The role of a maternity waiting area (MWA) in reducing maternal mortality and stillbirths in high-risk women in rural Ethiopia


Dr J Kelly, Public Health, Epidemiology and Biostatistics, College of Medical and Dental Sciences, University of Birmingham, Birmingham B15 2TT, UK. Email


Please cite this paper as: Kelly J, Kohls E, Poovan P, Schiffer R, Redito A, Winter H, MacArthur C. The role of a maternity waiting area (MWA) in reducing maternal mortality and stillbirths in high-risk women in rural Ethiopia. BJOG 2010;117:1377–1383.

Objective  To describe maternal mortality and stillbirth rates among women admitted via a maternity waiting area (MWA) and women admitted directly to the same hospital (non-MWA) over a 22-year period.

Design  Retrospective cohort study.

Setting  Hospital in rural Ethiopia, which provided comprehensive emergency obstetric care and has an established MWA.

Population  All women admitted for delivery between 1987 and 2008.

Methods  Data on maternal deaths, stillbirths, caesarean section and uterine rupture were abstracted from routine hospital records. Sociodemographic characteristics, antenatal care and other data were collected for 2008 only. Rates and 95% confidence intervals were calculated for maternal mortality and stillbirth.

Main outcome measures  Maternal mortality and stillbirth.

Results  There were 24 148 deliveries over the study period, 6805 admitted via MWA and 17 343 admitted directly. Maternal mortality was 89.9 per 100 000 live births (95% CI, 41.1–195.2) for MWA women and 1333.1 per 100 000 live births (95% CI, 1156.2–1536.7) for non-MWA women; stillbirth rates were 17.6 per 1000 births (95% CI, 14.8–21.0) and 191.2 per 1000 births (95% CI, 185.4–197.1), respectively; 38.5% of MWA women were delivered by caesarean section compared with 20.3% of non-MWA women, and none had uterine rupture, compared with 5.8% in the non-MWA group. For the 1714 women admitted in 2008, relatively small differences in sociodemographic characteristics, distance and antenatal care uptake were found between groups.

Conclusions  Maternal mortality and stillbirth rates were substantially lower in women admitted via MWA. It is likely that at least part of this difference is accounted for by the timely and appropriate obstetric management of women using this facility.


The role of maternity waiting areas (MWAs) in reducing maternal mortality and stillbirth has received little attention in the literature.1 Antenatal risk assessment cannot predict all women who will need emergency obstetric care (EmOC),2 but certain high-risk groups who require delivery in or near EmOC facilities can be identified: for example, women who have had a caesarean section or uterine rupture are at higher risk of rupture in labour, resulting in almost inevitable stillbirth and frequently maternal death.3 In rural parts of developing countries, the distance and difficulty in travelling to EmOC facilities in labour continue to contribute to high maternal death and stillbirth rates.4 Although some countries in rural parts of the developing world have maternity waiting facilities, there have been very few evaluations of these. A new Cochrane review of MWAs for improving maternal and neonatal outcomes found only six cohort studies (five hospital and one community based), all of which were relatively small, and no randomised controlled trials, and thus concluded that there is insufficient evidence on this at present.1 The review commented that appropriate risk selection for attendance at an MWA is critical, as is an appropriate level of skilled supervision and location adjacent to the hospital. Acceptance by the local community, including cost, is also important.

We present data from a large study of the outcomes of women admitted to an established MWA, which fulfils these important criteria, compared with those of women admitted directly to a rural hospital in Ethiopia over a 22-year period.


The aims of this study were to investigate the frequency of maternal mortality and stillbirth in all women admitted via an MWA, and to compare these rates with those for women admitted directly to the same rural hospital serving the same population catchment area.

The study design was a retrospective cohort analysis of routinely collected data from all women who delivered in Attat Hospital in rural Ethiopia over a 22-year period from 1987 to 2008.

Attat Hospital (and its integrated health service) is located 180 km south-west of Addis Ababa, in the southern region of Ethiopia, Gurage Zone. It has been operative since 1969 and its catchment area has increased over the years to cover a population of about 800 000. The hospital has 65 beds and provides comprehensive EmOC 24 hours per day, 7 days per week. Home delivery is the norm in the area, occurring in about 95% of women.5 Although there has been a recent slight increase in women without risk factors seeking hospital confinement, when admitted in labour most women would come with actual or anticipated complications of labour.

The hospital has had an established MWA (now 40 beds) in the hospital grounds since 1973, where high-risk women spend the last few weeks of pregnancy prior to delivery in the hospital. Indications for admission to MWA include risks relating to previous obstetric history and those pertaining to the current pregnancy. Previous history includes previous caesarean section, uterine rupture, perinatal death, habitual abortion or repaired obstetric fistula. Current pregnancy factors include short primigravida, young or elderly primigravida, as well as complications arising or detected during the current pregnancy, such as antepartum haemorrhage, pre-eclampsia, multiple pregnancy and severe anaemia. As routine HIV testing in pregnancy was offered 6 years ago, there are also an increasing number of women ill from HIV who are admitted. A few women with an antenatal diagnosis of intrauterine death (IUD) or severe abnormality incompatible with life (e.g. anencephaly) are also admitted. There are no directs costs of MWA to a woman, except for those pertaining to childcare and other tasks whilst she is away and to a companion who will reside at MWA, find firewood and food, and cook for her. In 1982, the hospital introduced a scheme to further encourage women to accept advice to reside in MWA, such that the fee was reduced to one-third of the usual cost for an instrumental delivery and one-tenth of the usual cost for a caesarean section, if required.

The hospital has an ongoing programme of training and updating community health agents and trained traditional birth attendants, who are given clear guidelines which detail all possible reasons (as described earlier) for referral to MWA. Women in MWA attend the hospital antenatal clinic weekly or, in case of emergency, are seen immediately by a midwife and/or doctor.

Hospital records for all women delivered in the hospital are abstracted each year and reviewed as part of the routine audit; data on delivery and outcomes for the years 1987–2008 were collated for the purposes of this study.

Given that some characteristics of the women in MWA and those admitted directly to the hospital will be different, we obtained data on sociodemographic characteristics for the two groups for 2008, such data not being routinely recorded. This was explained to the women and verbal consent was obtained by the admitting midwife. This year was considered to be typical, except for a higher than usual number of women admitted to MWA because of IUD (four) or anencephaly (six). Most of this information was collected by the midwife, in the local language, when the woman was admitted to the labour ward; however, in a few very ill women, it was collected later. Most women do not know their age, and so they are categorised as ‘young’ if they look under 18 years; ‘old’ generally suggests an age over 35 years. Parity was categorised into ‘0’, ‘1–3’ or ‘4 or more’. It is difficult to obtain gestation in all women, because most are unsure of dates, and none present before 20 weeks of gestation. Non-MWA women were asked by the admitting midwife whether they had received antenatal care, about the cost of transport to the hospital and whether the journey was by day or by night. The main reason for admission to MWA was recorded.

Maternal mortality per 100 000 live births and stillbirth per 1000 births were calculated separately for the MWA and non-MWA groups, both with 95% confidence intervals.


Maternal mortality and stillbirth

During the study period, there were 24 148 deliveries at Attat Hospital, 28.2% (n = 6805) of which were to women admitted via MWA. The maternal mortality ratio (MMR) was 89.8 per 100 000 live births (95% CI, 41.1–195.7) for women in MWA and 1333.1 per 100 000 live births (95% CI, 1156.2–1536.7) for non-MWA women. This is shown in Table 1, together with the rates for each year.

Table 1.   Delivery factors and outcomes for women admitted via maternity waiting area or directly to the same hospital, 1987–2008
YearMaternity waiting area (MWA)Women admitted directly (non-MWA)
DeliveriesCaesarean section (%)Stillbirth (stillbirth rate)*Maternal deaths (MMR)**DeliveriesCaesarean section (%)Stillbirth (stillbirth rate)*Uterine rupture (%)Maternal deaths (MMR)**
  1. *Stillbirth rate = stillbirths/1000 births (live births and stillbirths).

  2. **Maternal mortality ratio = maternal deaths/100 000 live births.

  3. ***Unusually high because 10 patients were admitted to MWA because of intrauterine death or anencephaly.

198715440 (26.0)4 (26.0)062392 (14.8)161 (258.4)45 (7.2)13 (2813.9)
198818857 (30.3)2 (10.6)058186 (14.8)145 (249.6)51 (8.8)9 (2064.2)
198922069 (31.4)6 (27.3)050846 (9.1)138 (271.7)49 (9.6)8 (2162.2)
199020548 (23.4)3 (14.6)052067 (12.9)140 (269.2)47 (9.0)17 (4473.7)
199123280 (34.5)6 (25.9)044173 (16.6)137 (310.7)48 (10.9)9 (2960.5)
199220269 (34.2)5 (24.8)056782 (14.5)140 (246.9)50 (8.8)4 (936.8)
199317777 (43.5)5 (28.2)1 (581.4)612109 (17.8)152 (248.4)60 (9.8)9 (1956.5)
1994291135 (46.4)1 (3.4)1 (344.8)757134 (17.7)180 (237.8)63 (8.3)6 (1039.9)
1995288121 (42.0)2 (6.9)0727166 (22.8)152 (209.1)51 (7.0)8 (1391.3)
1996276120 (43.5)3 (10.9)1 (366.3)865148 (17.1)192 (222.0)48 (5.5)9 (1337.3)
1997282111 (39.4)9 (31.9)1 (366.3)775138 (17.8)138 (178.1)41 (5.3)8 (1255.9)
1998302110 (36.4)5 (16.6)0821176 (21.4)164 (199.8)45 (5.5)16 (2435.3)
199925698 (38.3)5 (19.5)0740175 (23.6)146 (197.3)49 (6.6)4 (673.4)
2000306112 (36.6)4 (13.1)0823193 (23.5)163 (198.1)51 (6.2)3 (454.5)
2001353151 (42.8)5 (14.2)1 (287.4)887193 (21.8)135 (152.2)48 (5.4)3 (398.9)
2002386151 (39.1)3 (7.8)1 (261.1)902160 (17.7)158 (175.2)41 (4.5)8 (1075.3)
2003391161 (41.2)6 (15.3)0859200 (23.3)155 (180.4)41 (4.8)10 (1420.5)
2004387187 (48.3)2 (5.2)0805183 (22.7)141 (175.2)34 (4.2)10 (1506.0)
2005385160 (41.6)9 (23.4)01045272 (26.0)162 (155.0)43 (4.1)9 (1019.3)
2006431167 (38.7)7 (16.2)01122259 (23.1)153 (136.4)28 (2.5)11 (1135.2)
2007478163 (34.1)9 (18.8)01264373 (29.5)141 (111.6)48 (3.8)11 (979.5)
2008615236 (38.4)19*** (30.9)01099195 (17.7)123 (111.9)25 (2.3)2 (204.9)
Total68052623 (38.5)120 (17.6)6 (89.8)173433520 (20.3)3316 (191.2)1006 (5.8)187 (1333.1)

A high proportion of MWA women required delivery by caesarean section (38.5%), indicating their high-risk status, relative to 20.3% among those admitted directly. Uterine rupture (complete) did not occur in the MWA group, but occurred in 5.8% of direct admissions and was the cause of death for 62 (33.2%) of the maternal deaths among the non-MWA group. Three of the MWA women had incomplete rupture: all three mothers and two babies did well, and the third baby, a case of preterm abruptio, died.

The stillbirth rate for MWA women was 17.6 per 1000 births (95% CI, 14.8–21.0) compared with 191.2 per 1000 births (95% CI, 185.4–197.1) in non-MWA women (Table 1). This included 10 women with a diagnosis of IUD admitted via MWA because of this.

The main reasons for admission to MWA in 2008 are given in Table 2. The most common indication was previous caesarean section.

Table 2.   Main reasons for admission to the maternity waiting area (MWA) in 2008
ReasonNo. (%)
Previous caesarean section178 (28.9)
Short primigravida73 (11.9)
Suspected twins/malpresentation68 (11.1)
Previous stillbirths65 (10.6)
Grand multipara60 (9.8)
Pre-eclampsia/medical reasons58 (9.4)
Young/elderly primigravida58 (9.4)
Antepartum haemorrhage/history of postpartum haemorrhage22 (3.6)
Habitual abortion15 (2.4)
Intrauterine death/anencephaly10 (1.6)
Previously ruptured uterus3 (0.5)
Repaired vesico-vaginal fistula2 (0.3)
Not known3 (0.5)

Characteristics of women in the MWA and non-MWA groups

For the year 2008, most women were wives of poor subsistence farmers in both the MWA (584/615, 94.9%) and non-MWA (949/1099, 86.4%) groups. The women were also actively involved in farming. Only 2% in both groups had received any schooling. Literacy rates in the area are low, even for Africa, and a signature on the consent form at admission was used as an indication of literacy. This was similar in MWA and non-MWA groups, with a fingerprint being used in over one-half of women. Both primiparity and a younger age were more common in the non-MWA group. The receipt of some form of antenatal care in the current pregnancy was reported by 62% (681/1099) of non-MWA women. All MWA women were considered to have received some antenatal care, even those who were self-referred to hospital without previous antenatal care, as they then had care before delivery.

Women lived an average of about 40 km from the hospital, with a large range, but the distances were similar in the two groups (Table 3). The costs of transport for an average (40-km) and a long (≥ 200-km) journey by day were $US6 and $US30, respectively, whereas similar journeys by night cost $US36 and $US180, respectively. Many women who did not live near a road were carried by stretcher, sometimes aided by a donkey or mule until they reached a road or track that could be used by a vehicle. Seventeen women said that part of their journey was at night, and another 22 indicated that they ‘took shelter’ when darkness fell; 73% (807/1099) of non-MWA women reported that the cost of transport was the cause of their delay in coming to hospital.

Table 3.   Sociodemographic characteristics (2008)
 MWA (n = 615)Non-MWA (n = 1099)P
No. (%)No. (%)
  1. MWA, maternity waiting area; NS, not significant.

Patient signature117 (19)253 (23)0.020
Husband signature142 (23)286 (26)
Fingerprint356 (58)560 (51)
Young230 (37)490 (45)0.009
Average149 (24)228 (21)
Old229 (37)360 (33)
Not known7 (1)21 (2)
0193 (31)525 (48)<0.001
1–3294 (48)366 (33)
4 or more118 (19)177 (16)
Not known10 (2)31 (3)
Distance home to hospital (km)

The detailed modes of delivery of the MWA and non-MWA groups in 2008 are shown in Table 4. Vacuum delivery was more common in the non-MWA group and there were more caesarean sections in the MWA group. The causes of maternal death for the 10 years to 2008 are shown in Table 5. Obstructed labour, haemorrhage, hypertensive diseases and malaria were the most common.

Table 4.   Modes of delivery (2008)
 MWA (n = 615)Non-MWA (n = 1099)
No. (%)No. (%)
  1. MWA, maternity waiting area.

Spontaneous vaginal delivery (including twins)237 (38.5)488 (44.4)
caesarean section (including twins)236 (38.4)195 (17.7)
Vacuum113 (18.4)333 (30.3)
Breech26 (4.2)46 (4.2)
Craniotomy (baby dead)1 (0.2)25 (2.1)
Forceps2 (0.3)12 (1.1)
Table 5.   Causes of maternal death for 10 years to 2008 (1999–2008)
  1. MWA, maternity waiting area.

Haemorrhage (postpartum haemorrhage)114
Hypertensive disease011
Obstructed labour (including rupture)018
Abortion (complications)04
Malaria (cerebral)011
Severe anaemia06
Cardiac disease11
Relapsing fever01


This very large study, of over 24 000 deliveries in a rural hospital in Ethiopia, shows that maternal mortality among high-risk women admitted via an MWA, at 89.9 per 100 000 live births, was substantially lower than both the 1333 per 100 000 live births among women admitted directly and the 720 per 100 000 live births estimated by the World Health Organization for Ethiopia as a whole. Stillbirth rates were similarly much lower. With the inclusion of over 6800 women, this is by far the largest series of attendees at an MWA; the largest among the few previous studies included just over 1500.6 These large numbers allow more precise estimates of mortality.

The women admitted via MWA are there because they are at high risk, either from their previous history or from a factor identified in the current pregnancy, but they may not actually develop an obstetric problem. As most deliveries in the area are at home, by definition the women admitted directly to the hospital in or after labour will be there because they have developed a problem. Thus, higher maternal mortality and stillbirth rates in the non-MWA group relative to the MWA group would be expected. The differences in rates found between the two groups, however, were extremely large.

The true comparator of the MWA group would be all non-MWA women with a problem in their previous history and/or current pregnancy delivering within the catchment area, both at hospital or at home. Women with a normal history and normal current pregnancy would be excluded. This is, of course, not known. Indeed, such data from large, well-conducted, population-based studies anywhere in sub-Saharan Africa are sparse. Data from a large household survey in the capital of Ethiopia in 1981–3, of 9315 pregnant women, found an MMR of 566 per 100 000,7 and this is likely to be lower than that in rural areas. A smaller study, of 5832 households, in another major urban area in south-west Ethiopia found an MMR of 402 per 100 000.8 A household health survey in Sudan,9 of over 24 000 randomly selected households, calculated an MMR of 1107 per 100 000 live births.

The high-risk status of MWA women in this population was confirmed in many, with almost 40% requiring delivery by caesarean section in a clinical situation in which a vaginal delivery is attempted if at all possible. Vacuum delivery, as in most of the developing world, was used more often than forceps in both groups. The non-MWA group had a higher proportion of vacuum deliveries than the MWA group. This was partly because of the increased rate of IUDs which would be delivered by this method, and partly because delay in labour would be much greater.

There were 298 women with a history of ruptured uterus who made use of MWA. There were no complete uterine ruptures in this group, whereas the rate in those admitted directly was 5.8%, causing 62 maternal deaths. MWA women, especially with a history of uterine rupture, are advised to present at the labour ward early in labour. In our earlier follow-up study on the repair of ruptured uterus in the same hospital, 30.8% of babies were able to be delivered vaginally.10 Although caesarean section is generally the advised mode of delivery with a history of ruptured uterus, the women who delivered vaginally either wished to do so and/or came to the delivery ward progressing well in advanced labour.

To compare other features of the women and their care, additional data were obtained for deliveries in 2008. These showed household occupation and distance from the hospital to be similar for both groups. The lower proportions of primiparous and younger women in the MWA group can be explained by the fact that advice to use MWA was often based on a problem in a previous pregnancy and/or delivery. The significantly higher literacy rate, in particular a woman’s ability to sign her name, in the non-MWA group might be accounted for by women who planned to deliver in hospital in this group. Antenatal care attendance may vary between the groups and non-MWA women were asked about this on admission, with almost two-thirds reporting the receipt of some antenatal care. By definition, all MWA women must have received some antenatal care, even if only for the duration of their MWA stay. However, it cannot be assumed that all women were at MWA through antenatal care referral in this pregnancy. Some will have arrived because they had been advised at their last delivery that they would need a hospital birth and to attend MWA, and others may have been referred by their traditional birth attendant. An attempt was made for 2008 to ask the MWA women about this. There were 68% who were recorded as having received previous antenatal care. For the remaining 32% of women, receipt of previous antenatal care was not recorded, but it is considered that, in some of these cases, the data were missing rather than antenatal care had not been received. Nevertheless, receipt of antenatal care between the groups would not seem to represent a major difference.

The countryside of Ethiopia has difficult terrain with mountains and gorges, which contributes to problems with road communications. Almost three-quarters of women admitted directly reported that the cost of transport was an important cause of delay; thus, it is likely that this was a major contributor to maternal mortality and stillbirth in our study in women who did not use MWA. Few of the women or their relatives were willing to travel at night because of security problems and, even if willing, the costs were many times greater than during the day. Some women believed that, as they had delivered normally previously, they would do so again and could avoid the expense of transport by trying to deliver at home. Some women in labour, which in retrospect was prolonged and obstructed, reached a ‘health centre’ in the evening, but waited until the next day before they found transport to Attat Hospital. Several women gave a history of ‘the pains having stopped’ (a classic feature of uterine rupture) in the evening, and did not start the journey to Attat until after daylight the next morning!

The need for coordination between community and appropriate secondary care facilities in operating an effective MWA has been recognised,11 and the success of Attat’s MWA is founded on its community links, its acceptability to women and their husbands, and the provision of effective EmOC 24 hours per day, 7 days per week. Because of local demand, the size of MWA has been increased twice. The original facility was built in the local style with a thatched roof, which caught fire in 1999, and was replaced with a corrugated roof. Local opinion, especially that of women, is sought before any new building or project is commenced. One MWA project in another African country failed because the community was not involved.12 For the building of a new MWA, the local women wanted hot showers even though no homes in this part of Ethiopia have running water. This request was instigated by women who had already delivered in Attat, where a hot water shower had been installed in the delivery unit several years earlier. The attractive reduction in fees for any subsequent delivery intervention in those accepting MWA care has proved to be acceptable to women and their husbands.

The causes of maternal deaths for 1998–2008, with haemorrhage, mainly postpartum, and hypertensive disease being common, are in accordance with the causes of maternal mortality worldwide.13 The maintenance of any form of blood bank in this part of the world is almost impossible, and so if blood is required it generally needs to be obtained from relatives. This is more likely to be possible in MWA, where it will already have been discussed. In addition, a closer watch is kept on haemoglobin levels and any necessary iron therapy is given. Although treatment with magnesium sulphate is available at the hospital for women who develop eclampsia, effective treatment of some subsequent sequelae, such as anuria, is unsatisfactory. In MWA, because of monitoring, eclampsia is much less likely to develop and, if it occurs, it is diagnosed and treated promptly. Obstructed labour and uterine rupture also figure highly as causes of death, and cerebral malaria is still significant in the area, all of which are more effectively managed through MWA.

We acknowledge that the best form of study design to answer the question of whether an MWA results in a reduction in maternal mortality and stillbirth is a randomised controlled trial. Ethically, in the locality of Attat, this could not be undertaken because it is already believed to be of benefit. Such a study could be undertaken elsewhere; however, it would need to be very large to evaluate maternal mortality, and follow-up to obtain high-quality data on deaths in women randomised to non-MWA care who deliver at home would be difficult to achieve in rural areas. Provision of effective EmOC for all who require it is accepted as a critical component of reducing maternal mortality,14 and universal and prompt access for unpredicted antenatal and intrapartum problems is essential. However, for a subset of women at high risk, living in difficult and remote locations, MWAs offer the possibility of immediate access to timely and life-saving obstetric intervention, including caesarean section where necessary. We consider that the findings of this study provide the best evidence to date that a well-organised and locally accepted MWA is likely to reduce maternal mortality and stillbirth in rural areas of the developing world.

Disclosure of interests


Contribution to authorship

PP, RS, AR, JK were at various times clinicians at Attat and EK is hospital administrator, and all contributed to data collection. HW, JK and CM completed parts of the analysis and wrote various drafts of the manuscript, with input and comments from EK, PP, RS and AR.

Details of ethics approval

This study was approved by the Governing Body of the Attat Hospital and Integrated Health Services, and was considered to comprise an audit of routinely collected data.


No specific external funding.


We acknowledge the team work and patient care of all staff at Attat Hospital. We acknowledge Robert Lancashire who assisted with the statistical tests.

Heather Winter – In Memoriam

Heather Winter, one of the listed authors on this paper on maternity waiting areas (MWAs), was involved in its development, but sadly died in November 2007, before its completion. Heather had been diagnosed with ovarian cancer 10 years earlier, but her relentless efforts on behalf of women’s health were as unaffected by this as it was possible to be, even though she had to undergo many episodes of debilitating treatment.

Heather had already obtained her membership of the Royal College of Obstetricians and Gynaecologists when, in 1990, she took a 2-year clinical research fellowship in Public Health at the University of Birmingham to work on a cohort study of the natural history of cervical neoplasia. At the end of this, she decided not to go back to obstetrics and gynaecology, where she was regarded highly, because she thought it possible to make more of a difference to improving the health of others by remaining in Public Health. She retrained, obtaining her doctorate in 1995, and her membership of the Faculty of Public Health Medicine and a Senior Lectureship in 1996. In 1997, she and her husband Eugene had a much loved son, Anthony.

Heather never looked to furthering her own career, focusing on collaborating with others when she thought the topic was important, and her outstanding ability contributed to the production of research of high quality. She continued to collaborate with the cohort study project team and her intellectual engagement enriched their thinking and outputs (Woodman et al. The Lancet 2001;357:1931–6; Collins et al. BJOG 2002;109:96–8; Woodman et al. The Lancet 2003;361:40–3). In postnatal care, she used the knowledge gained from her dual training to instigate the design and evaluation of a model of care aiming to be maximally effective in managing women’s health problems (MacArthur et al. The Lancet 2002;359:378–85; Bick et al. Postnatal Care; Evidence and Guidelines for Management. London: Churchill Livingstone, 2008).

It is impossible to record here all of Heather’s achievements, but the area most important to her was finding ways of reducing maternal and perinatal mortality in the developing world. Her work on MWAs is one example, as are the numerous smaller studies in which she supported the involvement of intercalated and elective medical students. A notable achievement began in 1996, when Dr Abdul Jokhio came from Pakistan to undertake a PhD on this subject, Heather being one of his supervisors. She devoted much time to supporting his work, with, once again, her clinical skills and knowledge in both obstetrics and gynaecology and public health being invaluable. This collaboration resulted in the design of a randomised controlled trial training traditional birth attendants in clean delivery, linking them into the health services and showing a 30% reduction in perinatal mortality (Jokhio et al. N Engl J Med 2005;352:2092–9). This␣research has become a platform for ongoing work as Heather wanted. She continued to be frustrated, however, by the prevailing view that the focus on ‘a skilled birth attendant for all’ meant that further research on traditional birth attendants should not be pursued, especially as, in rural areas, such attendants are likely to continue in practice for many years and to continue to deliver large numbers of women (Jokhio et al. The Lancet 2006;368:2122).

Heather was generous with her support of others, even at difficult times. As a result of her desire to carry on as normally as possible, most colleagues did not know that she needed ongoing treatment. She never complained about her illness to those who did know, and retained her strong faith throughout. What Heather most wanted in her work was to improve the health of others, especially mothers living in adverse circumstances. Her work stands as clear testament to her success in achieving this.

C MacArthur
Public Health, Epidemiology and Biostatistics, School of Health and Population Sciences,
College of Medical and Dental Sciences, University of Birmingham, Birmingham