The prevalence of maternal near miss: a systematic review

Authors


  • The World Health Organization retains copyright and all other rights in the manuscript of this article as submitted for publication.

Dr Ö Tunçalp, Department of Population, Family and Reproductive Health, Johns Hopkins Bloomberg School of Public Health. 615 N Wolfe St MD 21205, Baltimore, USA. Email: otuncalp@jhsph.edu

Abstract

Please cite this paper as: Tunçalp Ö, Hindin MJ, Souza JP, Chou D, Say L. The prevalence of maternal near miss: a systematic review. BJOG 2012;119:653–661.

Background  Severe maternal morbidity or ‘near miss’ is a promising indicator to improve quality of obstetric care.

Objectives  To systematically review all available studies on ‘near miss’.

Search strategy  Following a pre-defined protocol, our review covered articles between January 2004 and December 2010. We used a combination of the following terms: near miss morbidity, severe maternal morbidity, severe acute maternal morbidity, obstetric near-miss, maternal near miss, obstetric near miss, emergency hysterectomy, emergency obstetric hysterectomy, maternal complications, pregnancy complications, intensive care unit.

Selection criteria  Nearly 4000 articles were screened by title and abstract, and 153 articles were retrieved for full text evaluation. There were no language restrictions.

Data collection and analysis  Data extraction was performed using an instrument that included sections on study characteristics, quality of reporting, prevalence/incidence and the definition and identification criteria. Univariate analysis and meta-analysis for sub-groups were performed.

Main results  A total of 82 studies from 46 countries were included. Criteria for identification of cases varied widely. Prevalence rates varied between 0.6 and 14.98% for disease-specific criteria, between 0.04 and 4.54% for management-based criteria and between 0.14 and 0.92% for organ-based dysfunction based on Mantel criteria. The rates are higher in low-income and middle-income countries of Asia and Africa. Based on meta-analysis, the estimate of near miss was 0.42% (95% CI 0.40–0.44%) for the Mantel (organ dysfunction) criteria and 0.039% (95% CI 0.037–0.042%) for emergency hysterectomy. Our meta-regression results indicate that emergency hysterectomy rates have been increasing by about 8% per year.

Authors’ conclusions  There is growing interest in the application of the maternal near-miss concept as an adjunct to maternal mortality. However, in the literature published before 2011 there was still important variation in the criteria used to identify maternal near-miss cases. The World Health Organization recently published criteria based on markers of management and of clinical and organ dysfunction which would enable systematic data collection on near miss and development of summary estimates. Comparing the rates over time and across regions, it is clear that different approaches are needed to lower the rates of near miss and that interventions must be developed with the local context in mind.

Introduction

Progress in the reduction of maternal mortality has been slow. Over 1000 women still die from pregnancy-related causes every day around the world and the vast majority of these deaths occur in developing countries.1 Low-income countries are heavily affected by the burden of maternal deaths and maternal mortality is still an important public health problem among middle-income countries. Strengthened health systems and effective maternal health care (particularly to those women experiencing acute pregnancy-related complications) are considered the key factors for reducing maternal mortality.2

Many countries are encouraging pregnant women to deliver in health facilities. On the one hand, this policy favours the reduction of delays in the identification and management of peripartum complications. On the other hand, such a policy may lead to overloading of health services, which are already insufficient in many of these settings, thereby adversely affecting the quality of care. In this context, quality of care has been identified as a central element in the United Nations Global Strategy for Women and Children Health.3

Confidential enquiries into maternal deaths have been in use for many years in the identification of quality of care and health systems issues.4 However, in low mortality settings or at the health service level, the number of maternal deaths is generally insufficient or not representative enough to allow reliable policy guidance. In the last 20 years, the concept of maternal near miss has been explored in maternal health as an adjunct to maternal-death confidential enquiries. Among other positive characteristics, near-miss cases occur more frequently than maternal deaths and can directly inform on problems and obstacles that had to be overcome during the process of health care, providing more robust conclusions and rapid reporting on maternal care issues.5,6 Hence, the identification of cases of severe maternal morbidity has emerged as a promising complementary or alternative strategy to reduce maternal mortality. In particular, near misses have been viewed as a useful outcome measure for the evaluation and improvement of maternal health services in developing countries.4

In 2003/04, the World Health Organization (WHO) conducted a systematic review on the prevalence of severe maternal morbidity and maternal near-miss cases. The substantial heterogeneity observed in the pre-2004 literature led WHO to establish a technical working group comprising obstetricians, midwives, epidemiologists and public healthcare professionals to develop a standard definition and uniform identification criteria for maternal near-miss cases.7 In April 2009, a paper was published by the WHO working group defining maternal near miss morbidity as ‘a woman who nearly died but survived a complication that occurred during pregnancy, childbirth or within 42 days of termination of pregnancy’. With a view to achieving a reasonable balance between the burden of data collection and useful information, the WHO working group targeted the identification of only very severe cases—i.e. primarily those presenting with features of organ dysfunction.7 The near-miss identification criteria developed by the technical working group have been tested and validated as being able to provide robust and reliable data. Detailed information about the near-miss concept and its development is published elsewhere (Souza JP et al., manuscript in preparation).7,8

Methods

A WHO systematic review published in 2004, spanned the literature on severe maternal morbidity and maternal near miss from 1997 to 2004.9 The current review provides an update based on a search for new articles between January 2004 and December 2010. We have included the electronic databases Pubmed, Embase, Lilacs, Popline, IndMed and WHO regional databases. The regional databases are as follows: Index Medicus for the Eastern Mediterranean Region (IMEMR), African Index Medicus (AIM), Western Pacific Region Index Medicus (WPRIM). We have also searched for relevant articles in the WHO Library by hand, and contacted experts in the field.

We used a similar search strategy to that in the previous study with a combination of the following terms: near miss morbidity, severe maternal morbidity, severe acute maternal morbidity, obstetric near-miss, maternal near miss, obstetric near miss, emergency hysterectomy, emergency obstetric hysterectomy, maternal complications, pregnancy complications, intensive care unit. To verify the compatibility between the current and previous search strategies, we applied this search strategy to the timeframe covered by the previous systematic review and compared the results.

The inclusion criteria for the current review were as follows: (1) articles with near-miss incidence or prevalence data, (2) published between January 2004 and December 2010, (3) included data from 1990 onwards, (4) sample size ≥200 and (5) clearly described methodology. There were no language restrictions.

Data extraction was performed using an instrument that included sections on the general study level characteristics (such as design, population, setting), quality of data reporting, prevalence/incidence of maternal near miss and the definition and identification criteria. Authors have also been contacted in cases where further information or clarification was required. The denominator used was either deliveries or live births (only five studies used live births as a denominator).

We describe the included studies with an emphasis on the different definitions used and criteria for identification of the cases. We performed univariate analyses and the ranges of near-miss prevalence are reported based on identification criteria and region. Studies reporting very high near-miss prevalence that was outside the ranges formed by most of the studies in the review were identified as outliers. We also reported on near miss over mortality ratio to assess the care that near-miss cases receive.

We conducted meta-analysis for the sub-groups management (emergency hysterectomy) and Mantel-based organ dysfunction criteria, where more homogeneous criteria were established among the studies. Also we conducted multivariable meta-regression for the emergency hysterectomy group. We included the studies from the 2004 review in our final analyses and observed the historical trends.9 Unless indicated otherwise all of the analyses include the 82 studies from the current review. We used Stata 10 (StataCorp LP; StataCorp., College Station, TX, USA) for our analyses.

Results

Nearly 4000 articles were initially screened by title and abstract, and 153 articles were retrieved for full text evaluation. Data have been extracted from 82 articles (Figure 1). A total of 82 studies from 46 countries were included in this review. Studies were mainly retrospective cross-sectional and except for one study in Brazil,10 all of the studies used data from facilities, mainly tertiary-care hospitals. A majority of the studies included describe the characteristics of the setting and participants as well as reporting definitions and procedures of identification of the cases. More detailed information on each study is presented in the Table S1. To validate our current strategy, we tested it on the articles included in the 2004 review and identified 27 out of the 30 articles included in the previous review strategy. The three articles not found initially were identified after a complementary reference lists search.

Figure 1.

 The flow diagram of identification of studies.

Except for the studies reporting on emergency hysterectomies and intensive-care unit (ICU) admissions, a majority of the studies defined near miss as a woman who almost died but survived through chance or as a result of good care received. Overall, there were three major approaches to the identification: (1) disease-specific criteria (i.e. severe pre-eclampsia, severe postpartum haemorrhage), (2) management-based criteria (i.e. admission to ICU, need for a blood transfusion), or (3) organ system dysfunction-based criteria. The majority used management-based criteria including 33 studies that used emergency hysterectomies, and 18 that used ICU admissions to define near miss. Seven studies used disease-specific criteria, nine of the articles used organ system dysfunction as the criteria for near miss and 14 used a combination of disease, organ and management-based criteria.

Near miss by identification criteria

Study-specific near-miss rates differed based on the method of identification and region (Table 1). Studies using disease-specific criteria reported a higher percentage of near-miss cases, and a wider range of estimates compared with the other criteria, 0.6–14.98%. Case identification criteria varied for disease-specific criteria, which included, but was not limited to, hypertensive disorders, haemorrhage, uterine rupture, sepsis and anaemia. Most of the studies established a certain criteria for the degree of severity for the selected clinical conditions.11–16

Table 1.   Near-miss rates by identification criteria and region
Identification criteriaRange of near misses (%)No. of studies (= 81)
  1. *Near-miss rate by Amorim et al.17 is reported separately as it is an outlier in the group.

  2. **Near-miss rate by Adisasmita et al.19 is reported separately as it is an outlier in the group.

  3. ***Regions have been allocated based on UN Classifications.25

Disease-specific0.6–14.987
Management-specific
Emergency hysterectomy0.04–0.2634
ICU sdmission0.04–4.54 (15.8)*18
Organ dysfunction0.14–2.3 
Mantel or Modified Mantel criteria0.14–0.92 (17.8)**8
Other2.31
Mixed criteria
Disease/Organ/Management0.04–4.437
Disease/Management0.09–3.427
Region***
Africa0.05–14.9814
Asia0.02–5.07 (17.8)**31
Latin America and Caribbean0.34–4.93 (15.8)*9
Europe0.04–0.7918
North America0.07–1.3810
Oceania (Australia)1.251

Near-miss rates identified by management-specific criteria ranged between 0.04 and 4.54%. However, it should be noted that within this group, studies identifying emergency hysterectomies reported lower percentages (0.04–0.26%) than studies using ICU admissions (0.04–4.54%). The Amorim et al.17 study in Brazil reported a near miss rate of 15.8%—much higher than the rest of the studies in this category and considered one of the two outliers identified in this review. The hospital in this study serves a very large geographic area as the main maternal ICU unit and therefore a significant proportion of the women admitted to the ICU did not enter the denominator—deliveries at the hospital.

Organ dysfunction-based criteria were used in nine studies and the near-miss rate reported ranged between 0.14 and 2.3%. Eight of nine studies used either Mantel or modified Mantel criteria in this category and reported rates between 0.14 and 0.92%. These criteria were first introduced in South Africa and combine organ dysfunction and certain management markers such as intensive-care admission, emergency hysterectomy to identify near-miss cases.18 This criterion is arguably the most stable compared with others in this review. Depending on the level of the facility and resources available, some studies used a modified version of these criteria.19–23 One study by Adisasmita et al.19 in Indonesia reported a near-miss rate of 17.8%—much higher than the other studies in this category and considered the second outlier in this review. This study included a large number of early pregnancy losses in the numerator and included a number of clinical diagnoses (eclampsia, uterine rupture and ectopic pregnancy) in their near-miss criteria. Using deliveries as the denominator and including these clinical diagnoses may have inflated the near-miss rate. However, even if the obstetric-related admissions were used as the denominator and only near miss cases with organ dysfunction are included, the near miss rate is 11.3%—still very high compared with all the other studies.

Near miss by region

Based on their income, all African and Asian countries (where there were near-miss studies) except Saudi Arabia and Kuwait are considered low-income or middle-income countries.24Table 2 shows the near-miss rates in each region by different identification criteria. The upper near-miss rate ranged from 4.93% in Latin America and the Caribbean, through 5.07% in Asia to 14.98% in Africa (excluding outliers). In contrast, studies from high-income countries (Europe, North America and Australia) reported an upper near-miss rate from a low of 0.79% in Europe to a high of 1.38% in North America: the lowest rates across all the criteria compared with those from low-income and middle-income countries.

Table 2.   Near miss rates in each region by different identification criteria (# signifies the number of studies)
Identification criteriaAfricaAsiaLACEuropeN. America and Australia
% Near missn% Near missn% Near missn% Near missn% Near missn
  1. *Near-miss rate by Amorim et al.17 is reported separately as it is an outlier in the group.

  2. **Near-miss rate by Adisasmita et al.19 is reported separately as it is an outlier in the group.

Disease-specific0.6–14.9833.21–5.0724.9310.791
Management-specific
Emergency hysterectomy0.05–0.4160.02–0.49160.8710.04–0.2080.07–0.163
ICU admission0.2410.12–0.9090.34–4.54 (15.8)*30.04–0.7340.29–1.332
Organ dysfunction
Mantel or Modified Mantel0.41–0.8430.9 (17.8)**20.6310.14–0.352
Other2.31
Mixed Criteria
Disease/Management3.1513.4212.10–3.0420.09–0.7221.251
Disease/Organ/Management2.12–4.4330.27–1.384
Total0.05–14.98140.02–5.07310.34–4.92110.04–0.79170.07–1.3810

The most commonly used criterion was emergency hysterectomies in all regions except North America, and Latin America and the Caribbean (LAC) (6/14 in Africa, 16/31 in Asia, 8/17 in Europe, 3/10 in North America and Australia and 1/11 in LAC). The second most common criterion across regions was ICU admission, for all regions except Africa. In North America, several of the studies identified near misses using criteria combining disease, organ and management markers, using International Classification of Disease (ICD) codes and national and regional databases. Depending on the resources of a facility or a country overall, the criteria used for identification of near misses vary. For example, in high-income countries where facility deliveries and systematic, national level data are the norm, more sensitive markers have been used to identify the near-miss cases, whereas in resource-poor settings, management-based criteria are more commonly used.

In low and middle-income countries, approximately 1% of the women experienced a near-miss event before, during or after delivery as identified by organ dysfunction criteria. It was around 0.25% in higher-income countries. Management-specific criteria using ICU admissions and emergency hysterectomies were under 1% across all regions, except the two studies from the LAC region.17,25 Using mixed criteria combining different markers, the rate ranges between 2.10 and 4.43% in low-income and middle-income countries and 0.09 and 1.38% in higher-income countries.

Historical trends in near miss

In the 6 years since the 2004 WHO review, more articles have been published focusing on near-miss maternal morbidity; thereby increasing the number of articles included from 30 to 82 articles. Table 3 shows a comparison of the two reviews based on identification criteria and region. The ranges of near-miss cases based on different criteria reported were similar between the two time points. However, it should be noted that in the current review the ranges are wider for each category, there are more studies using mixed criteria overall and a larger number of studies from low-income and middle-income countries report on organ dysfunction criteria.

Table 3.   Comparing the rates by criteria and region between 2004 and current review
 2011 Review (N = 82)2004 Review (N = 30)
  1. *Near-miss rate by Amorim et al.17 is reported separately as it is an outlier in this group.

  2. **Near-miss rate by Adisasmita et al.19 is reported separately as it is an outlier in this group.

  3. ***Near-miss rate by Kaye et al.32 is reported separately as it is an outlier in this group.

  4. ****Near-miss rate by Noor et al.33 is reported separately as it is an outlier in this group.

Identification criteria
Disease-specific0.6–14.980.8–8.23
Management-specific
 Emergency hysterectomy0.04–0.260.01–0.21 (2.99)****
 ICU admission0.04–4.54 (15.8)*0.08–1.02
Organ dysfunction
Mantel or Modified Mantel0.14–0.92 (17.8)**0.38–1.09 (10.61)***
Other2.3
Mixed criteria0.04–4.431.34
Region2011 Review2004 Review
Africa0.05–14.980.08–10.61
Asia0.02–5.07 (17.8)**0.01–4.37
Latin America and Caribbean0.34–4.93 (15.8)*0.14–0.24
Europe0.04–0.790.04–1.20
North America and Australia0.07–1.380.08–0.27
Oceania1.25

Meta-analysis of near miss

Although we could not conduct meta-analysis of near miss as a comprehensive category because of the variety of identification criteria, we have conducted meta-analyses for the two categories, emergency hysterectomy and Mantel-based organ-based dysfunction, where more homogeneous criteria were used among the studies compared with the other criteria. As mentioned above, Mantel criteria include a set of organ-based dysfunction and management-based criteria, the detailed description of indicators by organ system and degree of severity can be found elsewhere.18 For this analysis, we have also included 2004 review articles. All of the studies included in this analysis described the characteristics of the setting, participants, definitions and procedures of case identification (11 studies for Mantel criteria and 40 studies for emergency hysterectomy). For the Mantel-based criteria, the estimate of near miss was 0.42% (95% CI 0.40–0.44%). For the emergency hysterectomy criteria, the near-miss rate was 0.039% (95% CI 0.037–0.42%). Despite the very narrow range of the confidence intervals, the I-squared was high: 98.3% for the Mantel-based criteria and 95.5% from the emergency hysterectomy criteria, suggesting significant heterogeneity between studies (Figures 2 and 3).

Figure 2.

 Meta-analysis among studies using Mantel-based organ dysfunction criteria (= 11).

Figure 3.

 Meta-analysis among studies using emergency hysterectomy—a management based criteria (= 40).

We explored the heterogeneity in a multivariable meta-regression model for near miss identified by emergency hysterectomy by using median data collection year, study region and gross national income for the respective median data collection year (results not shown).26 For this analysis we combined North America and LAC region, as there was only one study from LAC. Our analysis showed that emergency hysterectomy rates were higher in more recent years of data collection (an approximate 8% increase per year, P = 0.02, 95% CI 0.01–0.15). Near miss identified by emergency hysterectomy was significantly more common in North America/LAC compared with Europe, mostly because of the one study in the LAC region from Mexico.27 We found that about one-third of the variation between studies in emergency hysterectomy rates (R2 = 27.96%) could be explained by these variables.

Discussion

We have included 82 studies in this systematic review. All of the included studies have used a variety of near-miss criteria ranging from disease-specific to organ dysfunction and a mix of different systems. The near-miss rates have not significantly changed between the review in 2004 and the current 2011 review, although the ranges are wider for each category in the current review. In the current study, disease-specific criteria produced higher rates than both management-specific and organ dysfunction criteria. Under the management-based criteria, studies using ICU admission produced larger variation than emergency hysterectomy studies.

Studies using management-based criteria were less likely to specifically include the discussion on maternal near miss; rather, they focused on emergency obstetric care and the case series in their facilities. We should underline the fact that the rates produced by these individual studies do not include all the near-miss cases in a facility. Admission to ICU relies on the availability of physical and human resources as well as the criteria for admission used in different institutions. For example Amorim et al.17 reported a very high ICU admission rate because the facility in the study was a referral institution, therefore it was almost impossible to identify the right denominator with all the deliveries.

Definitions of emergency hysterectomy and Mantel criteria were fairly uniform across studies; however, as shown in our meta-analysis there was very high heterogeneity between studies, 95.5% and 98.3% respectively. Although some of this heterogeneity is the result of the year of data collection, which was statistically significant, gross national income and region of the country, it can also be explained by various reasons including several unmeasured variables such as the capacity of the hospital to identify the cases within their chosen criteria, availability or the patient load of the referral institutions in which many of our studies are conducted and the general structure of the health system. It should be noted that although there is heterogeneity, near miss is a condition of very low prevalence and our data still show a narrow range in estimates.

This study has some limitations that should be noted. First, although we searched for unpublished data, it is very likely that we missed some unpublished studies. Second, despite the fact that the number of studies included in this review almost tripled compared with the 2004 review, we still had relatively sparse data globally. Despite these limitations, the study has a number of strengths. First, it is comprehensive in its scope and we made special efforts to reach out to the authors for further information and clarification. Second, this review is very timely as we move towards standardised criteria, as it summarises the current trends in global near-miss rates. Since the 2004 review, the number of studies examining maternal near miss almost tripled, underlining the growing emphasis on the issue globally. Next, we used the previous systematic review data both to assess our search criteria and to assess changes over time. Finally, the study adds to the previous systematic review by including meta-analysis to get overall near-miss prevalence rates for two of the criteria.

It is our hope that the recent WHO criteria for identifying maternal near-miss cases will stimulate researchers and clinicians to carry out near-miss assessments around the world using the same criteria. We are aware of a number of ongoing studies, including the large multi-country study that WHO is implementing in 29 countries.28 To use near-miss cases as a way to improve quality of care in the facilities, WHO has developed a systematised approach to implement near-miss criterion-based clinical audits.29 For these types of audits a set of evidence-based, explicit, measurable criteria for case management are agreed that can then be used to monitor practice and determine if standards of care have been met, by reviewing case notes.30 The expected results include, among others, understanding local patterns of maternal mortality and morbidity, strengths and weaknesses in the referral system, and the use of clinical and other healthcare interventions. In addition, implementing a surveillance strategy on women with life-threatening conditions being managed at the healthcare facility can foster a culture of early identification of complications and promote better preparedness for acute morbidities. It is necessary to go beyond surveillance and implement interventions to improve the quality of maternal care, Facility-based interventions can include the implementation of evidence-based guidelines, the use of reminders, opinion leaders’ endorsement, and continued audit and feedback to achieve behavioural and process changes.29,31 It would also lead to interventions aimed at increasing the awareness of danger signs among providers at lower level facilities as well as community-based providers to minimise the delays in referrals to prevent both maternal morbidity and mortality.

Conclusions

There is growing interest in the application of the maternal near-miss concept as an adjunct to maternal mortality. However, in the literature published before 2011 there was still important variation in the criteria used to identify maternal near-misses. An organ-system dysfunction approach remains as the most epidemiologically sound set of criteria. WHO recently published criteria based on markers of management, and clinical and organ dysfunction, which is currently being adopted by researchers and organisations around the world. By using the uniform criteria, clear data on near miss can be systematically collected, thereby facilities, countries and regions can monitor the near-miss rate over time to develop better interventions to improve the quality of obstetric care. Our review also suggests that it may be beneficial to explore why emergency hysterectomy rates are higher in more recent years of data collection (an approximate 8% increase per year). Although it is well known, our results indicate that more resources are needed in low-income and middle-income countries. Comparing the rates over time and across regions, it is clear that different approaches are needed to lower the rates of near miss and that interventions must be developed with the local context in mind.

Disclosure of interests

None.

Contribution to authorship

All of the authors participated in the formulation of the methodology for this review. OT performed the literature search and reviewed all abstracts and full text articles with assistance from JPS and DC. OT wrote the first draft of the manuscript and MJH, JPS, DC and LS assisted in the writing and editing of the manuscript.

Details of ethics approval

This is a systematic review of previously published data and therefore does not require ethical approval.

Funding

None.

Acknowledgements

The views expressed in this paper are those of the authors as individuals, and do not necessarily represent the views of the WHO and its member states.

Ancillary