Background: Poor access to institutional delivery services has been known as a significant contributory factor to adverse maternal as well as newborn outcomes. Previous studies measured access in terms of utilization while it has different dimensions (geographic accessibility, availability, affordability, and acceptability) that requires to be measured separately. Therefore, this study was conducted to assess the four dimensions of access and factors associated with each of these dimensions. Methods: Community-based cross-sectional study design was used, employing both quantitative and qualitative methods. A simple random sampling technique was used to select 605 mothers who had given birth in the last 6 months preceding the study. Multi-variable binary logistic regression was used to select factors associated with the four dimensions of access by using AOR with 95% CI. Ethical approval was secured from Jimma University Institutional Review Board. Results: Five hundred and ninety-three mothers involved in this study, resulting in a response rate of 98%. Four hundred five (68%), 273(46%), 279(47%), and 273(46%) had geographic, perceived availability, affordability, and acceptability access to institutional delivery services, respectively. Antenatal care [AOR = 3.74(1.56, 8.98)], occupation of mother [AOR = 5.10(1.63, 15.88)], and residence [AOR = 1.93(1.13, 3.29)] were independently associated with geographic accessibility. Household graduation [AOR = 1.46(1.03, 2.06)], residence [AOR = 1.74(1.17, 2.59)], and ANC [AOR = 3.80(1.38, 10.50)] were independently associated with perceived availability. Moreover, wealth quintile [AOR = 11.60(6.02, 22.35)], ANC [AOR = 3.48(1.36, 9.61)], and occupation of husband [AOR = 3.63(1.51, 8.74)] were independently associated with affordability. Lastly, mother’s education [AOR = 2.69(1.42, 5.09)], residence [AOR = 2.60(1.66, 4.08)], and household graduation [AOR = 3.12(2.16, 4.50)] were independently associated with acceptability of institutional delivery services. Conclusions: Moderate proportions of mothers have geographic accessibility to institutional delivery services, but access to the other three dimensions was low. ANC visits of 4 or above, occupation of husband, urban residence, graduation of mother’s household as a model family, higher wealth quintiles, and maternal educational level significantly affect access to institutional delivery services. Thus, it was recommended that concerned bodies should give due attention to ANC services, female education, training of model families, and enhancement of household wealth through job creation opportunities to increase access to institutional delivery services.
This study was conducted in six randomly selected Woredas of Jimma Zone (Mencho, Gomma, Nano Benja, Seka Chokorsa, Gumay, and Kersa), Southwest Ethiopia, from March 16 to April 15, 2018. Jimma Zone is one of the 17 Zones found in Oromia Regional State, located 350 km Southwest of Addis Ababa, the capital city of Ethiopia. There are 20 Woredas and two town administrations with a total of 561 Kebeles (the smallest geographic administrative unit in Ethiopia) in Jimma Zone. According to Jimma Zone Health Department annual estimate for the year 2017, the total population of the Zone was 3,312,914, of which 603,222 were reproductive age women. Concerning health facilities, there was 1 tertiary hospital, 8 primary hospitals, 122 health centers, and 513 health posts in the Zone. A community-based cross-sectional study design was used, employing both quantitative and qualitative methods, to undertake this study. The qualitative method was used to support the findings of quantitative results in order to explore and understand what is behind mere quantitative figures. The source population for this study was all women of reproductive age group (15–49 years), and the study population was randomly selected mothers who lived for at least 6 months in the selected Kebeles (6 months stay is considered as permanent residence in Ethiopia), and who had given birth in the last 6 months preceding the study (6 months time period preceding the study was preferred to minimize recall bias). A single population proportion formula was used to determine sample size with assumptions of P = 50% (proportion of mothers having access to IDS), 95% confidence level, and 5% margin of error. A design effect of 1.5 was used to account for the sampling variability of multi-stage sampling. The calculated sample size was 576 and 5% added for an expected non-response rate to have a final sample size of 605 for the quantitative study. For qualitative data, four focused group discussions (FGDs) were carried out: two with women health development army leaders, one with home-delivered mothers and one with health facility-delivered mothers. The FGD participants were identified by the suggestion of Health Extension Workers (HEWs) and community elders. Each FGD consisted of 8–10 mothers. A multi-stage sampling technique was employed, taking the Woredas as primary sampling units (PSU), Kebeles from the selected Woredas as secondary sampling units (SSU), and mothers at households from the selected Kebeles as tertiary sampling units (TSU). Thirty percent of the Woredas and Kebeles were selected by lottery method. Then, depending on the number of mothers who had given birth in the last 6 months preceding the study, the sample size was proportionally allocated to the selected Kebeles. Finally, a simple random sampling (SRS) method was used to get mothers for an interview at a household (HH) level. Data collection tools were prepared in English and then translated to the local language, i.e. Afan Oromo, by a language expert, and then back to English by another language expert to ensure consistency. Training was given to data collectors and supervisors before data collection about the aim of the study and data collection procedures. Six data collectors who were fluent in speaking and writing the local language participated in the data collection process using a pretested interviewer-administered structured questionnaire for the quantitative part. For the qualitative part, an interview guide was used for the FGD, and the discussions were recorded using a tape recorder. In addition, hand-written notes were taken. The FGDs were moderated by supervisors who were fluent in speaking and writing the local language. To ensure data quality, a pretest was conducted in a Woreda which was not included in the study, on 10% of the sample size to determine the clarity of the items and the consistency of the responses. During actual data collection, the filled questionnaires were checked for consistency and completeness each day and submitted to the supervisors. Data were entered into epi-Data version 3.1 by two persons separately and the two files were validated for consistency, and necessary corrections were made before exporting the data into SPSS for analysis. The dependent variables of measurement were geographic accessibility, perceived availability, affordability, and acceptability, whereas the independent variables were socio-demographic variables (age, religion, ethnicity, residence,marital status), socio-economic variables (HH wealth index, mother’s education, husband’s education, mother’s occupation, husband’s occupation), obstetric and related variables (parity, ANC use), and household-level variables (HH headship, family size, HH model status). With regard to data processing and analysis, the collected data were checked for consistency, then coded and entered into epi-Data version 3.1. After validation of duplicated files for consistency, it was exported to SPSS version 20 for cleaning and analysis. When using PCA, all assumptions of PCA such as metric level or dummy coded variables, minimum sample size of 50, cases to items ratio of 5 to 1 or more, two or more correlations of 0.3 or more on the correlation matrix, removing items with sampling adequacy less than 0.5, and significance of Bartlett test of sphericity were checked. Varimax rotation was used to get better loading variables for the components. Additionally, variables with communality less than 0.5 were removed, and then variables with complex structures (having loadings of 0.4 or more on more than one component) on the rotated component matrix were removed. Components with only one variable loading on them were also checked and removed. Finally, seven components were formed, and explained 67% of the total variability. The reliability of variables within each component was assessed, and all of the variables in the components had Cronbach alpha of 0.6 and above. For the descriptive statistical part, means, proportions, tables, graphs, and charts were used to summarize and present the results of this study. To identify associated factors, first, the chi-square test was run for individual study variables for each of the access dimensions. Variables that fulfilled the chi-square assumptions were passed for simple logistic regression. Then, variables with a p-value ≤0.25 on simple logistic regression were taken as candidates for multivariable logistic regression using 10% for probability of entry and 15% for probability of removal. A backward likelihood ratio method was used for variable selection. Finally, four separate models were fitted to show the effect of significant predictor variables on the outcome variables at a p-value ≤0.05. The adequacies of the models were checked by the Hosmer and Lemeshow test of goodness of fit at p-value ≥0.1. Nagelkerke pseudo R-squared and the overall prediction powers of the models from the classification tables were also reported. Using an adjusted odds ratio (AOR) with 95% confidence interval, the associations of dependent and independent variables were interpreted. Qualitative data were first transcribed, coded, and categorized to form primary themes based on the objectives of the study. Besides, quotes of participants that supported or contradicted the key quantitative findings were triangulated with the quantitative findings during the discussion of the results.
N/A