Full Terms & Conditions of access and use can be found athttp://www.tandfonline.com/action/journalInformation?journalCode=ijog20

Download by: [Ghent University Library] Date: 24 April 2017, At: 23:13

Journal of Obstetrics and Gynaecology

ISSN: 0144-3615 (Print) 1364-6893 (Online) Journal homepage: http://www.tandfonline.com/loi/ijog20

The magnitude and factors related to facility-based maternal mortality in Mozambique

Leonardo Chavane, Martinho Dgedge, Olivier Degomme, Osvaldo Loquiha,Marc Aerts & Marleen Temmerman

To cite this article: Leonardo Chavane, Martinho Dgedge, Olivier Degomme, Osvaldo Loquiha,Marc Aerts & Marleen Temmerman (2017) The magnitude and factors related to facility-basedmaternal mortality in Mozambique, Journal of Obstetrics and Gynaecology, 37:4, 464-470, DOI:10.1080/01443615.2016.1256968

To link to this article: http://dx.doi.org/10.1080/01443615.2016.1256968

Published online: 07 Feb 2017.

Submit your article to this journal

Article views: 38

View related articles

View Crossmark data

ORIGINAL ARTICLE

The magnitude and factors related to facility-based maternal mortalityin Mozambique

Leonardo Chavanea , Martinho Dgedgeb, Olivier Degommec, Osvaldo Loquihad, Marc Aertse andMarleen Temmermanc

aJhpiego – Mozambique, Maputo, Mozambique; bFaculty of Medicine, Eduardo Mondlane University, Maputo, Mozambique; cInternationalCenter for Reproductive Health, Ghent University, Ghent, Belgium; dDepartment of Mathematics and Informatics, Eduardo MondlaneUniversity, Maputo, Mozambique; eInteruniversity Institute for Biostatistics and Statistical Bioinformatics, (I – Bisotat) Hasselt University,Hasselt, Belgium

ABSTRACTFacility-based maternal mortality remains an important public health problem in Mozambique. A num-ber of factors associated with health system functioning can be described behind the occurrence ofthese deaths. This paper aimed to evaluate the magnitude of the health facility-based maternal mortal-ity, its geographical distribution and to assess the health facility factors implicated in the occurrence ofthese deaths. A secondary analysis was done on data from the survey on maternal health needs per-formed by the Ministry of Health of Mozambique in 2008. During the study period 2.198 maternaldeaths occurred out of 312.537 deliveries. According to the applied model the availability of Maternaland Child Health (MCH) nurses performing Emergency Obstetric Care functions was related to thereduction of facility-based maternal mortality by 40%. No significant effects were observed for the avail-ability of medical doctors, surgical technicians and critical delivery room equipment.

IMPACT STATEMENT� Is largely known that the availability of skilled attendants assisting every delivery and providing

Emergency Obstetric Care services during the pregnancy, labor and Childbirth is key for maternalmortality reduction.

� This study add the differentiation on the impact of different cadres of health services providersworking on maternal and child health services on the facility based maternal mortality. In this set-ting the study proven the high impact of the midlevel skilled maternal and child health nurses onthe reduction of maternal mortality. Another important add from this study is the use of facilitybased maternal mortality data to inform the management process of maternal healthcare services.

� The findings from this study have potential to impact on the decision of staffing prioritization insetting like the study setting. The findings support the policy choice to improve the availability ofmaternal and child health nurses.

KEYWORDSMaternal mortality; healthfacility-based; geographicaldistribution

Introduction

Maternal death is the death of a woman while pregnant orwithin 42 days of the termination of pregnancy, irrespectiveof duration and site of the pregnancy from any cause relatedto or aggravated by the pregnancy or its management (ICD-10). Although maternal mortality at the global level hasdecreased by 45% between 1990 and 2013, it continues tobe a large burden and a cause of concern worldwide, particu-larly among the least developed countries.

Data obtained from the 2011 Demographic and HealthSurvey (DHS) estimated the maternal mortality ratio as 408per 100 thousands live births in Mozambique (INE and MISAU2011). In developing countries, the measurement of maternalmortality presents specific challenges because maternaldeaths are not only rare but also difficult to identify (Hillet al. 2006). Accurate estimates of the number of maternaldeaths in both the community and facility levels are highly

important to inform policymakers and reproductive healthprogramme implementers (Qomariyah et al. 2009). On thebasis of the data obtained from the 2007 general populationcensus, it was estimated that 46% of all maternal deaths inMozambique occur within a health facility (Sa�ude et al. 2009).Factors related to health system organisation and the provi-sion of health services have been identified elsewhere to playan important role in maternal deaths within the health facili-ties (Kongnyuy et al. 2009; Cavallaro and Marchant 2013;Knight et al. 2013).

In 2008, the Ministry of Health of Mozambique (MoH) per-formed a nation-wide maternal health care needs assessment,which included an evaluation of facility-based maternaldeaths. In this study, we aimed to perform a secondary ana-lysis of this maternal health care needs assessment. We eval-uated the magnitude of facility-based maternal deaths, itsgeographical distribution, and health facility influencing fac-tors for maternal death.

CONTACT Leonardo Chavane leochavane@gmail.com Jhpiego – Mozambique, Maputo, Mozambique� 2017 Informa UK Limited, trading as Taylor & Francis Group

JOURNAL OF OBSTETRICS AND GYNAECOLOGY, 2017VOL. 37, NO. 4, 464–470http://dx.doi.org/10.1080/01443615.2016.1256968

Context

In 2007, Mozambique had an estimated population of �20.2million, with a total fertility rate of 5.8 children per womanand an infant mortality rate of 118 per 1000 live births (INE2009). In addition, 54.6% of all births occur in public healthservices, while the private sector accounts for 0.2% of birthsand the remaining 45.2% occurs at the community level(INE and MISAU 2011).

The MCH nurses are the main provider of maternal andchild health services at the health facility in the country. Theyare trained to provide emergency obstetric care (EmOC) serv-ices and if needed, they refer patients to medical doctors(General Practitioner and Obstetrician and Gynaecologists) orto the surgical technician known as ‘Tecnico de Cirurgia’, amid-level non-physician clinician trained to perform emer-gency surgical interventions (Pereira et al. 2007). In 2013, theratio of MCH nurse per population was estimated to be0.32/1000 inhabitants, which is below the minimum of1 nurse/1000 inhabitants as recommended by the WHO(Dgedge et al. 2014).

Materials and methods

The survey on maternal health needs included 427 out of atotal of 1277 health facilities in the country. Facilities withless than 1 delivery in the previous year were excluded. All ofthe major facilities, Central, General, Provincial and DistrictHospitals and the Health Centre Type I were included in thesample: these facilities were attributed to a probability ofselection factor p¼ 1 and the Health Centres Type II (a per-ipheral and smallest health centre) factor p¼ .4.

Core questionnaires from the Averting Maternal Death andDisability needs assessment toolkit (AMDD 2014) were adaptedto the country’s context. For this study, we selected the mod-ules related to: the general information on the health facility,including the type of facility, its geographical location and thenumber of maternity beds; the availability of material, medica-tion and equipment for the provision of Emergency ObstetricalCare (EmOC); and the availability of human resources of mater-nal health care service and general statistics on Maternal health.

A group of health care providers were selected andtrained for one week in data collection. Fieldwork was con-ducted from November 2007 to January 2008. An electronicdatabase was designed using SPSS version 17 (SPSS Inc.,Chicago, IL) and used for both data entry and analysis. Theanalysis included frequencies, cross tabulation of the percen-tages, means and medians. As a result, a report was pro-duced to inform the MoH.

A secondary analysis was performed to examine the factorsrelated to health facility-based maternal deaths. Maternal deathwas considered to be the primary outcome and grouped intodifferent potential influencing factors as the following:

Characteristics of health facility: type of facility according to thenational classification of health facilities and size based on thenumber of maternity beds.

Availability of human resources and provision of signal functionsof EmOC: The availability of key staff (Medical Doctors, MCH Nursesand Surgical Technician) for the provision of EmOC signalsfunctions (Box 1) was measured. If a health care provider did notprovide or prescribe one of the interventions according to thenational norms for his category, then he was classified as non-providing EmOC signal function interventions. According to thenational norms, the MCH nurse must offer Basic EmergencyObstetric Care (BEmOC) and the Medical Doctor, and the SurgicalTechnician must provide the Comprehensive Emergency ObstetricCare (CEmOC).

The workload was measured as a proxy of the number ofdeliveries per MCH nurse per day. This was classified as low if thehealth facility had less than three deliveries per MCH nurse perday, medium if the health facility had between three and fivedeliveries, and high if there were more than five deliveries perMCH nurse per day.

Availability of essential drugs for EmOC

A list of essential medicines for EmOC was formed from thenational list of essential medicines in use in the country. Asillustrated in Box 2, six categories of drugs were consideredfor evaluation. The criteria for evaluation of the availability ofmedicines were set as the following: for antibiotics, at leasttwo antibiotics should be available; magnesium sulphatestands alone; anti-hypertensive drugs, at least one should beavailable; intravenous solutions, at least one should be avail-able; quinine stands alone and uterotonic drugs, at least oneshould be available.

The medicine availability was then classified as bad if onlyzero to two types of medicines were available, regular if threeto four types of medicines were available, good if five medi-cines were available, and very good if all six types of drugswere available in the health facility.

The availability of delivery kits, resuscitator and aspiratorwas used as a proxy to measure the availability of criticalequipment and equipment for EmOC provision. It was clas-sified as bad if only the delivery kit or some componentsof the kit were available, regular if the delivery kit andresuscitator were available and good if the delivery kit,resuscitator and aspirator were all available in thematernity.

Box 1. List of interventions according to the EmOC signal functions.

1. Administer parental antibiotics2. Administer uterotonic drugs3. Administer parental anticonvulsants4. Manually remove the placenta5. Remove retained products6. Perform assisted vaginal delivery7. Perform surgery8. Perform blood transfusion

Box 2. List of essential medicines for EmOC.

Essential drugs for EmOC CategoriesAmpicillinþ PenicillinþGentamicinþ

Metronidazolþ Cloranfenicolþ CefalosporineAntibiotics

Magnesium Sulphate Anti-convulsantDihidralizineþNifedipine Anti-hypertensivesDextroseþ Saline Solutionþ Ringer Lactate Intravenous SolutionsQuinine Anti-malarialsMetylergometrineþOxytocin Uterotonics

JOURNAL OF OBSTETRICS AND GYNAECOLOGY 465

Modelling

A zero-inflated Poisson model was used to assess the influenceof different factors on the mortality. When the quantified dataexhibited an excess number of zeros relative to the Poissondistribution, an extension of the ordinary Poisson model, suchas the Zero-inflated Poisson or Zero-inflated Negative Binomialhas been extensively discussed in the literature (Dankmaret al. 1995; Bohning 1998; Famoye and Singh 2006). If the dataare correlated by clustering or longitudinal designs, then hier-archical or mixed effects versions of the above models provedto accommodate the sources of heterogeneity present in thedata (Hall 2000; Hall and Zhang 2004). A zero-inflated modelassumes that for each observation, there are two potentialdata-generating processes with different probabilities: onegenerates the zero and the other generates the Poisson orNegative binomial counts. A Bernoulli model is used to deter-mine which of the two processes was used.

We accounted for excess zeros in the distribution ofmaternal deaths via health centres and clustering over dis-tricts as suggested in Loquiha et al. (2013). To control for thesize effect and type of services offered, the facilities were div-ided into two groups, Hospitals and Health Centres.

This model was then extended with the inclusion of ran-dom effects to accommodate the district-specific variation.The facility-based maternal mortality ratio was estimated inan indirect manner for each district of the country. The crudematernal mortality ratio estimates provided by the modelwere plotted on the country map. Before starting the study,

approval was received from the National Ethical Committeefor Health of Mozambique.

Results

The survey identified 312.537 deliveries and 2.198 maternaldeaths from 1 November 2006 to 31 October 2007. Table 1summarises the results of the maternal health care servicescollected in the health facilities. This table presents the infor-mation accumulated by the type of health facilities.

Figure 1 shows the maternal mortality estimated by thetype of health facility. The central hospitals had the highestratio with 1.701 maternal deaths per 100.000 deliveries. We

Table 1. Abstract of maternal health indicators collected in the health facilities surveyed.

Type of health facilities

Centralhospital

General/provincialhospital

Districthospital Health centre Total

Total of health facilities 3 13 26 385 427Total obstetric admissions 30,736 67,876 52,805 211,241 362,658Total Deliveries 21,983 52,883 47,668 190,003 312,537Maternal deaths 374 824 404 596 2198Maternity beds 443 559 631 2245 3878Number of MCH nurses working in maternity 131 117 140 709 1097Number of Medical doctors working in maternity 42 31 27 48 148Number of Surgeon assistant working in maternity 4 4 18 3 29Medical doctor available and providing EmOC signal functions

No 0 1 1 331 333Yes 3 12 25 54 94

MCH nurse available and providing EmOC signal functionsNo 0 0 0 95 95Yes 3 13 25 279 320

Surgeon assistant available and providing EmOC signal functionsNo 2 5 5 372 384Yes 1 8 20 2 31

Availability of medicines (grouped)Bad 1 3 8 203 215Regular 1 1 7 93 102Good 0 7 8 36 51Very good 1 2 3 15 21

Delivery room equipmentVery bad 1 0 2 67 70Bad 1 2 4 109 116Moderate 1 8 15 93 117Good 0 3 4 32 39

Number of deliveries per MCH nurses (grouped)Low 0 0 0 117 117Moderate 0 0 0 0 0High 3 9 22 190 224

1701

1558

848

314

703

0 500 1000 1500 2000

Central Hospitals

Gen./Prov Hospitals

District Hospitals

Health Centers

Total

Figure 1. Facility based maternal mortality ratio by the type of health facility.This ratio is defined as the number of maternal deaths by total deliveries multi-plied by 100.000 (Deliveries as proxy of live births).

466 L. CHAVANE ET AL.

estimated a national health facility-based maternal mortalityof 703 deaths per 100 thousands deliveries.

Figure 2 presents the distribution of the estimated facility-based maternal deaths by district. This map shows a concen-tration of districts with high facility-based maternal mortalityin the southern and central regions of the country comparedto the northern region.

In order to control the effect of the size of the health facil-ity and the type of services and resources existing in the

health facilities, we did aggregate and analyse hospitals onone side and Health Centres and Health Posts on the otherside. Table 2 illustrates the distribution of the different char-acteristics in the two groups of comparison. In total, 39Province and District hospitals accounting for 1.228 maternaldeaths and 385 health centres and health posts accountingfor 596 maternal deaths were included.

Table 3 presents the modelling results. At the level ofhealth centres, an increase in size (number of maternity beds)

Figure 2. Estimated institutional maternal mortality ratio by District, Mozambique.

JOURNAL OF OBSTETRICS AND GYNAECOLOGY 467

Table 2. Abstract of maternal deaths on provincial/district hospital and health centre.

Provincial/District hospitals Health centres/HEALTH posts

n Maternal deaths n Maternal deaths

Total 39 1228 385 596Medical doctor available and providing EmOC

No 2 94 331 379Yes 37 1134 54 217

MCH nurse available and Providing EmOCNo 0 – 95 49Yes 38 1220 279 505

Surgical technician available and providing EmOCNo 10 241 372 541Yes 28 979 2 13

Availability of medicinesBad 11 145 203 176Regular 8 86 93 172Good 15 551 36 128Very good 5 446 15 86

Delivery room equipmentVery bad 2 32 67 55Bad 6 79 109 138Moderate 23 904 93 239Good 7 181 32 141

Deliveries per MCH nurseLow 0 – 117 68High 31 843 190 396

Average # MCH nurses (S.D) 8.29 (4 27) 2.31 (2.42)Average # medical doctors (S.D) 1.71 (1.45) 0.16 (0.41)Average #surgeon assistants (S.D) 0.69 (0.54) 0.01 (0.09)Average # beds in maternity ward (S.D) 30.51 (16.51) 6.02 (5.31)Average # deliveries (S.D) 2578.23 (1898.94) 510.76 (653.67)

S.D: Standard Deviation.

Table 3. Estimates of the relative risk and standard errors for the weighted negative binomial (NB) model andweighted zero-inflated negative binomial (ZINB) with random effects.

Provincial/District hospital Health centres

Effects (reference) Estimate (s.e) Estimate (s.e)Beds in maternity wardb – 1.455 (0.091)a

Model for counts of maternal deathRegion (South)North 4.3417 (2.7292)a 0.4107 (0.2226)b

Central 5.4125 (3.6387)a 0.2890 (0.1596)b

Health facility type (General/Provincial)District hospital 0.6885 (0.3237) –Beds in maternity ward 1.0262 (0.0146)b 0.8983 (0.03018)a

Location (Inside district capital)Outside district capital 0.7954 (0.5553) 0.1383 (0.05846)a

Malaria cases (%) 0.0003 (0.0021) 269.74 (1170.29)Medical doctor available and saving lives (No)Yes 0.7939 (0.8306) 0.5039 (0.2195)

Surgeon assistant available and saving lives (No)Yes 1.2798 (0.5713) –

MCH nurse available and saving lives (No)Yes – 0.6117 (0.2363)

Availability of medicines (Bad)Regular 3.1057 (2.1144)b 1.2340 (0.4181)a

Good 2.5422 (1.2399)b 0.7444 (0.3919)a

Very good 35.681 (9.775)a 0.3890 (0.2636)Delivery room equipment (Very bad/bad)Moderate 1.3553 (0.5593) 1.2490 (0.3926)a

Good 1.3335 (1.0082) 0.9261 (0.5149)b

Deliveries per MCH nurses (Low)High – 0.2161 (0.2501)Proportion of referrals from (%) 1158.303 (8036) 0.0006 (0.0026)

The NB was used to analyse the provincial/district hospitals and the ZINB was used to analyse the health centres/health posts. Weights were calculated based on the sampling probabilities of the health facilities.

ap < .005.bp < .1.

468 L. CHAVANE ET AL.

yielded an increase of �53% in the odds of an institutionalmaternal death (p<0.001). We observed significant regionaldifferences of facility-based maternal mortality ratio in healthcentres, with the highest mortality rates in the southernregion (p< .1).

However, the provincial/district hospitals’ maternal mortal-ity ratio was higher in the northern and central regions com-pared with the southern regions (p <.05) with the risk ofmaternal death being at least 4 times higher than in thesouth.

The risk of maternal deaths was reduced by �10% witha unit increase in the number of beds in the maternityward of health centres (p< .05), while increases of 2.6%with a unit increase in beds at provincial/district hospitalswas observed (p< .1). The number of maternal deaths was68% lower in health centres located outside the districtcapital compared with the centres located in the Districtcapital (p¼ .0026).

The risk of maternal death was reduced by 40% in healthcentres with a high workload if the MCH nurses also performedEmOC signal functions (p¼ .0202). No significant effects onmaternal deaths were observed for the availability of medicaldoctors, surgical technicians or delivery room equipment.

Discussion

This study found a high facility-based mortality ratio that wasestimated to be �703 per 100.000 deliveries. The overallmagnitude of facilities in maternal death is higher in thesouth region. This is very well represented in the distributionof the estimated facility-based maternal mortality mapping.Results also suggest lower performance of hospitals from thecentral and northern regions. The overall distribution ofmaternal death corresponds to the level of data availability inthe health facilities from the south region. The pattern of dis-tribution of facility-based maternal deaths matches with thelevel of utilisation of health services in the country and notthe actual distribution of the maternal deaths at the popula-tion level. The last population census indicates high maternalmortality in north and centre of the country. Efforts toincrease the utilisation of the healthcare facilities by the preg-nant woman in the north and centre of country are recom-mended. There is a need for improved quality of care toavoid the preventable death of the woman that manages tocome into the health facilities and to ensure accurate mortal-ity report. The number of maternal deaths increases from thelower to upper level of healthcare facilities. This may reflectthe delays and later arrival of the woman at the referral facili-ties due to reference process deficiencies.

The provision of EmOC is considered to be the main inter-vention at the referral facility to reduce the level of maternaldeaths (Paxton et al. 2005; Kongnyuy et al. 2009). A system-atic evaluation on the interventions for the reduction ofmaternal mortality performed by Nyamtema et al. suggestedthe need to ensure the implementation of integrated inter-ventions 24 hours a day and 7 days per week to have signifi-cant changes on mortality (Nyamtema et al. 2011). Theavailability of skilled human resources to manage obstetriccomplications, the availability of essential drugs, equipment

and a conducive environment are factors that have beendescribed as pre-conditions for a facility to offer EmOC serv-ices and an intervention performed in a few hospitals inMozambique have been able to demonstrate this effect(Jamisse et al. 2004). The importance of availability and well-distributed skilled attendant to reduce maternal deaths wasdescribed elsewhere (Robinson and Wharrad 2001; Guptaet al. 2011; Moyer et al. 2013). In this study, we observed thatthe presence of MCH nurses at the health facility offeringcontinuous EmOC interventions was strongly correlated withthe reduction of maternal deaths (p¼ .0202). However, wedid not observe a significant effect on the availability of med-ical doctors and surgical technicians. This finding might becorrelated to the distribution of this type of staff in thesampled health facilities. This suggests the need for continu-ous investment to improve the availability and deploymentof highly qualified MCH nurses that can secure the provisionof EmOC services. A recent multi-country survey on maternaland newborn health conducted by the WHO (Souza et al.2013), emphasise the need of universal coverage of lifesavinginterventions and an overall improvement in the quality ofmaternal care(Cavallaro and Marchant 2013). A systematicreview of the factors implicated in the institutional maternaldeath by Knight et al. indicated that the lack of humanresources were the most reported barrier: 41 of 43 studiesreferred to this barrier and particularly, the lack of training(Knight et al. 2013).

In this analysis, we found that the availability of drugs andminimal kits for delivery and emergency care was poor.According to our classification, 55% of all health facilities,including 1 central hospital, had poor availability of drugs foremergencies and only 13% were classified as good and 5%were classified as very good. This finding suggests that dur-ing the survey, several facilities were experimenting drugsstock outs. However, the result from our model did not showany significant difference of maternal deaths in relation tothe drugs availability. Facility-based maternal mortality con-tinues to be notably high in Mozambique and MCH nursesplay an important role for its reduction. Further analysis isneeded to clarify the relative importance of the different fac-tors related to health system functioning for the occurrenceof facility-based maternal deaths.

Disclosure statement

The authors declare that they have no conflict of interest.

ORCID

Leonardo Chavane http://orcid.org/0000-0001-7382-0665

References

AMDD, 2014. AMDD needs assessment ToolKit [Internet]. [cited 2015 Jan15]. Available from: http://www.amddprogram.org/content/emoc-needs-assessments.

Bohning D. 1998. Zero-inflated poisson models and C.A. MAN: a tutorialcollection of evidence. Biometrical Journal 40:833–843.

JOURNAL OF OBSTETRICS AND GYNAECOLOGY 469

Cavallaro FL, Marchant TJ. 2013. Responsiveness of emergency obstetriccare systems in low- and middle-income countries: a critical review ofthe ‘third delay’. Acta Obstetricia Et Gynecologica Scandinavica92:496–507.

Dankmar B, Dietz E, Schlattmann P. 1995. Zero-inflated count models andtheir applications in public health and social science. In Rost J,Langeheine R, editors. Applications of latent trait and latent classmodels in the social sciences. New York: Waxmann Publishing.p. 333–344.

Dgedge M, Mendoza A, Necochea E, Bossemeyer D, Rajabo M, FullertonJ. 2014. Assessment of the nursing skill mix in Mozambique using atask analysis methodology. Human Resources for Health 12:5.

Famoye F, Singh KP. 2006. Zero-inflated generalized Poisson regressionmodel with an application to domestic violence data. Journal of DataScience 4:117–130.

Gupta N, Maliqi B, Franca A, Nyonator F, Pate M, Sanders D, et al. 2011.Human resources for maternal, newborn and child health: from meas-urement and planning to performance for improved health outcomes.Human Resources for Health 9:16. DOI: 10.1186/1478-4491-9-16

Hall DB. 2000. Zero-inflated Poisson and binomial regression with randomeffects: a case study. Biometrics 56:1030–1039.

Hall DB, Zhang Z. 2004. Marginal models for zero inflated clustered data.Statistical Modelling 4:161–180.

Hill K, El Arifeen S, Koenig M, Al-Sabir A, Jamil K, Raggers H. 2006. Howshould we measure maternal mortality in the developing world? Acomparison of household deaths and sibling history approaches.Bulletin of the World Health Organization 84:173–180.

INE. 2009. Dados do Censo geral da Populac~ao Mocambique 2007 –Apresentac~ao Sum�aria, Maputo. [cited 2015 Jan 15]. Available from:http://www.ine.gov.mz/

INE, MISAU. 2011. Mocambique, Inquerito Demogr�afico e de Sa�ude.Maputo.

Jamisse L, Songane F, Libombo A, Bique C, Fa�undes A. 2004. Reducingmaternal mortality in Mozambique: challenges, failures, successes andlessons learned. International Journal of Gynaecology and Obstetrics85:203–212.

Knight HE, Self A, Kennedy SH. 2013. Why are women dying when theyreach hospital on time? A systematic review of the ‘third delay’. PLoSOne 8:e63846.

Kongnyuy EJ, Hofman JJ, van den Broek N. 2009. Ensuring effectiveessential obstetric care in resource poor settings. BJOG: AnInternational Journal of Obstetrics and Gynaecology 116:41–47.

Loquiha O, Hens N, Chavane L, Temmerman M, Aerts M. 2013. Modelingheterogeneity for count data: A study of maternal mortality in healthfacilities in Mozambique. Biometrical Journal 55:647–660.

Moyer CA, Dako-Gyeke P, Adanu RM. 2013. Facility-based delivery andmaternal and early neonatal mortality in sub-Saharan Africa: a regionalreview of the literature. African Journal of Reproductive Health17:30–43.

Nyamtema AS, Urassa DP, van Roosmalen J. 2011. Maternal health inter-ventions in resource limited countries: a systematic review of packages,impacts and factors for change. BMC Pregnancy and Childbirth 11:30.

Paxton A, Maine D, Freedman L, Fry D, Lobis S. 2005. The evidence foremergency obstetric care. International Journal of Gynaecology andObstetrics 88:181–193.

Pereira C, et al. 2007. Meeting the need for emergency obstetric care inMozambique: work performance and histories of medical doctors andassistant medical officers trained for surgery. BJOG: An InternationalJournal of Obstetrics and Gynaecology 114:1530–1533.

Qomariyah SN, Siti Nurul B, Jacqueline S, Pambudi E, Anggondowati T,Latief K, et al. 2009. A practical approach to identifying maternaldeaths missed from routine hospital reports: lessons from Indonesia.Global Health Action 2:1–5.

Robinson JJ, Wharrad H. 2001. The relationship between attendance at birthand maternal mortality rates: an exploration of United Nations’ data setsincluding the ratios of physicians and nurses to population, GNP percapita and female literacy. Journal of Advanced Nursing 34:445–455.

Sa�ude IN, Minist�erio da Sa�ude. 2009. Mortalidade em MocambiqueInqu�erito Nacional sobre Causas de Mortalidade, 2007/8. Maputo.

Souza JP, G€ulmezoglu A, Vogel J, Carroli G, Lumbiganon P, Qureshi Z,et al. 2013. Moving beyond essential interventions for reduction ofmaternal mortality (the WHO Multicountry Survey on Maternal andNewborn Health): a cross-sectional study. Lancet 381:1747–1755.

470 L. CHAVANE ET AL.