Distant and proximate factors associated with maternal near-miss: A nested case-control study in selected public hospitals of Addis Ababa, Ethiopia

Background: Ethiopia is one of the sub-Saharan Africa countries with the highest maternal mortality. Maternal near-misses are more common than deaths and statistically stronger for a comprehensive analysis of the determinants. The study aimed to identify the factors associated with maternal near-miss in selected public hospitals of Addis Ababa, Ethiopia. Methods: We conducted a nested case-control study in five selected public hospitals of Addis Ababa, Ethiopia from May 1, 2015 to April 30, 2016. Participants were interviewed by well-trained data collectors using pre-tested questionnaire. Medical records were also reviewed to gather relevant information. World Health Organization criteria were used to identify maternal near-miss cases. A total of three controls matched for age and study area was selected for each maternal near-miss case. Bivariate and multivariable conditional logistic regressions were performed using Stata version 13.0. Results: A total of 216 maternal near-miss cases and 648 controls were included in the study. The main factors associated with maternal near-miss were: history of chronic hypertension (AOR=10.80,95% CI; 5.16-22.60), rural residency (AOR=10.60,95% CI;4.59-24.46), history of stillbirth (AOR=6.03,95% CI;2.09-17.41), no antenatal care attendance (AOR=5.58,95% CI;1.94-16.07) and history of anemia (AOR=5.26,95% CI;2.89-9.57). Conclusions: There is a need for appropriate interventions in order to improve the identified factors. The factors can be modified through a better access to medical and maternity care, scaling up of antenatal care in rural areas, improve in infrastructure to fulfill referral chain from primary level to secondary and tertiary health care levels, and health education to pregnant women. We conducted a study in five selected public hospitals of Addis Ababa, capital of Ethiopia from May 1, 2015 to April 30, 2016. The selection of hospitals was based on the number of deliveries conducted per year. In addition, presence of an Intensive Care Unit (ICU), maternity ward, blood transfusion service and availability of cesarean section (CS) delivery were considered in the selection of hospitals. Accordingly, Tikur Anbessa, St. Paul’s Hospital Millennium Medical College, Zewditu Memorial, Yekatit 12 and Gandhi Memorial Hospitals were selected. A total of 29,697 live birth deliveries took place in the participating hospitals during the study period. The details of settings with location map has been described elsewhere [21]. A nested case-control study design, matched for age and study setting was employed. Age was categorized in five year interval. Participants were followed from admission till discharge. Women who were admitted to the selected hospitals during the study period for treatment of pregnancy-related complications (irrespective of gestational age), who delivered, or were within 42 days of termination of pregnancy, and fulfilled at least one of the conditions that is indicated in the WHO criteria presented in Table 1 [9] were included as cases. Identification criteria of maternal near-miss as used by the WHO 2011 Those women who have been admitted for reasons not related to pregnancy, delivery or 42 days after termination of pregnancy were excluded. Women who came to the same hospital where the cases happened and, having a similar age- interval category with that of the cases and delivered without any complications were enrolled as a control. For each near-miss case, three controls that occurred within the same day of the near-miss event were included. The sample size was estimated using Epi Info 7 software using sample size determination for unmatched case-control studies. The parameters that were used to estimate the sample size were: confidence level of 95%, power of 80%, case-control ratio of 1:3, expected frequency of exposure in control to be 4.11%, and percent exposure among cases, 10.78%. It was estimated from one study in Ethiopia taking no ANC follow-up as one of the main exposure variable for maternal near-miss that provide the maximum sample size [20]. Accordingly, these yield a minimum sample size of 166 cases and 497 controls. Adding a 10% non-response rate, the final sample size required for the study was 183 cases and 547 controls. To increase the power of the study, all cases observed during one year period (collected for a different objective to determine the incidence of maternal near-miss, which was described elsewhere) [21], along with the corresponding three controls were included in the study. Women with a maternal near-miss condition and those without any complications during delivery were interviewed by a well-trained midwives and nurses using structured questionnaire. In addition, medical records were reviewed to gather relevant information. Information on socio-economic and demographic characteristics, reproductive health and obstetric history, and pre-existing medical conditions of the women were obtained from the participant’s record. The questionnaires were prepared following a thorough review of literatures. Obstetrics and Gynecology Ward, ICU and Emergency Gynecology Outpatient Department (OPD) of each hospital were visited to collect data. The questionnaires were pre-tested prior to the commencement of data collection to determine the appropriateness of the tool. Data collectors were given a three day training in order ensure consistency of data collection. The data were entered using Epi Info 7 software and analyzed using Stata version 13.0. The data were cleaned before analysis. The outcome variable of the study was maternal near-miss. The independent variables which were identified from literatures includes: (i) socio-economic and demographic characteristics (educational level, place of residence, ethnicity, religion, marital status, maternal occupation), (ii) reproductive health and obstetric history of the women (antenatal care booking, parity, history of caesarian section delivery, multiple pregnancies, history of abortion, history of stillbirth, early marriage, female genital cutting) and (iii) pre-existing medical conditions (previous hypertension, previous anemia, human immunodeficiency virus (HIV), history of cardiac problems, history of diabetes mellitus (DM)). Bivariate logistic regression was performed to examine whether there is a significant association between each individual independent variable and maternal near-miss. For each individual variable, the P-value, and unadjusted odds ratio (OR) with its 95% confidence interval, and the number and proportion of each variable of case and control were calculated. Multivariable conditional logistic regression model was used to examine the independent effect of the factors on the occurrence of maternal near-miss. The variables that were mentioned as factors of maternal near-miss in our literature review were classified as either distant or proximate factors. Socio-economic and demographic variables were taken as a distant factors. Whereas, the rest such as, reproductive health and obstetric history of the women and pre-existing medical conditions were considered as proximate factors. Since distant factors are conceptually related with the proximate factors for the occurrence of maternal near-miss, hierarchical model for the analysis is recommended [22]. Based on this hierarchical order, we have developed two models. All socio-economic and demographic variables with p <  0.2 in the bivariate logistic regression analysis were fitted with model 1. Those variables that were significant in model 1 (p <  0.05) were fitted with model 2. Model 2 contained those significant variables from model 1 and proximate variables. For each model and variables their adjusted OR, its 95% CI and P-value were calculated. The model fitness was estimated using stata’s fitstat command. Good fit was indicated by a significance value less than 0.05. Both models which were used to determine the factors associated with maternal near-miss were shown to be significant (p <  0.0001), which shows the models were best fit. Multicollinearity among independent variables was assessed by calculating variance inflation factors (VIF). No multicollinearity was suggested during the current analysis as all the calculated VIF were less than 10. We also defined some of the important independent variables. Educational level was categorized into illiterate (no formal education), primary (grade 1–8), secondary (grade 9–12), and higher education (> 12). Antenatal care visit was considered to be present if a woman reported to have ANC during current pregnancy. Monthly income was categorized into the lowest 25 percentile (below 68 USD), between 25 and 75 percentile (68–181 USD), and above 75 percentile (greater than 181 USD). Marriage before age of 18 was considered as early (based on jurisdiction). Pre-existing medical conditions such as chronic hypertension, anemia, HIV, maternal cardiac disease and DM were considered as present if the women reported their presence before the current pregnancy.