Bulletin of the World Health Organization

Maternal near miss and maternal death in the World Health Organization’s 2005 global survey on maternal and perinatal health

João Paulo Souza a, Jose Guilherme Cecatti b, Anibal Faundes b, Sirlei Siani Morais b, Jose Villar c, Guillermo Carroli d, Metin Gulmezoglu a, Daniel Wojdyla d, Nelly Zavaleta e, Allan Donner f, Alejandro Velazco g, Vicente Bataglia h, Eliette Valladares i, Marius Kublickas j, Arnaldo Acosta k & for the World Health Organization 2005 Global Survey on Maternal and Perinatal Health Research Group

a. Department of Reproductive Health and Research, World Health Organization, Geneva, Switzerland.
b. Department of Obstetrics and Gynaecology, University of Campinas, R Alexander Fleming 101, 13083-881, Campinas-SP, Brazil.
c. Nuffield Department of Obstetrics and Gynaecology, University of Oxford, Oxford, England.
d. Centro Rosarino de Estudios Perinatales, Rosario, Argentina.
e. Instituto de Investigación Nutricional, Lima, Peru.
f. Faculty of Medicine and Dentistry, University of Western Ontario, London, ON, Canada.
g. Hospital Docente Ginecobstétrico América Arias, Havana, Cuba.
h. Department of Obstetrics and Gynaecology, Hospital Nacional de Itauguá, Itauguá, Paraguay.
i. Universidad Nacional Autónoma de Nicaragua, León, Nicaragua.
j. Karolinska Institute, Stockholm, Sweden.
k. Department of Obstetrics and Gynaecology, Universidad Nacional de Asunción, Asunción, Paraguay.

Correspondence to Jose Guilherme Cecatti (e-mail: cecatti@unicamp.br).

(Submitted: 15 August 2008 – Revised version received: 15 January 2009 – Accepted: 07 June 2009 – Published online: 11 September 2009.)

Bulletin of the World Health Organization 2010;88:113-119. doi: 10.2471/BLT.08.057828


Approximately 15 000 women die every year in Latin America and the Caribbean of causes related to pregnancy. The maternal mortality ratio (MMR) for the region, which is around 130 maternal deaths per 100 000 live births, lies somewhere between the MMRs observed in developed and underdeveloped countries.1

When maternal deaths are infrequent, the information they provide is not generalizable and they become poor sources of information on which to base measures to improve maternal health. Thus, in this situation severe maternal morbidity, which continues to be a public health problem in Latin America, has been proposed as a proxy for maternal death.2,3

Women who survive severe complications during pregnancy, childbirth and the postpartum period could serve as surrogates to help us gain a better understanding of the set of conditions and preventable factors that together contribute to a maternal death.2 This is known as the concept of maternal near miss4 recently defined by the World Health Organization (WHO), after some controversy,5 as the near death of a woman from a complication during pregnancy, childbirth or within 42 days after the termination of pregnancy.6

Assessing maternal near misses for their values as proxies may be especially useful in Latin America, where MMRs are moderately low. In this paper we describe the occurrence of severe maternal morbidity in large hospitals in Latin America and test the usefulness of a pragmatic definition of maternal near miss for predicting maternal deaths. We also checked for associations between a maternal near miss on the one hand, and maternal factors and perinatal outcomes on the other.


The WHO Global Survey on Maternal and Perinatal Health, whose methods have been described in detail elsewhere,7,8 is a multicountry and multicentre study designed by WHO in 2004 to explore the relationship between rates of Caesarean delivery and maternal and perinatal outcomes in selected medical institutions. It has also been used to gather information on severe maternal complications in Africa, Asia and Latin America. Briefly, the WHO Global Survey was implemented through a worldwide network of health institutions that were selected by means of a stratified multistage cluster sampling design. The present study is a secondary analysis, performed in 2005, of the Latin American database. Anonymous maternal and perinatal data were collected from the hospital records of all women admitted for delivery over a period of two to three months to 120 randomly selected hospitals located in eight randomly selected Latin American countries. In each hospital data were collected over two to three months and entered in a MedSciNet AB (Stockholm, Sweden) online data system.

Definitions and outcomes

In 2005, WHO criteria for maternal near miss had not yet been defined. Nonetheless, we used the information available in the database to establish the predictive value of five factors with respect to maternal death during pregnancy, childbirth and the first week postpartum. The five factors were admission to the intensive care unit (ICU), blood transfusion, hysterectomy, eclampsia, or cardiac and renal complications. We assumed that a combination of factors indicative of severe maternal morbidity would identify those women who, having narrowly escaped maternal death, could be considered maternal near misses.

We also examined the association between maternal near miss and certain maternal and institutional characteristics, namely age, marital status, years of schooling, parity, number of antenatal visits, obesity, Caesarean section in the previous pregnancy, and type of health facility in which the delivery took place (public, social security or private). We also looked at the following perinatal outcomes: length of maternal postpartum stay, low and very low birth weight, admission of neonate to ICU, stillbirth, early neonatal death and mode of delivery.


We first tested the usefulness of clinical diagnosis and management indicators for predicting maternal death by calculating, for each indicator, its sensitivity, specificity, likelihood ratio and its 95% confidence intervals (CIs), and the ratio of cases to deaths. The association between maternal death and each indicator was considered strong if the likelihood ratio was greater than 10.9 The ratio of cases to deaths was calculated as the number of cases of a particular maternal morbid indicator for every maternal death in a woman who had the indicator, from the date of hospital admission to the seventh day postpartum. Once we had identified the indicators most strongly associated with maternal death, we combined them into a single indicator of severe maternal morbidity when at least one of them was present.

We studied the occurrence of maternal near misses in each country and derived summary estimates proportional to each country’s population size. By using simple and multiple logistic regression, we calculated crude and adjusted odds ratios (ORs) to assess the association between maternal and institutional characteristics and maternal near miss. Maternal body mass index was excluded from the analysis due to missing data.

Finally, we used a cohort approach – with maternal near miss as an exposure and maternal and perinatal outcomes as effects – to examine the crude association between maternal near miss and perinatal outcomes, expressed as risk ratios (RRs) and their 95% CIs. A logistic regression model that included all possible predictors was then developed for each outcome. We excluded 61 women for whom there was no information on life status at discharge. The mode of delivery was included in the model only as an outcome. Statistical analyses were carried out with SAS, version 9.1.3 (SAS Institute Inc., Cary, NC, USA).

The WHO Global Survey research project was approved by the national ethics committee of each country studied as well as the WHO Scientific and Ethical Review Group and the Ethics Review Committee. Some large hospitals in Mexico and Argentina and all hospitals in Brazil independently approved the protocol.


A total of 120 institutions from eight Latin American countries contributed to the 97 095 deliveries included in this study, of which 96 026 culminated in a live births. There were 25 maternal deaths in hospital, from admission to the seventh day postpartum. Table 1 shows some of the indicators used to develop a pragmatic definition of maternal near miss. Hysterectomy, ICU admission, blood transfusion, cardiac or renal complication and eclampsia yielded the highest likelihood ratios. In our sample this high-risk group, which comprised 2964 women with any of these conditions, had a combined likelihood ratio of 21.1 (95% CI: 15.7–28.4). In other words, for a woman who had died, the likelihood of having had any of these conditions was about 21 times higher than for a woman who had survived. On the other hand, of 25 women who had died, 16 had presented at least one of these conditions, and the ratio of cases to deaths was thus 2964:16, or 185:1. According to our pragmatic definition, women who survived these conditions during pregnancy or childbirth were classified as maternal near misses.

Table 2 shows estimates of maternal near misses and its components, as defined, in the Latin American countries selected. Near misses were more frequent in Brazil and Cuba and less frequent in Paraguay. Admission to the ICU during pregnancy was more common in Brazil and rare in Nicaragua and Paraguay. Blood transfusions were more frequently administered in Cuba and Mexico. For all eight countries in the sample, the mean proportional ratio of near misses was around 34 per 1000 deliveries (Fig. 1).

Fig. 1. Estimates of maternal near misses per 1000 deliveries in referral hospitals in a study of eight Latin American countries, 2005
Fig. 1. Estimates of maternal near misses per 1000 deliveries in referral hospitals in a study of eight Latin American countries, 2005

Table 3 compared the characteristics of maternal near miss cases with those of women who had no severe complications. Being > 35 years of age, not having a partner, and being a primipara or para > 3 were independently associated with the occurrence of a near miss. Social security institutions showed a stonger independent association with near misses than private institutions. Near misses were less frequent among women with less than 12 years of schooling.

Women who had undergone Caesarean section in their previous pregnancy were at increased risk of severe maternal morbidity, whether or not their current delivery was by Caesarean. The association between Caesarean section in the previous pregnancy and severe maternal morbidity remained statistically significant after adjustment for current mode of delivery (data not shown). Table 4 shows the crude and adjusted RRs for selected maternal and perinatal outcomes. In both analyses, the occurrence of a near miss was positively associated with low and very low birth weight, admission of the neonate to the ICU, stillbirth, early neonatal death and a prolonged maternal postpartum stay.


In this study, a pregnant woman admitted to the ICU, or undergoing a hysterectomy, or receiving a blood transfusion, or presenting a cardiac or renal complication, or having eclampsia was found to be at increased risk of dying during pregnancy, childbirth or in the early postpartum period. Survivors of the above-mentioned conditions were then pragmatically labelled as near miss cases.

It is not surprising that age, marital status and parity, well known predictors of maternal death, were also found to be associated with the occurrence of maternal near miss. This finding supports the use of maternal near miss as a proxy or surrogate for maternal death in assessments of maternal health interventions. Maternal near miss could be used for auditing health facilities where maternal deaths rarely occur.

Low maternal education was found to be protective against the occurrence of a maternal near miss. Furthermore, a previous Caesarean section was independently and positively associated with the occurrence of a maternal near miss. Caesarean section has been reported to increase maternal morbidity in Latin America,7,10 where women with lower education are known to undergo fewer Caesarean sections. The worldwide trend towards an increase in Caesarean section rates may be linked to iatrogenic maternal morbidity and maternal death.7,10

The association between maternal near miss and poor perinatal outcomes was expected and is very strong.11 Babies delivered to women who are near misses are smaller, require intensive care more frequently and are at higher risk of dying in the first week of life. In addition, women who are near misses have more stillbirths.


Some study limitations should be addressed. The proposed pragmatic definition of near miss included both management indicators (admission to the ICU, blood transfusion, and hysterectomy) and clinical diagnoses. The application of management criteria is influenced by the availability and use of the corresponding resources. In addition, these findings are based mostly on the data recorded in medical charts, which may not have been fully standardized. Another limitation is the small number of maternal deaths on which we based our pragmatic definition of a near miss. All these maternal deaths occurred in hospital and some women may have died after being transferred or discharged and thus been omitted. This may explain the very low MMR observed.


Despite its limitations, this study is based on what is probably the world’s largest data set to date of severe maternal complications, such as peripartum hysterectomy, blood transfusions and admission to the ICU during pregnancy and childbirth. Because most deliveries in Latin America and the Caribbean occur in health facilities, the survey results are likely to represent the state of care during childbirth in that region. In addition, the strength of the association between indicators of severe maternal morbidity and maternal death makes the concept of near miss even more relevant for efforts to improving maternal health. In the future, these findings can also be compared with those from Africa and Asia to try to gain an understanding of severe maternal morbidity and near miss worldwide.

Ideally, the definition of maternal near miss should be based on organ dysfunction.5 However, a more pragmatic approach is needed to perform macroanalyses, which are often secondary analyses of large sets of data routinely collected from health systems. This study provides further evidence to support the use of selected conditions, such as hysterectomy, ICU admission and blood transfusion, to explore the concept of maternal near miss. Furthermore, the indicators currently reported or even the stricter, more precise criteria recently proposed by WHO could be applied in future epidemiologic studies or similar surveys on maternal and perinatal health.6


We thank P Bergsjö, EO Akande, and D Oluwole, who participated during the preparatory phase of the survey and provided advice and support during its implementation; S Marthinsson for providing technical support to the online data entry system; ME Stanton and PFA Van Look for their continuous support during the survey and subsequent activities. We also thank Ana Langer, Alberto Narvaez, Archana Shah, Liana Campodonico, Mariana Romero, Sofia Reynoso, Karla Simona de Pádua and Daniel Giordano for having actively participated in activities in the countries.

Funding: The study was funded by USAID and the Department of Reproductive Health and Research, World Health Organization.

Competing interests: None declared.