Considerable under-reporting has been found from a number of Provinces. However, the findings of this report are consistent with the Demographic and Health Survey for 1998 which demonstrates a Maternal Mortality Ratio (MMR) of 150/100 000 live births. The vast majority (92.6%) of women who died were African. Women of 30 years and older or of a parity of 5 or more were at greater risk of death during pregnancy and the puerperium. Most deaths (61.5%) occurred after delivery of the fetus, with 8.7% occurring during labour, 19.5% antenatally after 24 weeks, and 8.5% occurring before 24 weeks. Obstetric haemorrhage was the most common primary obstetric cause of death at level 1 hospitals, and complications of hypertension were the most common cause in level 3 hospitals. Relatively few women (40%) were referred during the process that led to the maternal death and there are very few (2%) cross-Provincial border referrals. Better collection of maternal deaths is urgently required. An education programme on the risks of age and parity in maternal death aimed at older women, women of higher parity and health workers working in family planning is required. Clear referral criteria and routes must be established and audited.
The World Health Organisation and UNICEF estimate that no less than 585 000 maternal deaths occur each year1. Most of these deaths are attributable to lack of access to health care or the provision of poor quality care. These alarming figures have given rise over the years to numerous programmes including the Safe Motherhood Initiative, and the setting an international target of reducing maternal mortality ratio in developing countries to approximately 124/100 000 live births by 2015.
In South Africa, available data reflect stark differences in the maternal mortality by population group, which is strongly suggestive of socio-economic status differences and differing levels of access to care. Measures of maternal mortality are critical as they reflect not only a woman's access to and use of essential health care services during pregnancy and child birth, but also broader underlying factors including women's general health and nutritional status, access to reproductive care services including family planning, financial resources and education. The most recent national statistics on maternal mortality for South Africa that were obtained from the recently completed Demographic and Health Survey (SADHS)2, suggest that the maternal mortality ratio remains high at 150 deaths per 100 000 live births.
Confidential inquiries are an important tool for identifying causes of maternal deaths and developing measures to prevent further maternal deaths. Whereas confidential inquiries are an excellent tool for improving quality of care in this respect their methodology is less ideal as a public health tool for estimating maternal mortality rates or ratios (MMR). This is primarily because of under-reporting of deaths. This is mainly due to the reporting being health institution based. The data presented in this section on MMR should therefore be interpreted with caution.
The total number of deaths (676) reported by province through the inquiry in 1998 are shown in Table 2.1.
KwaZulu-Natal and Gauteng reported the highest number of maternal deaths (188 and 131 respectively), whilst the Northern Cape and Northern Province reported only 22 and 27 maternal deaths respectively each in the same period. Table 2.1 gives the monthly reporting of maternal deaths per Province. It is clear from the Table 2.1, that there has been considerable under-reporting in some Provinces. Provinces where there is a fair degree of confidence that the vast majority of deaths were recorded were Free State Province, Gauteng Province and Western Cape.
Maternal mortality ratio is an important health status indicator as well as a critical measure of human development that is internationally comparable. The maternal mortality ratio (MMR) is defined as the annual number of maternal deaths per 100 000 live births.
Utilising the NCCEMD data (Table 2.2) the MMR for Provinces where there is thought to be adequate reporting of maternal deaths are as follows; Free State 135, Gauteng 67.4, Western Cape 49.8 per 100 000 deliveries.
The value of NCCEMD, at this stage, lies more in providing insight into the underlying causes of maternal deaths in South Africa
Most deaths occurred in African women (92.6%). The remainder 4.4%, 0.9% and 0.7% occurred in Coloured, Whites and Indian races. In 1.4% the race was not recorded.
The age distribution of maternal deaths is shown in Table 2.3. The number of women giving birth in 1998 is unknown, as is data concerning the age distribution of the mothers. Data has been obtained from extrapolating data from Census '91 where the age distribution of women is known and calculating an estimated number of deliveries per age category using the fertility rate. This information is compared with observed data from two sites, namely Atteridgeville, in Gauteng and Regions A and B in the Free State. The proportions are fairly similar. The proportion of the age group 25-29 years is considerably lower than observed and that age of the groups 40 years and above is considerably more than observed. The calculated number of deliveries is way above observed data (Table 2.1b).
A calculation of the MMR per age category is shown in Table 2.5. This is shown only to get an indication of the risk maternal age has on maternal death. The MMR is not valid because of under-reporting of maternal deaths and the denominator is very uncertain. However, it does give a comparison of the risks of maternal death per age category when compared with another age category. The proportion of maternal deaths started rising from 30 years and older when compared to the data sets of the two populations shown. It was more than twice as high for the categories 40 and older. There were no differences in the teenage group, (women <20 years of age). The age of lowest risk for death during pregnancy was between 15 and 25 years of age (Tables 2.3-2.5). The maternal age was not known in 6% of the maternal deaths. More details are shown in Appendix 1, Tables A4 and 5.
Women duringtheir first pregnancy or who had 5 or more pregnancies were also a greater risk of maternal death (Table 2.6). The parity was not known in 11.3% of deaths. More details are shown in Appendix 1, Tables 6 and 7.
The average gestational age at delivery or time of death of the women was 32.6 weeks ((7.1 weeks). The gestational age was not recorded in 23.9% of cases. Death occurred before 24 completed weeks in 8.5% of cases, in 19.5% in the antenatal period, in 8.7% in labour and in 61.5% in the postpartum period.
Details of the primary obstetric causes of deaths are given in Appendix 1, Table A3. The major categories of the primary obstetric cause of death are described in detail in Chapters 3-11. This chapter concentrates on patterns of disease at the various levels of care and geographic areas. More details on the causes of death within each Province are given under the relevant Province (Chapter 12) and in Chapter 12, Table 12.1.
Autopsies were requested on 15.5% of all maternal deaths, in 10% it was unknown whether it had been requested and in 74.5%, no autopsy had been requested. Extremely few autopsy results were forwarded to the NCCEMD. Attention will need to be paid to improving the autopsy service, if a more accurate cause of death is to be ascertained. This is especially the case in acute collapse, deaths outside the health facilities and deaths related to anaesthesia.
In this report, level 1 care was defined as hospitals staffed by doctors but without any full-time specialist staff. Level 2 care is a hospital with full-time specialists. Level 3 care was defined as a hospital with sub-specialists and full intensive care facilities. The Provinces have all classified their hospitals in this way. In some cases of level 2 hospitals, there are specialist posts, but they have not been filled.
The majority of maternal deaths occurred in the Level 2 hospitals (35.0%), Level 3 having 29.6% of maternal deaths and Level 1 hospitals 27.3% of maternal deaths. Home deaths accounted for 2.8% (n=16), community health centres for 2.3% (n=13), private hospitals for 1.8% (n=10) and the place of death was not known in 2.7% (n=15). Home deaths were defined as any death that occurred outside of a health facility.
It is not known what the total numbers of live births were at the various levels of care. In Gauteng Province, relatively few deliveries occur at level 1 hospitals, followed by level 2 and the majority occurring in level 3 hospitals. In other Provinces the distribution is very different. For example, Mpumalanga Province does not have a level 3 hospital, and only 2 level 2 hospitals, hence most of its deliveries occur in level 1 hospitals, and patients are referred to Gauteng Province for level 3 care. This fact must be taken into account when interpreting the data on levels of care.
Considerable differences in pattern of disease occurred at the various levels as illustrated in Tables 2.7 and 2.8. Obstetric haemorrhage was the commonest cause of death in the Level 1 hospitals whereas non-pregnancy related infections and AIDS were most common in Level 2 hospitals and complications of hypertension in pregnancy in Level 3. Deaths due to AIDS occurred most commonly at level 2 hospitals accounting for 1 in 5 of all deaths, whereas it accounted for only 1 in 10 of all deaths in level 1 hospitals and 1 in 6 deaths in level 3 hospitals. The areas with the most potential for preventing maternal deaths are the direct maternal deaths. Almost half of all direct causes of maternal deaths occurring in level 3 hospitals are as a result of complications of hypertension in pregnancy, whereas just under 10% of deaths due to obstetric haemorrhage occurred in level 3 hospitals. Almost a third of direct maternal deaths in level 1 hospitals were due to complications of hypertension in pregnancy and another third due to obstetric haemorrhage. Anaesthetic deaths accounted for almost 1 in 7 direct maternal deaths at level 1 hospitals. (See Table 2.8).
To understand the true relationships between patterns of disease and levels of care, a comparison of the number of deaths per number of deliveries for each level of care should be made. Unfortunately the numbers of deliveries per level of care is not known. Therefore, Table 2.9 might be misleading. A MMR per level of care could be determined if the number of deliveries for each level of care were available. However, a pattern does emerge with the acute emergencies (e.g. obstetric haemorrhage) occurring more frequently in Level 1 hospitals and the more chronic problems (e.g. complications of hypertension) occurring at level 3 hospitals.
Deaths occurring at the various levels of care and the referral patterns within and outside the Provinces are shown in Tables 2.10 and 2.11. Mpumalanga, North West and Northern Cape do not have level 3 hospitals in their regions and refer patients to neighbouring Provinces. All the referrals from Mpumalanga went to the Pretoria Region of Gauteng.
Reporting of maternal deaths was patchy, and must be considerably improved. Very few deaths (10) were reported from Private Hospitals. This apparently low number is probably due to poor reporting. It must be stressed that a Maternal Death Notification Form must be filled in for all maternal deaths and sent to the Maternal, Child and Women's Health Unit for that Province. This applies to all maternal deaths whether in public or private institutions. The Health Professions Council regards failure to report a maternal death as professional misconduct. Much greater attention will need to be paid to collection of all maternal deaths by the Provincial Maternal, Child and Women's Health Units in subsequent years. A mechanism linking the deaths reported to the NCCEMD and deaths reported by the death certification process is essential. This will provide cross checks of both systems. Efforts must also be made to encourage the community to report deaths of women who were pregnant to the local clinics or hospitals. In this way, the number of deaths outside the health services can be determined.
However, in three Provinces there is sufficiently good information to calculate a MMR. These are Free State Province, Gauteng Province and the Western Cape. Gauteng Province (MMR 67.4/100 000 live births) and the Western Cape (MMR 49.8/100 000 live births) are the country's most developed Provinces with regard to health services. This is reflected in their relatively low MMR for a developing country. Free State Province is more representative of the country as a whole and has a MMR of 135/100 000 live births (see Chapter 12b). The 1998 South African Demographic and Health Survey (SADHS) estimates a MMR for South Africa at 150 per 100,000 live births. The NCCEMD findings are fairly consistent with that observation. Therefore, accepting that there was a significant amount of under-reporting in 6 of the 9 Provinces, a MMR of 150/100 000 live births for South Africa is most likely to be accurate. If this is correct this is a relatively low MMR for a developing country.
The increased risk of death during pregnancy and the puerperium for older women and for those a with higher parity, is not a new finding. Most studies on risk factors for maternal death find the same thing. It is surprising therefore that there is a significant lack of knowledge in women over 34 years of any risk to themselves in pregnancy3. The vast majority (92%) of women over 34 years who were pregnant and interviewed thought they were at no increased risk for their pregnancy. Even women over 34 years who were not pregnant, but using contraceptives, were also ignorant of their risks. An education message needs to be sent to the general population as well as those working in the family planning services about the risks of pregnancy in older women and women with higher parity.
Forty percent of women who subsequently died were transported either within the Province or across Provincial borders. It is surprising that no deaths were reported in Gauteng from women transferred from Northern Province. This may indicate a severe break down in transport services, or lack of referral from doctors. GaRankuwa Hospital serves as the referral hospital for Pietersburg. Pietersburg functions as the tertiary hospital for Northern Province, but in 1998 was just getting fully established. More than half the women transported in Mpumalanga were transported across Provincial boundaries, this due to the lack of a tertiary hospital in that Province. Despite this, women were transported in only 30% of cases and 29 of 59 women died at community health centres or level 1 hospitals.
Furthermore, it is disturbing to see that less than 42% of women dying of complications due to hypertension in pregnancy, died in level 3 hospitals and more than 25% died in level 1 hospitals. This might indicate poor care and lack of referral or might be an indication that the vast majority of deliveries occurred in level 1 or 2 hospitals. Only when the MMR per level of care is known will this problem of interpretation be solved. Furthermore, it might be an indication that perhaps the referral system is not functioning optimally (Table 2.6). This is supported by the relatively high proportion of maternal deaths due to complications of hypertension that occurred at the lower levels of care, (20% at levels 1 and 2). Another possible indication of problems at the level 1 hospitals is the large proportion of women dying due to obstetric haemorrhage at those institutions. Only about 20% of women dying due to AIDS died at level 1 institutions.
In an ideal health system, women with severe life threatening diseases, which are curable, should be managed in level 3 hospitals where the expertise and resources are available. Theoretically they would first attend a Community Health Centre or level 1 hospital and be referred up the line. Thus ideally close to 100% of women with complications of hypertensive disease who die would die in level 3 institution. This is far from the case in South Africa as shown above. Women requiring more palliative care should be cared for in level 1 hospitals or hospices. For this to work, an effective referral system and good communication should exist between all levels of care to allow for the management of women at the most appropriate level of care. Implicit in this is the ability to refer women up and down the line to avoid bottlenecking at various levels of care.
There are many possible reasons for the poor referral findings shown above in Tables 2.10 and 2.11. Firstly, the woman may have arrived at the level 1 hospital extremely ill and died shortly thereafter, allowing no time for referral. Another possibility, is that the doctors at level 1 hospitals might not have recognised the severity of the disease or miss diagnosed the disease and managed her at the level 1 hospital until it was too late. If the severity condition of the patient was recognised, the doctors could have decided they could manage the women themselves, or from past experience not refer the patient due to a hostile reception from the level 2 or 3 hospital. The hostile reception could take the form of abusive or belittling language from the doctor accepting the patient, to just stating the referring hospital does not fall within their referral area or stating that the hospital is full and they must phone another hospital. Lack of acceptance of patients by the receiving doctors may be due their lack of knowledge of their geographical area's responsibilities, laziness or a truly full hospital. There might also be a lack of transport available to transfer the woman. Assessors of maternal deaths came across examples of all of these scenarios whil7e assessing cases of maternal death. There are some illustrative examples in this report. Research will have to answer the relative magnitude of these problems.
To solve these problems, clear guidelines must be established around referring of patients. The routes of referral must be clearly demarcated between Community Health Centres, level 1, level 2 and level 3 hospitals. The indications for referral must be clearly delineated and the level of care at which various patients with various complications should be managed clearly defined. The responsibilities of the referring and receiving doctor must be clearly defined. For example, if a doctor from a level 1 hospital refers, as specified, to their level 2 hospital for a valid reason, can the receiving doctor refuse to accept that patient? If the hospital is "full", is it the responsibility of the referring or receiving doctor to find an appropriate bed? Does a women with AIDS require level 3 treatment, or should they be cared for at level 1 or 2 institutions? These questions must be answered in each area and the Maternal, Child and Women's Health Units in each Province should ensure that the guidelines are available and known to each site dealing with pregnant women.
A further worrying aspect of the care at Level 1 hospitals is that 60% of all anaesthetic related deaths occurred at those institutions. (This is discussed in Chapter 7).
Much has been learnt from the first full year of the confidential enquiry into maternal deaths. More emphasis will have to placed on collecting all maternal deaths. The dangers of pregnancy in older women and women of higher parity need to be explained to the population and health workers involved in family planning. Referral criteria and routes need to be established and audited, to ensure the women is managed at the most appropriate level of care.
Organisation for Economic Co-operation and Development (OECD). Core Set of Indicators of Development Progress, 1998.
Department of Health.1998 Demographic and Health Survey: Preliminary Findings, 1999.
Davies R, Keti V, Pattinson RC. Knowledge of women, more than 34 years of age, on their risks of pregnancy. 18th Priorities in Perinatal Care Conference, Buffelspoort, North West Province, March 1999.