The effect of birth preparedness and complication readiness on skilled care use: a prospective follow-up study in Southwest Ethiopia

This community based prospective follow-up study was conducted in Jimma Zone, located 346kms Southwest of Addis Ababa, the capital city of Ethiopia, from September 2012-April 2013. Jimma Zone is one of the 17 Zones of the Oromia Regional State of Ethiopia, which was named for the former Kingdom of Jimma and absorbed into the former province of Kaffa in 1932. The Zone has a total of 17 rural districts called ‘Woredas’ and two town administrations with an estimated total population of about 2.6 million and a male-to-female ratio of 1.01:1. The great majority (89%) of the population of the Zone were rural residents. The Zone has a total of 521,506 households with an average household size of 4.77 persons per a household. The potential health service coverage of the zone for the year 2011 was about 52%[13, 14].

The source or target population considered for this study was all women who had given birth during the study period, from September 2012-April 2013 in Jimma Zone. However, as this was a prospective follow-up study, the study participants were all pregnant women identified based on the sampling procedure and enrolled in the study at a base line and had been followed till 28 days post-partum period. At the baseline, the status of BP & CR and affecting factors were studied which is under review for publication on Pan African Medical Journal. This study was the intermediate one collected during the second phase just at the end of neonatal period. The final study, determinants and causes of neonatal mortality, is again on peer review process for publication on Plos One.

The minimum required sample size for this study was determined by using Epi-Info V.3.5.1. by considering two sample comparisons of proportions based on the following assumptions. The outcome variable was skilled care use and the explanatory variable was BP & CR plan. As there was no similar study conducted in the country to be used as a base to determine the sample size, study from other developing country was used. In a study done in India, BP & CR was found to increase the skilled care use by 80%[11].

In Ethiopia, the proportion of women attended by skilled delivery attendant was 10% and this was taken as the proportion among non-exposed group (P₁ = 0.1)[8]. The proportion of women attended by skilled attendant among exposed (prepared for birth and its complication) was estimated to be 18% (P₂ = 0.18) to detect 8% difference or 80% increment. A level of confidence of 95% and power of 90% were considered. In addition, the coverage of exposure from the general population was estimated to be 22%[15]. Thus, a ratio of 1:4 (r = 4) was used for exposed-to-non-exposed. A design effect of 2 was also considered because of the multistage clustered sampling techniques. Finally, 10% was added for non responses and miss to follow-up and the final sample size became 2603 mothers. However, this study was part of a big longitudinal study in which 3612 pregnant women had been on follow-up to look at the determinants of neonatal mortality. As a result, after excluding lost-to-follow-up and abortion cases, 3472 mothers were included in the analysis of this study.

Multi-stage clustered sampling technique was used to identify pregnant women for the study. Initially, the Zone was stratified as Town Administrations (2 in number) and rural districts (17 in number). Then, at first stage, 5 districts were selected randomly from the 17 rural districts (‘Woredas’). At second stage, all the selected 5 districts were stratified in to urban and rural ‘Kebeles’ (A ‘kebele’ is the smallest administrative unit having 5000 population in average and considered as clusters in this study). Then, by simple random sampling method, 9 rural ‘Kebeles’ and 2 urban ‘Kebeles’ were selected from each selected district. Jimma town administration and Agaro town administration have 13 and 5 ‘Kebeles’ respectively and all were included purposefully. Then, for all the selected ‘kebeles’, pregnant women were enumerated by using house-to-house visit and all obtained and registered were included in the study. All women who reported to have pregnancy of 12 weeks or above as defined by loss of three consecutive menses were considered as eligible and enrolled in the study.

Pre-tested interviewer administered structured questionnaire was adapted from the safe motherhood questionnaire developed by maternal and neonatal health program of JHPIEGO to measure the composite variable for birth preparedness and complication readiness[9]. The indicators for the wealth index were adapted from EDHS[8]. To control the quality of data, training, pretest, supervision and use of local languages were made.

The collected data were coded and entered into EpiData V.3.1 to minimize logical errors and design skipping patterns. Then, the data were exported to SPSS for windows version 20.0 for cleaning, editing and analysis. Descriptive analysis was made by computing proportions and summary statistics. Socio-economic quintiles were determined by using Principal Component Analysis (PCA). As Jimma and Agaro town administrations were both purposefully included, the status of skilled care use was estimated by calculating weighted percentage based on the complex sample survey procedure.

Bivariate analysis was done by using cross-tabulation to see associations between each independent variable and skilled care use. All variables having P < 0.25 were considered as candidates for the final model. As multistage clustered sampling method was used because of the different levels of factors, mixed-effects multilevel logistic regression model was used by using STATA 13 to identify factors having significant association with skilled care use. ‘Kebeles’ were considered as clusters and kebele level variables such as place of residence and access to basic emergency obstetric care (BEmOC) and comprehensive emergency obstetric care (CEmOC) were taken as higher level (level-2). Mothers were nested within their households and the community. As a result, maternal individual variables including socio-demographic, economic, obstetric and BP & CR were taken as lower level (level-1). Goodness of fit of the multilevel model was tested by the log likelihood ratio (LR) test. Multicollinearity between the independent variables was assessed by using variance inflation factor (VIF). As all included variables had VIF < 10, no multicollinearity was detected. In addition cross-level two-way interactions were checked; but, no significant interaction was detected.

Ethical approval was obtained from the Institutional Review Board (IRB) of College of Health Sciences of Addis Ababa University. Necessary permission was secured from all local administrators. Written informed consent was obtained from each respondent before actual data collection. Issues of confidentiality were maintained by removing any identifiers from the questionnaire. To protect vulnerable group, data collectors were trained to maintain confidentiality and provide necessary health information based on the need of the participants, but not an intervention.

A total of 3472 respondents were included in the analysis for this study. Majority, 2618 (75.4%), were rural residents. Most, 2227 (64.2%), were in the age group of 20-29 years with the mean age of 26.5 ± 5.0. Oromo was the dominant ethnic group, 3041 (87.6%) and Muslim was the leading religion, 3032 (87.3%). The great majority, 3289 (94.8%), were housewives by occupation and more than half, 1878(54.1%), didn’t attend any formal education (Table 1).

Of all the 3472 respondents, 2634 (75.9%) had at least one ANC visit during pregnancy and 1830 (52.7%) had used either from hospital or health center attended by skilled attendant. However, only 1228 (35.4%) had four or more visits and 572 (16.5%) started during first trimester. Of the respondents, 1064 (30.6%) used skilled care during delivery (74.1% in urban and 16.5% in rural). However, the weighted coverage of skilled care use was much lower 17.5% (95% CI: 16.2%, 18.8%). The major reasons for giving birth in hospital or health center were pre-planed (43.5%), need skilled attendant (41.0%), problem during labor (28.9%) and need clean/safe place (15.1%). Whereas, the most common reason for home delivery was lack of transport (31.1%) followed by home delivery is the usual place (24.0%) and more comfortable (23.1%) (Table 2).

A total of 1858 (53.5%) had planned to have 4 or more ANC visits from skilled attendants. However, only 1228 (35.4%) actually had 4 or more ANC visits. Using 4 or more ANC visit was 46.0% among those who planned to have it as compared to 23.2% among who didn’t plan which is statistically significant (COR = 2.82; 95%CI: 2.43, 3.27). Similarly, 1136 (32.7%) had planned to use skilled care during labor and 1064 (30.6%) actually used it. About 61.7% of those who had already planed actually used as compared to only 15.5% of who didn’t plan which is statically significant (COR = 8.76; 95% CI: 7.44, 10.32).

Four hundred thirty five (38.3%) of those who had already planned to use skilled care didn’t use it actually. The main reasons were the labor was not associated with problems 382 (87.8%), lack of transport 64 (14.7%) and lack of money for transport and services 58 (13.3%). On the other hand, 363 (15.5%) of those who didn’t plan to use skilled care actually used it. The main reason was occurrence of unanticipated problems at labor 174 (47.9%) among which prolonged labor was the leading one 152(41.9%). The other reason was family insisted at labor to have safe and clean delivery 99 (27.3%). This indicated that intention significantly implies action in skilled care use; however, not 100% as complication during labor is unpredictable (Table 3).

To evaluate the applicability of the mixed-effects multilevel logistic regression model, the Intra-class Correlation Coefficient (ICC (ρ)) was calculated in the empty model and it was found to be 0.489 indicating that 48.9% of the variation was contributed by between cluster variation. The test of the preference of log likelihood Vs logistic regression was also strongly significant (P < 0.001). Then, the full model was run by including all the cluster level and individual level variables and the ICC (ρ) became 0.113 indicating that 11.3% of the variation was attributed to cluster level variables again suggesting the preference of multilevel analysis. The preference of log likelihood Vs logistic regression was still strongly significant (P < 0.0001).

After adjusting in the final two-level mixed-effects logistic regression model, both the cluster level variables and individual level variables were found to be important determinants of skilled care use. Among the higher (cluster) level variables, place of residence and access to BEmOC were found to have statistically significant association with skilled care use. Women in urban areas were more than two times more likely to use skilled care as compared to rural women (OR = 2.38; 95%CI: 1.74, 3.24). Similarly, women in clusters having BEmOC (health center) within 2 hours distance on foot in average were more likely to use skilled care (OR = 1.61; 95%CI: 1.12, 2.33). However, having CEmOC (hospital) within 2 hours distance on foot in average had non-significant effect on skilled care use (OR = 1.36; 95%CI: 0.74, 2.51).

Among the lower (individual) level variables, maternal education, husband’s occupation, wealth quintiles, gravida, inter-birth interval, knowledge of key danger signs during labor, ANC visit and BP & CR were found to have statistically significant association with skilled care use. Having primary (OR = 1.37; 95%CI: 1.10, 1.71), secondary (OR = 3.48; 95%CI: 2.22, 5.45) or tertiary (OR = 3.97; 95%CI: 1.38, 11.46) educations were found to increase the likelihood of skilled care use significantly as compared to not having formal education. Similarly, women whose husbands were employed (OR = 3.26; 95%CI: 2.15, 4.94) or merchants (OR = 2.27; 95%CI: 1.60, 3.23), were more likely to use skilled care as compared to those whose husbands were farmers. Women in the 3rd wealth quintiles (OR = 1.73; 95%CI: 1.12, 1.84) and 4th quintiles (OR = 1.28; 95%CI: 1.11, 1.73) were more likely to use skilled care as compared to those in the lowest quintile.

Gravida and inter-birth interval were among the individual level obstetric factors found to have significant association with skilled care use. Mothers with experience of 1-4 pregnancies (OR = 0.41; 95%CI: 0.26, 0.66) and 5 or more (OR = 0.45; 95%CI: 0.26, 0.76) were less likely to use skilled care as compared to primi-gravida mothers. Inter-birth interval of >48 months was found to increase the likelihood of skilled care use as compared to interval of <24 months (OR = 2.18; 95%CI: 1.30, 3.63).

Knowledge of key danger signs during labor was the other determinant of skilled care use. Women who knew 3 or more key danger signs during labor were more likely to use skilled care as compared to those who didn’t know any key danger signs (OR = 1.59; o5%CI: 1.05, 2.43). Similarly, having 1-3 ANC visits (OR = 1.63; 95%CI: 1.23, 2.18) and ≥4 visits (OR = 3.10; 95%CI: 2.31, 4.16) during pregnancy were found to increase the likelihood of skilled care use significantly. After controlling all the necessary variables in the mixed-effects multilevel model, birth preparedness and complication readiness plan had significant effect on skilled care use. Women who were well-prepared during pregnancy were more likely to use skilled care as compared to those who were not well-prepared (OR = 1.32; 95%CI: 1.03, 1.68) (Table 4).