Maternal and neonatal mortality in south-west ethiopia: Estimates and socio-economic inequality

Introduction: Ethiopia has achieved the fourth Millennium Development Goal by reducing under 5 mortality. Nevertheless, there are challenges in reducing maternal and neonatal mortality. The aim of this study was to estimate maternal and neonatal mortality and the socio-economic inequalities of these mortalities in rural south-west Ethiopia. Methods: We visited and enumerated all households but collected data from those that reported pregnancy and birth outcomes in the last five years in 15 of the 30 rural kebeles in Bonke woreda, Gamo Gofa, south-west Ethiopia. The primary outcomes were maternal and neonatal mortality and a secondary outcome was the rate of institutional delivery. Results: We found 11,762 births in 6572 households; 11,536 live and 226 stillbirths. There were 49 maternal deaths; yielding a maternal mortality ratio of 425 per 100,000 live births (95% CI:318-556). The poorest households had greater MMR compared to richest (550 vs 239 per 100,000 live births). However, the socio-economic factors examined did not have statistically significant association with maternal mortality. There were 308 neonatal deaths; resulting in a neonatal mortality ratio of 27 per 1000 live births (95% CI: 24-30). Neonatal mortality was greater in households in the poorest quartile compared to the richest; adjusted OR (AOR): 2.62 (95% CI: 1.65-4.15), headed by illiterates compared to better educated; AOR: 3.54 (95% CI: 1.11-11.30), far from road (≥6 km) compared to within 5 km; AOR: 2.40 (95% CI: 1.56-3.69), that had three or more births in five years compared to two or less; AOR: 3.22 (95% CI: 2.45-4.22). Households with maternal mortality had an increased risk of stillbirths; OR: 11.6 (95% CI: 6.00-22.7), and neonatal deaths; OR: 7.2 (95% CI: 3.6-14.3). Institutional delivery was only 3.7%. Conclusion: High mortality with socio-economic inequality and low institutional delivery highlight the importance of strengthening obstetric interventions in rural south-west Ethiopia. © 2014 Yaya et al.

The Ethical Review Committee for the Health Research of Southern Nations Nationalities and Peoples’ Regional State (SNNPRS) Health Bureau in Ethiopia, and the Regional Committee for Health Research Ethics of North Norway (REK Nord) approved the study. We obtained informed verbal consent from all respondents and the response was recorded on the questionnaire as “accepted” or “declined” to participate. Almost all approached households were willing to be interviewed, and written consent was not considered because a large number of the respondents were illiterates. The study involved only interview and the ethics committee approved the verbal consent procedure. Additionally, minors were not included in this study. A method for finding out the medical causes of death and ascertaining the factors that may have contributed to the death in women who died outside of a medical facility. It consists of interviewing people (family members, neighbours, traditional birth attendants) who had knowledge about the events leading to the death [18]. A death within 28 days of an alive born baby. A death of a woman while pregnant, in labour, or within 42 days of the termination of pregnancy, irrespective of the duration and site of the pregnancy, from any cause related to or aggravated by the pregnancy or its management, but not from accidental or incidental causes (ICD-10) [19]. A birth of a dead fetus after 28 weeks of gestation. We did not use the baby-weight criteria of classifying stillbirths, as it was not possible to measure weight in the rural area. A person or a group of people living in a room or rooms and sharing common things together. In cases of polygamy (more than one wife for a man, we considered each wife as a separate household as they culturally have separate houses. Is the number of maternal deaths in a population during a given time period per 100,000 live births during the same period. Is the number of newborn deaths (within 0–28 days) in a population per 1,000 live births in the same population. Is the number of births of dead fetuses after 28 weeks of gestation per 1,000 births. We conducted this study in Bonke, one of the 15 woredas in the Gamo Gofa zone in south-west Ethiopia. In 2010, the woreda had a population of 173,240 people [11]. A kebele is the lowest administrative structure with 5,000 to 7,000 residents in the Ethiopian government system. Bonke has 31 kebeles and one of these, the administrative centre, has a town status with a population of 6,347 people in 2007. Table 1 shows the profile of the 15 kebeles included in this study. Bonke is 618 km from Addis Ababa, and 68 km from Arba Minch (zonal capital) where the nearest hospital is situated. Nevertheless, over half of the remote areas of Bonke are more than 100 km (20 hours walking distance) away from the hospital in Arba Minch. An estimated three-fourths of the population also live in villages far from the motorable road (≥6 km). The only road to the woreda is the road from Arba Minch to Kamba, which crosses parts of Bonke. Overflowing rivers during the rainy season often interrupt the road. So, often people have to carry critical patients or use transport animals such as horses and mules to go to the hospital. Health care is provided by a health centre in the town (Geresse) as well as three other rural health centres. There are no medical doctors working in the woreda; a few health officers (people with a bachelor’s degree in medical training), nurses, and midwives staff the health centres. In Bonke, there is no access to lifesaving comprehensive essential obstetric care that can provide caesarean sections, blood transfusions, and effective care to sick and low birth-weight newborns. This study was part of an implementation project to reduce maternal mortality in Gamo Gofa. The project trains health officers in emergency obstetric services, community health workers in identifying and referring high-risk mothers, in addition to equipping health centres and hospitals with essential instruments. The study was a cross-sectional household survey with a five-year recall of events prior to the data collection. We collected the data in February 2011 from all households that had births and pregnancy outcomes between January 2006 and December 2010. We purposely selected January 2006 as the starting reference period for the recall because it was the immediate period after the 2005 National Election, an event well known to all respondents. We based our sample size calculation on the assumptions of a crude national birth rate of 35 per 1,000 population, and a neonatal mortality ratio of 35 per 1,000 live births [20]. We aimed to detect a minimum of 350 neonatal deaths to make empirical estimates and assess the household risk factors associated with neonatal deaths. To find 350 neonatal deaths we needed to find at least 10,000 live births in five years (an average of 2,000 per year) within the population. With a fertility rate of 35 births per 1,000 people, 2,000 live births per year could be obtained from an estimated population of 57,143. Assuming a constant birth rate over the five years, we projected the population of 57,143 in 2006 to be 67,244 in 2010 (half the rural population in Bonke). We used OpenEpi, open source calculator (www.openepi.com), and calculated the minimum sample needed based on information that three-fourths of the study households resided far from the motorable road (≥6 km), thereby expecting a neonatal mortality prevalence twice that among households far from the road. We used a statistical power of 80%, a 95% confidence interval, and an assumption that 4% of households far from motorable road could experience neonatal mortality to calculate the number of households needed for the study. This provided 5,187 households with expected births in the five years before the survey. On average, we expected two births per household over five years yielding 10,374 births. The neonatal mortality rate from the estimated number of households was also assumed to provide enough power to detect other risk factors (wealth, education, non-spaced births). We also assumed that a number of maternal deaths among the estimated 10,374 births would give an optimum MMR estimate, and a similar assumption was applied for stillbirths. Taking into account a potential 10% of non-responders, we decided to study 50% of the rural population in the Bonke woreda and we randomly selected 15 of the 30 rural kebeles in Bonke. Data collectors visited all of the households in the selected 15 kebeles asking about any pregnancy and birth outcomes (abortion, alive and stillbirths, neonatal and maternal deaths) in the households over the previous five years. Enumerators noted the number of households in each kebele, collecting data from the households that had pregnancy and birth outcomes during the stated time period. As projected from the Ethiopian 2007 census [11], the selected kebeles had a population of 78,181 people in 2010. The primary outcome variables were maternal and neonatal mortality and stillbirth rates while a secondary outcome variable was rate of skilled delivery service utilization. We also used the following household predictor variables: household wealth (assets index), educational level of heads of households, the number of births in a household over five years, and household distance to a motorable road. Survey questions included: where each delivery took place (home, health post, health centre, and hospital) and who attended the birth (family member, community health extension worker, skilled health professional-doctor, midwife). We also asked about the place where the maternal death occurred (at home or within a health facility) as well as what had happened to the fetus (stillbirth, neonatal death, or alive at the time of data collection). Respondents were asked about whether the newborn was alive or dead at the end of the 4th week after delivery. If the response was “dead”, then we asked about the timing of death (in weeks) in relation to the birth. However, we did not investigate the causes of deaths for neonatal deaths and stillbirths, assuming it would be difficult for rural respondents to answer it properly. In the households that had deaths of women -in reproductive age (15–49 years), we used questions modified from the WHO manual for verbal autopsy for maternal death to investigate the causes of deaths [18]. The questions included: whether the mother was pregnant, in the process of giving birth, or in postnatal period after birth, what main medical condition or symptom was associated with her death, what assistance she received, and from whom she received help. A nurse decided on pre-coded choices of the major causes of maternal deaths (bleeding, prolonged labour, fever and convulsions, including the option of “others”) based on quick algorithmic analysis of information provided by the respondents. Sensitive questions related to abortion deaths were placed at the end of the interview to minimize the intentional hiding of information. We collected information on the estimated walking distance (in hours) from each house to the nearest health centre, the nearest motorable road, and the nearest hospital. Based on the local experience of one hour of walking time per 5 km for an average person, we converted the walking distance into kilometers. We recruited 15 natives from the respective study villages who had completed the 12th grade for data collection. The purpose of selecting data collectors from their respective kebele of data collection was to reduce the potential recall bias by the respondents. Data collectors are aware of many vital events in the villages they collected data by living and participating in social events such as birth celebrations, mourning rituals, and burials at the time of the deaths. Five diploma graduates who had a thorough knowledge of the culture and language of the area supervised the data collection. The data collectors were trained for two days on pre-testing field interviews, translating the questions from “Amharic” (the official Ethiopian state language) to “Gamotho” (the language of the ethnic “Gamo” community) and how to introduce the simplified verbal autopsy questions. Depending whoever was present at home during the visit, the respondent was the father or mother for a recently deceased newborn. In cases of death of a married woman, we interviewed a husband while in the absence of a husband, an adult relative or an adult child of the deceased was interviewed. For those who were unmarried, we asked parents or siblings. If the respondents were not present at home during the first visit, the data collectors re-visited the next day in the early morning. Less than 1% of households were missed after two visits. For the wealth index, we selected 10 variables of household assets with the highest standard deviation (>0.20), as recommended by Seema Vyas and colleagues [21]. The types of asset variables and their standard deviations are presented in Table S1 in File S1. We transformed the categorical variables into dichotomous (0–1) indicators: 0 for indicators of poor wealth and 1 for indicators of good wealth. We examined the dichotomous variables by using the principal component analysis (PCA) to produce a factor score for each household with households being assigned a rank according to the factor score. Because of the low number of maternal deaths in the socio-economic classes for calculation, we divided households into four equal categories (quartiles), rather than the widely used five classes. Each category was comprised of 25% of the households studied. Table S2 in File S1 shows the mean score, standard deviations, communalities, and correlations of the variables to the first (main) component. The total variance explained by the first component was 20.58%, with an eigenvalue of 2.06. We used two units of analysis (household and birth). By using births as the unit of analysis, we presented descriptive tabulations of outcomes in the form of rates and ratios. By using the household as a unit of analysis, and applying logistic regression, we present household risk factors associated with the mortality outcomes (Tables 2, ​,3,3, and ​and4).4). We used SPSS 16 (Statistical Package for Social Sciences) for the data entry and analysis [22]. Data are freely available from the corresponding author on request.

• yaya, yaliso
• eide, kristiane tislevoll
• norheim, ole frithjof
• lindtjørn, bernt