Objective To systematically review articles describing complications of abortion in settings where abortions are thought to be unsafe and to determine the incidence of severe acute maternal morbidity (SAMM) attributed to abortion at the population level.
Methods We searched relevant databases using search terms related to abortion and complications. We included population-representative studies that listed complications of abortion. We extracted data on the definitions and numbers of severe complications and SAMM, and we report abortion complication rates (per 100 000 women of reproductive age) and ratios (per 100 000 live births) for SAMM, severe complications and any complications.
Results We included 15 studies representing eleven countries (six in Africa, four in Asia and one in Latin America). We found a median abortion ratio of SAMM of 237 (range 91–1892) per 100 000 live births and a median abortion ratio of severe complications of 596 (range 435–5298). There was a great degree of heterogeneity between definitions and study populations.
Conclusions The burden of SAMM attributed to abortion is much greater than what is reported for deaths caused by abortion. However, the great heterogeneity in definitions makes it difficult to draw firm conclusions. We call for future work on the burden of unsafe abortion to use strict definitions of SAMM.
Objectif: Analyser systématiquement les articles décrivant les complications de l’avortement dans des contextes où l’avortement est considéréà risque et déterminer l’incidence de la morbidité maternelle sévère aiguë (MMSA) due à l’avortement à l’échelle des populations.
Méthodes: Nous avons effectué des recherches dans des bases de données pertinentes en utilisant des termes de recherche liés à l’avortement et aux complications. Nous avons inclus des études représentatives de populations qui ont énuméré les complications liées à l’avortement. Nous avons extrait les données sur les définitions et le nombre de complications sévères et la MMSA et nous rapportons un taux de complications liées à l’avortement (pour 100.000 femmes en âge de procréer) et le ratio (pour 100.000 naissances vivantes) pour la MMSA, les complications sévères et toutes formes de complications
Résultats: Nous avons inclus 15 études représentant onze pays (six en Afrique, quatre en Asie et un en Amérique latine). Nous avons trouvé un ratio médian d’avortements à MMSA de 237 (intervalle de 91 à 1892) pour 100.000 naissances vivantes et un ratio médian d’avortement à complications sévères de 596 (intervalle de 435 à 5298). Il y avait une hétérogénéité importante entre les définitions et les populations étudiées.
Conclusion: La charge de MMSA due à l’avortement est beaucoup plus grande que ce qui est rapporté pour les décès dus à l’avortement. Cependant, la grande hétérogénéité dans les définitions rend difficile de tirer des conclusions claires. Nous lançons un appel aux futures études sur la charge de l’avortement à risque, d’utiliser des définitions strictes de la MMSA.
Objetivo: Realizar una revisión sistemática de las publicaciones en las que se describen complicaciones debidas a un aborto en lugares en los que estos se consideran inseguros, y determinar la incidencia de la morbilidad materna aguda y severa (MMAS) por abortos a nivel de la población.
Métodos: Hemos realizado una búsqueda en bases de datos relevantes, utilizando términos de búsqueda relacionados con abortos y complicaciones. Hemos incluido estudios poblacionales representativos en los que se enumeran las complicaciones debidas al aborto. Extrajimos datos sobre las definiciones y números de complicaciones severas y MMAS y reportamos la tasa de complicaciones por aborto (por 100,000 mujeres en edad reproductiva) y la tasa (por 100,000 nacidos vivos) de MMAS, complicaciones severas y cualquier complicación.
Resultados: Hemos incluido 15 estudios que representan a once países (seis en África, cuatro en Asia y uno en América Latina). Encontramos una tasa media de MMAS por aborto de 237 (rango 91 a 1892) por 100,000 nacidos vivos y un tasa media de complicaciones severas por aborto de 596 (rango 435 a 5298). Había un alto grado de heterogeneidad en las definiciones y poblaciones de estudio.
Conclusiones: La carga de MMAS debida al aborto es mucho mayor de lo que se reporta por muertes atribuibles al aborto. Sin embargo, la gran heterogeneidad de las definiciones hace que sea difícil poder sacar conclusiones sólidas. Hacemos un llamamiento para que en trabajos futuros sobre la carga por abortos inseguros se utilicen definiciones estrictas de MMAS.
The worldwide burden of unsafe abortion is high (WHO 2007). In developing countries, five million women are admitted to hospital for treatment of complications relating to unsafe abortion every year (Singh 2006). The disability burden of unsafe abortion is thought to be substantial: deaths and illness from abortion may be responsible for the loss of 5 million years of productive life worldwide (WHO 2002). However, the evidence base is weak, leading to likely underestimation of the actual burden (Singh 2006).
The WHO defines unsafe abortion as ‘a procedure for terminating an unintended pregnancy either by individuals without the necessary skills or in an environment that does not conform to minimum medical standards, or both’ (WHO 1992), but agreeing on standard criteria for the adequacy of skills or the enabling environment is not straightforward. Unsurprisingly, the shortage of reliable information on the incidence of unsafe abortion is most acute in the poorest countries, where either abortion is legally restricted or access to safe abortion services is poor. In these settings, national estimates of the annual number of unsafe abortions are generally derived from the number of admissions to health facilities for abortion-related complications. The number of abortion-related admissions is obtained from a variety of sources, including routine hospital statistics (WHO 2007), direct counts by researchers imbedded in hospitals (WHO 2007) or provider’s perceptions of the number of abortion complications occurring in their facilities (WHO 2007).
Few studies (Rees et al. 1997; Gebreselassie et al. 2005, 2010; Jewkes et al. 2005; Fetters et al. 2008) define the nature and severity of complications owing to abortions, and no studies estimate the incidence of near-miss or severe acute maternal morbidity (SAMM) caused by abortion at the population level. A near-miss is ‘a woman who nearly died but survived a complication that occurred during pregnancy, childbirth or within 42 days of termination of pregnancy’ (Say et al. 2009). A near-miss, by definition, survives the complication, and the term SAMM is used when near-miss and maternal deaths are combined (Say et al. 2009). These criteria aim at identifying women at the very severe end of the morbidity spectrum, in a way that the woman is unlikely to survive if she does not make it to hospital. This concept is attractive, because hospital cases can be assumed to represent all cases of near-miss in the population, even if not all women give birth in hospital (Ronsmans et al. 2009).
The aim of this study is to apply the near-miss concept to the measurement of the burden of abortion. We report the findings of a systematic review to determine the incidence of SAMM associated with abortion at the population level and discuss methodological strengths and weaknesses of the available evidence.
We searched Pubmed, Embase, Popline, Lilacs, Emro and African Index Medicus for studies published between 1 January 1991 and 1 July 2010. A search on abortion, incorporating thesaurus and free-text terms including abortion and pregnancy termination, was combined with thesaurus and free-text terms for complications of abortion including puerperal infection, septic infection, septicaemia, haemorrhage, shock, embolism, organ failure, uterine perforation, gut injury, SAMM, near-miss and death. These complications were chosen based on the complications listed under ‘pregnancy with abortive outcome’ in ICD 10 (WHO 1993) and published near-miss criteria (Mantel et al. 1998; Ronsmans et al. 2009; Say et al. 2009). We also conducted a comprehensive search of near-miss studies and included papers that reported near-miss complications due to abortion. Reference lists from all relevant articles were checked. The full search strategy is available upon request.
Inclusion and exclusion criteria
We included studies from countries where abortion is considered unsafe. As there is no internationally accepted list for this, we classified countries based on WHO estimates of the incidence of unsafe abortion by WHO regions (WHO 2004). We excluded countries within the AMRO A, EURO A and WPRO A regions, where incidence of unsafe abortion and deaths caused by unsafe abortion are defined as ‘negligible’ (WHO, 2004). All studies from other regions were included, except for trials comparing effectiveness of abortion methods, which were assumed safe.
We included population-representative studies conducted since 1990. A study was considered representative if women were sampled randomly from a defined population or if health facilities were sampled such that population-representative estimates of the frequency of abortion complications could be inferred. Single-hospital studies were excluded unless they were the sole hospital for a well-defined catchment population. When the same study was reported in two papers, only the paper reporting the most comprehensive results was included, using supplementary information from the other paper to augment the data. Studies with <30 women with complications of abortion were excluded to ensure reasonable precision for incidence estimates.
Studies grouping induced and spontaneous abortions were included because induced abortions in countries where abortion is illegal may be recorded as spontaneous abortion and thus exclusion of recorded spontaneous abortions could underestimate the incidence of unsafe abortion (Singh 2006). Studies reporting complications associated only with spontaneous or therapeutic abortion were excluded.
Studies were included if they provided numbers of near-miss associated with abortion. We did not use the WHO criteria of near-miss (Say et al. 2009) because most studies were published before these criteria were available and few studies provided the required clinical or laboratory detail to do so. Rather, we based our near-miss definitions on clinical diagnoses representing organ failure or dysfunction as suggested by Mantel et al. (1998) and Adisasmita et al. (2008), including vascular dysfunction (low blood pressure or hypovolemic shock), immunological dysfunction (sepsis, septicaemia, septic shock and septic abortion) or other organ dysfunction. We included peritonitis, tetanus, gangrenous uterus and severe trauma with uterine perforation, gut injury or bowel injury as near-miss criterion because we assumed that such conditions were life-threatening in the settings included in the review. Studies only reporting broad and undefined clinical entities such as ‘infection’ or ‘haemorrhage’ and studies that relied on self-reported complications were excluded.
Information was extracted by one reviewer (AJA) on study period, region, study design, sampling, hospital type, study population, complication definition and sample size. We extracted information provided by authors on assumptions made to derive population-based estimates and numbers of live births or women of reproductive age in the catchment area of the hospitals.
We report three categories of complications from abortion. First, we report the number of women with SAMM (near-miss and deaths) caused by abortion. Second, because five of the studies used a standard definition of ‘complications of high severity’ (Rees et al. 1997; Gebreselassie et al. 2005, 2010; Jewkes et al. 2005; Fetters et al. 2008), we created a ‘severe complications’ category based on criteria proposed in these studies. These criteria were broader than the near-miss criteria proposed earlier, as they additionally included fever, tachycardia or sign of interference with the pregnancy. When possible, similar conditions were extracted from the ten studies not reporting an explicit ‘severe complications’ category. Third, we report the number of abortion complications in women admitted to health facilities, regardless of severity.
We derived the annual number of abortion complications from health facilities and derived projected population-level numbers based on the sampling fraction of the health facilities, using three scenarios. First, where all or nearly all health facilities offering abortion care were sampled, no adjustments were made. Second, where random or stratified samples of health facilities were taken and authors extrapolated numbers in the paper to the population level, we accepted these numbers. Third, when samples of health facilities were used and no extrapolated data were provided, we multiplied the observed number of cases by the inverse of the sampling fraction of the health facilities.
For each study, we identified the number of women with near-miss or death caused by abortion, the number of women with severe abortion complications and the number of women with any abortion complications. We calculated six indicators of the incidence of abortion complications, expanding on two commonly used indicators of the burden of unsafe abortion: the abortion ratio and the abortion rate (WHO 2004). We calculated an abortion complication ratio (per 100 000 live births per year) and rate (per 100 000 women of reproductive age (15–49) per year) for SAMM, severe complications and any complications. We reported SAMM abortion ratios separating studies with and without stringent criteria for near-miss and, where possible, separating induced and spontaneous abortion. We did not provide estimates of abortion complications as a per cent of total abortions in the population because estimates for number of abortions in the population calculated by WHO were derived from the number of women presenting in hospitals for complications of abortion (WHO 2004, 2007; Singh 2006). Data on live births and women of reproductive age in the catchment population were obtained from the articles, a demographic database (http://esa.un.org/wpp/unpp/panel_indicators.htm) and other references (Jewkes et al. 2002; Etemadifar et al. 2006; Singh 2006; Hashemipour et al. 2007; Ronsmans et al. 2009).
We used box plots to describe the range of abortion ratios and rates for SAMM, severe complications and any complication across studies. Boxes represent the interquartile range (IQR), and any data that lie more than 1.5-fold IQRs below the first quartile or 1.5-fold IQRs above the third quartile are considered outliers (Cox 2009).
Our search identified 14 745 references (Figure 1). After screening all abstracts, we reviewed the full text of 1069 potentially eligible papers; 43 articles described abortion complications, and 12 were regionally representative. We added three population-representative near-miss studies located through the near-miss review. These studies represented 11 countries of which six were in Africa, four in Asia and one in Latin America.
All public hospitals (71) and random sample of health centres (115) stratified by hospital capacity and ability to provide abortion services; sampling proportional to size
Private hospitals, hospitals not offering maternity services
Complications of induced or spontaneous abortion <22 weeks gestation
We extracted near-miss data from all 15 studies, but the nature and range of conditions varied (Table 2). The five near-miss studies provided stringent criteria for all major organ dysfunctions including cardiac, vascular, immunological, respiratory, renal, liver, coagulation or cerebral dysfunction (Mantel et al. 1998; Cochet et al. 2003; Adisasmita et al. 2008; Ronsmans et al. 2009; Roost et al. 2009). Criteria for vascular dysfunction generally included hypovolemic shock (with a strict cut-off for blood pressure) as well as the need for blood transfusion, while immunological dysfunction generally relied on the presence of signs of septicaemic shock.
Table 2. Definitions of near-miss and severe complications resulting from abortion in studies retained in the review
Data on ‘severe abortion complications’ were available for nine studies (Table 2). The diagnoses listed under ‘organ failure’ differed for near-miss and severe complications in three of the five studies reporting standard categories for ‘severe complications’ because organ failure was not reported separately in one study (Gebreselassie et al. 2005), another study reported generalised peritonitis in the ‘severe complications’ category but only reported actual numbers for localised peritonitis (Fetters et al. 2008), and one study had too broad a definition of infection to be included as near-miss (Gebreselassie et al. 2010).
Table 3. Observed and annual projected number of women with near-miss, severe complications, any complication and deaths from unsafe abortion in studies retained in the review
Number of near-miss
Number of severe complications
Number of all abortion admissions
Number of deaths
Projected number per year in catchment (% of all abortion admissions)
Projected number per year in catchment (% of all abortion admissions)
Projected number per year in catchment
Projected number per year in catchment (% of all abortion admissions)
Annual projections based on (*) projected numbers reported in paper (i.e. hospitals were sampled, but authors calculated and reported annual projections); (†) observed number (i.e. paper reported all hospitals in one area for 1 year); (§) annualised observed number (i.e. paper reported all hospitals in one area for less or more than 1 year); (‡) annualised observed number multiplied by inverse of sampling fraction of hospitals.
The median proportion of abortion complications that were near-miss was 6.3% (range 1.9–33.2%); 27.9% of abortion complications were considered severe (range 9.7–75.0%), and 0.3% resulted in death (range 0.0–2.8%; Table 3).
Table 4 and Figures 2 and 3 show projected annual numbers of live births and women of reproductive age, and the six indicators of the incidence of abortion complications. The annual rate of abortion complications varied substantially depending on the indicator used: the median abortion rate (per 100 000 women of reproductive age) was 18 for SAMM (range 14–121), 90 for severe complications (range 18–367) and 366 (range 50–867) for any complication. Similarly, the median annual abortion complication ratio (per 100 000 live births) was 237 for SAMM (range 91–1892), 596 for severe complications (range 435–5298) and 4195 (range 1667–10 335) for any complication. Removing deaths from the SAMM estimates did not substantially alter the findings (data not shown).
Table 4. Abortion complication incidence rates and ratios for SAMM, severe complications and total complications in studies retained in the review
Annual number of live births and women of reproductive age (from closest year)
Abortion complication incidence rate per 100 000 women of reproductive age
Abortion complication incidence ratio per 100 000 live births
The median abortion ratio for SAMM was 136 per 100 000 births (range 106–271) in the five studies using stringent near-miss criteria compared to 378 per 100 000 births (range 91–1892) in the other studies. Only two studies separated induced and spontaneous abortions. In Thailand, 29% of induced and 5% of spontaneous abortion complications were associated with near-miss (Warakamin et al. 2004). In Iran, 52% of induced and 30% of spontaneous abortion complications were near-miss (Majlessi et al. 2008). Admissions owing to spontaneous abortions were more common in both countries, and the resulting SAMM ratio was 1.0 and 3.8 per 1000 live births for induced abortion in Iran and Thailand, respectively, compared to 4.5 and 1.8 per 1000 for spontaneous abortion. The high incidence of SAMM in spontaneous abortions in Iran reflects the 7-fold higher incidence of reported spontaneous abortions compared to induced abortions.
This study reports the findings of a systematic review of the annual incidence of SAMM associated with abortion in poor- and middle-income countries. Results from 15 studies between 1990 and 2010 across three continents show that every year, on average, 237 women experience a SAMM from abortion for every 100 000 live births. This is much higher than the abortion-related maternal mortality ratio of 37 and 12 abortion deaths per 100 000 live births estimated for 2000 in Africa and South Asia, respectively (Ronsmans & Graham 2006).
Caution is required in the interpretation of these findings: no standard methods were used across studies; in addition, the heterogeneity in study methods prevented us from summarising the results in a single or region-specific estimate of the annual incidence of SAMM following abortion (Egger et al. 2001). Major sources of heterogeneity relate to the representativeness of health facilities and definitions of abortion and near-miss. All studies were facility based, and many assumptions had to be made to extrapolate the findings to the general population. The studies covered the public sector only, and most studies excluded health centres or clinics. While this is likely to result in an underestimation of the incidence of SAMM, ascertaining the magnitude of this bias is difficult. In Ethiopia, for example, 45% of all admissions for severe abortion complications were in public health centres and 17% only in private or NGO facilities (Gebreselassie et al. 2010). In such a context, excluding health centres or private facilities would result in a substantial underestimation of the incidence of SAMM following abortion.
Most studies included spontaneous and induced abortions, and only two studies provided breakdowns by type of abortion. These two studies suggest that SAMM may be more common among induced than among spontaneous abortions. The high proportion of near-miss among spontaneous abortions in the Iran study is surprising, however, and may suggest that induced abortions are misclassified as spontaneous, supporting the argument that abortion studies should include both induced and spontaneous abortions.
The advantage of focussing on near-miss complications is that women with near-miss are unlikely to survive in the absence of hospital care; the numbers admitted can be assumed to represent the numbers occurring in the catchment population of the hospital (Ronsmans et al. 2009). The latter is true on two conditions, however. First, the use of near-miss requires adherence to strict criteria of severity. This was not the case in the abortion studies reviewed here, most of which relied on recorded clinical diagnoses without providing strict criteria. Many studies focused on the immunological dysfunction associated with severe infection, using terms such as sepsis, septicaemia and septic shock interchangeably. The terms sepsis, severe sepsis and septic shock refer to conditions of progressively greater severity (as measured by higher case fatality rates), while septicaemia spans various degrees of severity (Lever & Mackenzie 2007). Few studies reported the numbers of abortions associated with vascular dysfunction such as haemorrhagic shock, even though severe bleeding is a likely complication of abortion. The higher incidence of SAMM after abortion in studies not using strict criteria suggests that such estimates may be inflated.
Second, all cases of near-miss admitted to hospital need to be counted. As very ill women seeking care from hospitals after an induced abortion may not report the abortion on admission, they may end up in intensive care or in a general non-obstetric ward, and such records may be missed. However, the reported median incidence of SAMM after abortion is consistent with that reported in obstetric near-miss studies. A systematic review found an incidence of 380–1090 near-miss per 100 000 births in studies using organ system–based criteria (Say et al. 2004). Given that abortions only represent a fraction of all obstetric near-miss, we expect incidence of near-miss associated with abortion to be lower than the incidence of near-miss for all obstetric causes.
The use of broader categories of morbidity is more problematic, particularly for estimating a population-based incidence. Such categories of morbidity are useful in describing the profile of complications among women admitted to hospital after an induced abortion. Women seeking care from a hospital after an abortion may have no clinical signs of infection or suspicious findings on evacuation (Gebreselassie et al. 2005), and distinguishing these women from those with more severe complications on admission is necessary to gain insights into the safety of induced abortions. In Ethiopia, for example, the smaller proportion of severe morbidity among women admitted after abortion in urban compared to rural areas led the authors to suggest that access to health facilities or safe abortion services is greater in urban areas (Gebreselassie et al. 2010). However, broad and undefined clinical diagnoses such as ‘infection’ or ‘haemorrhage’ cannot be used to arrive at the incidence of population-level abortion complications, because women may experience such complications without accessing a hospital. Infections presenting only with a temperature of 38 °C or more, for example, can be treated with antibiotics in the community and will be missed in studies relying on hospital admissions. The greater range in the incidence of severe abortion complications compared to abortion complications resulting in SAMM may reflect the variability in case definition, but also variability in the extent to which the so-called severe abortion complications are successfully treated in the community.
The search strategy we used was extremely inclusive: we did not restrict on regions of the world so as not to miss studies that did not include the study setting in the title or key words. All major relevant databases as well as several smaller regional databases were searched. While every effort was made to include all studies discussing abortion within the defined parameters, it is possible that some studies were missed. Because the term ‘abortion’ was included in the search strategy, our search was limited to studies that had abortion in the title, abstract or keywords. To account for this gap, we searched the literature for near-miss studies, and it is unlikely that major studies reporting on SAMM following abortion were missed.
The WHO definition of unsafe abortion (WHO 1992) refers to procedures performed by unskilled individuals or conducted in inadequate environments. However, it is possible for a skilled provider in a safe hospital environment to perform an abortion that results in serious complications for the woman, or for a provider in a so-called unsafe environment to perform an abortion not resulting in any complications for the woman. Studying near-miss or SAMM associated with any abortion procedure is a way of catching all women who have life-threatening complications from abortion and removes the need for a definition of what set of skills and what environment is needed for that abortion to be considered safe.
Measuring the burden of SAMM associated with abortion is a promising and powerful way of ascertaining the burden of unsafe abortion. Major efforts have gone into the measurement of the burden of hospital admissions resulting from induced abortion (Singh & Wulf 1994; Singh et al. 1997, 2005, 2006; Henshaw et al. 1998; Huntington et al. 1998; Juarez et al. 2005; Prada et al. 2005; Singh 2006), and such studies should now aim at ascertaining the severity of complications from all abortions using strict criteria drawn from the obstetric literature. Standard definitions of severe haemorrhage as >1000 ml (Carroli et al. 2008) are available, and clear criteria for the definition of severe anaemia (WHO 1994), severe sepsis and septic shock exist (Lever & Mackenzie 2007). The use of stringent criteria for the ascertainment of severe abortion complications, such as the near-miss criteria developed by WHO (Say et al. 2009), will greatly enhance our understanding of the burden of unsafe abortions in populations.
This research was carried out through funding from the Child Health Epidemiology Reference Group (CHERG). The authors would like to thank Susheela Singh and Elisabeth Aahman for their help and advice. We would also like to thank Karen Wagner, Bert Peterson, Bob Black, Lale Say, Doris Chou and Clara Calvert for comments.