Maternal mortality: near-miss events in middle-income countries, a systematic review

Abstract Objective To describe the incidence and main causes of maternal near-miss events in middle-income countries using the World Health Organization’s (WHO) maternal near-miss tool and to evaluate its applicability in these settings. Methods We did a systematic review of studies on maternal near misses in middle-income countries published over 2009–2020. We extracted data on number of live births, number of maternal near misses, major causes of maternal near miss and most frequent organ dysfunction. We extracted, or calculated, the maternal near-miss ratio, maternal mortality ratio and mortality index. We also noted descriptions of researchers’ experiences and modifications of the WHO tool for local use. Findings We included 69 studies from 26 countries (12 lower-middle- and 14 upper-middle-income countries). Studies reported a total of 50 552 maternal near misses out of 10 450 482 live births. Median number of cases of maternal near miss per 1000 live births was 15.9 (interquartile range, IQR: 8.9–34.7) in lower-middle- and 7.8 (IQR: 5.0–9.6) in upper-middle-income countries, with considerable variation between and within countries. The most frequent causes of near miss were obstetric haemorrhage in 19/40 studies in lower-middle-income countries and hypertensive disorders in 15/29 studies in upper-middle-income countries. Around half the studies recommended adaptations to the laboratory and management criteria to avoid underestimation of cases of near miss, as well as clearer guidance to avoid different interpretations of the tool. Conclusion In several countries, adaptations of the WHO near-miss tool to the local context were suggested, possibly hampering international comparisons, but facilitating locally relevant audits to learn lessons.


Introduction
Women are at risk of developing severe morbidity and mortality during pregnancy, childbirth and postpartum, especially in low-income and middle-income countries where 99% of all maternal deaths occur. 1 Improvement of maternal health is urgently needed and one of the sustainable development goals is to reduce the global maternal mortality ratio to less than 70 per 100 000 live births by 2030. 2 In addition to maternal mortality, severe maternal morbidity is used as an indicator of quality of maternity care. 3,4 Measuring and comparing outcomes of severe maternal morbidity studies have been difficult because of the use of different identification criteria. 5,6 In 2009, the World Health Organization (WHO) developed the maternal near-miss tool to introduce a universal approach to comparing the quality of maternity care between different countries. [6][7][8][9][10] Maternal near miss is defined by WHO as "a woman who nearly died but survived a complication that occurred during pregnancy, childbirth or within 42 days of termination of pregnancy." [6][7][8][9][10] Maternal near miss occurs more frequently than maternal death and by evaluating the condition, more robust lessons may be learnt about quality of care. 5,6 Several studies, however, have demonstrated difficulties in applying the tool. [11][12][13] Box 1 shows the WHO maternal nearmiss criteria for determining life-threatening conditions and additional criteria for baseline assessment of quality of care. Among the requirements to meet the various criteria of the tool are: advanced laboratory diagnostic tests; large numbers of units of blood in transfusion as the threshold to identify severe haemorrhage; and intensive clinical monitoring. Some of these requirements cannot easily be met in low-resource settings due to limited diagnostic capacity and reduced options for medical intervention in these settings, which may lead to underestimation of the incidence of maternal near miss. 13 Researchers in sub-Saharan Africa have suggested adaptations of the maternal near-miss tool for use in low-income countries. 14,15 But even in high-income countries, where sufficient resources should be available, there has been discussion about what the appropriate inclusion criteria for maternal near miss should be. 16 Identification of maternal near miss was found to be compromised by incomplete documentation in the medical records to establish whether maternal near-miss criteria were met.
Reports about the incidence of maternal near miss have been published for several high-and low-income countries, and the applicability of the WHO maternal near-miss tool has been evaluated in several of these. However, data are lacking about maternal near miss in middle-income countries. We therefore made a systematic review of the incidence and main causes of maternal near miss in middle-income countries. We also aimed to evaluate qualitative findings documented by researchers with regard to applicability of the tool and suggest possible adaptations of the WHO maternal near-miss approach for middle-income settings.

Methods
We conducted the review according to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses guideline, 17 and registered with the International Prospective Register of Systematic Reviews (CRD42021232735).

Study selection
We performed a search of online databases for articles on maternal near miss in middle-income countries published between 1 January 2009 and 12 November 2020 without language restrictions. The earlier date was chosen since 2009 is the year when the WHO maternal near-miss approach was first published. 6,9 Retrospective studies that used data from before 2009 were included only if they made use of the WHO definition for maternal near miss.
We used the keywords "severe acute maternal morbidity," "maternal near miss" and "middle income country." Since PubMed® does not provide medical subject headings terms for country income groups, we first determined which countries were classified as middleincome and inserted each country name as a separate term in the search strategy. The search was last run in November 2020 in the online databases PubMed®, Embase®, Web of Science, Cochrane Library, Emcare and Academic Search Premier. In addition, we searched the following regional databases: Index Medicus for the Eastern Mediterranean Region; Index Medicus for South-East Asia Region, Latin America and the Caribbean; African Index Medicus; Ind-MED; and Global Health Library. More details of the search strategy are in the data repository. 18 We included studies that met all four inclusion criteria: (i) articles about maternal near miss as defined by WHO; (ii) data on the incidence of maternal near miss per 1000 live births and the main causes; (iii) describing countries meeting the World Bank classification for middle income; 19,20 and (iv) reporting the specific criteria used to identify maternal near miss and experiences with applying the WHO maternal near-miss criteria, including possible modifications of the WHO maternal near-miss tool for local use. We included studies containing multiple countries only if outcomes per country could not be found elsewhere. If multiple studies published data on the same country, all of them were reviewed and included. We used the World Bank classifications by gross national income per capita to determine country income groups. 19,20 As the classification of several countries changed over the search dates, we included studies if countries were middle-income in the year of publication, as classified by the World Bank at that time.
We excluded studies that: (i) did not apply WHO maternal near miss definitions; (ii) only focused on one specific disease or risk factor without providing overall data on maternal near miss; (iii) were comments, abstracts, secondary analysis or surveys of existing studies; (iv) only focused on neonatal outcomes; or (v) only described the process of health care or methods of identifying maternal near miss without providing incidence or most frequent causes, and without providing qualitative findings with regard to applicability and adaptations of the tool.
Two independent researchers screened all citations initially for relevance based on title and abstract and selected studies for inclusion after reading the full-text papers. Disagreements were resolved in a discussion between these two reviewers to reach a consensus. In case no consensus could be reached, the reviewers consulted a third researcher to reach an agreement on inclusion of articles.

Data extraction
We extracted data on the number of live births, number of cases of maternal near miss and number of maternal deaths. Where available, we noted the following indicators: maternal near-miss ratio (number of cases of maternal near miss per 1000 live births), maternal mortality ratio (number of maternal deaths per 100 000 live births), ratio of maternal near miss to maternal death (number of cases of maternal near miss ÷ the number We also extracted data on the most frequent organ dysfunction and the most frequent cause of maternal near miss. When studies included qualitative comments on the methods of using the WHO maternal near-miss approach, we noted any modifications to the WHO tool applied in the studies and any problems reported by the study researchers. When articles described the use of multiple methods to identify maternal near miss, we only reported data concerning use of the WHO maternal near-miss tool.

Data analysis
We subdivided the countries for analysis into lower-middle income and uppermiddle income according to the World Bank categories. 19,20 We report the number of studies and the frequency of causes of near miss as numbers and percentages. We calculated the median values and interquartile range (IQR) of the maternal indicators if the data were not normally distributed. We performed statistical analysis using SPSS version 24.0 (IBM Corp., Armonk, United States of America).
We estimated risk of bias in individual studies by quality assessment of studies. Studies were considered to be of acceptable quality if: (i) there was a clear description of the study population with a minimum of 100 live births over a period of at least 3 months; (ii) new cases of maternal near miss were identified in daily audits or rounds by trained medical staff; and (iii) the setting was an entire hospital rather than only one intensive care unit. The two reviewers who selected the studies did the quality assessment. We amended the Newcastle-Ottawa scale 21 for this study by coding the item Selection of the non-exposed cohort as not applicable (NA). The maximum quality score was therefore 8 instead of the original score 9 in the Newcastle-Ottawa Scale; more details are in the data repository. 18

Ethical approval
Ethical approvals were obtained from the Health Research Ethics Committee (HREC), Faculty of Health Sciences, Stellenbosch University, on 3 October 2018 (Project ID: 1427, HREC Reference #: S18/02/023) and from the Provincial Health Authority, the chief executive officer of Tygerberg Hospital and the heads of respective departments.

Results
The search resulted in 996 records. After removal of duplicates, we screened 973 articles based on title and abstract, after which 138 articles were retrieved for fulltext evaluation. Of these, we excluded 76 articles (39 of which did not apply the WHO maternal near-miss tool; Fig. 1). For the final review we included 62 articles.  Our quality assessment of the articles showed the following scores: eight articles with score 4; 15 articles with score 5; 26 articles with score 6 and 13 articles with score 7. No articles described possible missing data in the follow-up period which resulted in none of the articles having a maximum score of 8.
The included articles reported data from 69 studies in 26 countries (12 lower-middle-income countries and 14 upper-middle-income countries). Two of the articles 30,83 presented data on multiple countries. Of the 69 studies, 40 (58%) were done in lower-middleincome countries and 29 (42%) in upper-middle-income countries. Half (35 studies) of them, were conducted in one or more tertiary health-care facility. General descriptions of the studies and differences in methods are summarized in Table 1 (available at: https:// www .who .int/ publications/ journals/ bulletin/ ). Four retrospective studies described data from before 2009 using the WHO definition for maternal near miss. [23][24][25][26]

Incidence
The incidence and causes of maternal near miss in middle-income countries are presented in Table 2. The studies reported a total of 50 552 maternal near misses out of the total live births of 10 450 482. Overall, the median maternal near-miss ratio in these middle-income countries was 9.6 per 1000 live births (IQR: 7.0-23.3). In lower-middle-income countries the median maternal near-miss ratio was 15.9 per 1000 live births (IQR: 8.9-34.7), ranging from 4.0 in an Indian government tertiary care centre 27 to 198.0 in a private tertiary care centre in Nigeria. 28 For upper-middle-income countries, the median maternal near-miss ratio was 7.8 per 1000 live births (IQR: 5.0-9.6), ranging from 2.2 in two Malaysian tertiary hospitals 29 to 54.8 in Brazil. 34 Studies reported a total of 2917 maternal deaths. The median maternal mortality ratio for all middle-income countries was 163 per 100 000 live births (IQR: 52-367), with a median of 306 per 100 000 live births (IQR: 162-666) in lower-middle-income countries versus 62 per 100 000 live births (IQR: 9-105)

Causes
Hypertensive disorders of pregnancy and obstetric haemorrhage were the commonest causes of maternal near miss. In the lower-middle-income countries, the most frequent cause of near misses was haemorrhage (including reported severe postpartum haemorrhage, obstetric haemorrhage, postpartum haemorrhage, haemorrhage and placenta praevia), reported in 18 out of 40 studies (45%) from 10 countries. Hypertensive disorders of pregnancy (including severe preeclampsia and eclampsia) were the cause of near miss in 15 studies (38%) from four countries. In the upper-middle-income countries, hypertensive disorders of pregnancy were the commonest cause of maternal near miss in 15 out of 29 studies (52%) from six countries. Obstetric haemorrhage was reported as the commonest cause in eight studies (29%) from seven countries. In both lower-middleand upper-middle-income countries, the main identified organ failure was coagulation or haematological dysfunction (which included haemorrhage with a minimum of 5 units of blood for transfusion and a platelet count < 50 000 platelets/mL). Cardiovascular organ dysfunction (shock, cardiac arrest) was the second most common organ failure.

Adaptations
Adaptations to the maternal near-miss tool were suggested in 33 out of 69 (48%) studies. These modifications and difficulties in applying the WHO maternal near-miss tool are described in Table 3. Seven studies recommended reducing the threshold for defining major haemorrhage from 5 units of blood required for transfusion to 4 units, 38,39 3 units 30,40,41 or even 2 units, 22,42 to account for limited availability of blood. Other additions to the maternal near-miss tool suggested by researchers were: a definition of shock and sepsis (obstetric and non-obstetric); estimation of blood loss; bedside clotting time; severe anaemia; use of vasoactive drugs; assessing keto-acids in urine; and application of an oxygen face mask. In five studies, researchers recommended inclusion of admission to an intensive care unit as a criterion. 32,34,40,41,43 Moreover, additional diagnoses to the current six life-threatening conditions criteria were advised, such as: placental abruption; medical and surgical disorders; diabetic keto-acidosis; acute collapse or thromboembolism; and non-pregnancyrelated infections. 37,38,44,45 Some studies reported problems with applying the tool, including underestimation of maternal near miss by using only criteria based on organ dysfunction; 35,84 and difficulties with identifying women with near miss because the necessary equipment and facilities were unavailabile 14 or due to time pressure in clinical emergencies. 36 Researchers also reported that difficulties with categorization of the WHO maternal near-miss criteria and different interpretations of the tool would make comparisons problematic. 37

Discussion
The WHO maternal near-miss tool facilitated evaluation of the maternal near-miss ratio in 26 middle-income countries. The main reported causes of maternal near miss were hypertensive disorders in pregnancy and obstetric haemorrhage. The maternal near-miss ratios were considerably higher in lowermiddle-than upper-middle-income countries (median: 15.9 versus 7.8 per 1000 live births). This finding is not unexpected due to differences in countries' resources, but is an important finding about the validity of the maternal nearmiss approach. Lower-middle-income countries also had considerably higher maternal mortality ratios and mortality indices than upper-middle-income countries.
The median maternal near-miss ratios per 1000 live births in middleincome countries in our study were higher than those in previous studies of high-income countries (for example, 1.8 in Ireland and 2.0 in Italy) 85,86 and lower than those in low-income countries (for example, 17.0 in Ethiopia, 88.6 in Somalia and 23.6 in United Republic of Tanzania). 15,87,88 These differences might in part be explained by differences in quality of care, reflected by the mortality index, where the higher the index, the more women with life-threatening conditions die. Comparisons of maternal near-miss ratios and sharing lessons learnt from audits in different regions or countries might benefit maternal health worldwide.
Monitoring maternal near misses and maternal deaths showed differences not only among middle-income countries but also across different settings of the same countries. Differences between rich and poor or urban versus rural populations are often large in middleincome countries. Outcomes will differ depending on the quality of care and socioeconomic circumstances in different regions. 19,20 Adaptations to the WHO maternal near-miss tool have previously been considered for high-and low-income countries. [14][15][16] We found that various adaptations of the WHO tool were also suggested by researchers in middleincome countries, depending on the setting. Adaptation of the tool hampers comparisons across different settings, but may sometimes be necessary to prevent under-reporting of severe morbidity. Several of the included studies recommended reducing the threshold for defining major haemorrhage, or making additions to the WHO criteria. Researchers in our study mentioned the limitations of under-reporting maternal near miss using the current WHO criteria based on organ dysfunction. These limitations, however, have also been reported in both low-and high-income countries. 11,13,15,16 While some studies limited the organdysfunction criteria only to life-threatening conditions, other studies added up to six diagnoses of severe maternal complications or critical interventions from the list of WHO criteria in Box 1. Moreover, in the original search, we had to exclude 39 studies applying different criteria that were too far from the original WHO criteria and seven studies whose criteria were unclear.
The issues mentioned above show that the maternal near-miss tool is helpful in recognizing severe morbidity, but may benefit from adaptations to be locally applicable. The major aim of the tool is that lessons for clinical care are drawn. Only including cases of maternal near miss that occur in tertiary level hospitals does not provide a comprehensive picture of maternal near miss in a country. Especially in middle-income countries, differences in quality of care in facilities are large between richer and poorer populations, those living in urban versus rural areas and those using public versus private facilities. 89 The WHO criteria can be seen as a package of minimum criteria that should be in place to provide appropriate care. These minimum criteria Systematic reviews Maternal near miss in middle-income countries Anke Heitkamp et al. Conclusions about maternal near miss are dependent on the quality of data and challenges to this should be acknowledged. Researchers recommended adhering to the WHO criteria (adjusted to specific settings as needed) to enable meaningful comparison between similar reference populations. Jayaratnam et al., 2019 71 Timor-Leste Not modified Determining a clear diagnosis in a woman with maternal near miss is difficult due to presence of multiple symptoms, lack of diagnostics due to fast deterioration of the woman and lack of laboratory-based markers. Researchers concluded that maternal near-miss criteria must be modified to the local context to enhance incorporation of cases (e.g. requiring lower transfusion requirements) in future studies.

Author Setting Modifications applied in study Comments and problems reported by study researchers
Oppong et al., 2019 47 Ghana Addition to definition of coagulation in organ dysfunction criteria (bedside clotting time of > 7 mins) Organ system-based criteria are regarded as the most specific means of identifying maternal near miss. However, researchers argued that these criteria require ready availability of laboratory tests and medical technologies, thus impeding their use in many low-resource local settings. Owolabi

Systematic reviews
Maternal near miss in middle-income countries Anke Heitkamp et al. Norhayati et al., 2016 29 Malaysia NA Researchers noted that use of the WHO criteria was limited in smaller health facilities. Laboratory-based markers (e.g. pH, PaO 2 , lactate) and management-based markers (e.g. vasoactive drugs and hysterectomy) were less likely to be applicable in these health facilities. Akrawi

Brazil, Alagoas
Researchers noted that intensive care unit admission was not included in the WHO criteria but was an important marker of maternal severity in their study (identified in 94.5% of pregnant women) Researchers noted that, in contrast to laboratory and management criteria, clinical criteria are important for low-income regions, because no complex laboratory and hospital infrastructures are required. Limitations of laboratory and management criteria are that most of these criteria require high-complexity units, wards, equipment or facilities for their use. Women experiencing near miss may therefore be missed. Lowering the numbers of packed red blood cell units or including disease-based criteria was necessary in low-resource settings to classify women as near miss. Mu et al., 2019 79 China NA Lack of high-quality medical institutions in rural areas is a problem for maternal health. In recent years, China has strengthened management of women with severe complications so that they must give birth in tertiary hospitals. The researchers argued that the lack of tertiary hospitals in rural areas will affect accessibility of pregnant women to high-quality health care. Heemelaar  A limitation of our study is that small differences in methods of identification of maternal near miss between countries could result in major differences in outcomes. Moreover, we had to exclude a considerable proportion of studies that used different criteria to identify maternal near miss. This underlines the complexity of the challenge when aiming to compare maternal near miss across different countries and settings. An additional list of diagnoses would be a valuable contribution to reflect actual health problems in different settings. 37,38,44,45 This issue was also discussed in a study published by our team after this search in 2021. 90 Our search was performed without any language restriction and in large databases, but it is still possible that the search may have missed studies.
A strength of our study was the relatively large number of publications that allowed us to obtain a comprehensive overview of maternal near miss in middle-income countries and to make robust comparisons between different regions and countries. We only report data about maternal near miss from 26 of the world's 105 middle-income countries. We excluded some studies of near miss from our review because they used different criteria from the WHO near-miss criteria or did not clearly report the criteria used. Nevertheless, the countries analysed here reported large numbers of live births as denominator populations, providing a relatively robust and comprehensive overview of maternal near-miss ratios.
We found multiple studies for Brazil and India, with India showing a particularly broad range of outcomes. These data for India reflect the large differences within this large country, indicating that smaller studies might not be representative for the entire territory. [31][32][33][34] We conclude that instead of adapting the WHO maternal near-miss tool, the foremost important aim of the tool should be to improve the quality of maternity care from lessons learnt by performing audits of cases of maternal near miss. ■

Acknowledgement
We thank JW Schoones, Leiden University Medical Center, the Netherlands. Omission of items from laboratory criteria (glucose and keto-acids in urine) Lowering the threshold for use of blood products to 3 units of blood NA NA: not applicable; PaO 2 : oxygen arterial pressure; PaO 2 /FiO 2 : ratio of arterial oxygen partial pressure to fractional inspired oxygen; WHO: World Health Organization. a Severe pre-eclampsia (blood pressure of 170/110 mmHg measured twice); proteinuria of 5 g or more in 24 hours; and HELLP syndrome (haemolysis, elevated liver enzymes and low platelets) or pulmonary oedema or jaundice or eclampsia (generalized fits without previous history of epilepsy) or uncontrollable fits due to any other reason. b Sepsis or severe systemic infection, fever (> 38 °C), confirmed or suspected infection (e.g. chorioamnionitis, septic abortion, endometritis, pneumonia), and at least one of the following: heart rate > 90 beats per minute, respiration rate > 20 breaths per minute, leukopenia (white blood cells < 4000/μL), leukocytosis (white blood cells > 12 000/μL). c See the supplementary files of the original article for the complete list. 61 d Anaemia was defined by the researchers as haemoglobin level of < 60 g/L or clinical signs of severe anaemia without acute haemorrhage. e Abnormal or difficult childbirth or labour for more than 24 hours. f Low haemoglobin level (< 6 g/dL) or clinical signs of severe anaemia in women without severe haemorrhage. Note: See Box 1 for the WHO inclusion criteria.