Health facility delivery service utilization and its associated factors among women in the pastoralist regions of Ethiopia: A systematic review and meta‐analysis

Abstract Background and Aims Utilizing health facility delivery services is one of the pillars of lowering maternal mortality. However, the coverage of health facility delivery service utilization continues to be uneven around the world. In Ethiopia, particularly among pastoralist regions, health facility delivery service utilization is less common. Therefore, the purpose of this study was to determine the pooled prevalence of health facility delivery service utilization and identify the associated factors among women in the pastoralist regions of Ethiopia. Methods A comprehensive systematic search was carried out in PubMed/MEDLINE, Hinary, Cochrane Library, Google Scholar, Google, and Ethiopian online university repositories. Studies were appraised using the JBI appraisal checklist. The analysis was done using STATA version 16. The pooled analysis was conducted using DerSimonian and Laird random‐effects model. I 2 test and Eggers & Begg's tests were used to assess the heterogeneity and publication bias, respectively. p < 0.05 was set to determine the statistical significance of all the tests. Results The pooled prevalence of health facility delivery service utilization was 23.09% (95% CI: 18.05%−28.12%). Have ANC visit during pregnancy (OR = 3.75, [95% CI: 1.84−7.63]), have information regarding maternal health service fee exemption (OR = 9.51, [95% CI: 1.41−64.26]), have a nearby health facility (OR = 3.49, [95% CI: 1.48−8.20]), and women attend secondary and above education (OR = 3.06, [95% CI: 1.77−5.29]) were found to be significant associated factors. Conclusions Health facility delivery service utilization is very low in pastoralist regions of Ethiopia, and ANC follow‐up, distance from the health facility, women's educational status, and information regarding maternal health service fees were identified as significant associated factors. Consequently, strengthening ANC services, introducing free health services to the community, and constructing health facilities for the nearby residents are recommended to improve the practice.

with big regional variations. 5 Almost 60% of African women give birth at home without the assistance of trained birth attendants, compared to less than 1% in developed nations. 6 In Ethiopia, the prevalence of health facility delivery service utilization varies between regions and location. [7][8][9] According to the 2016 Ethiopian Demographic and Health Survey (EDHS), the national prevalence of home delivery was 73%, with considerable regional variances ranging from 3% in Addis Ababa to 85% in Afar. 7 In more than 122 districts throughout Ethiopia, pastoralist regions occupy 61% of the total area. 10 According to studies, Ethiopia's pastoralist groups have a higher proportion of home deliveries than other populations. For instance, research in pastoralist areas of Dubti, Afar Region, Ethiopia revealed that 92.6% of births occurred at home. 11 Similar to this, 83.3% of deliveries were made at home as reported in the semi-pastoralist village of Malie District, Southern Ethiopia. 12 Despite the importance of giving birth in a health facility, pastoral women in Ethiopia use the service at a very low level. Therefore, it is imperious to determine the pooled prevalence and investigate factors associated with utilization of health facility delivery service among women in pastoralist regions of Ethiopia. Decision-makers and program managers might use it as a gate to design evidence-based strategies to overcome barriers hindering predominantly pastoralist women from giving birth in health facilities.

| Search strategy
A comprehensive systematic search for all relevant studies was carriedout in PubMed/MEDLINE, Hinary, Cochrane Library, Google Scholar, and Google. Besides, a search of online university repositories (University of Gondar and Addis Ababa University) and the reference list of already identified studies to retrieve additional articles were done.
Throughout the comprehensive literature search, the following search terms were used: "institutional delivery" OR "skilled delivery" OR "health facility delivery" OR "home delivery" OR "home birth" OR "giving birth at home" AND "associated factors" OR "determinants" OR "predictors" AND "women" OR "mother" AND "pastoralists" OR "pastoral community" AND "Ethiopia." Boolean operators "AND" and "OR" were used to combine the search terms as appropriate. The PRISMA standards were followed in the selection of publications, data extraction, and reporting of results for the review. 14

| Study selection and eligibility criteria
Every article that was found by the search approach was exported to Endnote version 7 to remove duplicate studies. Two sets of reviewers independently examined titles and abstracts for inclusion in full-text evaluation after duplicate papers were deleted. The defined criteria for choosing articles determined the resolution of the reviewers' disagreement.

| Exclusion criteria
Studies that are unavailable because they have not been published, are not retrievable from the internet, or have not had an email response from the respective authors have been omitted. After a full text examination, a study that did not provide enough information to determine the outcome of interest, non-full text papers, and secondary studies were also excluded from this analysis.

| Quality assessment and critical appraisal
The qualities of each article selected for inclusion in the systematic review (i.e., those that meet the inclusion criteria described above) were assessed by using the Joanna Briggs Institute (JBI) quality appraisal checklist adapted for cross-sectional, case-control, and cohort studies. 15 The assessment was done independently by two set of reviewers to assess the methodological quality of a study and to determine the extent to which a study has addressed the possibility of bias in its design, conduct, and analysis. The third reviewer dealt with any disagreements, and the reviewers discussed their differences until a consensus was established. Studies considered low risk whenever fitted to 50% and or above quality assessment score. 16

| Data extraction
A clear data extraction format was prepared by the principal author in Microsoft Excel. It included names of author, year of publication, the target population, the study region, study design, sample size, response rate, prevalence, number of success and failure in exposed and unexposed groups, and crude odds ratio with confidence Interval (CI) of associated factors. Data was extracted using this structured data extraction form by two separate sets of reviewers. The phase was repeated whenever differences in the extracted data were observed. As long as there were disagreements amongst the data extractors, a third reviewer was brought in, and the reviewers talked things out until they all agreed.

| Data synthesis and analysis
The extracted data in Microsoft Excel were imported to STATA version 16 for analysis. Pooled analysis was conducted using a DerSimonian and Laird random-effects model which assumes heterogeneity across studies. 17,18 I 2 test was used to check the heterogeneity of included studies. The values of 25%, 50%, and 75% were declared as low, moderate, and high heterogeneity, respectively. 19 With the evidence of heterogeneity, subgroup analysis was computed by considering study region as a grouping variable to further explore it. In addition, the sensitivity analysis was also performed using both metaninf and metaplot command to assess whether the pooled prevalence estimates were affected by single studies. Publication bias across studies was also checked using funnel plot and more objectively through Eggers and Begg's tests. Trim and fill analysis was conducted to overcome the publication bias. The effect of selected determinant variables was analyzed using separate categories of meta-analysis. The findings of the meta-analysis were presented using forest plot and odds ratio (OR) with its 95% CI. All the tests were two-sided, and a p < 0.05 was set to determine the statistical significance of the tests.

| Prevalence of health facility delivery service utilization
One case-control study was excluded on the prevalence estimation (Wondimu & Woldesemayat). 23 Then, using a DerSimonian and Laird random-effects model, the overall pooled prevalence of health facility delivery service utilization among women in pastoralist regions of Ethiopia was 23.09% (95% CI: 18.05%−28.12%) with significant heterogeneity between studies (I 2 = 95.36%, p < 0.001). The overall pooled prevalence of health facility delivery service utilization was presented using a forest plot (Figure 2).

| Heterogeneity and publication bias
The Cochrane I 2 values for this meta-analysis were 95.36%, p < 0.001, suggesting the presence of noticeably high heterogeneity. Subgroup and sensitivity analyses were performed to further investigate the evidence DESSIE ET AL. | 3 of 9 of heterogeneity. Based on a subgroup analysis using study regions, the prevalence of health facility delivery service utilization was high in Somali region 26.40% (95% CI: 18.76−34.05, I 2 = 84.39, p < 0.001) and low in Oromia region 13.90% with 95% CI: 11.49−16.31. A leave-oneout sensitivity analysis was also executed using both metaninf and metaplot command to investigate the influence of a single study on the overall magnitude estimate and suggesting that the finding was not unduly influenced by a single study. Thus, the point estimate of its omitted analysis lies within the CI of the combined analysis and the overall heterogeneity was not significantly changed by excluding a particular study (Figure 3).
The presence of publication bias was examined using funnel plots and tests (Egger and Begg). A funnel plot showed an asymmetrical distribution ( Figure 4A,B). Besides, the results of the Egger and Begg tests showed significant evidence of publication bias (p < 0.05).
Therefore, trim and fill analysis was conducted. After two studies were imputed using run R0 estimator for the number of missing study, the trim and fill analysis gave a pooled prevalence of 19.21 (95% CI: 13.93−24.50).

| Factors associated with health facility delivery service utilization
Factors such as women's level of education, ANC follow-up during pregnancy, knowledge about obstetric complications & health facility F I G U R E 1 Flow chart of the study selection process for this systematic review and meta-analysis.

T A B L E 1 Characteristics of included studies in this systematic review and meta-analysis.
Authors Publication

| DISCUSSION
This systematic review and meta-analysis is conducted to estimate the best available evidence for the prevalence and associated factors of health facility delivery service utilization in pastoralist regions of Ethiopia. As far as is known, no previous systematic reviews/metaanalyses were conducted to address this issue. Accordingly, the finding indicated that pooled prevalence of health facility delivery service utilization in pastoralist regions of Ethiopia was 23.09% (95% CI: 18.05%−28.12%). Despite efforts to achieve the World Health Organization's recommendation that every pregnant woman give birth at a health facility with trained birth attendants, the rate in Ethiopia is still low, especially among pastoralist regions. This implies that the Ethiopian government is expected to do more in this area to increase the use of institutional delivery services and improve maternal and child health in pastoralist regions and the country at large.
F I G U R E 4 (A) A funnel plot with a pseudo 95% confidence limit used to test for publication bias. (B) A funnel plot with a pseudo 95% confidence limit after a trim-and-fill analysis in which two studies have been imputed.
This study's finding were much lower than those of a study based on the 2016 Ethiopia Demographic and Health Survey (32.8%) 28  This study demonstrated that there were a significant association between health facility delivery service utilization and ANC followup. Women who have ANC visit during her pregnancy were 3.75 times more likely to give birth at health facility than their counterparts. This finding is consistent with a systematic review from Ethiopia, 30,31 and primary studies conducted in Kenya 32 & Eretria. 33 This could be because of ANC is the most convenient way for mothers to learn more about pregnancy and delivery. Consequently, women who had no ANC visits may be less aware of birth preparedness and complication readiness plans, pregnancy danger signs, and the dangers of giving birth at home, which may make them to give childbirth at home. Furthermore, this study also revealed that women with poor knowledge of obstetric complications and health facility delivery service utilization, as well as had no information on maternal health service free exemption, were less likely to give birth in a health facility.
As this study indicates, women who had access to a nearby health facility were also 3.49 times more likely to give birth in health facility than women who did not have access to a nearby health facility. This might be due to women who reside in nearby health facilities will have no problem with transportation to attend health facility delivery at any time. Since there may be unpredictable labor and a lack of support from their spouse due to seasonal migration out of their house to feed and get water for their cattle in pastoralist regions, women who do not have access to a nearby health facility may have difficulty having a health facility delivery.
Besides, women resided to the nearby health facility might have different access of maternal health services such as health education and ANC services.
The educational status of women was also found to be a significant predictor of their health facility delivery service utilization practice in this study. Women with a secondary and above education had higher odds of giving birth in health facility than those with no formal education. This is in line with studies from Uganda, 34 Bangladesh, 35 and Haiti. 36 The reason behind might be educated women are more aware of the risks connected with home delivery and have a better understanding of service availability. Furthermore, learned women may be more concerned for their health & have a good decision-making capacity making them more likely to give birth at health facility.
Significant heterogeneity and publication bias had been reported in this study. Nevertheless, the sensitivity analysis in both Metaninf and Metaplot command showed that no single study had a significant effect on the pooled prevalence. In addition, studies indicated that a high I 2 value is not always synonymous with high heterogeneity, and, like this study, when only a small number of studies with no true heterogeneity are included in the meta-analysis, I 2 will overestimate heterogeneity. 37,38 It could not be therefore difficult to utilize this paper for different purposes. The publication bias has also been overcome by conducting trim and fill analysis. Hence, the pooled prevalence of 19.21% that resulted from the trim and fill analysis can also be considered.

| LIMITATIONS OF THE STUDY
It is important to interpret the review's findings by considering the following limitations. This systematic review and meta-analysis includes almost exclusively cross-sectional articles. The factors and the outcome variables cannot therefore be linked to one another in time. Furthermore, as a result of the included studies' differing T A B L E 2 Factors associated with health facility delivery service utilization among women in pastoralist regions of Ethiopia.