The effect of antenatal care on perinatal outcomes in Ethiopia: A systematic review and meta-analysis

Background The estimated annual global perinatal and neonatal death is four million. Stillbirths are almost equivalent to neonatal mortality, yet they have not received the same attention. Antenatal care is generally thought to be an effective method of improving pregnancy outcomes, but its effectiveness as a means of reducing perinatal mortality has not been evaluated in Ethiopia. Therefore, we will identify the pooled effect of antenatal care on perinatal outcomes in Ethiopia. Methods Medline, Embase, Cinahl, African journal online and Google Scholar was searched for articles published in English language between January 1990 and May 2020. Two independent assessors selected studies and extracted data from eligible articles. The Risk of Bias Assessment tool for Non-Randomized Studies was used to assess the quality of each included study. Data analysis was performed using RevMan 5.3. Heterogeneity and publication bias were assessed using I2 test statistical significance and Egger's test for small-study effects respectively. The random effect model was employed, and forest plot was used to present the risk ratio (RR) with 95% confidence interval (CI). Results Thirteen out of seventeen included studies revealed antenatal care utilization had a significant association with perinatal outcomes. The pooled risk ratio by the random-effects model for perinatal death was 0.42 (95% CI: 0.34, 0.52); stillbirth 0.34 (95% CI: 0.25, 0.46); early neonatal death 0.85 (95% CI: 0.21. 3.49). Conclusion Women who attended at least one antenatal care visit were more likely to give birth to an alive neonate that survives compared to their counterpart. Therefore, the Ethiopian Ministry of health and other stakeholders should design tailored interventions to increase antenatal care utilization since it has been shown to reduce perinatal mortality.

Introduction was registered on the PROSPERO prospective register of systematic reviews after piloting the study selection process (registration number PROSPERO 2020: CRD42020188340).

Eligibility criteria
Assessment for eligibility was conducted and studies were included in this review if (i) the study involved a delivering/laboring women or newborn babies or women of child-bearing age or pregnant women or postpartum women; (ii) the study reported the outcomes (perinatal death, stillbirth, early neonatal death); (iii) the ANC utilization was considered as factors/exposure for the outcomes; (iv) the study was done in the perinatal period and the author(s) defined perinatal outcomes (perinatal mortality) as death of newborn between 28 weeks' of gestation and seven days postpartum; (v) it was an observational study design (cross-sectional, case-control or cohort study design) and (v) English language article.
We excluded studies from the review that focused only on the number of ANC visits based on full-text assessment.
PICO. Population: Newborn after 28 weeks' gestation and survived seven days postpartum.
Intervention: utilized at least one ANC visit. Comparison: Newborns whose mothers received at least one ANC service as compared to newborns whose mothers did not.
Outcome: Newborn death during perinatal period (from 28 weeks' of gestation to 7 days postpartum).

Study selection
The study selection involved several steps. First, the title and abstract were selected independently by the review authors using the inclusion criteria. Second, after removing the duplicates, the full reports of all titles that met the inclusion criteria were independently identified by review authors. Third, the review authors screened the full text reports to decide whether the studies meet the eligibility criteria. Finally, any disagreements among review authors were resolved through discussion or review authors who did not participate in step one thru three decided whether to include or exclude the article. An attempt was made to meet study authors for additional information by email and in order to have put reasons for excluding studies (Fig 1). third or fourth author was consulted to decide on the disagreement. For each study, the first author's last name, publication year, design, setting, sample size, study period, sample age, the definition of outcomes, population, outcome and comparison groups were documented.
In this review, our evaluation of perinatal outcomes related to the death of the newborn from 28 weeks' of gestation to seven days postpartum (i.e., fetal death, stillbirth, and early

Individual study's risk of bias
The review authors assessed all selected studies rigorously for inclusion in the review. The Risk of Bias Assessment Tool for Non-Randomized Studies (RoBANS) [65] was used to assess the quality of each included study. Studies were evaluated across six groups (selection bias, attrition bias, detection bias, performance bias, confounding bias and reporting bias). Each domain was assigned one of three possible groups for each of the involved studies: 'low risk', 'high risk' and 'unclear'. RoBANS is shown in S4 Table. Synthesis and analysis of data Statistical analysis was carried out in RevMan version 5.3. A DerSimonian and Laird random effects model [66] was used to measure ANC's overall effect on perinatal mortality and the risk ratio was measured with a 95% confidence interval. We calculated the I 2 statistic which describes the percentage of total variation among studies to assess heterogeneity among studies. An I 2 statistical value of 25%, 50% and 75% representing low, moderate and high heterogeneity respectively [67]. A p-value less than 0.05 was considered as statistically significant both for risk ratio and heterogeneity. Sensitivity analysis was conducted to assess the stability of results and test individual study effects on the meta-analysis using leave one out method. Furthermore, possible sources of heterogeneity were explored using subgroup analysis. Egger's test for small-study effects was used to investigate potential publication bias (p-value > 0.1) [68].

Characteristics of studies
The review included studies from all regions in Ethiopia; the majority were from Amhara and Oromia. Nine cross-sectional, six case-control and two cohort studies were included in the meta-analysis. The sample size of the studies ranged from 300 to 12560. Among the included studies, 5 and 12 were community-based and facility-based, respectively. A total of 51729 study samples were included, of which 2951 newborns died during the perinatal period, making the perinatal mortality rate 41 per 1000 total births (total deliveries, total stillbirths and total early neonatal deaths), excluding case-control studies in which total numbers of live births at the time of the study were unknown. Similarly, the stillbirth rate and early neonatal mortality rate were 38 per 1000 total births (stillbirths and live births) and 19 per 1000 live births. Table 1 displays the characteristics of the 17 included primary studies.

Individual study's risk of bias
The risk of bias assessment for all included studies is shown in Table 2. The risk of bias in selection of participants into the study was low for all studies. The bias due to missing or incomplete data was low in most of the studies, although a few studies have unclear explanation. The performance bias during measurement of exposure variable was low in fourteen and unclear in three studies. However, the risk of detection bias was high in all studies. The risk of confounding bias was low in thirteen, high in three and unclear in one study. The bias due to reporting of results was low in fifteen and unclear in two studies. See S4 Table. Pooled effect size of ANC on perinatal outcomes Among the seventeen studies included in the analysis, thirteen studies with at least one ANC visit showed statistically significant associations with perinatal outcomes, whereas four studies had no statistically significant association. Similarly, the pooled effect size for perinatal death by the random-effect model was 0.42 (95% CI: 0.34, 0.52) for babies born to women who received at least one ANC follow-up as compared to newborns whose mothers did not receive any ANC follow-up (Fig 2). Furthermore, the pooled stillbirth and early neonatal death effect size by random effects model was 0.34 (95% CI: 0.25, 0.46) and 0.85 (95% CI: 0.21. 3.49) respectively.

Heterogeneity of the studies
There was overall substantial heterogeneity across studies (I 2 = 87%, p-value < 0.001), as well as within subgroups for sample size, design and place. Heterogeneity that was present in the overall meta-analysis was partially explained with stratification by study design and place. For example, in a subgroup analysis, cohort studies' (RR = 0.83[95% CI: 0.67-1.02]; p-value = 0.45 for heterogeneity test, I 2 = 0%) and community-based studies (RR = 0.64[95% CI: 0.51-0.80]; p-value = 0.23 for heterogeneity test, I 2 = 29%) were not statistically heterogeneous (pvalue > 0.10); however, heterogeneity was present when the subgroup analysis was performed by sample size (Table 3).
Sensitivity analysis was performed for the outcome variable to observe a significant change in risk ratio and confidence interval. The meta-analysis resulted in no substantial difference in the overall risk ratio during the sequential removal of each study from the analysis. For instance, when a statistically insignificant study [14] and those study with wide confidence interval were excluded [101], the risk ratio of the effect of ANC did not change significantly or are within the confidence interval of pooled effect of ANC (0.32, 0.52). An Egger's test for small-study effects showed no publication bias (p-value = 0.49). Therefore, there was no significant threat to the validity of the review.

Discussion
The purpose of this review was to evaluate the effectiveness of focused ANC as a means of reducing perinatal mortality among women (pregnant, delivering, postpartum, and mothers) in Ethiopia. Seventeen eligible primary studies were identified evaluating ANC with a range of populations including pregnant women, laboring women and postpartum mothers and their perinatal outcomes. Literature throughout Ethiopia support the benefits of ANC's that provided by skilled attendants for the health of newborns. To improve ANC's effectiveness, numerous approaches and strategies have been employed in LMICs [106][107][108]. The focused ANC approach, developed in the 1990s by WHO has been implemented by most LMICs including Ethiopia [109,110].
The perinatal mortality and stillbirth rate were 41 and 38 per 1000 total births respectively in this meta-analysis which were slightly higher than the perinatal mortality rate in SSA (34.7 per 1000 total births) [4] however, lower than the pooled perinatal mortality rate (51.3 per 1000 total births) and slightly higher than stillbirth rate in Ethiopia (37 per 1000 total births) [6]. The review in SSA utilized only demographic health survey data whereas the pooled perinatal mortality in Ethiopia included both demographic health survey and study data. The difference may be attributed to not only a variation in the study nature, sample size, and setting but also maternal and child health utilization and access to quality maternal and newborn health services [6]. However, early neonatal mortality rate was 19 per 1000 live births in this review which was lower than systematic reviews found in Ethiopia (30 per 1000 live births). A global multipartner movement to end preventable maternal and newborn deaths and stillbirths, setting a target for national stillbirths less than 12 per 1000 live births and will reduce death and disability continuously, ensuring no newborn is left behind in all countries by 2030 [2,19]; however, this review, along with the EDHS [5] and another systematic review in Ethiopia [6] revealed that the perinatal mortality has remained stable for two decades. Using this study's perinatal mortality rate as a benchmark, the annual rate of reduction (ARR) must increase to achieve The Every Newborn Action Plan.

PLOS ONE
The effect of antenatal care on perinatal outcomes in Ethiopia Pregnant women's attendance of at least one ANC follow-up had a statistically significant effect on perinatal mortality. This study found a 58% and 66% lower risk of perinatal mortality and stillbirth among women who attended at least one ANC by a skilled attendant in Ethiopia. The basic finding of this study was even limited ANC (as little as one visit) leads to better newborn outcomes compared with no ANC, and encouraging pregnant women to seek ANC would significantly impact perinatal mortality rate (PMR) and would be an important strategy to incorporate in planning initiatives aimed at reducing PMR; this appears to be consistent with studies from another countries [56,111]. The finding was also in line with the global network's population-based birth registry results in Africa, India, Pakistan and Guatemala [19]. A review in Asia also revealed a protective effect on perinatal mortality for women who used ANC and health facility delivery [39].
Antenatal care utilization and delivery at a health facility by a skilled attendant [112] who provides quality care are established as an intervention to reduce perinatal mortality [113][114][115]. This may be due to the women receiving interventions during her pregnancy, [116][117][118] which have a positive effect on lowering mortality; ANC also has an indirect impact since those women attending ANC are more likely to have a skilled birth attendant [39,112,119,120] hence, their newborns have access to basic neonatal resuscitation [121,122] which prevent perinatal mortality. Therefore, receiving high quality and an accessible health care services to reduce perinatal mortality is critical for pregnant women [123]. Skilled training of health care providers and resources of local primary healthcare facilities should be strengthened [124].
The factors associated with perinatal mortality (preterm labor, hypertensive disorders of pregnancy, intrauterine growth restriction, gestational diabetes) can be identified in the prenatal period, thus reinforcing the need to upgrade the continuum of care from initiation of ANC to complication management at health facilities [113,125].
A comprehensive database search was conducted to include all pertinent studies, and subgroup analysis was conducted to determine whether any specific study level factor described the outcomes. The large sample size of the analysis, could detect the effect of ANC on perinatal outcomes since the review included all studies conducted in Ethiopia. As a limitation, the systematic review and meta-analysis were based on English language and observational studies associated with inherent biases. We were unable to pool the overall effect of ANC for those studies that were based on the number of visits, since they did not define zero visits and therefore that were excluded. The study authors defined stillbirth and early neonatal death based on gestational age and the days of life of the newborn. The future research should focus on visits and specific ANC interventions that may affect perinatal outcomes.

Conclusion
This review showed that women who received at least one ANC follow-up by a skilled attendant were less likely to experience perinatal mortality than those who did not. Thus, increasing a woman's ANC utilization by a skilled attendant is mandatory in Ethiopia to reduce perinatal mortality. Furthermore, to address perinatal mortality in the country, strategies should focus on women's mobilization to seek ANC services and facility-based deliveries.
Supporting information S1