes from health facility records for women dying in health facilities, but studies that only reported births from health facilities were excluded. This restriction aimed to minimise non-representativity given low institutional delivery in sub-Saharan Africa and was consistent with the previous review in this area (Khan et al. 2006). Included studies had to estimate mortality through counting rather than modelling contain at least 25 deaths and be published from 2003 (to not overlap with the previous systematic review (Khan et al. 2006)). Conference abstracts were excluded.

Data extraction and analysis

One author extracted data on: study setting and population; completeness of death ascertainment; source of data on deaths; method of determining cause-of-death; method for identifying HIV as cause-of-death; number of deaths; and numbers of deaths attributed to HIV, ‘unknown’ causes, and other categories which may contain HIV deaths.

Summary estimates of the proportion of deaths due to HIV and assessment of heterogeneity between studies were carried out using meta-analysis of proportions in R 3.0.1 (R Core Team, 2013). Random effects were assumed, as the study populations and methods varied greatly. Heterogeneity in the proportion of deaths attributed to HIV was assessed by the I-squared statistic (Huedo-Medina et al. 2006) and the P-value of the tau-squared estimate (Rucker et al. 2008). A stratification by subregion was carried out using the classification used by Hogan et al. (2010).

Quality assessment

In line with the PRISMA guidelines, the risk of bias was assessed, focusing on the quality of death ascertainment, completeness of cause-of-death assignation, quality of method of assigning cause-of-death and quality of the method of assigning HIV as cause-of-death. Studies were categorised as having high, medium or low quality. Table 1 outlines the study attributes corresponding to these quality categories.

Table 1. Quality-assessment criteria
Level of qualityQuality of method of death ascertainmentCompleteness of cause-of-death assignationQuality of method of assigning cause-of-deathQuality of method of assigning HIV as cause-of-death
  1. DSS, demographic surveillance system; VA, verbal autopsy; RAMOS, reproductive age mortality survey.

High

DSS with frequent rounds

RAMOS/survey using multiple data sources or household ascertainment

Census

Causes assigned to ≥90% of deathsClinical diagnosis or post-mortem examinationWHO staging criteria (HIV-positive & clinical/immunological criteria)
Medium

DSS with infrequent rounds/round length unreported

RAMOS/survey using limited sources (e.g. facility records or key informants)

Causes assigned to 75–90% of deathsVA with recall period ≤1 year

Clinical or other method based on symptomatic assessment

Known HIV-positive status

LowRoutine reporting with no assessment of completeness possibleCauses assigned to <75% of deaths

VA with recall period >1 year/unreported

Method unreported

Unreported

Results

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Conclusions
  7. Acknowledgements
  8. References
  9. Appendix

The search gave 7025 publications for screening, of which 535 were retained for full text review (Figure 1). One study was suggested by an expert (Lewis et al. 2011). Ultimately, data were extracted from 19 studies from 17 publications: one publication contained three study populations (Mswia et al. 2003). Table 2 describes the study settings, study populations and methods of ascertaining deaths and causes of death.

Table 2. Methods of population-based studies
Author (year)SettingStudy populationData source on deathsMethod of assigning cause of death
  1. DSS, demographic surveillance system; COD, cause of death; VA, verbal autopsy; InterVA-M, probabilistic model for interpreting VA data, for pregnancy-related deaths; TBA, traditional birth attendant; CHN, community health nurse; ICD-10, International Classification of Diseases, 10th Revision; RAMOS, reproductive age mortality survey; UNHCR, United Nations High Commission for Refugees.

West Africa
Bell et al. (2008)Ouargaye and Diapaga, Burkina Faso; largely ruralICD-10 pregnancy-related deaths, 2002–2006CensusVA by InterVA-M, no further detail, recall ≤5 years
Cham et al. (2007)Central & Upper River divisions, Gambia; largely ruralICD-10 maternal deaths among women reaching health facilities, January–September 2002

In facilities: facility records

In community: interviews and record review with TBAs and community health nurses

By physicians using facility records, TBA/CHN records, interviews & VA, recall not stated
Mills et al. (2008)Navrongo, Ghana; largely ruralICD-10 maternal deaths among female deaths with VA (99%), 2002–2004DSSVA by physician review: 3 independently, COD assigned when 2 agree or consensus after discussion, recall ≤4 years
Zakariah et al. (2009)Accra, Ghana; urbanICD-10 pregnancy-related deaths, 2002RAMOS using multiple facility records and mortuary log books

Facility deaths: as assigned in facility

Non-facility deaths: post-mortem examination by pathologist

Issah et al. (2011)Upper West Region, Ghana; largely ruralMaternal deaths (definition not given), 2009

In facilities: maternal death audits

In community: routine reports of community health volunteers

Facilities: as assigned in facility audit

Community: VA, no further detail, recall ≤11 months

Asamoah et al. (2011)Ghana; urban & rural; National (sample)ICD-10 maternal deaths, 2000–2005Cross-sectional mortality surveyVA to ICD-10, method unclear, recall ≤5 years
Lori and Starke (2012)One rural county, LiberiaICD-10 maternal deaths, 2008

In facilities: audit with daily ward visits

In community: reports from rural clinics, traditional midwives and family members; mechanism not described

Facilities: as assigned in facility

Community: VA, no further detail, recall usually ≤7 days

Adamu et al. (2003)Kano State, Nigeria; urban & ruralMaternal deaths (definition not given), 1990–1999Routine reports from health facilities and village/local government councils to state health ministry

Facilities: As assigned in facility

Community: Not stated

Etard et al. (2003)Niakhar, Senegal; ruralICD-10 maternal deaths, 1984–1997DSSVA by physician review: 2 obstetricians independently + third in discrepant cases, recall ≤13 years
Ba et al. (2003)Bandafassi, Senegal; ruralICD-10 maternal deaths including late maternal, 1988–1997DSSVA to ICD-10 by physician review, by two obstetricians + third in discrepant cases, recall not stated
East Africa
Ziraba et al. (2009)Nairobi, Kenya; urbanICD-10 maternal deaths among female deaths with VA (86%), 2003–2005DSSVA by physician review: 3 independently, COD assigned when 2 agree or consensus after discussion, recall within 6 weeks of death registration
Hynes et al. (2012)Four refugee camps >25,000 population, KenyaICD-10 maternal deaths, 2008–2010Routine reports from UNHCR health information system and additional audit, methods unclearBy research team, from community reports; method unclear
Lewis et al. (2011)Mozambique; urban & rural; National (sample)ICD-10 maternal deaths, 12 months prior to 2007 censusCensusVA to ICD-10 by physician review, no further detail, recall ≤1 year
Mohammed et al. (2011)Kassala state, eastern Sudan, urban & ruralICD-10 maternal deaths among female deaths with VA (88%), 2004–2006Cross-sectional mortality survey (RAMOS using household reports)VA reviewed by three physicians, no further detail, recall ≤3 years
Mswia et al. (2003a)Morogoro District, Tanzania; ruralICD-10 maternal deaths, 1992–1999DSSVA to ICD-10 by ‘a panel of physicians’, recall usually ≤4 weeks
Mswia et al. (2003b)Hai District, Tanzania; ruralICD-10 maternal deaths, 1992–1999DSSVA to ICD-10 by ‘a panel of physicians’, recall usually ≤4 weeks
Mswia et al. (2003c)Dar es Salaam, Tanzania; urbanICD-10 maternal deaths, 1993–1999DSSVA to ICD-10 by ‘a panel of physicians’, recall usually ≤4 weeks
Southern Africa
Garenne et al. (2013)Agincourt subdistrict, South Africa; ruralICD-10 pregnancy-related deaths, 1992–2010DSSVA to ICD-10 by physician review, by 2 obstetricians + third in discrepant cases, recall not stated
Munjanja et al. (2007)Zimbabwe; urban & rural; National (sample)ICD-10 pregnancy-related deaths recorded (77%), 2007–2008

In facilities: records from multiple wards

In community: regular reports from multiple community members

Facilities: as assigned in facility

Community: VA by physician review (COD when 2/3 agreed) and InterVA-M, no further detail, recall ≤1 year

image

Figure 1. Search process for selection of studies.

Download figure to PowerPoint

Setting

Ten studies were from west Africa (four from Ghana (Mills et al. 2008; Zakariah et al. 2009; Issah et al. 2011; Asamoah et al. 2011), two from Senegal (Etard et al. 2003; Ba et al. 2003), one each from Burkina Faso (Bell et al. 2008), Gambia: (Cham et al. 2007), Liberia: (Lori & Starke 2012) and Nigeria (Adamu et al. 2003)). Seven were from east Africa (three from Tanzania: (Mswia et al. 2003), two from Kenya: (Ziraba et al. 2009; Hynes et al. 2012) and one each from Mozambique: (Lewis et al. 2011) and Sudan: (Mohammed et al. 2011)). Two were from southern Africa (Zimbabwe: (Munjanja et al. 2007) and South Africa: (Garenne et al. 2013)).

Three studies represented national populations (Lewis et al. 2011; Asamoah et al. 2011; Munjanja et al. 2007). Of the sixteen subnational studies, ten were in rural areas, three were urban, two were mixed and one was in refugee camps. The time periods in which deaths occurred varied greatly from 1984–1997 to 2008–2010.

Study population

Most studies (15/19) reported maternal deaths: 12 according to the ICD-10 definition (Lewis et al. 2011; Mswia et al. 2003; Mills et al. 2008; Asamoah et al. 2011; Etard et al. 2003; Cham et al. 2007; Lori & Starke 2012; Ziraba et al. 2009; Hynes et al. 2012; Mohammed et al. 2011), one (Ba et al. 2003) including late maternal deaths and two without a definition (Issah et al. 2011; Adamu et al. 2003). Four studies reported ICD-10 pregnancy-related mortality (Zakariah et al. 2009; Bell et al. 2008; Munjanja et al. 2007; Garenne et al. 2013).

Sources of data on deaths

Twelve studies ascertained deaths solely from household or community reports, enumerating deaths using demographic surveillance systems (n = 8) (Mswia et al. 2003; Mills et al. 2008; Etard et al. 2003; Ba et al. 2003; Ziraba et al. 2009; Garenne et al. 2013), cross-sectional surveys (n = 2) (Asamoah et al. 2011; Mohammed et al. 2011) or a census (n = 2) (Lewis et al. 2011; Bell et al. 2008). Six studies supplemented community-based reports with records from health facilities (Zakariah et al. 2009; Issah et al. 2011; Cham et al. 2007; Lori & Starke 2012; Adamu et al. 2003; Munjanja et al. 2007). Of these, one (Munjanja et al. 2007) commented on reconciliation of reports from the two sources. The study from urban Accra (Zakariah et al. 2009) was treated as a census, as reporting non-facility deaths to mortuaries was legally required. The study in Kenyan refugee camps used community reports, and it is unclear whether facility records were also used (Hynes et al. 2012).

Methods of determining cause-of-death

Most studies (15/19) assigned causes of death based on verbal autopsies. Eleven interpreted the data using physician review (Lewis et al. 2011; Mswia et al. 2003; Mills et al. 2008; Etard et al. 2003; Ba et al. 2003; Cham et al. 2007; Ziraba et al. 2009; Mohammed et al. 2011; Garenne et al. 2013); two did not state how data were interpreted (Asamoah et al. 2011; Lori & Starke 2012). One used the InterVA-M model to interpret the data (Bell et al. 2008), and one used both physician review and InterVA-M without stating how these were combined (Munjanja et al. 2007). One study Hynes et al. 2012 used community reports of the circumstances of death but did not state how causes were arrived at.

Four studies (Issah et al. 2011; Lori & Starke 2012; Adamu et al. 2003; Munjanja et al. 2007) took the causes of deaths in facilities to be those assigned in the facility. Two of these did not state how causes were assigned to deaths in the community (Issah et al. 2011; Adamu et al. 2003). One study extracted causes from facility and mortuary records (Zakariah et al. 2009). One study assigned causes based on physician review of facility and health-worker records and interviews, and VA reports (Cham et al. 2007).

Table 3 describes how HIV was categorised and assigned as a cause of death, the numbers of deaths and the proportion attributed to HIV. The number of maternal or pregnancy-related deaths in the studies ranged from 28 (Lori & Starke 2012) to 4154 (Adamu et al. 2003); the median number of deaths was 107 (Mswia et al. 2003).

Table 3. Results of population-based studies
Author (year)Total maternal/pregnancy-related deathsCategory to which HIV deaths are assignedMethod of assigning HIV as COD% Deaths attributed to HIV% of Deaths in other pertinent categories
  1. COD, cause of death.

West Africa
Bell et al. (2008)396Indirect maternalInterVA-M using VA data. Category is ‘HIV/AIDS-related death’5.3

28 obstetric sepsis

3 TB

0.25 unknown

Cham et al. (2007)42Indirect maternalMethod not given. Category is ‘HIV/AIDS’2.4

4.8 sepsis

2 unknown

Mills et al. (2008)45Indirect maternalMethod not given. Category is ‘HIV/AIDS’2.2

6.7 post-partum sepsis

17.8 ‘Other indirect’

Zakariah et al. (2009)179Indirect maternalMethod not given. Category is ‘HIV/AIDS’1.1

7.3 genital tract sepsis

1.1 pneumonia

1.1 cervical cancer

Issah et al. (2011)47No distinction madeMethod not given. Category is ‘HIV/AIDS’8.5

21.3 sepsis

4.3 meningitis

Asamoah et al. (2011)605Did not report HIV0

4.3 unspecified/other infectious

0.03 TB

6.9 sepsis

7.3 unspecified/other obstetric

Lori and Starke (2012)28Did not report HIV0

14.3 sepsis

17.9 unknown

Adamu et al. (2003)4154Did not report HIV0

7 sepsis

20 ‘Other’ including cervical cancer

Unclear unknown

Etard et al. (2003)87Did not report HIV032.2 ‘Other causes’ including all indirect
Ba et al. (2003)36 (inc. 3 late maternal)Did not report HIV0

23.8 obstetric infection

11.1 ‘Indirect’

30.6 unknown

East Africa
Ziraba et al. (2009)29Indirect maternalMethod not given. Category is ‘HIV/AIDS/TB’13.8

10.3 post-partum sepsis

13.8 other indirect

Hynes et al. (2012)68Indirect maternalMethod not given. Category is ‘HIV/AIDS’0

13.2 pregnancy-related sepsis

7.4 unknown

Lewis et al. (2011)213Indirect maternalMethod not given. Category is ‘HIV/AIDS-related indirect obstetric death’18.1
Mohammed et al. (2011)64Did not report HIV0

10.9 sepsis

4.7 ‘miscellaneous’ indirect

Mswia et al. (2003a), Morogoro224Indirect maternalClinical signs and symptoms of advanced HIV or AIDS0.4

6.7 obstetric sepsis

4.0 ‘Other indirect’

Mswia et al. (2003b), Hai110Indirect maternalClinical signs and symptoms of advanced HIV or AIDS3.6

7.3 obstetric sepsis

10.9 ‘Other indirect’

Mswia et al. (2003c), Dar107Indirect maternalClinical signs and symptoms of advanced HIV or AIDS6.5

14.0 obstetric sepsis

15.9 ‘Other indirect’

Southern Africa 
Garenne et al. (2013)137Non-obstetricalMethod not given. Category is ‘HIV/TB’27.0

0.1 other infections

13.9 undetermined

Munjanja et al. (2007)243Indirect maternalMethod not given. Category is ‘AIDS-defining conditions’25.5

7.8 puerperal sepsis

9.9 unknown

Categorising and assigning HIV as a cause of death

Six studies did not mention HIV at all. Of the 13 studies that reported HIV as a cause, 11 described it as an indirect maternal cause (one of which assigned no deaths to HIV (Hynes et al. 2012)). One categorised HIV as a ‘non-obstetrical’ cause (Garenne et al. 2013) and one did not place HIV deaths into categories (Issah et al. 2011). Eight did not describe criteria used to assign HIV as a cause of death. One of the four studies reporting criteria used the probabilistic model InterVA-M (Bell et al. 2008). The authors of the other three (Mswia et al. 2003) used clinical signs and symptoms of advanced HIV or AIDS, but the signs/symptoms used were not reported.

Other pertinent causes

Most studies (17/19) included cause-of-death categories in which deaths due to HIV could have been hidden: 16 reported deaths due to obstetric sepsis; eight reported categories called ‘other indirect’ or ‘indirect’; four reported infectious causes of death that appear in stages three or four of the WHO clinical staging criteria (pneumonia, meningitis, tuberculosis, cervical cancer) (WHO 2007); and eight reported an ‘unknown’ category.

Quality assessment

The methods used to ascertain deaths were of mixed quality. Only two studies commented on the completeness of their death recording, estimating, respectively, that 77% of eligible deaths (Munjanja et al. 2007), and ‘all possible deaths’ were identified (Zakariah et al. 2009). Three studies reported few deaths outside facilities – 4/47, 2/42 and 8/28, respectively – despite low institutional delivery rates (Issah et al. 2011; Cham et al. 2007; Lori & Starke 2012). Only three studies reported the coverage of their VA interviews – 99%, 86% and 88% of eligible deaths, respectively (Mills et al. 2008; Ziraba et al. 2009; Mohammed et al. 2011). Ten studies reported no deaths with unknown causes. The proportion of deaths with unknown causes was <10% in six studies (Asamoah et al. 2011; Bell et al. 2008; Cham et al. 2007; Lori & Starke 2012; Hynes et al. 2012; Munjanja et al. 2007,14% of deaths in one study (Garenne et al. 2013, 31% in one study (Ba et al. 2003) and unclear in one study (Adamu et al. 2003) (Table 4).

Table 4. Quality assessment of retained studies
 Quality of death ascertainmentCompleteness of cause-of-death assignationQuality of method of assigning cause-of-deathQuality of method of assigning HIV as cause-of-death
Bell et al. (2008)HighHighLowMedium - InterVA-M
Cham et al. (2007)LowHighMediumLow
Mills et al. (2008)HighHighLowLow
Zakariah et al. (2009)HighHighHighLow
Issah et al. (2011)MediumHighMediumLow
Asamoah et al. (2011)HighHighLowNo deaths attributed to HIV
Lori and Starke (2012)LowMediumMediumNo deaths attributed to HIV
Adamu et al. (2003)LowLowMediumNo deaths attributed to HIV
Etard et al. (2003)HighHighLowNo deaths attributed to HIV
Ba et al. (2003)MediumLowLowNo deaths attributed to HIV
Ziraba et al. (2009)HighHighMediumLow
Hynes et al. (2012)LowHighLowLow
Mohammed et al. (2011)MediumHighLowNo deaths attributed to HIV
Mswia et al. (2003a)HighHighMediumMedium
Mswia et al. (2003b)HighHighMediumMedium
Mswia et al. (2003c)HighHighMediumMedium
Lewis et al. (2011)HighHighMediumLow
Garenne et al. (2013)HighMediumLowLow
Munjanja et al. (2007)HighHighMediumLow

Quality of methods of assigning cause of death was generally medium or low. Of twelve studies that assigned HIV as a cause of death, only four stated the method for doing this (Mswia et al. 2003; Bell et al. 2008), none of which was of high quality.

The proportion of deaths assigned to HIV in studies that reported HIV as a cause ranged from 0.0% in Kenyan refugee camps in 2008–2010 (Hynes et al. 2012) to 27.0% in rural South Africa in 1992–2010 (Garenne et al. 2013). Individual-country findings varied widely: in Ghana, a national sample of deaths in 2000–2005 assigned none to HIV (Asamoah et al. 2011); this proportion was 8.5% in the Upper West Region in 2009 (Issah et al. 2011); 2.2% in Navrongo in 2002–2004 (Mills et al. 2008), and 1.1% in Accra in 2002 (Zakariah et al. 2009). Tanzanian studies in the 1990s reported HIV as cause of 0.4% and 3.6% of deaths in two rural areas, and 6.5% in an urban area (Mswia et al. 2003).

Figure 2 shows the proportion of deaths attributed to HIV in the 19 studies, overall and stratified by subregion. The summary proportion was 3.4% (95% confidence interval: 1.8–6.3), with very high heterogeneity (I-squared = 90.1%, tau-squared P-value < 0.0001). The summary proportion was lowest in West Africa (1.1%; 95%CI 0.4–3.3), higher in East Africa (4.5%; 95%CI 1.7–11.2) and highest in Southern Africa (26.1%; 95%CI 21.9–30.7). Heterogeneity was high in West Africa (I-squared = 76.9%, tau-squared P-value<0.0001) and East Africa (I-squared=83.4%, tau-squared P-value<0.0001), and very low between the two studies from Southern Africa (I-squared=0%, tau-squared P-value=0.750).

image

Figure 2. Proportion of deaths attributed to HIV in population-based studies, by subregion.

Download figure to PowerPoint

Excluding the six studies that did not mention HIV, the summary estimate for sub-Saharan Africa overall was 6.7% (95%CI: 3.7–11.6) with high heterogeneity (I-squared=90.0%, tau-squared P-value<0.0001) for West Africa 3.9% (2.0–7.6%, I-squared 44.1%, tau-squared P-value=0.128) and for East Africa 5.3% (2.0–13.1%, I-squared 84.4%, tau-squared P-value<0.0001) (data not shown).

Conclusions

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Conclusions
  7. Acknowledgements
  8. References
  9. Appendix

Our systematic review found 19 studies reporting cause-specific pregnancy-related or maternal mortality in national or subnational populations in sub-Saharan Africa. Twelve of the 19 studies attributed deaths to HIV. The summary proportion of deaths attributed to HIV was 3% (95%CI: 2–6), but there was substantial subregional variation: the proportion was 1% across ten West African studies (95%CI: 0–3), 4% across seven East African studies (95%CI: 2–11) and 26% in two Southern African studies (95%CI: 22–31). With the exception of the estimate for Southern Africa, high heterogeneity across studies means that the pooled estimates must be treated with great caution.

Our summary estimate of the proportion of deaths attributed to HIV is consistent with the findings from the previous systematic review of cause-of-death studies in Africa, as the 6.2% of maternal deaths previously attributed to HIV (Khan et al. 2006) falls inside the 95% confidence interval of the present summary estimate. However, our estimate is substantially lower than those from global models or the HIV population attributable fraction of pregnancy-related mortality found in six demographic surveillance sites in Africa (Zaba et al. 2013). No study in our review had a proportion as high as the 32% estimated across sub-Saharan Africa in 2008 by Hogan et al. (2010) or the 45% population attributable fraction estimated in East and Southern Africa (Zaba et al. 2013). The subregional estimates in the Hogan et al. model were also much higher: 26% in West Africa, 39% in East Africa and 80% in Southern Africa. The 9% estimated by the MMEIG model (WHO, 2010) is closer to the proportions reported in this review, but still more than twice the present estimate. The subregional estimates for 2008 in the MMEIG model were also higher: 4% in West Africa, 11% in East Africa and 48% in Southern Africa.

As our review includes data collected across a range of time periods, close correspondence with estimates for other time periods would not be expected, but this alone cannot explain the huge discrepancies. The distinction between deaths ‘with HIV’ and deaths ‘from HIV’ may account for some of the observed difference. Hogan et al. considered all deaths in HIV-positive pregnant/post-partum women as maternal, while population attributable fractions assume that all mortality in HIV-positive pregnant/post-partum women beyond the level observed in HIV-negative pregnant/post-partum women is due to HIV. Physicians in cause-of-death studies would only consider a death as HIV-related if the verbal autopsy or clinical history suggests the presence of AIDS. Data on the proportion of deaths in HIV-positive pregnant/post-partum women due to AIDS are scarce, but data from facility-based studies suggest that advanced HIV disease and AIDS cause a substantial proportion of pregnancy-related deaths among HIV-positive women. The South African 2008–2010 Confidential Enquiries into Maternal Deaths (CEMD), for example, reported that 66% of HIV-positive pregnant/post-partum women who died had AIDS, although AIDS was only the cause of 75% of those deathsNCCEMD 2012. A study in Mozambique using autopsies assigned 28% of pregnancy-related deaths in HIV-positive women to ‘HIV/AIDS-related conditions’ (Menendez et al. 2008). The much higher MMEIG estimate is therefore particularly surprising, as only 50% pregnancy-related deaths in HIV-positive women were considered to be maternal.

Our review was comprehensive, covering all population-based studies reporting cause-specific pregnancy-related or maternal deaths in sub-Saharan Africa published since 2003. We excluded studies solely based on health-facility data, as the causes of deaths in facility births compared with home births are known to differ (Directorate of Maternal & Child Health Care, Ministry of Health & Population 2001). We excluded the South African CEMD, which include overwhelmingly facility-based deaths (NCCEMD 2012). Including the 2008–2010, CEMD data would have raised the summary estimate for Southern Africa slightly to 27.7% (95% CI: 26.6–29.0) (data not shown), although the definition of AIDS deaths used by the CEMD (HIV-positive and CD4 count < 200) was far stricter than in any of the included studies. As we included the studies that did not report HIV and could have hidden deaths due to HIV in other categories, our estimates should be treated as the lower estimates from our data; however, using the higher estimates obtained by excluding those studies would not change our conclusions.

The overall quality of the included studies was poor. Few studies estimated the proportion of eligible deaths actually recorded, and three studies raised substantial doubts about the completeness of their recording of deaths outside health facilities (Issah et al. 2011; Cham et al. 2007; Lori & Starke 2012). Very little information was given on how deaths were assigned to particular causes, including HIV, except for stating that physicians reviewed the VA forms. Verbal autopsies based on physician review have substantial problems of reliability (Mills et al. 2008; Todd et al. 1994) and validity (Lozano et al. 2011) and must be interpreted with caution. HIV was most often reported as an indirect cause of maternal death; the question of how to classify HIV in pregnancy-related/maternal mortality was only mentioned in one study critical of the concept of ‘indirect’ causes (Garenne et al. 2013).

The inclusion of maternal as well as pregnancy-related deaths in our summary estimates may have affected the proportion of deaths attributed to HIV. This is unlikely to have introduced bias, for the following reasons. Distinguishing indirect from coincidental causes is not easy when relying on interviews with the relatives of the deceased, and maternal mortality estimates based on verbal autopsies may include some – if not all – coincidental causes. Clinical data on the contribution of coincidental causes to pregnancy-related mortality in sub-Saharan Africa are scarce, but among deaths notified to the South African Confidential Enquiries in 2008–2010, only 2–3% of pregnancy-related deaths were due to coincidental causes (NCCEMD 2012). As accidents and injuries constitute a small proportion of adult female mortality (Fottrell et al. 2007; Hosegood et al. 2004), it seems unlikely that inconsistencies in reporting ‘maternal’ or ‘pregnancy-related’ deaths would introduce bias to the results.

The subregional patterns are consistent with variations in background HIV prevalence and epidemic maturity. Two studies from southern Africa had very similar results, while heterogeneity persisted after subregional stratification in West and East Africa. The small number of studies within each stratum, and the lack of methodological rigour, however, precluded a meaningful analysis of the reasons for this residual heterogeneity.

There were limitations to our criteria for assessing study quality: some items reflect the quality of the reporting rather than of the actual methods used; quantitative and qualitative cut-offs are arbitrary; and for judging, the quality of verbal autopsy studies we considered only length of recall, on which there are guidelines (WHO 2012a), and not methods of VA interpretation. We consider our criteria reasonable in the absence of valid guidelines, but did not feel it appropriate to stratify results based on study quality.

The lack of transparency in how HIV was assigned as a cause of death underlines the general uncertainty around the classification of HIV-related maternal or pregnancy-related deaths. HIV was almost universally treated as an indirect cause, suggesting that aggravation is widely assumed – if indeed, any consideration is made of the criteria defining an ‘indirect’ maternal death. Few studies reported the basis on which a death was regarded as attributable to HIV, and those using clinical signs/symptoms did not describe these. Evidence for aggravation of HIV disease by pregnancy is scant and inconclusive (McIntyre 2005); a systematic review suggests the only direct obstetric complication clearly associated with HIV infection is puerperal sepsis (Calvert & Ronsmans). It is not clear that deterioration in health due to pregnancy would lead to death during or within 6 weeks of pregnancy (Van der Paal et al. 2007).

Where HIV is highly prevalent, AIDS deaths are common among HIV-positive pregnant/post-partum women (NCCEMD 2012; Menendez et al. 2008; Black et al. 2009), and knowing the proportion of pregnancy-related deaths associated with AIDS will provide valuable information for the monitoring and evaluation of safe motherhood programmes. There is substantial experience on how to assign deaths to HIV/AIDS in adults, and clear signs and symptoms of HIV disease have been identified (WHO 2007). Recent efforts to improve the reliability of verbal autopsies, such as InterVA (Byass et al. 2006) and ‘simplified symptom pattern’ (Murray et al. 2011), have improved the capacity for verbal autopsies to produce standard cause-of-death categories, including for AIDS. Concerns are currently being raised about the inclusion of AIDS deaths in maternal mortality statistics (Calvert & Ronsmans 2013; Garenne et al. 2013; Bradshaw & Dorrington 2012). Categorising pregnancy-related deaths by direct obstetric and non-obstetric causes (broken down by cause, including AIDS) (Garenne et al. 2013; Menendez et al. 2008) avoids the discussion on attribution. Such a classification, building on reproducible methods, will provide valuable information on how the HIV epidemic affects pregnancy-related mortality, and how actions can be taken accordingly.

Countries in sub-Saharan Africa are making least progress in reducing maternal mortality, and some countries have seen a reversal in their progress (WHO, 2010). Despite the high HIV-attributable mortality in pregnant or post-partum women in this region, HIV has received little attention as a cause of maternal or pregnancy-related mortality, and methods have been poor. Clear guidelines on classifying pregnancy-related deaths as attributable to HIV are urgently needed, so that the effect of the HIV epidemic on pregnancy-related mortality can be adequately monitored.

Acknowledgements

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Conclusions
  7. Acknowledgements
  8. References
  9. Appendix

We thank Clara Calvert for screening a 50% sample of the titles and abstracts from the original search. CG was funded by an Economic and Social Research Council PhD studentship. Part of the work was supported by the Child Health Epidemiology Reference Group (CHERG), funded by the World Health Organization.

References

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Conclusions
  7. Acknowledgements
  8. References
  9. Appendix
  • Adamu YM, Salihu HM, Sathiakumar N & Alexander GR (2003) Maternal mortality in northern Nigeria: a population-based study. European Journal of Obstetrics, Gynecology and Reproductive Biology 109, 153159.
  • Asamoah BO, Moussa KM, Stafstrom M & Musinguzi G (2011) Distribution of causes of maternal mortality among different socio-demographic groups in Ghana; a descriptive study. BMC Public Health 11, 159.
  • Ba MG, Kodio B & Etard JF (2003) Verbal autopsy to measure maternal mortality in rural Senegal. [French]. Journal de Gynecologie Obstetrique et Biologie de la Reproduction 32, 728735.
  • Bell JS, Ouedraogo M, Ganaba R, Sombie I & Byass P (2008) The epidemiology of pregnancy outcomes in rural Burkina Faso. Tropical Medicine and International Health 13(Suppl 1), 3143.
  • Black V, Brooke S & Chersich MF (2009) Effect of human immunodeficiency virus treatment on maternal mortality at a tertiary center in South Africa: a 5-year audit. Obstetrics and Gynecology 114, 292299.
  • Bradshaw D & Dorrington R (2012) Maternal mortality ratio - trends in the vital registration data. South African Journal of Obstetrics and Gynaecology 18, 3842.
  • Byass P, Fottrell E, Dao LH et al. (2006) Refining a probabilistic model for interpreting verbal autopsy data. Scandinavian Journal of Public Health 34, 2631.
  • Calvert C & Ronsmans C (2013) HIV and risk of direct obstetric complications: a systematic review and meta-analysis. PLoS ONE 8, e74848.
  • Calvert C & Ronsmans C (2013) The contribution of HIV to pregnancy-related mortality: a systematic review and meta-analysis. AIDS 27, 16311639.
  • Cham M, Vangen S & Sundby J (2007) Maternal deaths in rural Gambia. Global Public Health 2, 359372.
  • Directorate of Maternal and Child Health Care, Ministry of Health and Population (2001) Egypt National Maternal Mortality Study 2000. Directorate of Maternal and Child Health Care, Ministry of Health and Population, Egypt [last accessed 20th September 2013]; Available from: http://www.mofa.go.jp/policy/oda/category/health/sympo0306/study.pdf.
  • Etard JF, Kodio B & Ronsmans C (2003) Seasonal variation in direct obstetric mortality in rural Senegal: role of malaria? American Journal of Tropical Medicine and Hygiene 68, 503504.
  • Fottrell E, Byass P, Ouedraogo TW et al. (2007) Revealing the burden of maternal mortality: a probabilistic model for determining pregnancy-related causes of death from verbal autopsies. Population Health Metrics 5, 1.
  • French R & Brocklehurst P (1998) The effect of pregnancy on survival in women infected with HIV: A systematic review of the literature and meta-analysis. British Journal of Obstetrics and Gynaecology 105, 827835.
  • Garenne M, Kahn K, Collinson M, Gomez-Olive X & Tollman S (2013) Protective Effect of Pregnancy in Rural South Africa: Questioning the Concept of “Indirect Cause” of Maternal Death. PLoS ONE 8, 5.
  • Graham W & Hussein J (2003) Measuring and Estimating Maternal Mortality in the Era of HIV/AIDS. Un/Pop/Mort/2003/8. United Nations Department of Economic and Social Affairs, Population Division, New York.
  • Hogan MC, Foreman KJ, Naghavi M et al. (2010) Maternal mortality for 181 countries, 1980-2008: a systematic analysis of progress towards Millennium Development Goal 5. Lancet 375, 16091623.
  • Hosegood V, Vanneste AM & Timaeus IM (2004) Levels and causes of adult mortality in rural South Africa: the impact of AIDS. AIDS 18, 663671.
  • Huedo-Medina TB, Sanchez-Meca J, Marin-Martinez F & Botella J (2006) Assessing heterogeneity in meta-analysis: Q statistic or I2 index? Psychological Methods 11, 193206.
  • Hynes M, Sakani O, Spiegel P & Cornier N (2012) A study of refugee maternal mortality in 10 countries, 2008-2010. International Perspectives on Sexual and Reproductive Health 38, 205213.
  • Issah K, Nang-Beifubah A & Opoku CF (2011) Maternal and neonatal survival and mortality in the Upper West Region of Ghana. International Journal of Gynaecology and Obstetrics 113, 208210.
  • Khan KS, Wojdyla D, Say L, Gülmezoglu AM & Van Look PF (2006) WHO analysis of causes of maternal death: a systematic review. Lancet 367, 10661074.
  • Kumar RM, Uduman SA & Khurrana AK (1997) Impact of pregnancy on maternal AIDS. Journal of Reproductive Medicine 42, 429434.
  • Lewis R, West L & Mazive E (2011) Maternal mortality in Mozambique: findings on the timing of death and the role of HIV/AIDS. Paper prepared for the annual meeting of the Population Association of America, Washington DC, March 31-April 2, 2011. [Last accessed 5th November 2012] Available from: http://paa2011.princeton.edu/download.aspx?submissionId=111467
  • Lori JR & Starke AE (2012) A critical analysis of maternal morbidity and mortality in Liberia, West Africa. Midwifery 28, 6772.
  • Lozano R, Lopez AD, Atkinson C, Naghavi M, Flaxman AD & Murray CJ (2011) Performance of physician-certified verbal autopsies: multisite validation study using clinical diagnostic gold standards. Population Health Metrics 9, 32.
  • McIntyre J (2005) Maternal health and HIV. Reproductive Health Matters 13, 129135.
  • Menendez C, Romagosa C, Ismail MR, Carrilho C & Saute F (2008) An autopsy study of maternal mortality in Mozambique: the contribution of infectious diseases. PLoS Medicine 5, e44.
  • Mills S, Williams JE, Wak G & Hodgson A (2008) Maternal mortality decline in the Kassena-Nankana district of northern Ghana. Maternal and Child Health Journal 12, 577585.
  • Mohammed AA, Elnour MH, Mohammed EE, Ahmed SA & Abdelfattah AI (2011) Maternal mortality in Kassala State - Eastern Sudan: community-based study using reproductive age mortality survey (RAMOS). BMC Pregnancy and Childbirth 11, 102.
  • Mswia R, Lewanga M, Moshiro C, Whiting D & Wolfson L (2003) Community-based monitoring of safe motherhood in the United Republic of Tanzania. Bulletin of the World Health Organization 81, 8794.
  • Munjanja S , Nystrom L, Nyandoro M & Magwali T (2007) Maternal and perinatal mortality study. Available from: http://www.unicef.org/zimbabwe/ZMPMS_report.pdf.
  • Murray CJ, James SL, Birnbaum JK, Freeman MK, Lozano R & Lopez AD (2011) Simplified Symptom Pattern Method for verbal autopsy analysis: multisite validation study using clinical diagnostic gold standards. Population Health Metrics 9, 30.
  • NCCEMD (2012) Saving Mothers 2008–2010: Fifth report on the confidential enquiries into maternal deaths in South Africa. Available from: http://www.doh.gov.za/docs/reports/2012/Report_on_Confidential_Enquiries_into_Maternal_Deaths_in_South_Africa.pdf.
  • R Core Team (2013). R: A Language and Environment for Statistical Computing. R Foundation for Statistical Computing, Vienna.
  • Rosen JE, de Zoysa I, Dehne K, Mangiaterra V & Abdool-Karim Q (2012) Understanding methods for estimating HIV-associated maternal mortality. Journal of Pregnancy 2012, 958262.
  • Rucker G, Schwarzer G, Carpenter JR & Schumacher M (2008) Undue reliance on I(2) in assessing heterogeneity may mislead. BMC Medical Research Methodology 8, 79.
  • Todd JE, De Francisco A, O'Dempsey TJ & Greenwood BM (1994) The limitations of verbal autopsy in a malaria-endemic region. Annals of Tropical Paediatrics 14, 3136.
  • WHO (2010) Trends in maternal mortality: 1990 to 2008. Estimates developed by WHO, UNICEF, UNFPA and The World Bank. Available from: http://apps.who.int/iris/handle/10665/44423.
  • Van der Paal L, Shafer LA, Mayanja BN, Whitworth JA & Grosskurth H (2007) Effect of pregnancy on HIV disease progression and survival among women in rural Uganda. Tropical Medicine and International Health 12, 920928.
  • WHO (2007) WHO Case Definitions of HIV for Surveillance and Revised Clinical Staging and Immunological Classification of HIV-Related Disease in Adults and Children. WHO, Geneva.
  • WHO (2012a) Verbal Autopsy Standards: 2012 WHO Verbal Autopsy lnstrument. WHO, Geneva.
  • WHO (2012b) The WHO application of ICD-10 to deaths 5 during pregnancy, childbirth and the puerperium: ICD-MM. Issah K, Nang-Beifubah A & Opoku CF (2011) Maternal and neonatal survival and mortality in the Upper West Region of Ghana. International Journal of Gynaecology and Obstetrics 113, 208210.
  • Zaba B, Calvert C, Marston M et al. (2013) Effect of HIV infection on pregnancy-related mortality in sub-Saharan Africa: secondary analyses of pooled community-based data from the network for Analysing Longitudinal Population-based HIV/AIDS data on Africa (ALPHA). Lancet 381, 17631771.
  • Zakariah AY, Alexander S, Van Roosmalen J, Buekens P, Kwawukume EY & Frimpong P (2009) Reproductive age mortality survey (RAMOS) in Accra, Ghana. Reproductive Health 6, 1.
  • Ziraba AK, Madise N, Mills S, Kyobutungi C & Ezeh A (2009) Maternal mortality in the informal settlements of Nairobi city: What do we know? Reproductive Health 6, 6.

Appendix

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Conclusions
  7. Acknowledgements
  8. References
  9. Appendix

Search strategy

Embase & Medline
1(((“postpartum” or “antepartum” or “post-partum” or “ante-partum” or “uterine” or “vaginal” or “uterine” or “puerperal” or “peripartum” or “peri-partum”) and (“hemorrhage” or “haemorrhage”or “bleeding”)) or exp obstetric hemorrhage/or exp antepartum hemorrhage/or exp intrapartum hemorrhage/or exp postpartum hemorrhage/)
2exp Endometritis/or “endometritis” or exp Puerperal Infection/or “puerperal infection*” or exp intrauterine infection/or exp puerperal disorder/or “infectious pregnancy complications” or ((“maternal” or “pregnan*” or “childbirth” or “mother*” or “Parturition” or exp Pregnancy/or exp Pregnancy Disorder/or exp Pregnancy Complication/or exp expectant Mother/or exp mother/or exp Birth/or exp Childbirth/) and (“sepsis”or exp sepsis/or “septic*”)) or exp intrauterine infection/or exp puerperal disorder/or (“Pregnancy Complications” adj3 “Infectious”)
3(“abortion” or “menstrual regulation” or “pregnancy termination” or exp pregnancy termination/or exp Second Trimester Abortion/or exp Legal Abortion/or exp Abortion/or exp Septic Abortion/or exp Therapeutic Abortion/or exp Induced Abortion/or exp Illegal Abortion/or exp Missed Abortion/or exp Selective Abortion/or exp Surgical Abortion/or exp Spontaneous Abortion/or exp Medical Abortion/or exp Imminent Abortion/or exp Recurrent Abortion/or abortion.mp. or exp Incomplete Abortion/or “unsafe abortion”)
4((“eclamp*” or “pre-eclamp*” or “preeclamp*” or “(hypertensi*” and (“pregnan*” or “maternal”)) or “hellp” or exp Maternal Hypertension/or exp Eclampsia and Preeclampsia/or exp Eclampsia/or “eclampsia” or exp Preeclampsia/or exp Hellp Syndrome/)
5(((“obstructed” or “prolonged”) and (“labor” or “labour”)) or “dystocia” or (“uter*” and “rupture”) or exp Uterus rupture/)
6(((“ectopic” or “tubal”) and “pregnan*”) or “eccysis” or exp Ectopic Pregnancy/)
7((“Embolism” or “amniotic fluid embolism” or exp Amnion Fluid Embolism/or exp Embolism/) and (“maternal” or “pregnan*” or “childbirth” or “mother*” or “Parturition” or exp Pregnancy/or exp Pregnancy Disorder/or exp Pregnancy Complication/or exp expectant Mother/or exp mother/or exp Birth/or exp Childbirth/))
8((“hiv infections” or “hiv” or “human immunodeficiency virus” or “human immuno-deficiency virus” or “hiv-1” or “hiv-2” or “hiv1” or “hiv2” or “aids” or “hiv/aids” or “hiv-aids” or “acquired immune deficiency syndrome” or “acquired immunodeficiency syndrome” or exp HIV/or exp Human Immunodeficiency Virus Infection/) and (“maternal” or “pregnan*” or “childbirth” or “mother*” or “Parturition” or exp Pregnancy/or exp Pregnancy Disorder/or exp Pregnancy Complication/or exp expectant Mother/or exp mother/or exp Birth/or exp Childbirth/))
9((“malaria” or exp malaria/) and (“maternal” or “pregnan*” or “childbirth” or “mother*” or “Parturition” or exp Pregnancy/or exp Pregnancy Disorder/or exp Pregnancy Complication/or exp expectant Mother/or exp mother/or exp Birth/or exp Childbirth/or “placent*” or exp placenta/))
10((“tetanus” or exp tetanus/) and (“maternal” or “pregnan*” or “childbirth” or “mother*” or “Parturition” or exp Pregnancy/or exp Pregnancy Disorder/or exp Pregnancy Complication/or exp expectant Mother/or exp mother/or exp Birth/or exp Childbirth/))
11((“spouse abuse” or “battered women” or “domestic violence” or “domestic abuse” or “spous* abuse” or “violence” or “abuse*” or “battered women” or “violence against women” or “wife abuse” or “intimate partner violence” or “battered mother*” or exp battered woman/or exp partner violence/) AND (“maternal” or “pregnan*” or “childbirth” or “mother*” or “Parturition” or exp Pregnancy/or exp Pregnancy Disorder/or exp Pregnancy Complication/or exp expectant Mother/or exp mother/or exp Birth/or exp Childbirth/))
12((“iron deficien*” or “anaemia” or “anemia”) and (“maternal” or “pregnan*” or “childbirth” or “mother*” or “Parturition” or exp Pregnancy/or exp Pregnancy Disorder/or exp Pregnancy Complication/or exp expectant Mother/or exp mother/or exp Birth/or exp Childbirth/))
13(exp Pregnancy/or exp Pregnancy Disorder/or exp Pregnancy Complication/or exp expectant Mother/or exp mother/or exp Birth/or exp Childbirth/or “mother*” or “pregnan*” or “maternal” or “parturition” or “childbirth” or “ante-partum” or “antepartum” or “intra-partum” or “intrapartum” or “peri-partum” or “peripartum” or “post-partum” or “postpartum” or “delivery”)
14 (1-13) with OR, time limited
15(“Death*” or “mortali*” or “fatalit*” or exp Death/or exp Cause of Death/or exp Mortality/)
16((“suicide” or exp suicide/) and (“maternal” or “pregnan*” or “childbirth” or “mother*” or “Parturition” or exp Pregnancy/or exp Pregnancy Disorder/or exp Pregnancy Complication/or exp expectant Mother/or exp mother/or exp Birth/or exp Childbirth/or “post-partum” or “postpartum” or “peri-partum” or “peripartum” or “ante-partum” or “antepartum”))
17((exp Homicide/or “homicide*” or “killing*” or “murder*”) and (“maternal” or “pregnan*” or “childbirth” or “mother*” or “Parturition” or exp Pregnancy/or exp Pregnancy Disorder/or exp Pregnancy Complication/or exp expectant Mother/or exp mother/or exp Birth/or exp Childbirth/or “post-partum” or “postpartum” or “peri-partum” or “peripartum” or “ante-partum” or “antepartum”))
18(“maternal mortality” or exp Maternal Mortality/)
19 ((14 & 15)/16/17/18) - Full maternal death, year-limited
20(“hiv infections” or “hiv” or “human immunodeficiency virus” or “human immuno-deficiency virus” or “hiv-1” or “hiv-2” or “hiv1” or “hiv2” or “aids” or “hiv/aids” or “hiv-aids” or exp HIV/or exp Human Immunodeficiency Virus Infection/)
21(exp Cause of death/or “Cause of death” or “mortality review” or “death review” or “verbal autopsy” or (“confidential” and (“enquir*” or “inquir*”)))
22 (19 & (20/21) - Maternal Mortality and (HIV or COD)
23“Angola” or “Benin” or “Botswana” or “Burkina Faso” or “Burundi” or “Cameroon” or “Central African Republic” or “Chad” or “Comoros” or “Congo” or “Cote d'Ivoire” or “Democratic republic of the congo” or “Djibouti” or “Equatorial guinea” or “Eritrea” or “Ethiopia” or “Gabon” or “Gambia” or “Ghana” or “Guinea” or “Guinea-Bissau” or “Kenya” or “Lesotho” or “Liberia” or “Madagascar” or “Malawi” or “Mali” or “Mauritania” or “Mauritius” or “Mozambique” or “Namibia” or “Niger” or “Nigeria” or “Rwanda” or “Sao tome and principe” or “Senegal” or “Seychelles” or “Sierra Leone” or “Somalia” or “South Africa” or “Sudan” or “Swaziland” or “Tanzania” or “Togo” or “Uganda” or “Zambia” or “Zimbabwe” or angola/or benin/or botswana/or burkina faso/or burundi/or cameroon/or central african republic/or chad/or comoros/or congo/or cote d'ivoire/or “democratic republic of the congo”/or djibouti/or equatorial guinea/or eritrea/or ethiopia/or gabon/or gambia/or ghana/or guinea/or guinea-bissau/or kenya/or lesotho/or liberia/or madagascar/or malawi/or mali/or mauritania/or mauritius/or mozambique/or namibia/or niger/or nigeria/or rwanda/or sao tome and principe/or senegal/or seychelles/or sierra leone/or somalia/or south africa/or sudan/or swaziland/or tanzania/or togo/or uganda/or zambia/or zimbabwe/or “Luanda” or “Porto-Novo” or “Gaborone” or “Ouagadougou” or “Bujumbura” or “Yaounde” or “Bangui” or “N'Djamena” or “Moroni” or “Brazzaville” or “Abidjan” or “Kinshasa” or “Djibouti” or “Malabo” or “Asmara” or “Addis Ababa” or “Libreville” or “Banjul” or “Accra” or “Conakry” or “Bissau” or “Nairobi” or “Maseru” or “Monrovia” or “Antananarivo” or “Lilongwe” or “Bamako” or “Nouakchott” or “Port Louis” or “Maputo” or “Windhoek” or “Niamey” or “Abuja” or “Lagos” or “Kigali” or “Sao Tome” or “Dakar” or “Victoria” or “Freetown” or “Mogadishu” or “Pretoria” or “Khartoum” or “Mbabane” or “Dar es Salaam” or “Lome” or “Kampala” or “Lusaka” or “Harare” or africa south of the sahara/or “sub-saharan africa” or “developing countr*”
24 (22 & 23) - Total result
African Index Medicus 
Maternal AND mortalityMothers AND mortality
Maternal AND deathChildbirth AND mortality
Pregnancy AND deathChildbirth AND death
Pregnancy AND mortalityPregnant AND mortality
Mothers AND deathPregnant AND death
Popline
(maternal/pregnan*/childbirth/mother*/placenta*/parturition/ante-partum/antepartum/intra-partum/intrapartum/peri-partum/peripartum/post-partum/postpartum) & (death/mortalit*/fatalit*) & ((cause*/etiology/aetiology)/verbal autopsy/(confidential & (enquir*/inquir*))/(HIV/human immuno-deficiency virus/human immunodeficiency virus/hiv-1/hiv-2/hiv1/hiv2/aids/hiv/aids/hiv-aids/acquired immune deficiency syndrome/acquired immunodeficiency syndrome)) & (“Angola”/“Benin”/“Botswana”/“Burkina Faso”/“Burundi”/“Cameroon”/“Central African Republic “/“Chad”/“Comoros”/“Congo”/“Cote d'Ivoire”/“Democratic republic of the congo”/“Djibouti”/“Equatorial guinea”/“Eritrea”/“Ethiopia”/“Gabon”/“Gambia”/“Ghana”/“Guinea”/“Guinea-Bissau”/“Kenya”/“Lesotho”/“Liberia”/“Madagascar”/“Malawi”/“Mali”/“Mauritania”/“Mauritius”/“Mozambique”/“Namibia”/“Niger”/“Nigeria”/“Rwanda”/“Sao tome and principe”/“Senegal”/“Seychelles”/“Sierra Leone”/“Somalia”/“South Africa”/“Sudan”/“Swaziland”/“Tanzania”/“Togo”/“Uganda”/“Zambia”/“Zimbabwe”/“Luanda”/“Porto-Novo”/“Gaborone”/“Ouagadougou”/“Bujumbura”/“Yaounde”/“Bangui”/“N'Djamena”/“Moroni”/“Brazzaville”/“Abidjan”/“Kinshasa”/“Djibouti”/“Malabo”/“Asmara”/“Addis Ababa”/“Libreville”/“Banjul”/“Accra”/“Conakry”/“Bissau”/“Nairobi”/“Maseru”/“Monrovia”/“Antananarivo”/“Lilongwe”/“Bamako”/“Nouakchott”/“Port Louis”/“Maputo”/“Windhoek”/“Niamey”/“Abuja”/“Lagos”/“Kigali”/“Sao Tome”/“Dakar”/“Victoria”/“Freetown”/“Mogadishu”/“Pretoria”/“Khartoum”/“Mbabane”/“Dar es Salaam”/“Lome”/“Kampala”/“Lusaka”/“Harare”/“sub-saharan africa”/“developing countr*”)