Bulletin of the World Health Organization

Improved access to comprehensive emergency obstetric care and its effect on institutional maternal mortality in rural Mali

Pierre Fournier a, Alexandre Dumont b, Caroline Tourigny a, Geoffrey Dunkley c & Sékou Dramé c

a. Unité de Santé Internationale, Centre de Recherche du Centre Hospitalier de l’Université de Montréal, 3875 Saint-Urbain, Montreal, Quebec, H2W 1V1, Canada.
b. Centre Hospitalier Universitaire Ste-Justine, Université de Montréal, Montreal, Quebec, Canada.
c. Direction Régionale de la Santé, Kayes, Mali.

Correspondence to Pierre Fournier (e-mail: pierre.fournier@umontreal.ca).

(Submitted: 24 August 2007 – Revised version received: 22 April 2008 – Accepted: 24 April 2008 – Published online: 18 November 2008.)

Bulletin of the World Health Organization 2009;87:30-38. doi: 10.2471/BLT.07.047076


Maternal mortality is a major public health problem, particularly in sub-Saharan Africa, where half (50.4%) of all maternal deaths worldwide occur.1 One objective of the Millennium Development Goals is to reduce maternal mortality by 75% between 1990 and 2015.2 In 2005, the maternal mortality ratio in sub-Saharan Africa, estimated at 900 maternal deaths per 100 000 live births, was by far the highest in the world.1 Unlike other regions, sub-Saharan Africa has not seen improvements in indicators linked to maternal mortality, leading to fears that the Millennium Development targets will not be met.3 In response to this disquieting situation, many African countries have adopted measures towards reducing maternal mortality.

The context and causes of maternal mortality and morbidity are well known,4 and strategies to ameliorate them were recently reported.5 One proven effective strategy is to provide access to basic emergency obstetric services (parenteral oxytocics, antibiotics and anticonvulsants; assisted deliveries; manual extraction of the placenta; removal of retained products) and, if necessary, to comprehensive emergency obstetric services (basic services plus Caesarean sections and blood transfusions).6 Access to these services is a key element of the WHO Making Pregnancy Safer programme.7 Ensuring timely Caesarean delivery when needed is a priority in sub-Saharan Africa.

In western Africa, maternal mortality is highest in rural areas where access to emergency obstetric care is limited by large geographic distances to health facilities and scarce resources.8 While progress has been made in reducing maternal mortality rates in urban areas, the situation in rural areas is not improving. In Mali, for example, the population-based rate of Caesarean delivery in urban areas rose from 1.6% to 3.5% between 1991 and 1998, while in rural areas it remained unchanged (1.6% and 1.5%, respectively).9 Implementing emergency obstetric care programmes, and maternity referral systems in particular, is complicated in settings where resources are scarce.10 In western Africa, where the Bamako Initiative has made cost recovery in health care the standard,11 the costs of comprehensive emergency obstetric care represent a major outlay for households12,13 and several strategies have been attempted to reduce this financial burden.14,15

In 2002, the Government of Mali launched a nationwide maternity referral system16 aimed at improving the quality and accessibility of comprehensive emergency obstetric care services and at reducing the danger of death associated with obstetric complications. In this study, we aim to evaluate the effects of the system in a rural population of more than one million inhabitants.



Of 177 countries on the Human Development Index for 2005, Mali was ranked 173rd17 and had the 17th highest maternal mortality ratio1 The health region of Kayes, situated in the west of Mali, has nearly 1.7 million inhabitants unevenly distributed over 120 760 km². Rugged terrain and periods of intense rain with flooding make communications difficult. The study area consists of six of the region’s seven districts. The provincial district was excluded because it is more urban. The study area contains just over 1.25 million inhabitants, and the population density of the districts varies between 9.7 and 26.2 inhabitants per km² (mean of 14.7 inhabitants per km²).

Despite efforts in recent years, the geographic accessibility of health services remains poor: the distance from home to a primary health care centre is more than 5 km for 56% of the population and more than 15 km for 30%. The public health system, which is almost the only provider of modern health-care services, has few resources (one doctor per 28 000 and one midwife per 96 000 inhabitants). The study area has 101 community health centres (15 to 20 in each district) and six district health centres. Whereas the latter offer comprehensive emergency obstetric care, community health centres provide basic obstetric services, including assisted deliveries. When an emergency complication arises at the community health centre, the patient is referred to a district health centre. The overall utilization rate of reproductive health services in the study area resembles the national average for rural areas in Mali.18


The maternity referral system is a national programme launched in 2002 to reduce the risk of maternal death associated with obstetric complications.16 The system relies on three main components (Fig. 1). First, it seeks to improve communication and transport opportunities to eliminate delays in the delivery of emergency obstetric services. Funds from overseas donors are used to improve radio communications between community health facilities and district health services, as well as ambulance transport between them. Second, alternative funding options, including community cost-sharing schemes, are accessed to eliminate financial barriers to obstetric care. Community-funded schemes receive funds from the local government, local health services and community health associations and then reimburse health providers for all services they give to women, who contribute only a small co-payment. Third, training and equipment are provided to improve the clinical management of obstetric emergencies.

Fig. 1. Components and function of referral system for comprehensive emergency obstetric care, Mali, 2003–2006

The programme was designed in accordance with national guidelines but is implemented with adaptations for regional contexts in cooperation with local health partners. In the study area, the start-up period was between 2002 and 2005, depending on the district.

Two categories of women use the programme’s obstetric services: (i) those with obstetric complications who are referred by community health centres and have benefited from all components of the system, and (ii) those who are self-referred to the district health centre.

There are six categories of obstetric emergency, defined on the basis of the medical diagnosis or the reason for referral: haemorrhage, uterine rupture, pre-eclampsia/eclampsia, dystocic labour, infection and other (for other obstetric emergencies that cause maternal death directly, such as abortion, or indirectly, mainly malaria and anaemia).19

Study design and statistical analysis

In 2004, a system for ongoing registration of obstetric emergencies was set up in all districts of the Kayes region. During an initial pilot phase, data collection was supported and supervised by the Regional Health Authority of Kayes and the research team. The system allowed for the documentation of each patient deemed to be an obstetric emergency, and the data collected included the sociodemographic characteristics of the woman, the obstetric diagnosis and outcome, pregnancy follow-up, etc. Data were collected retrospectively for the period from 1 January 2003 to 30 June 2004, after which the data were collected prospectively until 30 November 2006. All cases with direct and indirect obstetric complications were recorded and classified in accordance with accepted standards.19

Because the maternity referral system is a national programme that all regions must implement, it was neither ethical nor practical to include a control group that would be denied access to this programme in our study design. Therefore, we used a quasi-experimental uncontrolled before-and-after study design. The main criterion to assess the effects of the intervention is the risk of death among obstetric emergency cases. To evaluate the efficacy of the maternity referral system, we considered four periods: before the intervention (P−1), the year during which the intervention was implemented (P0) and two 12-month post-intervention periods (P1 and P2). Data availability and start date of intervention by district is shown in Fig. 2.

Fig. 2. Data availability, by period and district, in referral system for comprehensive emergency obstetric care, Mali, 2003–2006
P-1, year before the intervention; PO, year of the intervention; P1, 1 year after the intervention; P2, 2 years after the intervention.

The maternity referral system was implemented on a different date in each of the six districts of the study area. As a result, the P−1 period lasted from 3 to 34 months, with a mean of 12.2 months; the P0, P1 and P2 periods all lasted 1 year, except in district 4, where P0 lasted 11 months, and for district 1, where P2 lasted 7 months. District 3 had no P1 or P2 while the study lasted since implementation of the programme was delayed because of the time it took to reach a consensus regarding community cost-sharing schemes.

We checked for group comparability for the main known risk factors for maternal death (age, cause of obstetric complications). Patients came from districts where access to health services varies greatly. Districts were classified according to the percentage of the population living within 15 km from a primary health care centre (good accessibility: > 85%; average accessibility: from 60 to 85%; poor accessibility: < 60%). We compared the risk of death among obstetric emergency cases across the different periods of the study by calculating odds ratios (ORs) and their 95% confidence intervals (CIs). To evaluate the potential effects of the intervention at different periods, adjusted ORs were calculated using various logistic regression models supported by Stata software, version 9.1 (Stata Corporation, College Station, TX, United States of America). Confounding variables included in the model were age, previous Caesarean section, diagnosis, district accessibility, Caesarean delivery and transfusion.


Table 1 presents data on obstetric activities in the study area. During the study, the rate of institutional deliveries and the number of obstetric emergencies treated in district health centres increased. The proportion of deliveries benefiting from major obstetric interventions for absolute maternal indications increased markedly: from 0.13% (95% CI: 0.10–0.16) at P−1 to 0.46% (95% CI: 0.26–0.66) at P2.

To take into account the sharp increase in the number of institutional deliveries during the study period, we performed calculations with institutional deliveries as a denominator. These calculations showed little change in the rate of obstetric emergencies treated (P−1 versus P2; 4.8% and 4.7%, respectively) while the rate of major obstetric interventions has increased (25% to 43%, respectively). Also of note is that there is little change in the percentage of major obstetric interventions performed for absolute maternal indications (55% in P−1 and 57% in P2).

During the study, 2617 obstetric emergency patients received care in district health centres and benefited from all or some components of the maternity referral system. Across the different periods, they differed significantly in terms of age, diagnosis, previous obstetric history and case management (Caesarean section), as well as in geographical access to primary care services in their district of residence (Table 2).

Table 3 shows the causes of maternal death in the different study periods. The crude case fatality rate (defined as the ratio of total deaths observed, regardless of cause among cases, to the total number of cases observed) also decreased from 10.1% to 5.13% between P−1 and P2. Nearly half the reduction in mortality could be attributed to fewer deaths from haemorrhage. Overall, the risk of death decreased, and among women with haemorrhage, the reduction in risk was nearly three-fold and was statistically significant (OR: 0.37; 95% CI: 0.17–0.79). The decrease in the risk of death from haemorrhage was greater among referred women and was also statistically significant (OR: 0.17; 95% CI: 0.04–0.68].

Table 4 presents the adjusted odds ratios between the pre-intervention, implementation and two post-intervention periods. The risk of death was reduced by about half for all women, and the reduction was statistically significant between P−1 and P1 (P = 0.027) and between P−1 and P2 (P = 0.002). The reduction was even more marked among referred women, for whom the risk of death was three times lower in P2 than in P−1 (P = 0.002). The statistical tests for trends in case fatality rates did not yield significant results.


This study had the advantage of being set within a national programme, so that the measurement of its effects and the analysis of processes allowed us to draw lessons that are directly applicable not only to Mali but also to other resource-poor countries in sub-Saharan Africa. Existing studies on the implementation of transport systems, emergency loans, community financing, communications, or various combinations of these elements2022 offer few or no empirical data on whether they reduce institutional maternal mortality. Furthermore, the usefulness of data from these studies is limited by study designs that are less than robust.

Our data show that Mali’s national maternity referral system increases the coverage of obstetric emergencies and reduces the risk of death among women delivering with obstetric complications. This risk reduction is achieved primarily in those with haemorrhage, whose prognosis is directly related to whether appropriate care is received within 2 hours or not.23 Furthermore, the reduction in risk among women with haemorrhage is distinctly more marked in those who are referred by community health centres than in those who come to the district health centre on their own. This point is particularly important because the referred women have benefited from all components of the intervention, particularly faster modes of communication and transport that reduce treatment waiting times. Thus, improved access to comprehensive emergency obstetric care and to Caesarean sections has contributed to the programme’s success.

The effects of the intervention were quick to be noted because there were considerable unmet needs in emergency obstetric care and baseline case fatality rates were very high. Over the study period, the number of women attending community health centres for normal deliveries increased progressively. Furthermore, obstetric complications became more likely to be diagnosed at an earlier stage in the labour or post-partum stages, allowing more women to benefit from the intervention sooner. Indeed, the number of obstetric emergencies and of major obstetric interventions performed in reference health centres increased dramatically during the study period. However, 2 years after implementation of the referral system, the rates of major obstetric interventions performed for absolute maternal indications remained low (0.46%) by comparison to rates reported in 2003 from a study in an urban setting in neighbouring Burkina Faso. In that study, the number of major obstetric interventions for absolute maternal indications increased significantly (from 0.75% to 1.42%) the year after implementation of a cost-sharing mechanism for emergency obstetric care.15

An analysis of the conditions in which the intervention described herein was implemented made it possible to identify its strengths and weaknesses.10,24 One of its strengths was the sustained political support it received at both the regional and national levels. Reproductive health is a major component of Mali’s national health plan, and the regional authorities have made the required investments. A further strength of the programme was its reliance on the establishment of community cost-sharing schemes, which not only reduce financial barriers to care, but also help to ensure programme sustainability by eliminating the need for financial contributions from outside the community. The last strength was the organization of the referral system and the collaboration among various partners. The programme was unique in that its general framework was defined at the national level, while regions supported the districts, which in turn designed their own local system. This method has the occasional drawback of slowing down implementation in areas where local dynamics are unfavourable; however, it generally has the advantage of producing a system that is supported by the community and local actors. In the Kayes region, implementation of these systems at the district level took 3 years (from December 2002 to November 2005). In Mali, the availability of Caesarean sections at the district level is made possible by a programme of surgical qualification for general practitioners that allows them to perform Caesarean deliveries. Neighbouring countries with greater resources have had difficulty achieving surgical training for general practitioners, resulting in reduced or non-existent access to surgical services in rural areas. The implementation of the system for ongoing registration of obstetric emergencies will yield the data necessary for continuous monitoring of the effectiveness of the maternity referral system. Data collection also serves as a stimulus and a performance incentive for those locally responsible for reproductive health services.

Areas for additional improvement include the availability and quality of basic emergency obstetric treatment at the first level of care. Most deliveries in community health centres are performed by poorly qualified birth attendants who fail to detect obstetric complications early enough. Furthermore, at district health centres more patients could be saved if the currently inadequate supplies of blood for transfusion were increased.

This study has some limitations. It had no control group or randomization and was constrained by the fact that the maternity referral system is a national programme whose effects should be evaluated under normal conditions of implementation. The absence of a control group made it impossible to control for other interventions or effects that could have modified the outcome of obstetric emergencies over time. However, the study area did not undergo any important social or economic changes during the study period. Changes to the health-care system in connection with obstetric emergencies were made within the framework of the maternity referral system being evaluated. The only notable modification to health services outside of the programme was the government’s decision to provide Caesarean sections free of charge. This decision completely eliminated the woman’s financial contribution, which had already been considerably reduced by community cost-sharing schemes associated with the intervention.

The second limitation of the study was the variability of data quality, which showed improvements over time in tandem with improvements in the monitoring system (for example, missing data for age decreased from 3.8% in P−1 to 1.1% in P2). Decreases in case fatality rates could be linked to an increase in the registration of less severe cases resulting from better case registration overall and the detection of complications at an earlier stage.

The limitations of the study were mitigated by the methods of analysis, which allowed us to control for the principal cofactors pertaining to patients’ personal characteristics and obstetric history (age, previous Caesarean section), case mix (diagnosis), case management (Caesarean section and transfusion) and geographical accessibility. Given the constraints linked to the evaluation of a non-pilot referral system, the design and methods of analysis chosen provided the most robust results possible.25

Our results show that in poor countries, programmes to reduce barriers to comprehensive emergency obstetric medical care can substantially decrease deaths associated with obstetric emergencies. Furthermore, they show that such programmes can be implemented on a large scale without major external funding. Our results were observed in a very poor country and in a region where the geographic accessibility of health-care services is among the poorest. Therefore, the system should be applicable to most rural African contexts, where most maternal deaths occur.

Furthermore, our results allow us to draw lessons not only about the intervention, but also about the research methods applied. More in-depth research should be carried out on the effects of this intervention on the population, especially related to geographical accessibility. The effects observed so quickly after this intervention were the result of the programme’s response to important unmet needs. However, efforts to satisfy such needs will be limited by geographic accessibility; any future effects of the intervention could be diminished unless patient recruitment is extended to those living far from community health centres. Studies are currently under way to model accessibility, not in terms of distance, but rather of time, taking into account the seasons and the modes of transport being used. A better understanding of the causes of the first delay in seeking services26 will also help improve access to the system.

In contexts where the rate of institutional deliveries is low, improvements in emergency obstetric care will enhance the effects of the maternity referral system on maternal mortality. However, to achieve a more comprehensive approach for monitoring deliveries27 that will ensure better maternal outcomes, the rate of institutional deliveries must be increased. In turn, this necessarily calls for improvements in service quality at all levels of the health system, some of which can derive from the successes of the maternity referral system. Our results suggest that part of the success of the intervention was linked to increased rates of institutional deliveries during the study period. ■


We thank Karim Sangaré, Aguissa Maïga, Diarrah Coulibaly, Odette Laplante and Sylvie Charron for their work in the implementation and follow-up of the intervention and their input into the research activities. We also thank Maria-Victoria Zunzunegui and Anna Kone for their important statistical advice, and the district teams, without whose help this work could not have been accomplished. We also thank the reviewers for their useful comments towards improving the paper.

Funding: The research was supported by the Governance, Equity and Health Initiative of the International Development Research Centre in Ottawa, Ontario, Canada. The Regional Health Authority of Kayes receives support from the Canadian International Development Agency.

Competing interests: None declared.