Dr H. Gebreselassie, Ipas Africa Alliance, Chania Avenue, Kilimani, PO Box 1192/00200 City Square, Nairobi, Kenya.
Objective To estimate and describe the magnitude of abortion complications presenting at public hospitals in Kenya.
Design Cross-sectional descriptive study.
Population Records of all women presenting prior to 22 weeks of gestation with abortion-related complications at selected hospitals during a three-week study period. All public tertiary and provincial hospitals were included; stratified random sampling was employed to select a subset of 54 district hospitals nationwide.
Methods Data collectors identified 809 patients with abortion complications on all hospital wards and completed a standardised questionnaire for each by extracting information from the patient's hospital record.
Main outcome measures Incidence, aetiology, morbidity and mortality of abortion complications.
Results Most women (80%) presented with incomplete abortion. Approximately 34% of the women had reached the second trimester of pregnancy. Adolescents (14–19 years old) accounted for approximately 16% of the study sample. Manual vacuum aspiration was used to manage 80% of first trimester cases. The projected annual number of women with abortion complications admitted to public hospitals in Kenya is 20,893. The case fatality rate was estimated to be 0.87% (95% CI 0.71–1.02%), so an estimated 182 (95% CI 148–213) of these women die annually. The annual incidence of incomplete abortion and other abortion-related complications per 1000 women aged 15 to 49 years is projected to be 3.03.
Conclusions The high rate of abortion-related morbidity and mortality documented in the study highlights the critical need to address the issue of unsafe abortion in Kenya.
Unsafe abortion constitutes one of the most neglected health care issues in Africa, accounting for an estimated 34,000 maternal deaths annually or 13% of all maternal deaths in the region.1 Based on the World Health Organization (WHO) definition, unsafe abortion is ‘a procedure for terminating an unwanted pregnancy either by persons lacking the necessary skills or in an environment lacking the minimal medical standards, or both’.2 Unsafe abortion is exacerbated by restrictive laws on termination of pregnancy (TOP), the social and cultural stigma associated with TOP and inequitable access to TOP services, especially among rural, poor, uneducated and young women. The mortality risk from an unsafe abortion in Africa is estimated at one in 150 procedures.1 For comparison, the risk of death from a TOP performed with modern methods in the United States is less than one per 100,000 procedures.1 In addition, abortion-related morbidity poses a serious concern to women throughout their reproductive years. Complications can include incomplete evacuation of the products of conception, sepsis, haemorrhage, genital damage and abdominal injury; as well as long term health effects, such as chronic pelvic pain, tubal blockage, infertility, increased risk of ectopic pregnancy, miscarriage or preterm delivery in subsequent pregnancies.
Complications from miscarriage account for an additional amount of abortion-related morbidity and mortality. The WHO adjusted their projected abortion/live birth ratios by using a biological estimation approach suggested by Singh and Wulf to account for miscarriage.1 Based on data from six countries in Latin America, they estimate that ‘the number of pregnancies ending in miscarriage within 13–22 weeks of gestation is the equivalent of 3.41% of all live births.3 Another approach suggested by Singh and Wulf,3 using Latin American data, estimates that 15–25% of women who present at hospital with abortion complications have miscarried.
Hospital-based studies have shown that unsafe abortion is a major cause of morbidity and mortality in Kenya,4–6 where abortion is illegal except in cases were it is required to save the life of the woman.7 The provision of safe, legal TOP in the public and private health sectors has not been well documented in Kenya where safe terminations (even if illegal) are, more often than not, provided at the discretion of the provider.
The relationship between legal status and safe TOP services is evident from the decrease in abortion-related complications and deaths that follows legalisation of abortion. In the United States, for example, the death rate due to abortion complications fell by more than 50% in the five years following legalisation.8 Liberalisation of the law in Romania also coincided with dramatic declines in maternal mortality.9–11 Abortion law in Kenya, like in most African nations, has its origin in the laws of the former colonial powers, who have since reformed their own laws.
Until now there have been no national studies of abortion-related morbidity/mortality in Kenya. A retrospective review of hospital records in the Kenyatta National Hospital attributed up to 60% of total gynaecological emergency admissions to abortion complications.4 In a prospective study of abortion patients in the same facility, approximately 62% of total abortion admissions were determined to have been induced or likely to have been induced.5 These two studies documented a high mortality rate of approximately two to three deaths per 1000 abortion admissions.
The present research was conducted to address the lack of national data on the magnitude and treatment of abortion complications at public hospitals in Kenya. To date, published epidemiological research on abortion has been limited to only a few facilities, mostly at the Kenyatta National Hospital, a large, tertiary-level facility in Nairobi. Furthermore, data from these studies are outdated as they were conducted in the 1980s.
Using methodology from similar research in South Africa,12–15 the study was conducted nationwide in Kenya in the two public tertiary-level hospitals, the seven provincial hospitals and a subset of the district hospitals (Fig. 1). Investigators selected a sample of 54 of the 89 district hospitals in Kenya by constructing a sampling frame of all district hospitals stratified by district. If the stratum consisted of only one facility, that facility was included in the study. If there was more than one district hospital in the stratum, one was randomly selected. Special types of district hospitals (e.g. mental and prison hospitals) that do not provide emergency obstetric care were excluded.
Study data came from case records of all women who presented at the participating hospitals during a three-week period between August and December 2002 with a diagnosis of incomplete abortion or other abortion-related complications resulting from an unsafe termination or miscarriage prior to 22 weeks of gestation. Study investigators did not attempt to differentiate TOP from miscarriage, because it is extremely difficult, if not impossible to confirm clinically and presents both ethical and methodological challenges in a patient interview. Women with threatened abortion or ectopic pregnancy were excluded from the study.
In the South Africa study, researchers developed a clinical typology, which was adopted by our team, based on the severity of abortion-related complications (Table 1).13 We use this schema to categorise the severity of abortion complications as well as to assess appropriate treatment of complications with antibiotics, replacement intravenous fluids and blood products.
Data collection was staggered over a four-month period in order to contain costs. To improve case identification and provide a check on the quality of record keeping, site-based data collectors evaluated all women of reproductive age (from all hospital wards) with an admission diagnosis compatible with complications of TOP or miscarriage to determine their study eligibility. Data collectors at each of the facilities completed a standardised questionnaire for all eligible cases by extracting information from patient records. Blood pressure cuffs and digital thermometers were provided to all sites to facilitate classification of the severity of complications. In addition, the research co-ordinator visited all sites to monitor and verify the completeness of the data collection forms by comparing study data to the ward admission and theatre records.
We imputed normal or negative values (i.e., low-severity category) for the 18 women with missing data for one or more of the variables needed to classify the severity of abortion complications; most missing values were for temperature. The national annual number of abortion-related complications was projected by summing the number of abortions projected to be treated at each level of the public health system as follows:
where Ai= the mean number of cases per week treated at ith-level hospitals; Bi= the total number of ith-level public hospitals in Kenya; Ci= the total number of cases treated at ith-level public hospitals in one year; C= the total number of cases treated at all public hospitals in one year.
Questionnaire data were entered and checked using EpiInfo 2000. Data were analysed using SAS version 8.01. Generalised estimating equation models were used to calculate all odds ratios in order to account for the correlation within each group of cases that attended the same facility. We adjusted for sampling and clustering when calculating the national case fatality rate. Population estimates of 6,895,132 women aged 15–49 years and 1,088,102 total births16 were used to project the national annual incidence rates.
The institutional review boards of Kenyatta National Hospital, Moi Teaching and Referral Hospital and the Department of Health Standards and Regulatory Services of the Ministry of Health provided ethical approval for the study. No patient names or addresses were recorded on the questionnaires.
Most (80%) of the 809 study cases had retained products of conception upon presentation for care. The remaining women were diagnosed with complete abortion (10%), inevitable abortion (5%), missed abortion (1%) and ‘other’ diagnoses (4%). ‘Other’ diagnoses included septic abortion (n= 33), septic abortion and perforated uterus (n= 1) and habitual abortion (n= 1).
Table 2 shows the demographic and reproductive history profile of the study participants by facility type. The mean clinical estimation of gestational age was 12.2 weeks (SD 3.6) on bimanual examination. Participants presenting with second trimester complications did not vary from those presenting with first trimester complications either by age, gravidity, rural or urban status or facility type.
Table 2. Demographic and reproductive profile of study participants by facility type. Values are presented as n (%).
Adolescents (<20 years old)
Second trimester abortion
Current use of contraception
Four or more
Of the women reporting contraceptive use at the time of conception (14%) and prior to conception (34%), the most commonly used methods were oral contraceptives (48% of current and 34% of past users) and injectable contraceptives (34% and 43%, respectively). Adult women (age ≥ 20) were more likely to report contraceptive use at the time of conception (OR 3.8, 95% CI 1.2–11.7) or a history of use prior to conception (OR 6.7, 95% CI 3.1–14.4) than were adolescent women. Self-reported current and past contraceptive use did not vary by urban versus rural status.
Table 3 shows the range of clinical symptoms at admission, stratified by facility type. Mechanical injury most commonly included vaginal or cervical lacerations (n= 35), uterine perforation (n= 3) and tenaculum marks (n= 2). Women with second trimester abortion complications were more likely to have signs of mechanical injury than women with first trimester complications (OR 3.6, 95% CI 1.9–6.8). Also, the odds of having evidence of mechanical injury among adolescents were twice that of adult women (OR 2.0, 95% CI 1.0–4.1). Women with second trimester abortion complications were more likely to have offensive products of conception than those with first trimester complications (OR 1.6, 95% CI 1.0–2.5). The odds of having offensive products of conception did not vary by age.
Table 3. Clinical symptoms at admission by facility type. Values are presented as n (%).
Participants could have >1 sign of infection or complication.
Systolic blood pressure
Evidence of foreign body in vagina, cervix or uterus
Signs of mechanical injury to vagina, cervix, uterus or intra-abdominal
Table 4 shows several treatments administered by clinical severity category. Almost all women were treated with antibiotics, which were generally given at admission (83%). Usually, antibiotics were given orally (72%) with the remainder administered either parenterally (14%) or a combination of the two routes (14%). Forty-four percent of the patients had medical indications for treatment with antibiotics (e.g. fever, clinical signs of infection, system or organ failure and/or suspicious findings on evacuation). Conversely, over half (56%) of the patients were given antibiotics where no medical need was indicated. The apparent overuse of antibiotics did not vary by facility type or trimester of pregnancy.
Table 4. Treatments administered by severity of abortion complication in percent (%).
Low (n= 456)
Moderate (n= 129)
High (n= 224)
Total (n= 809)
Blood products were usually administered as whole blood (95%) rather than packed cells (2%) or platelets (3%). Most women requiring blood products received one (75%) or two units (21%). Approximately 7% of women with first trimester abortion complications and 11% of those with second trimester complications received blood. Three percent of women in the low-severity category received blood products while 14% of women in the moderate- or high-severity category were given blood products. Almost all women with pallor (94%) were administered blood products.
Intravenous fluids were administered to one-third of the women, of whom 63% had no apparent indication (i.e. uterine perforation, gangrenous uterus or gut, gut injury, generalised peritonitis, sepsis, septicaemia, septicaemic shock, system or organ failure, weak pulse or low blood pressure) for the treatment. Conversely, 22% of the women with one or more of these indications did not receive intravenous fluids. Among the women with indications for intravenous fluids, 19% in tertiary facilities, 29% in provincial facilities and 22% in district facilities did not receive the treatment.
Manual vacuum aspiration (MVA), a portable, non-electric device for uterine evacuation of first trimester incomplete abortion and for induced abortion, was used to manage 80% of women admitted with incomplete abortions in the first trimester of pregnancy. Thirty percent of these women were provided analgesia for pain. Sharp curettage was used in 14% of the women and manual removal of products of conception in 6%. MVA with analgesia was more likely to be used in tertiary (OR 7.9, 95% CI 3.3–19.0) and district facilities (OR 2.7, 95% CI 0.9–8.8) than in provincial hospitals. Also, specialists and mid-level providers were more likely to use the procedure than were general practitioners (OR 4.9, 95% CI 1.7–14.5 and OR 3.1, 95% CI 0.9–9.8, respectively).
Three percent of women required a surgical procedure other than uterine evacuation. Laparotomy was performed on 14 women and hysterectomy was performed on 2 women. In addition, five women required a repeat MVA evacuation.
Most patients (94%) were discharged according to hospital protocol. Seven women with abortion-related complications died, resulting in a case fatality rate of 0.87% (95% CI 0.71–1.02%), or an estimated 182 (95% CI 148–213) abortion-related deaths per year. Two of the women who died were adolescents, three were adults, and the age of the remaining two was missing. Six deaths were the result of second trimester complications while one was from complications of first trimester abortion. Half of the women who died were from a rural location (n= 3; one case had a missing value for residence status). Three of the women who died presented at a tertiary-care facility, one attended a provincial facility and three attended a district facility.
Table 5 shows the estimated total annual number (and percentage) of abortion complications by severity category treated at different levels of the public health system. The odds of being in the moderate- or high-severity category did not vary by age, gravidity or type of facility. Second trimester abortion cases had higher odds of being in the moderate or high category than first trimester abortion cases (OR 1.7, 95% CI 1.1–2.7), which remained unchanged when adjusted for residence status.
Table 5. Projected total number (and percentage) by severity category of abortion complications treated annually in various levels of the public health care system.
After adjustment for sampling, the projected annual number of women with abortion-related complications admitted to public hospitals is 20,893. The annual incidence of hospital-based abortion-related complications per 1000 women aged 15 to 49 years is estimated at 3.03 and the projected abortion ratio per 1000 live births is 19.2.
Despite differences in study design and context between the Kenya and South Africa studies, several key findings provide interesting comparisons. The annual rate of abortion complications was twice as high in South Africa (39 vs 19 per 1000 live births)13 but the case fatality rate in Kenya (0.87%) was considerably higher than in South Africa (0.37%)13—despite a recent report citing an apparent decline in abortion-related mortality in Kenya.17 The abortion case fatality rate for the African continent is estimated at 0.7%.1 Also, although the percentage of women admitted with abortion complications classified as high severity was greater in Kenya than South Africa (27.7%vs 16.5%),15 the percentage of cases in the second trimester of pregnancy was slightly lower in Kenya than in South Africa (34%vs 39.5%).15
Tragically, over a third of the women admitted with abortion complications were in the second trimester of pregnancy. Furthermore, because the study was restricted to abortion complication cases less than 22 weeks of gestation, this estimate does not account for cases of later gestation, which could have been from either induced abortion or premature labour resulting in pregnancy loss. Abortion-related mortality increases as pregnancies progress, as was demonstrated in this study, where six of the seven reported deaths were from complications during the second trimester. These data suggest that women in Kenya face significant barriers to obtaining safe TOP and emergency obstetric services. Possible obstacles include delays in women's recognition of their pregnancy and symptoms of emergency complications, legal restrictions on TOP, the clandestine nature and high cost of available safe TOP services, long geographic distances to care and other factors.
Only 14% of the women in this study reported the use of a contraceptive method at the time of conception and only 34% reported use of a method prior to conception. The ever-contraceptive-use prevalence reported in the 2003 Kenya Demographic and Health Survey is 55%.18 Thus, the study findings demonstrate compelling need for post-abortion contraceptive services in Kenya. Ward-based post-abortion contraceptive services have been shown to dramatically reduce the percentage of repeat unplanned pregnancies and abortions and are the best way to reduce subsequent abortion-related mortality in a post-abortion care clinical setting.1–21
MVA was introduced in Kenya in 1987 to treat first trimester, incomplete abortion and has been shown to be safer than sharp curettage and less resource-intensive when utilised in an ambulatory setting.22,23 The widespread use of MVA for treating first trimester incomplete abortion provides evidence for the success of recent interventions to promote this technique in Kenya. However, although MVA was used to treat 80% of first trimester incomplete abortion cases, analgesia was used in only 30% of these cases. According to the WHO clinical management guidelines, the first step in MVA should be to assess the woman's need for pain control and administer it as necessary.24 Some health care providers do not give appropriate attention to pain management either due to negative attitudes toward women who they suspect of having had induced abortion or because pain medications are not readily available, whereas other providers might use general anaesthesia to have a more passive patient.25 Previous research in Kenya has shown that although only 3% of women undergoing MVA and 44% of those who had sharp curettage received pain medication, almost all of the women reported experiencing pain during their procedure, with 60% describing the pain as ‘extreme’.20
Almost all women (99%) were treated with antibiotics. Unnecessary use of antibiotics has important consequences both in terms of financial expenditures as well as the potential to contribute to antibiotic resistance. While the data are insufficient to identify the misuse of antibiotics in individual cases, excessive use seems likely, as 55% of the women given antibiotics had no apparent medical indication for the treatment. Similarly, there appeared to be common misuse of intravenous fluids, which were administered to a considerable number of women who had no medical indication; and, at least 22% of women who had a medical indication for intravenous fluids use received none. This finding has implications for financial as well as clinical management.
Despite the limitations of a hospital-based study, this design was preferable to a population-based study as substantial underreporting of abortion is likely to occur in research that relies on self-reported data.26 Also, the study design avoided the contentious ethical issue of asking women about the origin of their abortion while they seek medical care in a setting where induced abortion is heavily restricted.14,27 Investigators attempted to improve case identification by monitoring the case diagnoses from hospital records in the different departments to find unidentified abortion-related cases. An additional strength of the research was the high hospital participation rate (91%), which further reduces the potential for bias.
One study limitation is that data collection occurred over a three-week period. Unknown seasonal variations in rates could limit the validity of using the study results to calculate annual projections. In South Africa, investigators conducted a retrospective validation study in the two largest hospitals and found little or no seasonal variation in treatment13; however, seasonal variation in Kenya could occur.
It is difficult, if not impossible, to determine the true number and aetiology of abortion complications in a country with strong social and cultural taboos and legal restrictions on TOP. We suspect that many of the women in this study were treated for abortion complications that resulted from unsafe TOP; however, it is very difficult to link the most common clinical symptoms, such as haemorrhage and infection to induced abortion. Indeed, many women who experience unsafe TOP do not have complications and, therefore, do not present for post-abortion care; the WHO estimates that 50% to 90% of women who undergo unsafe abortion do not have medical complications.1 In a methodological critique of the South Africa research, the research team noted that 95% of the women in their study who admitted having an illegal/unsafe abortion were classified in the low-severity category.14
The rate of abortion complications reported is almost certainly an under-estimation of the true rate in Kenya as we did not follow women post-hospitalisation; we excluded women over 21 weeks of gestation; and we did not capture the many women who present in health centres, mission hospitals and private clinics. Prior studies conducted in Kenya reported lower abortion complication estimates, but these earlier studies were conducted in only one national hospital with limited generalisability.4,5
Although both private institutions and public-sector health centres in Kenya are thought to treat a substantial proportion of abortion cases,28 the study was restricted to women attending public hospitals. And, some women requiring medical treatment might not attend a hospital due to poor access to care or death before reaching the facility. Thus, the projected abortion complication rates probably substantially under-estimate rates for the nation as a whole. Furthermore, women who attend private facilities or public health centres for abortion complications or who never reach any facility could differ from public hospital attendees, which would limit generalising the study findings to the entire Kenya population.
Despite the limitations of projecting incidence of abortion complications using hospital-based data, this study fills a critical gap in knowledge regarding abortion complications presenting at public-sector hospitals in Kenya.
The high rate of abortion-related morbidity and mortality documented in this study suggests that unsafe abortion, especially during the second trimester of pregnancy, remains a major public health problem in Kenya. Programme managers should redouble their efforts to target women and their partners who have unmet contraceptive needs especially including the provision of post-abortion contraceptive counselling and methods. Continuing unmet contraceptive need and contraceptive failure will invariably be associated with high rates of unplanned and unwanted pregnancies, forcing many women to resort to unsafe TOP with a consequent unacceptably high rate of complications including infertility, long term morbidity and death. Also, until the abortion law is liberalised and women have access to safe TOP services, addressing the complications of unsafe abortion will continue to impose a great financial burden on the constrained resources of the Kenya health care system.
Funding for this study was provided by the Rockefeller Foundation. The authors would like to thank Sarah Onyango for co-ordinating the data collection process; Heather Brown for technical support with adapting the original South Africa study design; Mark Weaver for technical support with data analysis; and the Federation of Women Lawyers (Kenya) and Kenya Medical Association for their collaboration. The authors also thank all of the hospital-based data collectors; and Janie Benson, Eunice Brookman-Amissah and Barbara Crane for their comments on the draft manuscript.