Maternal death and delays in accessing emergency obstetric care in Mozambique

Delays in access to quality care have been identified as one of the important determinants of preventable maternal death [1, 2]. Thaddeus and Maine’s three delays model that describes the multiple factors that drive maternal mortality has proven to be an effective tool to evaluate the circumstances surrounding access to and appropriateness of emergency obstetric and newborn care (EmONC) [3]. The model has been helpful identifying barriers and potential points of intervention along the continuum from home to hospital for more than 20 years [4–6].

According to this framework 3 delays in access to quality emergency care are defined. The first delay occurs at the household and community level and reflects the delay in deciding to seek care for pregnancy complications. The second delay (delay II) refers to the delay to reach the facility that provides emergency obstetric care (EmONC) and, the third delay (delay III) refers to the delay that occurs in receiving care after arrival at the health facility [3].

In Mozambique, estimates based on the 2007 population census indicate that around 46% of maternal deaths occur within health institutions [7], a substantial proportion considering that these women reached a health facility.

Maternal mortality ratio in Mozambique has declined from more than 1600 estimated deaths per 100.000 live births in 1990 [8] to 403 in 2003 [9] and DHS 2011 [10]. Despite considerable reduction over time, the latest estimations available show that the ratio still is very high, even when compared with the neighboring countries [2].

Several factors have been identified at the health care facility level that interfere with the readiness to deal adequately with obstetric emergencies. Knight described six groups of factors, namely drugs and equipment, policy and guidelines, human resources, facility infrastructure, patient-related and referral-related aspects [11].

A near miss study, an approach to evaluate maternal quality of care and learning from women that survived severe maternal complications [12, 13] covered 564 survival women and 71 maternal deaths, conducted in Mozambique’s Maputo Province, found delay II in 21.3% and delay III in 69.7% (women could experience multiple delays [14]. In some cases these delays followed the woman’s path throughout the referral system from admission to a peripheral health facility and then on to the referral facility [14].

With this paper we aim to examine the relationship between the cause of maternal death and delays II and III.

Study setting

Mozambique is a low income country located in the Sub-Saharan Africa region with an estimated population of 26.4 million in 2016, based on projections of the 2007 population census. The public sector is the main provider of health care services [15]. Childbirth services are provided throughout 4 levels of care. At peripheral level services are available in the smallest type 2 health centers, with 1 MCH nurse as provider and 3 maternity beds. The different layers of care are connected by a network referral system based on ambulances managed at the referral health facilities. Usually a health facility covering an average population of 300.000 inhabitants has 1 to 2 ambulances to serve a network of 10 to 15 health facilities. In the district capitals maternity services are available in large type 1 health centers, with more than 10 maternity beds; at this level the staff include at least 3 MCH nurses and in some facilities, a general practitioner physician. The District, Provincial, General and Central hospitals are more complex facilities with different specialties, wards and operating rooms. Health centers type 1 and 2 should have capacity to offer Basic Emergency Obstetric Care (BEmOC). According to the Annual Joint Assessment Report in 2012, in average 2.3 health facilities were available with BEmOC services per 500 thousand inhabitants. When needed, the health centers refer patients to hospitals with capacity to offer Comprehensive Emergency Obstetric Care (CEmOC).

This paper draws on secondary data analyses of a large health facility needs assessment on maternal and neonatal health care services, which included 427 health facilities nationwide. The sample was designed to be representative of the mix of the health facilities at national and provincial levels. Data collection was done between November 2007 and January 2008 by health providers previously trained to undertake this task. Service statistics were collected from registers and logbooks and covered a period of 12 months, from November 2006 to October 2007.

The national survey identified a total of 2,198 maternal deaths in the 427 health facilities out of 312,537 deliveries, corresponding to an estimated facility based maternal mortality ratio of 703 deaths per 100.000 deliveries. At each facility, a detailed maternal death review was conducted if the patient’s records were available. A maternal death review was completed for 712 maternal deaths for which there was sufficient information in the files. Prior to data entry, these were reviewed by a committee of experts composed of 6 ObGyn Medical doctors. The committee reviewed the circumstances of death to determine: 1) the final cause of death, 2) if evidence existed to suggest that the woman experienced a delay in arriving at the health facility providing emergency obstetric care services (delay II). Women that faced delay in the referral process reaching the second health facility was considered as delay II. and 3) if evidence suggested a delay in receiving adequate health care services after reaching the health facility with emergency obstetric care services (delay III).

For each maternal death, the time of arrival at the initial health facility was collected as well as the time of treatment initiation, time of arrival at the referral (second) facility and time of death. The delay was defined based on assessment of each woman’s individual story since her arrival at the initial health facility. This included the combination of the clinical status on admission in the health facility, and the time taken for action at peripheral and referral facilities. Women presenting life threatening condition at admission of ether initial facility or at the referral facility were considered facing delay type II. In all these steps the quality of care was assessed relating the decision (diagnosis and or to refer to next level of care) and the treatment prescribed and or initiated. The 712 maternal deaths were found in 93 of the 427 facilities included in the survey, including health centers and hospitals, located in rural and urban areas of the 11 provinces of Mozambique. Approval from the Mozambican national bioethical committee was received before starting the survey implementation.

Statistical analysis

Data were analyzed using the SAS-STAT software, version 9.4. A bivariate binary mixed effects model was applied to examine if and how the probability for delay type II and delay type III depends on the main cause of maternal death and other risk factors and covariates, while accounting for the correlation between both types of delay. Specifically, as in Wilson and Lorenz (2015) [16],we used one logistic regression model for each outcome variable (delay type II and III), both models consisting of fixed effects related to observed covariates and a random intercept related to unobserved effects affecting both delay types specific to each woman.

The advantage of such model specification is that it allows the logistic regressions to be modelled simultaneously, for which, conditional on the latent variable, delay type II and III were assumed as independent outcomes. This allows the estimation of the correlation between the outcomes while also controlling for the main cause of death and other observed covariates such as age, parity, referrals, type of health facility and availability of human resources. The potential covariates were selected for the final model using logistic regression for each outcome separately in a backwards selection procedure with removal rob of 0.2. The initial model didn’t include any interaction between the risk factors.

Study limitations

The absence of information on survival limited the analysis that might have further clarified factors that were determinants of maternal deaths. Also the fact that only part of the facility based data was available, limits the generalization of the conclusions to all country but rather to the covered facilities.

The quality of hospital and health center registers and medical charts was poor, as were storage and retrieval systems, which reduced our access to the files of all maternal deaths and to more complete information on the circumstance surrounding maternal deaths and the actions taken. The analysis done by the expert group helped reduce the inconsistency in the data from the files that were accessible.

This survey found that only 55% of the maternal deaths registered in the health facilities were reported in the national health information system. This study covered the analysis of around 32% of total institutional maternal deaths during the evaluation period.