Characteristics of included studies
The search strategy retrieved 104 from PubMed, 23 from Cochrane library, 16 from Web of Science, 55 from Google, 69 from Google Scholar and 05 from other sources. After duplication removed, 97 remained. Finally, 43 studies were screened for full-text review and 24 were included to the prevalence and/ or associated factors analysis (Figure 2).Seven (7) Studies were found in Amhara region [22, 23] [24-28], Nine(9) in Oromia[29-37], two in Afar [38, 39], one in Bienishangul-Gumuz [40] ,three(3) in Southern Nation Nationalities and Peoples region (SNNPR) [41-43],one in Gambiela [44] and one in Tigray&Oromiya [45].
Twenty two of the studies were done by cross-sectional study design and the other two study was conducted by case-control and cohort study design. Regarding year of publication, five (5) studies were published in 2018, eleven (11) studies were published between 2015 and 2017, and eight (8) studies were between 2011 and 2014.A total of 19,969 women who gave birth in the last two years preceding the current survey were included in the study and among which 6,672 women gave birth in health institution (Table 1).The selection process is illustrated in Figure 2.After removing the duplication among 97 articles fifty four (54) were excluded because they did not meet the inclusion criteria following reading titles and/ or abstracts (because of their title are not related to review topic). Nineteen (19) more articles were excluded after reading the entire article because outcome variable and study subject are not the same with the review. Thus, twenty four (24) studies (Table 1) satisfied the specified inclusion criteria and constituted the basis of this investigation.
Quality of studies
The JBI quality appraisal criteria established for cross-sectional, case-control, and cohort studies were used. The studies included in this systematic review and meta-analysis had no considerable risk (has low risk). Therefore, all the studies were considered [22-45] (Table 1).
Studies and Participants
Table 1 summarizes the characteristics of the 19,969 women’s who gave birth within the past two years preceding the survey encompassed by the twenty four included studies. The year of study publication ranged from 2010 to 2018. The design of 22 of the studies examined was cross-sectional, one prospective follow up, and one case control. Seven studies were reported from Amhara, nine studies from Oromiya, three from SNNPR, two from Afar, one from Gambiela, one from Binishangul Gumuz and the remaining from Tigray&Oromiya. All twenty four studies had an enrollment period exclusively after 2010, when institutional delivery service utilization with at least one factor becomes available. All papers stated the proportions of institutional delivery service utilization. A large proportion of the participants (69 %) in these studies were not use institutional delivery for their last birth.
Meta-analysis
Heterogeneity of studies
Heterogeneity test for the proportion of the review indicated I2=0.0%, no variability was observed among the included studies hence fixed effect model was assumed in the analysis.
Prevalence of Institutional delivery service utilization
The estimated overall prevalence of institutional delivery service utilization (IDSU) is presented in a forest plot (Figure 3). The overall prevalence of IDSU was 31% (95% CI: 0.30-0.312; I2 = 0.000%).
Publication bias
A funnel plot showed a symmetrical distribution (Figure 4). Egger’s regression test p-value was 0.193, which indicated the absence of publication bias.
Subgroup analysis
The subgroup analysis based on the region and year of publication was done. Based on this, the prevalence of IDSU found to be 47%,39% and 24% in Amhara, SNNPR and Oromiya studies respectively, on the other hand the prevalence of IDSU found to be 37% between the year of 2015-2016 and 2017-2018 and 24% between the year of 2011-2014 (Table 2).
Sensitivity analysis
There is no any study out of the confidence bound mean that all study has almost equal influence on the pooled proportion (Figure 5).
Time-trend analysis
The time-trend analysis showed that the prevalence of IDSU is increase from 24% (95%CI 0.23-0.25) in 2011-2014 to 37 %( 95%CI 0.36-0.38) in 2017-2018. However, the pooled prevalence from year to year is increasing significantly (p-value = 0.00) (Figure 6 ).
Associated factors
Based on this review, IDSU in Ethiopian context is associated with the three dynamics factors of Andersen Healthcare Utilization conceptual model i.e. predisposing factors (The socio-cultural characteristics of individuals that exist prior to their illness.), enabling factors (the logistical aspects of obtaining care), and need (The most immediate cause of health service use, from functional and health problems that generate the need for health care services).
Predisposing factors
In a family where husbands had been a decision maker on a place where to gave birth is 54 %( AOR=0.46; 95% CI 0.18, 1.19) less likely to gave birth at health institution [31].
Attitude towards Institutional delivery
Women who had favorable attitude on institutional delivery service utilization were 2.8 times (AOR = 2.8; 95% CI 1.6, 4.8) more likely to use institutional delivery service as compared to those women having unfavorable attitude [35].
The pooled effects of three studies[27, 35, 38] showed that favorable attitude towards institutional delivery was a significant associated factors with IDSU.The pooled effect of this three studies revealed that those mothers who had favorable attitude towards institutional delivery service utilization were 2.8 times more likely to use the services. Heterogeneity test indicated moderate variability, I2=48.4% i.e. moderate heterogeneity, hence random effect model was assumed in the analysis. Sensitivity analysis was done, and no change was observed in the overall OR (Figure7).
Maternal Knowledge
Knowledge regarding danger signs during pregnancy and institutional delivery service increased the probability of utilizing health institution for delivery service. Women who had good knowledge on institutional delivery service utilization were 2 times (AOR = 2.1; 95% CI 1.32, 4.87) more likely to use institutional delivery service as compared to those women having poor knowledge [38] .
The pooled effect of eleven studies[24, 26-29, 34, 35, 38, 40, 41, 43] showed that women who were knowledgeable were 3 times (AOR=3.04;95%CI 1.76-5.24) more likely to give birth in health institution than those women who were not knowledgeable. Heterogeneity test indicated high variability, I2=77.3% i.e high heterogeneity, hence random effect model was assumed in the analysis. Sensitivity analysis was done, and no change was observed in the overall OR (Figure8).
Maternal age at first pregnancy
Women whose age 15-24 (AOR=4.02; 95%CI 2.07-8.55) and 25-34 (AOR=2.21; 95% CI 1.32-3.69) at first pregnancy were more likely to use institutional delivery service[32].
The pooled effect of four studies[26, 27, 32, 40] showed that women who had their first pregnancy between the age of 15-24 years were 3.6 times more likely to give birth in health institution than those who became pregnant after 35 years of age (AOR= 3.592; 95% CI 2.269 5.686 ). Heterogeneity test indicated I2=0.0%, no variability was observed among the included studies hence fixed effect model was assumed in the analysis. Sensitivity analysis was done and illustrated stability of overall OR (Figure 9).where as women who had their first pregnancy between the ages of 25-34 years were not became significantly associated factors (Figure 10).
Maternal Education
This review indicated that significant association between women’s’ educational status and utilization of institutional delivery service. Women who can read & write (AOR=1.75; 95% CI 1.21-2.54), Primary level (AOR=2.23;95%CI 1.39-3.59) and secondary & above(AOR=2.4;95%CI 1.09-5.52) more likely to use institutional delivery service [23].
The pooled estimate findings of five [23, 25, 26, 31, 34] review also indicated significant association between mothers’ educational status and utilization of institutional delivery service. Mothers who can read and write were 1.6 times more likely to give birth at health institution as compare to who can’t read& write(AOR=1.62;95%CI 1.178-2.238). Heterogeneity test indicated I2=0.0%, hence fixed effect model was assumed in the analysis. Sensitivity analysis did not bring significant change in the overall ORs (Figure 11).
Similarly women’s who attend Primary school were almost two times more likely to give birth at health institution than who can’t read and write (AOR=1.953; 95% CI 1.42-2.685). Heterogeneity test indicated I2=0.0%, hence fixed effect model was assumed in the analysis. Sensitivity analysis did not bring significant change in the overall ORs (Figure 12).
Women’s who attended secondary and above educational level were almost three times more likely to give birth at health institution as compare to women’s who can’t read and write (AOR=2.915;95% 1.881-4.517. Heterogeneity test indicated I2=0.0%, hence fixed effect model was assumed in the analysis. Sensitivity analysis did not bring significant change in the overall ORs (Figure 13).
Maternal occupation
Women’s who is not house wife in occupation (AOR=2.5; 95 % CI 0.8-8.4) were more likely to use institutional delivery service [25].
The pooled estimate findings of five [23, 25, 26, 32, 34, 36, 41] review showed insignificant association of utilization of institutional delivery service with mothers’ occupation (AOR=1.095; 95% CI 0.574 2.089). Heterogeneity test indicated I2=65.1%, moderate variability hence random effect model was assumed during analysis. Sensitivity analysis was done, and no change was noted on overall OR (Figure 14).
Parity
The combined estimate findings of five [26, 31, 32, 39, 42] review shows that the number of children the women delivered was not significantly associated with institutional delivery service utilization(AOR= 1.484; 95% CI 0.746 2.951) and (AOR=1.10;95%CI 0.871 1.392) for women who have one and 2-4children respectively. Heterogeneity test indicated for women who have one child is I2=81.2%, hence random effect model was assumed in analysis Sensitivity analysis was done, and no significant change was observed in overall OR (Figure 15), where as for women who have 2-4 children I2=0.0%, hence fixed effect model was assumed in analysis.
Enabling factors
Availability of Information Source
The chance of delivering in health institution among those women have the access of information source were 1.8 times higher than those women who have not the access of information source (AOR = 1.797; 95%CI 1.160 2.783). Heterogeneity test indicated I2=71.9%, hence random effect model was assumed in the analysis. Sensitivity analysis was done and revealed the stability of overall effect size (Figure 16).
Place of residence
The pooled estimated findings of seven [25, 27, 28, 33, 34, 40, 42] review showed that place of residence as defined rural and urban were the enabling factors that determined utilization of institutional delivery service. Women from urban area were 3.8 times more likely to deliver in health institution than women from rural area (AOR= 3.844; 95%CI 1.313 11.253). Heterogeneity test indicated I2=91.3%, hence random effect model was assumed in the analysis. Sensitivity analysis was done, and no change was noted in the overall OR (Figure 17).
Distance to health facility
The findings of five [26, 31, 32, 39, 42] review shows that distance to health facility was not significantly associated with institutional delivery service utilization(AOR= 1.451;95% CI 0.966 2.180). Heterogeneity test indicated I2=77%, hence random effect model was assumed in analysis. Sensitivity analysis was done, and no significant change was observed in overall OR (Figure 18).
Need factors
ANC follow up
The combined finding of thirteen study[23, 24, 27-29, 31-35, 38, 39, 45] review showed that ANC follow up was significantly associated with institutional delivery service utilization. Women who had ANC follow-up 2.6 times more likely to utilize the service than those who did not visited ANC (AOR= 2.574; 95%CI 1.460 4.540). Heterogeneity test indicated I2=93.3%, hence random effect model was assumed in the analysis. Sensitivity test was done, and no change was noted on overall OR (Figure 19).
Frequency of ANC follows up
The combined effect of four study [28, 34, 40, 41] shows that attending ANC follow up four or more was significantly associated with institutional delivery service utilization. Women who attended ANC four or more times were 4 times more likely to give birth in health institution than those women who attended ANC service below four or less than four times (AOR= 4.039 ;95%CI 1.212 13.462). Heterogeneity test indicated I2=90.6%, hence random effect model was assumed in the analysis. Sensitivity test demonstrated stability of the overall OR (Figure 20).
Place of birth the most recent birth
The pooled effects of four studies [26, 28, 30, 42] showed that women’s who have an experience a birth of the most recent birth in health institution were 8.4 times more likely to utilize the service again than those who gave the most recent birth in home (AOR=8.44;95% CI 5.748 12.393).Heterogeneity test indicated I2=0.0%, no heterogeneity, hence fixed effect model was assumed in the analysis. Sensitivity analysis did not bring significant change in the overall ORs (Figure 21).
Presence of complication during birth preceding the most recent birth
Finding of review from four studies [24, 37-39] indicated that presence of complication during birth preceding the most recent birth was not significantly associated with institutional delivery service utilization (AOR= 1.002; 95%CI 0.391 2.70). Heterogeneity test indicated I2=85.8%, high variability, hence random effect model was assumed in the analysis. Sensitivity analysis was done and illustrated stability of overall OR (Figure 22).