Magnitude and determinants of obstetric case fatality rate among women with the direct causes of maternal deaths in Ethiopia: a national cross sectional study

This study utilized a large public dataset available at the Ethiopian Public Health Institute (EPHI). The EPHI is a large-scale research institute, which undertakes various basic and operational researches on the priority health agenda to generate and disseminate scientific knowledge that help in improving the health of the population. The EPHI conducted a national cross-sectional survey titled ‘assessment of emergency obstetric and newborn care in Ethiopia’ in 2016. With this project, different data about maternal and neonatal health indicators were collected from health facilities’ logbooks.

In the 2016 national survey, the EPHI included all public and private health facilities found in the country, which were deemed functional and offered delivery services. There were 3804 health facilities that were included in the survey to collect data about maternal and neonatal health indicators. However, the current study analysed the data of all governmental/public and private hospitals, of which there were 293 at the time of data collection. The data that were collected from 30 (10.2%) specialized hospitals, 103 (35.2%) general hospitals and 160 (54.6%) primary hospitals were included.

The data about number and type of obstetric complications, maternal deaths and all deliveries conducted throughout 2015 among all hospitals found in Ethiopia were included in the analysis. The data of all public/government and private hospitals were included with the exclusion of maternal data collected from the lower level facilities including clinics and health centres. In Ethiopia, women who have experienced obstetric complications are often referred to hospitals where they can receive specialized treatment. Therefore lower level health facilities were not included in the analysis since information of the majority of women who sustained obstetric complications can be obtained from hospitals.

The EPHI collected data about the performance of the Emergency Obstetric and Newborn Care (EmONC) signal functions at a particular hospital using a standardized questionnaire. The EPHI collected the data from hospitals’ registers and records within the reference period of January–December 2015 inclusive. The available dataset included the number and mode of deliveries, number of women who experienced different types of direct obstetric complications and associated maternal deaths that occurred in each hospital. The dataset also had different facility-specific characteristics including type of hospital, managing authority, location of the hospitals and fulfilment of basic facility infrastructure.

The primary outcome variable is the DOCFR, the proportion of women who are admitted to a hospital or an emergency obstetric care facility with major direct obstetric complications, or who develop such complications after admission, and die before discharge [9]. Then hospitals’ specific characteristics such as type, managing authority, and location of the hospitals and fulfilment of basic facility infrastructure were assessed for whether they predict the magnitude of the DOCFR in hospitals or not.

Initially, descriptive statistics including means, percentages, frequency tables and ratios were used to describe facility specific characteristics and the magnitude of the DOCFR. Then logistic regression analysis was conducted to examine the association between the DOCFR and hospital-specific characteristics. Maternal deaths due to direct obstetric complications that were reported in all hospitals during 2015 were included to determine the magnitude of the DOCFR. For the analysis purpose, the magnitude of the DOCFR was calculated for each hospital according to the WHO definition, the proportion of women admitted to a hospital or an EmOC facility with direct obstetric complications, or who developed such complications after admission, and died before discharge [9]. Using the number of women who experienced direct obstetric complications as the denominator, we calculated the DOCFR, the numerator being the number of maternal deaths attributed to direct obstetric causes. Hospitals were dichotomized based on the WHO recommendation of DOCFR: hospitals with DOCFR ≥1 (hospitals with unfavourable outcomes) and hospitals with DOCFR < 1 (hospitals with favourable outcomes). First, we conducted univariate logistic regression analysis to assess the presence of association between the magnitude of the DOCFR and hospital-specific characteristics. Facility-level variables that were significantly associated with the magnitude of the DOCFR in the univariate analysis were successively included in a multivariate model, using a backwards-stepwise elimination to select variables. For all models fitted, we used generalised linear model for the multilevel analysis by using Stata version 15 statistical software. The degree of association between variables was measured using an adjusted odds ratio with 95% confidence interval and association was declared significant at p value < 0.05.

The Ethiopian national survey titled ‘Assessment of emergency obstetric and newborn care in Ethiopia’ was granted ethical approval from the Scientific and Ethical Review Office of the EPHI (approval number: EPHI-6-13-728) on 6-Jun-2016. Ethical approval for this analysis was obtained from the Human Research Ethics Committee (HREC) of The University of Newcastle, Australia (approval number: H-2018-0245) on 15-Aug-2018. The protocol developed to conduct this analysis had received ethical approval from the Scientific and Ethical Review Office of the EPHI before the data were accessed (Protocol number: EPHI-IRB-048-2018) on 25-July-2018.

The current study presented the magnitude of DOCFR and associated factors in 293 hospitals found in nine regional states and two city administrations of Ethiopia. Nearly a fourth (N = l73; 24.9%) of the hospitals were from Oromia region and 30 (10.2%) of the hospitals were referral specialized hospitals. Two hundred and thirty-five (80.2%) hospitals were public hospitals while the remaining were private facilities. Geographically, 272 (92.8%) hospitals were located in urban centres. The majority of the hospitals, (N = 207; 70.6%) had initiated the maternal death review and reporting system (MDRS); a health reform used to continuously notify, review, analyse and respond to maternal deaths (Table 1).

Overall, 335,054 deliveries were conducted in all hospitals in Ethiopia and 68,002 women experienced direct obstetric complications, giving the crude incidence rate of 203 per 1000 deliveries (68,002/335054). Major direct obstetric complications accounted for 40,080 (58.9%) of the cases while 27,922 (41.1%) of women experienced other direct obstetric complications (Fig. 1). Prolonged labour was the leading cause of major direct obstetric morbidity, accounting for 23.4% of the total cases. Hypertensive disorders was the second most frequent cause of major direct maternal morbidity as 11.6% of mothers experienced this condition (Table 2).

Of women who experienced direct obstetric complications, 435 died, resulting in crude national DOCFR of 0.64 (95% CI: 0.58–0.70) per 100 mothers (435/68,002). That is, a general ratio of 1 death for every 156 cases of direct obstetric complications was observed. Hypertensive disorders was the leading cause of direct maternal mortality, accounting for 27.8% of the overall maternal deaths. Obstetric haemorrhage was the second most frequent direct cause of maternal mortality, accounting for 104 (23.9%) of the overall maternal deaths. The majority of deaths due to obstetric haemorrhage (N = 81; 77.8%) were caused by postpartum haemorrhage. On the other hand, postpartum haemorrhage accounted for 18.6% of the overall maternal deaths. The DOCFR varied considerably with the complications that occurred. The highest DOCFR was observed in postpartum haemorrhage 2.88% (95% CI: 2.29–3.57%) followed by ruptured uterus 2.71% (95% CI: 1.95–3.65%). The DOCFR due to hypertensive disorders 1.53% (95% CI: 1.27–1.82%) and postpartum sepsis 1.16% (95% CI: 0.67–1.85%) were also in excess of the recommended level (Table 2).

Although the crude national DOCFR was within the recommended threshold of less than 1%, considerable regional variation occurred. The DOCFR ranged from 0.27% (95% CI: 0.20–0.37%) in Addis Ababa city administration to 3.82% (95% CI: 1.42–8.13%) in the Gambella region. The DOCFR was significantly higher in the Afar 1.39 (95% CI: 0.45–3.22), the Harari 1.08 (95%CI: 0.61–1.78) and the Somali 1.69 (95% CI: 1.13–2.42) regions (Table 3).

Bivariate and multivariate logistic regression analyses were conducted to analyse the hospital specific determinants of the magnitude of DOCFR. In the final multivariate analysis, whether formal payment was required for the service or not, types and managing authority of the hospitals were found to be significantly associated with the magnitude of DOCFR after possible confounders were controlled for. The odds of observing DOCFR≥1 was reduced by 58% in general hospitals as compared to primary hospitals [AOR = 0.42; 95%CI (0.21, 0.85)]. The magnitude of DOCFR was also lower in specialized hospitals compared to primary hospitals although the difference was not statistically significant. Regarding managing authorities, the odds of DOCFR≥1 was about two times higher in public/governmental hospitals than private counterparts [AOR = 2.46; 95% CI (1.28, 4.32)]. Payment was also identified as a factor in that the odds of DOCFR≥1 was 64% higher in hospitals where formal payment was not required for the service than in hospitals where mothers need to cover service fee [AOR = 1.64; 95%CI (1.04, 3.19)]. This might be explained by differences in quality of care between public and private hospitals. In private hospitals where clients need to cover costs associated with health service, better quality of obstetric care were provided than it was in public hospitals (Table 4).

This study has several strengths. In this study, we used a national representative data collected from all hospitals found in Ethiopia. Therefore, the findings of this study are highly generalizable and can apply to all regions of Ethiopia. These findings might also be useful for other low-income countries with similar demographic and economic characteristics. From the commencement of the survey, experts from national and international partners were involved in the data collection and management processes hence the analysed data were of high quality. In the meantime, this study was suffered from the usual limitation of a cross-sectional study in that the causal relationship could not be concluded. This study did not investigate the maternal morbidity and mortality happened at the lower level of care including health centers and clinics. Although the majority of women who experienced obstetric complications are often referred to hospitals, few women might sometimes seek care from lower level facilities. Therefore, the findings of this study can only be generalized to hospitals in the country. The maternal deaths reported in this study represent only the numbers, which were registered on the hospital’s logbook. Therefore, the number of deaths might be affected by maternal deaths that happened at home, which accounts for the majority of maternal deaths in low-income countries. Similarly, we faced difficulties in comparing our findings with similar reports since there were limited similar studies available in Ethiopia.