Menu
Menu
Menu
Menu
Background Maternal mortality continues to have devastating impacts in many societies, where it constitutes a leading cause of death, and thus remains a core issue in international development. Nevertheless, individual determinants of maternal mortality are often unclear and subject to local variation. This study aims to characterise individual risk factors for maternal mortality in Tigray, Ethiopia. Methods A community-based case-control study was conducted, with 62 cases and 248 controls from six randomly-selected rural districts. All maternal deaths between May 2012 and September 2013 were recruited as cases and a random sample of mothers who delivered in the same communities within the same time period were taken as controls. Multiple logistic regression was used to identify independent determinants of maternal mortality. Results Four independent individual risk factors, significantly associated with maternal death, emerged. Women who were not members of the voluntary Women’s Development Army were more likely to experience maternal death (OR 2.07, 95%CI 1.04—4.11), as were women whose husbands or partners had below-median scores for involvement during pregnancy (OR 2.19, 95%CI 1.14—4.18). Women with a pre-existing history of other illness were also at increased risk (OR 5.58, 95%CI 2.17—14.30), as were those who had never used contraceptives (OR 2.58, 95%CI 1.37—4.85). Previous pregnancy complications, a below-median number of antenatal care visits and a woman’s lack of involvement in health care decision making were significant bivariable risks that were not significant in the multivariable model. Conclusions The findings suggest that interventions aimed at reducing maternal mortality need to focus on encouraging membership of the Women’s Development Army, enhancing husbands’ involvement in maternal health services, improving linkages between maternity care and other disease-specific programmes and ensuring that women with previous illnesses or non-users of contraceptive services are identified and followed-up as being at increased risk during pregnancy and childbirth.
Tigray Region is located in northern Ethiopia, with a total population of more than 5.1 million. Details of the geography and population in the Region are described in more detail in a previous paper [5]. Most of the population live in scattered rural villages, some of which are quite remote in terms of access and facilities. As previously described [13], a study of maternal mortality was conducted in six rural districts of Tigray Region (Welkayat, Laelay Adiyabo, Tahtay Maychew, Saesi Tsaedaemba, Hintalo Wajirat and Alamata), which were randomly selected as a stratified sample of one District per Zone, as shown in Fig 1. The sampled districts included a total of 183,286 households, with a total population of 843,115, covering around 20% of the total population of rural Tigray. Of these, 166,515 were women of reproductive age (WRA), defined as 15–49 years, representing 19% of all women of reproductive age in rural Tigray. A community-based case-control study was designed, with geographical matching at the lowest administrative level (tabia), to assess the individual-level risks for maternal death while controlling for wider geographic variations. A case of maternal death was defined as a woman of reproductive age (15–49 years old) who died during pregnancy, childbirth or puerperium due to maternity-related causes. Controls were defined as women of reproductive age group (15–49 years old) who delivered a live child during the reference period and who were alive six weeks postpartum. Sample size was calculated using Epi Info Version 3.5.1 with the following assumptions: 95% confidence, 80% power, proportion of exposure among controls to key parameters as 20%, odds ratio of 2.5, a case to control ratio of 4:1 and a 10% contingency for non-responses. This gave a total sample size of 62 cases and 248 controls. Since the one-year survey period from May 2012 to May 2013 previously described [5] included 51 maternal deaths, this study included those 51 as cases, and recruited a further 11 cases in the three months following the one-year survey period. A census of all households in the six selected Districts was conducted in mid–2013 to identify all births and deaths among women aged 15 to 49 years. The causes of death were surveyed using the 2012 WHO VA instrument [14], processed using the InterVA-4 model [15], and all maternal deaths were included in the study as cases. A list was then drawn up of all the living women in the same reproductive age group who gave birth during the study period and lived in the same tabia as each case. This was used as a sampling frame for the selection of controls. Four controls were selected for each case, using simple random sampling, from these lists. Data were collected by trained Maternal and Child Health experts, responsible for the maternal and child health programmes at district level, with the help of health extension workers as community guides, using a structured questionnaire developed by reviewing similar studies and different relevant guidelines and protocols. The questionnaire was developed in English and then translated into the local language (Tigrigna). Adult respondents who were caregivers at the time of death of the mother were used to collect information about all cases. The information about controls was collected from the controls themselves. Supervisors were trained by the principal investigator for one day and the data collectors were trained by the supervisors for three days on the details of data collection tool, interviewing techniques, the importance of data quality and research ethics. Written consent was obtained from all the controls and adult respondents who were caregivers at the time of death of the mother. The consent was approved by the Institutional Review Board of Mekelle University Data were analysed using Stata 11. Bivariable logistic regression was used to analyse relationships between maternal death and independent variables, with crude odds ratios and 95% confidence intervals estimated for each parameter. A series of nine parameters relating to husbands’ involvement with maternity were appreciably inter-correlated, and consequently a score for husbands’ involvement was developed, based on the number of positive responses out of the nine items. This score was dichotomised as above or below the median number of positive responses and used as a single variable in the further analyses. Multivariable logistic regression was used to build an overall model from the factors that were significantly associated with maternal mortality on a bivariate basis, giving adjusted odds ratios and 95% confidence intervals. The dataset on which these analyses are based is available on request from the Corresponding Author at moc.oohay@yafedogh.
N/A
