Skip to main content
Advertisement
Browse Subject Areas
?

Click through the PLOS taxonomy to find articles in your field.

For more information about PLOS Subject Areas, click here.

  • Loading metrics

Measuring Unsafe Abortion-Related Mortality: A Systematic Review of the Existing Methods

  • Caitlin Gerdts ,

    cgerdts@obgyn.ucsf.edu

    Affiliations Advancing New Standards in Reproductive Health, University of California San Francisco, San Francisco, California, United States of America, Division of Epidemiology, Berkeley School of Public Health, University of California, Berkeley, Berkeley, California, United States of America

  • Divya Vohra,

    Affiliation Division of Epidemiology, Berkeley School of Public Health, University of California, Berkeley, Berkeley, California, United States of America

  • Jennifer Ahern

    Affiliation Advancing New Standards in Reproductive Health, University of California San Francisco, San Francisco, California, United States of America

Abstract

Background

The WHO estimates that 13% of maternal mortality is due to unsafe abortion, but challenges with measurement and data quality persist. To our knowledge, no systematic assessment of the validity of studies reporting estimates of abortion-related mortality exists.

Study Design

To be included in this study, articles had to meet the following criteria: (1) published between September 1st, 2000-December 1st, 2011; (2) utilized data from a country where abortion is “considered unsafe”; (3) specified and enumerated causes of maternal death including “abortion”; (4) enumerated ≥100 maternal deaths; (5) a quantitative research study; (6) published in a peer-reviewed journal.

Results

7,438 articles were initially identified. Thirty-six studies were ultimately included. Overall, studies rated “Very Good” found the highest estimates of abortion related mortality (median 16%, range 1–27.4%). Studies rated “Very Poor” found the lowest overall proportion of abortion related deaths (median: 2%, range 1.3–9.4%).

Conclusions

Improvements in the quality of data collection would facilitate better understanding global abortion-related mortality. Until improved data exist, better reporting of study procedures and standardization of the definition of abortion and abortion-related mortality should be encouraged.

Introduction

The true global burden of unsafe abortion-related mortality remains unknown. Employing the newest figures for global maternal mortality, the WHO estimates that in 2008 approximately 13% of maternal mortality worldwide, or 47,000 deaths were due to unsafe abortion. [1] Such estimates, however, are based on statistics from developing countries that are known to have unreliable data, [2] and because of the often sparse, poor quality data in countries where abortion is the least safe are, at best, thought to underestimate the true global incidence of mortality from unsafe abortion. [1][3].

Maternal deaths occur most often in settings where national vital registration systems are weak or non-existent. [1], [2] As such, measurement of maternal mortality relies on alternative methods of data collection; [4] estimates of all-cause maternal mortality can be derived from population-level surveys [5] or indirect estimation techniques. [6] Some recent methodological advances have been made in measurement techniques for all-cause maternal mortality, [6], [7] an issue that has received increased attention since the inclusion of a commitment to reductions in maternal mortality (reducing maternal mortality by 75% from 1990 levels by the year 2015) as a part of the Millennium Development Goals (MDGs) in the year 2000. Such improvements in the measurement of abortion related deaths, however, have been slow to develop. [8].

Cause-specific maternal mortality data, where cause of death is identified as one of the WHO specified direct or indirect obstetric causes of death, [9] can be captured through vital registration (death certificates), hospital or facility records (case notes and/or death certificates), verbal autopsy (a WHO validated tool for measuring cause-specific mortality at the community level through a structured questionnaire with family members of a recently deceased person, to assign cause of death (COD) in the absence of vital registration data), [10], [11] or Reproductive Age Mortality Studies (RAMOS), which combine vital registration data and verbal autopsy data. [12] Abortion-related mortality, a direct obstetric cause, is uniquely difficult to document for a number of reasons: 1) In countries where abortion is restricted or illegal altogether, women often seek abortion related services outside of the formal medical system; 2) In such settings, due to social and cultural stigma, and fear of legal consequences, women are often reluctant to seek medical services in the event of complications or reveal to family members the underlying cause of the complications; [13][20] 3) Because of legal consequences for patients and providers alike, clinicians who provide abortion-related services may be reluctant to report abortion-related deaths. [8], [14], [21].

The validity of existing estimates of unsafe abortion-related maternal mortality has been called into question, [1], [3] and the consequences of continuing to ignore measurement deficiencies in this field have real implications for the development of policy and implementation of programs that aim to reduce maternal mortality. However, to date there has been no assessment of the validity of existing studies that report estimates of the burden of abortion-related mortality with respect to the biases they may suffer from.

Our aim is to systematically review the available peer-reviewed evidence on unsafe abortion-related mortality published since the establishment of the MDGs (September, 2000). This review establishes criteria for evaluating the quality of research papers that cite estimates of abortion-related mortality, and presents a discussion of the methodological strengths and weaknesses of the current peer-reviewed evidence about abortion-related mortality.

Materials and Methods

Search Strategy

We followed a protocol adapted for the evaluation of observational studies from criteria established by the PRISMA statement. [22] Pubmed, Popline, Embase, Medline, and JStor were searched for studies published between September 1st, 2000 and December 1st, 2011. Although some relevant studies may have been excluded, due to the overwhelming majority of English language publications generated by the search, non-English language studies were excluded. Combinations of the following keywords were used in the search process: abortion, induced abortion, unsafe abortion, maternal mortality, maternal death, pregnancy related death, cause of death, verbal autopsy. Reference lists of relevant articles were reviewed for sources that may have been missed in the database search. The full, line-by-line search strategy for each database can be found in the appendix.

Inclusion and Exclusion Criteria

To be included, articles had to meet the following criteria: (1) published after September 1st, 2000 and before December 1st, 2011; (2) conducted in or use data from a country where abortion is “considered unsafe”; (3) enumerated causes of maternal death, and specified “abortion” as one of those causes; (4) enumerated at least 100 maternal deaths from all causes; (5) a quantitative research study; (6) published in a peer-reviewed journal. The justification for each criterion is elaborated below.

We established calendar date restrictions for the search strategy (Inclusion Criterion # 1) to examine evidence published since the establishment of the Millenium Development Goals (MDGs). The MDGs set a specific target for the reduction of maternal mortality by 75% from 1990 levels by the year 2015, sparking interest in improved measurement of maternal mortality and an infusion of new funds for maternal mortality research.

Included studies were restricted to studies conducted in countries where abortion is “considered unsafe” (Inclusion Criterion #2), using criteria developed by Adler et al. [23] While no international standard exists for the classification of such countries, Adler et al excluded regions of the world where the WHO classifies the incidence of unsafe abortion and associated deaths as “negligible”. [1] We followed the same classification system, resulting in the exclusion of studies from the AMRO A (Canada, Cuba, United States), EURO A (Andorra, Austria, Belgium, Croatia, Cyprus, Czech Republic, Denmark, Finland, France, Germany, Greece, Iceland, Ireland, Israel, Italy, Luxembourg, Malta, Monaco, Netherlands, Norway, Portugal, San Marino, Slovenia, Spain, Sweden, Switzerland, United Kingdom), and WPRO A (Australia, Brunei Darussalam, Japan, New Zealand, Singapore) regions. Studies which reported data in their text from in AMRO A, EURO A, and WPRO A were excluded, studies reporting data from all other regions of the world were considered for inclusion.

We included studies that enumerated the direct obstetric causes of maternal death in a study population (Inclusion Criterion #3), specified the cause “abortion”, and calculated the number and or proportion of maternal deaths that were due to abortion. Because the definition of abortion varies widely in the literature, various definitions were accepted including: clinical definitions of induced abortion and/or unsafe abortion; all definitions of induced abortion provided by the International Classification of Disease (Code #’s 632, 635–639, and 640.03). [9] There is compelling evidence to suggest that in low-resource settings, it is often difficult to distinguish between induced abortions spontaneous abortions, therefore, in much of the literature, abortion is defined as a combination of both the ICD definition of induced abortion (see above for code #’s) and the ICD definition of spontaneous abortion (ICD Code # 634). [24], [25] Given that this is an internationally accepted definition of abortion, definitions that combined induced and spontaneous abortion into one category were also accepted. Deaths from spontaneous abortion as an independent category were not included.

The sample size criterion (Inclusion Criterion #4) was established to ensure sufficient sample size for adequate precision of estimates of abortion related deaths and was based on the sample size calculations from past reviews of abortion-related sequelea. [23], [26].

We aimed to evaluate the current, quantitative evidence on the burden of abortion-related mortality. To that end, articles that did not consist of original, quantitative research (Inclusion Criteria #5) such as review articles, commentaries, opinion pieces, and case studies, were not included.

Finally, because this review is focused on evaluating the highest quality evidence available, only articles that had first undergone a peer-review process were eligible for inclusion (Inclusion Criteria #6).

Rating Criteria

Studies were evaluated for quality on a scale modeled after a rubric developed by Charles et al [26] and derived from five primary criteria: 1) study design; 2) diagnostic procedures for assigning cause of death; 3) definition of abortion; 4) study reporting; 5) risk of bias (Table 1). Studies were ranked on the scale from Excellent to Very Poor. Table 2 provides the rubric for study evaluation.

Methodological Considerations for Development of Evaluation Rubric

I. Sources of mortality data.

Nearly two-thirds of the worlds’ countries do not routinely register vital events and thus lack complete information about births and deaths. [5], [27] Maternal mortality is often more difficult to measure than other deaths due to unique challenges in identifying and classifying maternal deaths, and especially abortion-related deaths. [5], [7] Facility-based maternal deaths are often not classified as maternal deaths if women were not registered in the labor and delivery wards (for example, the death occurred in the emergency department), and can be missed if women are not identified as pregnant, which is more likely in case of abortion-related deaths because there may not be evidence of the pregnancy, or because of reporting errors due to legal concerns about treating patients with abortion related complications. [2], [3], [28] Despite the incomplete nature of the data, maternal mortality data in low-income countries can be extracted from numerous sources including medical-facility records, vital registries (when available), coroners’ records, churches, and community registries. For community-based studies to gather the most complete possible count of maternal deaths, multiple sources of data (facility records, and community-based sources) must be reviewed to identify of maternal deaths. [28], [29] For facility-based studies, records from multiple departments or wards must be reviewed to ensure comprehensive capture of maternal deaths in the facilities. [28], [29].

II. Study protocol.

Variations in protocol used to assign cause of death for maternal deaths are common, and the quality of data sources vary with regard to the quality of information available for cause of death assignment. [29] Nevertheless, standard clinical definitions of direct and indirect causes of maternal death exist, and international guidelines are provided by the International Classification of Disease. [9] Studies should provide a standardized definition of causes of maternal death, and should follow clinical or international standard protocol for cause of death attribution. Verbal autopsy studies must contend with an additional layer of complexity due to the non-clinical nature of the data collection process. Various algorithms based on ICD-10 definitions have been developed for clinicians and computer-based algorithms to assign cause of death from verbal autopsy data with the highest degree of validity possible. [30] While computer-based algorithms for cause of death assignment have been validated in facility-based settings, [31], [32] the generalizability of such algorithms, derived from cause of death distributions in facilities, may be limited in community settings. [33] Studies that assign cause of death from verbal autopsy data should establish the procedure used and should justify the choice of algorithm based on the study sample.

III. Selection bias.

When the aim of a study is to document the total and cause specific burden of maternal mortality for a general population (e.g. a city, a country), facility based studies may suffer from selection bias because women with abortion related complications face a range of barriers to the access of medical services, including regulations that restrict access to safe abortion, cultural practices that stigmatize abortion, and socio-economic conditions that often lead women to attempt unsafe abortion even in settings where abortion is safely available. Facility-based data from developing countries where access to health facilities may be limited by social, cultural, and economic factors, are rarely generalizable to populations outside of those seeking medical care in facilities. Nevertheless, studies often attempt to make inference from facility-based data to a larger target population (e.g., surrounding communities). Such interpretations compromise the internal validity of facility-based studies.

The obstacles to medical care for women who have abortions outside of the formal medical system may produce underestimates of abortion-related mortality in facility-based datasets. In some circumstances, selection bias could also cause over-estimation of abortion-related mortality; in facility-based studies that use datasets collected from referral hospitals, abortion-related deaths may be over-represented as a proportion of maternal deaths. This is because a) the most severe cases may get sent directly to referral hospitals and b) delays in seeking care may disproportionately affect women with abortion-related complications resulting in those cases arriving at referral facilities too late to save the women’s lives. [25].

IV. Misclassification.

Some women who experience complications from an unsafe abortion will seek care in health facilities; however, even among those who do, in settings where abortion is legally restricted or culturally stigmatized, women are often reluctant to disclose attempted abortion to providers. Such underreporting of abortion-related complications in facilities is a form of misclassification that almost surely leads to an underestimate of abortion related deaths. [1], [8] Verbal autopsy may provide some advantages over facility-based estimates in providing estimates of abortion related death at the community level, but the stigmatization of abortion often influences what information is reported by relatives, and may lead to misclassification. Mortality resulting from unsafe abortion is often a highly stigmatized event [13], [34] and the social, economic, and legal considerations surrounding abortion may lead to a reluctance among family members report abortion-related deaths. [20], [21].

Women who experience complications from unsafe abortion most often present to facilities with symptoms much akin to hemorrhage or sepsis. Physicians who assign cause of death may unintentionally misclassify abortion related deaths as death from hemorrhage, sepsis, or spontaneous abortion. [4], [19], [25] The risk of misclassification is heightened with verbal autopsy data, as physicians do not have the advantage of examining a patient and must rely on the accuracy of symptoms and contributing factors reported by non-clinicians. [35][37] Additionally, in settings where abortion is legally restricted, providers can face legal action if they provide medical care to a patient who has attempted to induce abortion. [8] Thus, in an effort to provide much needed care for their patients, providers may intentionally misclassify abortion-related complications and deaths, leading to differential misclassification that is almost certain to produce an underestimate of abortion related deaths. [37] Finally, when cause of death is unclear, it can be assigned as ‘unknown cause’, and evidence suggests that, because of its unique measurement challenges, abortion related death is more likely than the other obstetric causes to be classified as ‘unknown’. [37], [38].

All studies were evaluated with respect to the degree to which they achieved the five criteria outlined in Table 1. Emphasis was placed on the potential of study results to suffer from the various bias considerations outlined above, and the extent to which authors addressed these biases in analysis or interpretation of their findings. In addition, 10 studies were selected randomly and were reviewed by a second reviewer (DV) to determine inter-rater reliability. All studies were evaluated using the same rubric and with particular attention to the methodological considerations outlined above.

Results

Figure 1 summarizes the results of the search process. The initial search strategy identified 7,438 articles. After excluding all duplicate titles, and reviewing titles and abstracts for English language and relevance to the research question, the full text of 92 articles were reviewed for possible inclusion in the study. Of those articles whose full text was reviewed, 56 did not meet inclusion criteria. Two articles were review articles, synthesizing data from a variety of sources, forty-five articles did not meet the sample size inclusion criteria, five articles were not published in peer-reviewed journals, and three articles did not report any abortion related deaths in their sample. The total number of studies included in the review was thirty-six.

thumbnail
Figure 1. Depicts the search strategy for this review.

7,438 articles were initially identified. After excluding all duplicate titles, and reviewing titles and abstracts, the full text of 92 articles were reviewed for possible inclusion. Of those articles whose full text was reviewed, 56 did not meet inclusion criteria. The total number of studies included in the review was thirty-six.

https://doi.org/10.1371/journal.pone.0053346.g001

The thirty-six articles included in this review were conducted in a wide range of settings; the majority were conducted in Sub-Saharan Africa (n = 18), nearly one third of studies were conducted in Asia (n = 10), while four studies were conducted in Latin America and the Caribbean, and another four studies conducted in the Middle East. The articles can be divided into two types of studies: 1) facility-based studies (n = 22) where all data were collected at hospitals or medical facilities, and 2) community-based studies (n = 14) where data were collected from a variety of data sources in the community. Of the community-based studies, some included data from facilities (n = 8). A variety of study designs were used, not all of which conform to traditional epidemiologic designs. However, of thirty-six included studies, twenty three retrospective designs, three were ambi-directional designs, and ten were prospective designs. Sample sizes of the included studies ranged from 104–769 maternal deaths. Twenty-two out of thirty-six (61.1%) studies provided a clinical or international standard definition of abortion. No study presented confidence intervals, or any measure of precision for estimates of abortion-related mortality or any other cause of mortality. Table 3 summarizes the main findings of each of the studies reviewed by quality rating.

Quality Rating

No study received a rating of Excellent; this can primarily be attributed to poor evaluation for the criterion “Risk of bias”. To be considered “Excellent”, studies would have had to empirically demonstrate (through validation studies or other methods) that their data were free from major sources of systematic error, or, in the absence of such freedom from bias, perform a quantitative analysis of the effect of potential biases present in the data (through sensitivity analyses or other bias correcting techniques). No study attempted either strategy.

Of the 10 randomly selected studies that were reviewed by two raters (DV and CG), all ten were assigned to the same rating categories by both raters. Although the individual ranking components varied slightly across reviewers, the overall ratings were identical for both reviewers.

Meta-analysis of the data from the thirty-six studies was determined to be inappropriate due to the wide variation in context, study design, and measures. Findings, however, were qualitatively analyzed to determine whether any discernable pattern emerged by quality, geographic region, or type of study with regard to the proportion of abortion-related deaths reported by each study. Overall, studies receiving a “Very Good” rating found the highest estimates of abortion related mortality (median: 16%, range 1–27.4%). Studies receiving a “Very Poor” rating found the lowest overall proportion of abortion related deaths (median: 2%, range 1.3–9.4%). Table 4 shows the studies by quality level and proportion of abortion related deaths reported.

thumbnail
Table 4. Proportion of abortion related deaths reported by study quality.

https://doi.org/10.1371/journal.pone.0053346.t004

Ten of thirty-six studies received the rating of Very Good. All studies in the Very Good category used multiple data sources to identify maternal deaths, provided the international standard definition of abortion (ICD version 9 or 10), and clearly described the methods used to assign cause of death. Predominantly, studies that were categorized as Very Good were prospective in design. Despite the lack of quantitative bias assessment, all studies receiving a Very Good rating enumerated the biases thought to be present in their data, and provided a thorough discussion of potential study limitations and cautions to be taken in interpreting the results of the studies. The 2001 paper by Sloan, et al [39] provides a notable example of such a discussion. In this paper, the authors reanalyzed data from a verbal autopsy study conducted in three regions of rural Mexico in 1995, using multiple validated methods to determine cause of death from verbal autopsy. The paper aimed to assess variations of cause of death found through the various methods used. In their discussion, the authors discuss various limitations of verbal autopsy data, stating that

“In our rural study, many women delivered at home and the information given on death certificates was probably both incomplete and inaccurate, rarely being based on pathological examination or direct observation…”.

Additionally, the authors note that variations in the distribution of cause of death using different methods for assigning cause of death were at times so great that the data became un-interpretable.

Six out of thirty-six studies received a Fair rating. Studies in the Fair category varied in the sources of data reviewed; some reviewed multiple sources of data, others reviewed only hospital records. A mix of retrospective, prospective, and ambi-directional study designs were used. All studies, however, provided a definition of abortion, and most reported with sufficient detail the procedures used to assign cause of death. No study that received a Fair rating provided a detailed description of limitations or the potential for biases contained in the data. One typical “Fair” study is a nationally representative cohort study of maternal deaths in Egypt, conducted by Campbell, et al in 2005. This study reviewed official records of maternal deaths, collected through active surveillance of maternal deaths during two one-year periods (1992–1993, and the year 2000) and followed up with verbal autopsy to assign cause of death. Clear definitions of maternal death and all cause of death were provided based on international standards, and the citation for ICD-10 classification of cause of death was provided. A detailed description of physician training in verbal autopsy and cause of death assignment was given, and the procedure for validation of cause of death (repeating verbal autopsies in a percentage of cases to ensure validity of initial recording) was clearly articulated. Despite the large sample size (772 maternal deaths in the first year, 585 in the second year) and the nationally representative nature of the data, the authors provide no discussion of the general limitations of verbal autopsy for assigning cause of death nor do they provide any assessment of potential misclassification or underreporting that could have occurred with respect to abortion related deaths because of the legal status or stigma surrounding abortion.

Fourteen of thirty-six studies received a rating of Poor. These studies predominantly used retrospective study designs, most were facility-based studies, and no studies categorized as Poor used multiple sources of data to identify maternal deaths. Only three studies in this category provided a definition of abortion (two studies reported clinical definitions, one study reported ICD-9 definitions), few studies offered descriptions of the protocol followed or the process used to assign cause of death, and no study provided a thorough discussion of biases and limitations of their data. Additionally, some studies in the Poor category found smaller or larger proportions of maternal death attributable to abortion than what is suggested by the general literature or other studies in a similar geographic region. When such findings occurred, studies rated Poor were most likely to dismiss the results of other studies, or ignore the contradiction all together. One such discrepancy can be found in the paper by Mswia et al. [40] Despite the prospective nature of the study, and the explanation of protocol used to assign cause of death, significant variations in distribution of cause of death are found across study sites. Though the authors claim that the rural sites are similar in size and socio-economic make-up, no explanation is provided about factors that might be considered as driving the differences in distribution of cause of death across sites, nor is any discussion devoted to the discrepancy between the studies’ finding of abortion related deaths (7.4% of maternal deaths) and other studies that have suggested a higher proportion (up to 20% of maternal deaths [1], [28]) in East Africa.

Six out of thirty-six studies received a rating of Very Poor. All studies in this category were facility-based studies, though the directionality of the study designs varied, none of the studies receiving a Very Poor rating used multiple sources of data to identify maternal deaths. None of these studies reported any definition of abortion, and few provided any description of the process or protocol followed in the assignment of cause of death. The discussion sections of these papers were found to be severely lacking, and most of the studies in the Very Poor category failed to discuss any limitations of the study or the data.

Discussion

A few notable trends emerge with respect to the quality of studies in this systematic review. First, more than half (54%) of all studies reviewed were categorized in the lowest two possible categories of quality ratings, and not one study achieved the highest possible quality rating. Such results highlight the need for a thorough examination of data sources, data collection techniques, and study reporting in the maternal mortality literature. Second, even among studies receiving a Very Good rating, where maternal mortality estimates were determined to be more valid, the risk of bias in the data reported was moderate to high. While some studies acknowledged the presence of selection bias or misclassification only one study addressed potential biases by using multiple techniques in attempt to validate results [39] and not one study out of thirty-six presented any quantitative assessment of the role of potential biases on their results. Recent developments in analytic tools that allow for the evaluation of sensitivity to multiple potential sources of systematic error and bias [41][43], could be extremely productive when applied to estimates of abortion related mortality. Third, the majority of studies in this systematic review failed to provide a clear definition of abortion, or abortion-related mortality. Without a standard definition, it becomes nearly impossible to compare results across studies or draw conclusions regarding trends of abortion related mortality globally, regionally, or locally. Some controversy surrounding the definition does indeed exist; while the current ICD-10 standard is to separate induced abortion from spontaneous abortion [9] when measuring incidence of abortion as well as abortion-related death, some have suggested that the risk of misclassification, in both directions, indicates that induced and spontaneous abortions should be measured as one category [23]. Misclassification of induced vs. spontaneous abortion related deaths may be present in some of the studies reviewed here. It was beyond the scope of this systematic review to determine the direction or magnitude of such misclassification, however, regardless of which measure is ultimately chosen, it is imperative that the field settle on a clear and precise definition of abortion.

An additional trend emerges from the results of studies in this systematic review; on average, studies of higher quality reported estimates of abortion-related mortality that were higher than the estimates of abortion-related mortality reported by studies of lower quality. While meta analysis of the studies included in this review was not possible, this finding supports the widely stated position that current estimates of maternal mortality due to unsafe abortion, which are primarily estimated from resource poor settings where high quality data collection is most challenging, [1] are likely under-estimating the true burden of unsafe abortion-related mortality.

While many studies in the review had substantial limitations, this systematic evaluation allowed identification of key directions for improvement of future research. Improvements in the quality of data sources and data collection are the ultimate solution to better understanding global abortion-related mortality, and recent calls for investments from the global community in vital registration systems for all countries may go a long way to addressing such issues. [2], [11], [44] In the mean time, the field should encourage better reporting of study procedures and standardization of the definition of abortion and abortion-related mortality, and should support the use of multiple bias analysis techniques in the reporting of data, a method that could greatly aid the interpretation of results from studies seeking to quantify abortion related mortality.

Supporting Information

Appendix S1.

Provides a line by line search strategy for all databases searched.

https://doi.org/10.1371/journal.pone.0053346.s001

(PDF)

Acknowledgments

The authors would like to thank Dr. Sujit Rathod for his thoughtful comments and edits to the manuscript.

Author Contributions

Reviewed the entire content of the manuscript, provided edits, and comments: CG DV JA. Conceived and designed the experiments: CG. Performed the experiments: CG DV. Analyzed the data: CG DV JA. Contributed reagents/materials/analysis tools: CG DV JA. Wrote the paper: CG DV JA.

References

  1. 1. WHO (2010) Unsafe abortion: global and regional estimates of the incidence of unsafe abortion and associated mortality in 2008. WHO, Geneva, Switzerland.
  2. 2. AbouZahr C (2011) New estimates of maternal mortality and how to interpret them: choice or confusion? Reprod Health Matters 19(37): 117–128.
  3. 3. Ahman E, Shah I (2011) New estimates and trends regarding unsafe abortion mortality. International J Gynecol & Obstet 115(2): 121–126.
  4. 4. AbouZahr C (2006) Global burden of maternal death and disability. Br Med Bull 67: 1–11.
  5. 5. Graham W, Ahmed S, Stanton C, Abou-Zahr C, Campbell O (2008) Measuring maternal mortality: an overview of opportunities and options for developing countries. BMC Medicine 6(1): 12.
  6. 6. WHO (2010) Trends in Maternal Mortality 1990–2008. Estimates developed by WHO, UNFPA, UNICEF, and the World Bank. WHO, Geneva, Switzerland.
  7. 7. Hogan M, Foreman K, Naghavi M, Ahn S, Wang M, et al. (2010) Maternal mortality for 181 countries, 1980–2008: a systematic analysis of progress towards Millennium Development Goal 5. Lancet 375(9726): 1609–1623.
  8. 8. Sedgh G, Singh S, Shah I, Öhman E, Henshaw S, et al. (2012) Induced abortion: incidence and trends worldwide from 1995 to 2008. Lancet 379(9816): 625–632.
  9. 9. WHO (1993) ICD-10: International Statistical Classification of Diseases and Related Health Problems: Tenth Revision. WHO, Geneva, Switzerland.
  10. 10. Say L (2008) Maternal mortality and unsafe abortion: preventable yet persistent. IPPF Med Bull. 42(2).
  11. 11. Say L, Chou D (2011) Better understanding of maternal deaths‚ the new WHO cause classification system. BJOG 118: 15–7.
  12. 12. Betran A, Wojdyla D, Posner S, Gulmezoglu A (2005) National estimates for maternal mortality: an analysis based on the WHO systematic review of maternal mortality and morbidity. BMC Public Health 5(1): 131.
  13. 13. Singh SW. (2009) Abortion Worldwide: A decade of Uneven Progress. Guttmacher Institute, New York, New York.
  14. 14. Sedgh GH (2007) Legal Abortion Worldwide: Incidence and Recent Trends. Int Fam Plan Perspect 33(3): 106–116.
  15. 15. Rasch V, Kipingili R (2009) Unsafe abortion in urban and rural Tanzania: method, provider and consequences. Tropical Medicine & International Health 14(9): 1128–1133.
  16. 16. Kaye D (2006) Domestic violence as risk factor for unwanted pregnancy and induced abortion in Mulago Hospital, Kampala, Uganda. Tropical Medicine & International Health 11(1): 90–101.
  17. 17. Grimes D, Benson J, Singh S, Romero M, Ganatra B, et al. (2006) Unsafe abortion: the preventable pandemic. Lancet 368(9550): 1908–1919.
  18. 18. Glasier A, Gülmezoglu A, Schmid G, Moreno C, Van Look P (2006) Sexual and reproductive health: a matter of life and death. Lancet 368(9547): 1595–1607.
  19. 19. Cates W (1982) Mortality from abortion and childbirth: are the statistics biased? JAMA 248: 192–196.
  20. 20. Garenne M (1997) Risk Factors for Maternal Mortality: A Case-Control Study in Dakar Hospitals (Senegal). Afr J Reprod Health 1997;1(1).
  21. 21. Shah IA (2009) Unsafe Abortion: Global and Regional Incidence, Trends, Consequences, and Challenges. J Obstet Gynaecol Can 31(12): 1149–1158.
  22. 22. Moher D, Liberati A, Tetzlaff J, Altman D (2009) Preferred reporting items for systematic reviews and meta-analyses: the PRISMA statement. PLoS medicine 6(7): e1000097.
  23. 23. Adler A, Filippi V, Thomas S, Ronsmans C (2012) Quantifying the global burden of morbidity due to unsafe abortion: Magnitude in hospital-based studies and methodological issues. International J Gynecol & Obstet 118: S65–S77.
  24. 24. Rasch V (2011) Unsafe abortion and postabortion care–an overview. Acta Obstet Gynecol Scand 90(7): 692–700.
  25. 25. Rossier C (2003) Estimating induced abortion rates: a review. Stud Fam Plann 34(2): 87–102.
  26. 26. Charles V, Polis C, Sridhara S, Blum R (2008) Abortion and long-term mental health outcomes: a systematic review of the evidence. Contraception 78(6): 436–450.
  27. 27. Hill K (2006) Making deaths count. Bull World Health Organ 84(3): 162–162.
  28. 28. Khan K (2006) WHO analysis of causes of maternal death: a systematic review. Lancet 367: 1066–1074.
  29. 29. Cross S, Bell J, Graham W (2010) What you count is what you target: the implications of maternal death classification for tracking progress towards reducing maternal mortality in developing countries. Bull World Health Organ 88(2): 147–153.
  30. 30. Fottrell E, Byass P (2010) Verbal autopsy: methods in transition. Epidemiol Reviews 32(1): 38–55.
  31. 31. Kalter H, Salgado R, Babille M, Koffi A, Black R (2011) Social autopsy for maternal and child deaths: a comprehensive literature review to examine the concept and the development of the method. Population Health Metrics 9(1): 45.
  32. 32. Murray C, Lopez A, Black R, Ahuja R, Ali S, et al. (2011) Population Health Metrics Research Consortium gold standard verbal autopsy validation study: design, implementation, and development of analysis datasets. Population Health Metrics 9(1): 27.
  33. 33. Prata N, Gerdts C, Gessessew A (2012) An innovative approach to measuring maternal mortality at the community level in low-resource settings using mid-level providers: a fewaibility study in Tigray, Ethiopia. Reprod Health Matters 20(39): 1–10.
  34. 34. Shaw D (2010) Abortion and Human Rights. Best Practice & Research Clinical Obstet & Gynaecol 24(5): 633–646.
  35. 35. Shahidullah M (1995) A Comparison of Sisterhood Information on Causes of Maternal Death with the Registration Causes of Maternal Death in Matlab, Bangladesh. Int J Epidemiol 24(5): 937–942.
  36. 36. Kao S (1997) Underreporting and misclassification of maternal mortality in Taiwan. Acta Obstet Gynecol Scand 76: 629–636.
  37. 37. Walker D, Campero L, Espinoza H, Hernández B, Anaya L, et al. (2004) Deaths from Complications of Unsafe Abortion: Misclassified Second Trimester Deaths. Reprod Health Matters 12(24S): 27–38.
  38. 38. Gülmezoglu A, Say L, Betrán A, Villar J, Piaggio G (2004) WHO systematic review of maternal mortality and morbidity: methodological issues and challenges. BMC Medical Research Methodology 4(1): 16.
  39. 39. Sloan N, Langer A, Hernandez B, Romero M, Winikoff B (2001) The etiology of maternal mortality in developing countries: what do verbal autopsies tell us? Bull World Health Organ 79(9): 805–810.
  40. 40. Mswia R (2003) Community-based monitoring of safe motherhood in the United Republic of Tanzania. Bull World Health Organ 81: 87–94.
  41. 41. Lash T, Fink A (2003) Semi-Automated Sensitivity Analysis to Assess Systematic Errors in Observational Data. Epidemiology 14: 451–8.
  42. 42. Greenland S (2001) Sensitivity Analysis, Monte Carlo Risk Analysis, and Bayesian Uncertainty Assessment. Risk Analysis 21(4): 579–583.
  43. 43. Phillips C (2003) Quantifying and Reporting Uncertainty from Systematic Errors. Epidemiology 14(4): 459–466.
  44. 44. AbouZahr C (2001) Maternal mortality at the end of a decade: signs of progress? Bull World Health Organ 79(6): 561–58.
  45. 45. Zakariah A, Alexander S, Van Roosmalen J, Buekens P, Kwawukume E, et al. (2009) Reproductive age mortality survey (RAMOS) in Accra, Ghana. Reprod Health 6(1): 7.
  46. 46. Ramos S, Karolinski A, Romero M, Mercer R (2007) A comprehensive assessment of maternal deaths in Argentina: translating multicentre collaborative research into action. Bull World Health Organ 85(8): 615–622.
  47. 47. McCaw-Binns A, Lindo J, Lewis-Bell K, Ashley D (2008) Maternal mortality surveillance in Jamaica. Lancet 100(1): 31–36.
  48. 48. McCaw-Binns A, Alexander S, Lindo J, Escoffery C, Spence K, et al. (2007) Epidemiologic transition in maternal mortality and morbidity: new challenges for Jamaica. International J of Gynecol & Obstet 96(3): 226–232.
  49. 49. Sloan N, Langer A, Hernandez B, Romero M, Winkoff B (2001) The etiology of maternal mortality in developing countries: what do verbal autopsies tell us? Bull World Health Organ 79(9): 805–810.
  50. 50. Chowdhury M, Botlero R, Koblinsky M, Saha S, Dieltiens G, et al. (2007) Determinants of reduction in maternal mortality in Matlab, Bangladesh: a 30-year cohort study. Lancet 370(9595): 1320–1328.
  51. 51. Chowdhury M, Ahmed A, Kalim N, Koblinsky M (2009) Causes of maternal mortality decline in Matlab, Bangladesh. J Health Popul Nutr 27(2): 108.
  52. 52. Zakariah A, Alexander S, van Roosmalen J, Yao Kwawukume E (2006) Maternal mortality in the Greater Accra region in Ghana: assessing completeness of registration and data quality. Acta Obstet Gynecol Scand 85(12): 1436–1441.
  53. 53. Bartlett L, Mawji S, Whitehead S, Crouse C, Dalil S, et al. (2005) Where giving birth is a forecast of death: maternal mortality in four districts of Afghanistan, 1999–2002. The Lancet 365(9462): 864–870.
  54. 54. Jafarey S, Rizvi T, Koblinsky M, Kureshy N (2009) Verbal autopsy of maternal deaths in two districts of Pakistan–filling information gaps. J Health Popul Nutr 27(2): 170.
  55. 55. Campbell O, Gipson R, Issa A, Matta N, El Deeb B, et al. (2005) National maternal mortality ratio in Egypt halved between 1992–93 and 2000. Bull World Health Organ 83(6): 462–471.
  56. 56. Geelhoed D, Visser L, Asare K, Schagen van Leeuwen J, et al. (2003) Trends in maternal mortality: a 13-year hospital-based study in rural Ghana. European J of Obstet & Gynecol and Reprod Biology 107(2): 135–139.
  57. 57. Verma K, Thomas A, Sharma A, Dhar A, Bhambri V (2001) Maternal mortality in rural India: a hospital based, 10 year retrospective analysis. J Obstet & Gynaecol Research 27(4): 183–187.
  58. 58. Bell J, Ouédraogo M, Ganaba R, Sombié I, Byass P, et al. (2008) The epidemiology of pregnancy outcomes in rural Burkina Faso. Tropical Medicine & International Health 13(s1): 31–43.
  59. 59. Oyieke J, Obore S, Kigondu C (2006) Millennium development goal 5: a review of maternal mortality at the Kenyatta National Hospital, Nairobi. East Afr Med J 83(1): 4–9.
  60. 60. Mogobe K, Tshiamo W, Bowelo M (2007) Monitoring maternity mortality in Botswana. Reprod Health Matters 15(30): 163–171.
  61. 61. Aboyeji A, Ijaiya M, Fawole A (2007) Maternal mortality in a Nigerian teaching hospital-a continuing tragedy. Tropical Doctor 37(2): 83–85.
  62. 62. Fabamwo A (2009) Correlates of Abortion Related Maternal Mortality at the Lagos State University Teaching Hospital, Ijeka. Afr J Reprod Health 13(2): 139–146.
  63. 63. Panchabhai T, Patil P, Shah D, Joshi A (2009) An autopsy study of maternal mortality: A tertiary healthcare perspective. J Postgrad Medicine 55(1): 8.
  64. 64. Granja A, Machungo F, Gomes A, Bergström S (2001) Adolescent maternal mortality in Mozambique. J Adol Health 28(4): 303–306.
  65. 65. Puri A, Yadav I, Jain N (2011) Maternal mortality in an urban tertiary care hospital of north India. J Obstet & Gynecol India 61(3): 280–285.
  66. 66. Daramola A, Banjo A (2005) Medical audit of maternal deaths in the Lagos University Teaching Hospital, Nigeria. East Afr Med J 82(6).
  67. 67. Oi AF (2007) Abortion related maternal deaths at Lagos State University Teaching Hospital Ikeja, Nigeria: a four year review. Sex Health Matters 8(3): 59–63.
  68. 68. Jain V, Saha S, Bagga R, Gopalan S (2004) Unsafe abortion: A neglected tragedy. Review from a tertiary care hospital in India. J Obstet & Gynaecol Research 30(3): 197–201.
  69. 69. Abdel-Hady E, Mashaly A, Sherief L, Hassan M, Al‚Gohary A, et al. (2007) Why do mothers die in Dakahlia, Egypt? J Obstet and Gynaecol Research 33(3): 283–287.
  70. 70. Bergsjø P, Vangen S, Lie R, Lyatuu R, Lie-Nielsen E, et al. (2010) Recording of maternal deaths in an East African university hospital. Acta Obstet Gynecol Scand 89(6): 789–793.
  71. 71. Lema V, Changole J, Kanyighe C, Malunga E (2005) Maternal mortality at the Queen Elizabeth Central Teaching Hospital, Blantyre, Malawi. East Afr Med J 82(1).
  72. 72. Almerie M, Matar H, Almerie Y (2011) A 20-year (1989–2008) audit of maternal mortality in Damascus, Syria. BMC Pregnancy Childbirth 112(1): 70.
  73. 73. Supratikto G, Wirth M, Achadi E, Cohen S, Ronsmans C (2002) A district-based audit of the causes and circumstances of maternal deaths in South Kalimantan, Indonesia. Bull World Health Organ 80(3): 228–234.
  74. 74. Chhabra S, Sirohi R (2004) Trends in maternal mortality due to haemorrhage: two decades of Indian rural observations. J Obstet & Gynecol 24(1): 40–43.
  75. 75. Iyengar K, Iyengar S, Suhalka V, Dashora K (2009) Pregnancy-related deaths in rural Rajasthan, India: exploring causes, context, and care-seeking through verbal autopsy. J Health Popul Nutr 27(2): 293.
  76. 76. Ujah I, Aisien O, Mutihir J, Vanderjagt D, Glew R, et al. (2005) Factors contributing to maternal mortality in north-central Nigeria: a seventeen-year review. Afr J Reprod Health 9(3): 27–40.
  77. 77. Okonta P, Okali U, Otoide V, Twomey D (2002) Exploring the causes of and risk factors for maternal deaths in a rural Nigerian referral hospital. Lancet 22(6): 626–629.
  78. 78. Mariaga A, Saleh W (2009) Maternal mortality at the state specialist Hospital Bauchi, Northern Nigeria. East Afr Med J 86(1).
  79. 79. Kullima A, Kawuwa M, Audu B, Geidam A, Mairiga A (2010) Trends in maternal mortality in a tertiary institution in Northern Nigeria. J Obstet & Gynecol 8(4).