The impact of age on the epidemiology of incomplete abortions in South Africa after legislative change


Dr R. Jewkes, Medical Research Council, Gender and Health Group, 1 Soutpansberg Road, Private Bag X385, Pretoria, 0001, South Africa.


Objective  In 1996 termination of pregnancy was legalised in South Africa. This article examines the impact of age on the epidemiology of incomplete abortion after legislative change. It draws comparison with the findings of a similar study undertaken in 1994.

Design  Multicentre, prospective, descriptive study.

Setting  Forty-seven public hospitals in all nine provinces.

Sample  A stratified random sample of all hospitals treating gynaecological emergencies was drawn. All women of gestation under 22 weeks who presented with incomplete abortion during three weeks of data collection in 2000 were included.

Methods  A data capture sheet completed by a clinician from the case notes.

Main outcome measures  Demographic characteristics and clinical findings on admission by age of women.

Results  Overall, there was a significant increase in the proportion of cases with no signs of infection on admission (from 79.5% to 90.1%) and a significant decrease in evidence of interference on evacuation (4.5% to 0.6%) between 1994 and 2000. Substantial age differentials were seen. Women over 30 were significantly less likely than those 21–30 years or under 21 to be low severity (65.5%vs 75.2%vs 76.4%, P= 0.0087) and more likely to have offensive products (16.3%vs 6.0%vs 6.4%, P= 0.01) than the younger women.

Conclusions  Legalisation of abortion had an immediate positive impact on morbidity, especially in younger women. This is an important change as teenagers had the highest morbidity in 1994. The trend is supported by evidence from the 1999–2001 Confidential Enquiry into Maternal Deaths, which further suggested that abortion mortality dropped by more than 90% since 1994.


Liberalisation of termination of pregnancy legislation is known to be associated with dramatic declines in abortion-related maternal mortality and morbidity. In Romania, when abortion was legalised in 1990, there was a 67% reduction in maternal deaths from 445 in 1989 to 145 in 1990 due to illegal abortion.1 In Guyana, admissions for septic, incomplete abortions to the capital's largest maternity hospital dropped by 41% within six months of legalisation.2 There were similarly high expectations of the 1996 Choice in Termination of Pregnancy Act of South Africa, which is one of the most liberal examples of abortion legislation in the world. Its passage through Parliament was significantly influenced by the findings of a national study of the epidemiology of incomplete abortion which was undertaken in 1994.3,4 This provided a description of the magnitude of the problem of unsafe abortion and broadly followed methodology developed for a World Health Organisation multicountry study of unsafe abortion5 with certain modifications.6 One of its key findings was a differential burden of morbidity associated with incomplete abortion experiences by women under 21 years who had a substantially greater risk than older women of having offensive products found on uterine evacuation odds ratio 3.0.3

In the first three years after the legislation was enacted, services were established in the private and public sectors such that 40,000 legal terminations have been performed annually7 (compared with 800–1000 under the previous legislation8). The ability of provinces to establish this service so quickly in a resource constrained environment was substantially influenced by the promotion of low cost technology, that is, manual vacuum aspiration following cervical ripening with misoprostol under local anaesthesia.9 There has been a provincial clinical training programme10 and a values clarification programme aimed chiefly at securing participation of nursing staff in the service.11

In 1999, the Department of Health commissioned an evaluation of progress in implementation of the Act and its health impact. Three studies have been undertaken. A national survey of facilities has shown that services are available across the country and that in most provinces, most of the population lives within 50–100 km of some form of service for first trimester abortions. However, activity is nonetheless disproportionately concentrated in major urban centres, particularly Gauteng Province (Johannesburg and Pretoria), and in many places the case load is very small.12 A study of why women are still aborting illegally in Gauteng, where services are available, shows that many women lack specific knowledge of their rights under the Act.13 The old abortion practices are continuing in parallel with the new. Most women who abort outside the new designated facilities are still using the old methods of ‘backstreet’ (or illegal) abortion described in 1994,14 although there has been the introduction of safer modern methods with misoprostol being given illegally by nurses. In order to determine impact on morbidity and medical management, the 1994 incomplete abortion study was repeated in 2000. Papers presenting a brief overview and comparison of the epidemiology15 and comparison of the medical management16 of cases have been published. This article examines the impact of age on the epidemiology of incomplete abortion after legislative change. It presents the findings and draws comparison with the data from 1994.3,4


This hospital-based study collected descriptive data prospectively on all women presenting to selected public hospitals in a specific three-week period in South Africa with incomplete abortions under 22 weeks of gestation. All the public hospitals in the nine provinces of South Africa responsible for treating women with gynaecological problems in the year 2000 were included in the sampling frame. The sample design was stratified and random, with stratification by province and hospital category (district, regional, tertiary). Within each stratum, two hospitals were selected with the sampling probability proportionate to size (pps), using the number of beds in the hospital as a measure of size. The total sample was 47 hospitals as five of the provinces had only one or no hospital in the tertiary stratum. Women admitted with threatened abortions, or who said they had had legal abortions, were excluded from the study. Data were collected from the hospitals over three consecutive weeks between May and August 2000. These differ from the period used in the 1994 study (two weeks in September), but we have not observed seasonal variations in the number of incomplete abortion cases.3

A data capture sheet was completed by health personnel on all women who met the study criteria based on their hospital records. The data sheet included information on demographic characteristics, contraceptive use, clinical assessment on admission, treatment received and outcome. A clinician based at each hospital was responsible for data collection. The research team staff verified submissions from all hospitals. Data could not be collected on five cases due to missing files (0.7% of the total). Ethics approval was obtained from the University of Witwatersrand and the relevant committees of all participating institutions.

Three clinical severity categories were used for the purpose of data analysis and interpretation. These were the same used for the 1994 study and had been developed by a reference group of gynaecologists. Low severity was defined as a temperature below 37.3°C and no signs of infection. Medium severity was defined as a temperature of 37.3–37.9°C and/or any sign of mild infection of admission (a tender uterus, offensive discharge or localised peritonitis). High severity was defined as a temperature of 38.0°C or above, or pulse of 120 or above, or any sign of interference with the pregnancy or any sign of organ failure or peritonitis or death. These categories' strengths and limitations are discussed in detail elsewhere.6 Where appropriate, calculations are based on national population estimates for 1999 of 13,478,000 women aged 12–49 years, and 1,106,000 live births. The analysis of the data took into account the complex sample design, which was a stratified multistage sample and was not self-weighting. In comparing categorical variables with the previous study,3 the Rao–Scott F test17 was used. The main difference in methods between the two studies was the sampling of hospitals.

The main differences in methods between the 1994 and 2000 studies were the duration of data collection and the sampling of hospitals. In the former study, data were collected over two weeks and the sample included all hospitals over 500 beds in the country and a random sample of hospitals with under 500 beds. There were 61 hospitals in the 1994 sample and 56 participated in the study.


All of the hospitals sampled responded. In total, 761 data capture sheets were returned from the 47 hospitals, however, 3 of these had no cases of incomplete abortion (either spontaneous miscarriages or illegally induced) during the study period. The mean age of the sample was 27.0 years (SD 7.6) and the range was 13–48 years. We estimate that 49,653 women per year were admitted nationally to public hospitals with incomplete abortion (95% confidence interval [CI] 38,742 to 60,563), compared with 44,686 (95% CI 35,633 to 53,709) in 1994.3 The incidence of incomplete abortion per 100,000 women aged 12–49 years was 362 (282 to 441) compared with 375 (299 to 451) in 19943; 95% confidence limits for the difference in incidence rates are (−123 to 97). During the study period in 2000, there were no deaths from unsafe abortion and thus an estimate of abortion-related mortality cannot be reliably made.

Table 1 shows the demographic characteristics and clinical findings on admission by age of the women in the 2000 study. The women have been divided into three groups to enable comparison with the findings of the previous study: under 21 years, 21–30 years and over 30 years. The findings show an expected increase in parity with increasing age of the women. There were no significant differences in racial group or trimester status. In the 2000 study, a statistically significant (P= 0.005) decrease in the proportion of cases with signs of infection on admission was seen (from 20.5% to 9.1%, P= 0.005). There was no difference in signs of infection on admission or signs of organ failure in 2000 between age groups.

Table 1.  Demographic characteristics and clinical findings on admission by age of women, 2000 study.
ParameterAge < 21 years (n= 142)Age 21–30 years (n= 374)Age > 30 years (n= 244)Total (N= 760*)P**
  • *

    There was no age recorded for one woman, hence, N= 760.

  • **

    P values have been calculated only when a difference was apparent, which seemed as if it might have been statistically significant.

Race (%)
Interquartile range0–00–22–40–2 
Trimester status (%)
12 weeks and under67.168.665.067.0
Over 12 weeks32.931.435.033.0 
Signs of infection (%)
Offensive discharge5.
Tender uterus2.
Localised peritonitis0.
Generalised peritonitis00.300.14
Septicaemic shock00.400.2
Signs of organ failure (%)
Disseminated intravascular coagulation00.500.24
Respiratory distress00.500.2
Hypovolaemic shock1.
Renal failure00.300.14
Findings on evacuation (%)
Offensive products6.
Mechanical/chemical injury0.
Foreign body0000
Evidence of misoprostol1.20.200.4
Severity categories (%)

Significant changes in findings on evacuation were seen between 1994 and 2000. The proportion with evidence of interference had reduced significantly. Finding evidence of mechanical or chemical injury to the genitals declined from 3.2% to 0.6% (P= 0.002) and of foreign bodies from 1.3% to 0.1% (P= 0.03). This suggests a decline in the most unsafe forms of illegal abortion.Table 1 shows that there were highly significant differences between the age groups in findings on evacuation. Women over 30 were much more likely to have offensive products than those under 30. There was no difference between women under 21 and those in their 20s in the prevalence of having offensive products and a difference was not found for other observations at evacuation.

The severity categories provide a summary of the clinical picture and these show that the high severity group has declined from 16.5% to 9.7%, with an increase in the low severity category from 66.2% to 72.4%. Overall, these differences were not statistically significant (P= 0.11). However, again there were statistically significant differences between age groups with the proportion of cases in the low severity category lower among women over 30 years.


The results showed no statistical evidence of change in the incidence of incomplete abortion since 1994, but they suggest that there has been a differential reduction in the prevalence of unsafe abortion in younger age groups compared with women in their 30s. The 1994 study indicated that teenagers were most at risk of unsafe abortion, with 20.5% in the high severity category.3 Teenagers, when compared with women over 30 years, were most likely to have offensive products on evacuation (35.4%vs 15.0%, P < 0.0001), evidence of mechanical injury (9.7%vs 2.4%, 0.006) and a foreign body in their vagina (1.8%vs 0.8%, P= 0.59).3 By the year 2000, teenagers were more likely to be of low severity than women over 30 years. This indicates that there have been differential benefits for this vulnerable age group, although undeniably there is still scope for further reduction in morbidity.

This conclusion was substantially driven by the finding that offensive products were less commonly seen upon evacuation in younger women. Obviously, there is a need for caution when relying on a single indicator of ‘safety’. It is reassuring that the trend seen here was paralleled by an identical, if non-significant, decrease in other indicators of morbidity, including signs of infection and signs of organ failure. Age groups in 2000 could not be compared on signs of mechanical injury and presence of foreign bodies as so few, or none, were found. Probably the most compelling supporting evidence, however, is that on patterns of mortality. The Second Confidential Enquiry into Maternal Deaths in South Africa (1999–2001) found that 31.8% of pregnancies are in women over 30 years but 41.9% of abortion-related mortality is in this age group. In contrast, 12.0% of pregnancies occur in women under 20 and 7.1% of abortion-related mortality.18

When the 1994 study was repeated, we had expected to see a major impact of the new legislation on morbidity and mortality from incomplete abortion. We were disappointed not to be able to calculate an estimate of mortality that could be used for comparison. It was surprising to see that the incidence of incomplete abortion had not changed and that changes in morbidity were modest. However, there is evidence that the new legislation had a very substantial impact on mortality. In 1994, it was estimated that there were 425 (78–736) deaths in public facilities from unsafe abortion.3 In the 1998 Confidential Enquiry, 32 abortion-related maternal deaths (5.7% of total) occurred in 1998.19 The Second Report (1999–2001) found 40 abortion-related deaths per year.18 These reports are believed to have had very good ascertainment of deaths in hospitals. They may under-ascertain out-of-hospital deaths but the research-based estimate from the 1994 study did not count these either. Thus, although some caution is needed in comparing data from different sources, the main biases in the two data sources are in the same direction. Comparison of the 1994 research estimate and 1998–2001 mortality data (averaged) indicates that there has been a 91.06% reduction in deaths from unsafe abortion, with a possible range of 51.28–94.84%, depending on the position of the true figure in 1994 within the confidence intervals of the estimate. This reduction in mortality is even greater than that reported in other countries, such as Romania.1

In the light of this reduction in mortality, it is surprising that the incidence of incomplete abortion should be unchanged. The estimated annual number of incomplete abortion cases in public hospitals was only a little higher than the annual number of legal terminations. If many of the 40,000 legal terminations each year reflected women who would have previously aborted illegally and attended hospital with an incomplete abortion, a substantial reduction in incidence of incomplete abortion would have been expected. This was not seen. It seems unlikely that miscarriages have become more common or women's threshold for attending hospital after miscarriage has changed. Although ascertainment was more complete in 2000 and confidence intervals are wide, the findings suggest that a reduction in unsafe illegal abortions has been paralleled by an increase in ‘safer’ medical abortions, most likely through use of misoprostol in unregistered settings. These are illegal if nurses give the misoprostol, they are not illegal if general practitioners prescribe it.

Can the reduction in mortality and morbidity be attributed to the new designated facilities and the 40,000 legal terminations performed per year or is it the effect of widespread misoprostol use? This is an important question which cannot be answered from this study, but both practices have been contingent on legislative change. It seems likely that the widespread prescription of misoprostol for induction of abortion has been influenced by doctors knowing that their actions are not actually illegal. Similarly, the widespread discussion of misoprostol in the context of termination of pregnancy, which is an application for which the drug is not registered and for which it is not promoted by drug companies, has occurred within the context of its use in designated facilities but has also provided information to people who may use it outside these settings.


This study has demonstrated that the 1996 liberalisation of abortion legislation has had a positive impact on morbidity from incomplete abortion, especially among the young. There also seems to have been a dramatic reduction in abortion-related mortality, with a differential mortality risk seen in women over 30 years. This is encouraging as the medical benefits were seen relatively quickly after legislative change and the health sector has had considerable burdens on its resources over this period, with tight control on Government spending and the burgeoning HIV epidemic. The findings suggest that further reductions in morbidity could be achieved with improved access to services, including abortion facilities in less well-served areas and information in the community on abortion rights.


The authors would like to thank the South African National Department of Health, the Kaiser Family Foundation and UK Department for International Development (DFID) for funding this study. The authors would also like to thank Lindiwe Makubalo for commissioning it on behalf of the National Department of Health. Tebogo Gumede assisted with quality control. Alta Hansen and Roman Baloyi entered the data. The authors are very grateful to the staff in all the hospitals who collected the data.

Accepted 1 July 2004