Measuring equity in utilization of emergency obstetric care at Wolisso Hospital in Oromiya, Ethiopia: A cross sectional study

Introduction. Improving equity in access to services for the treatment of complications that arise during pregnancy and childbirth, namely Emergency Obstetric Care (EmOC), is fundamental if maternal and neonatal mortality are to be reduced. Consequently, there is a growing need to monitor equity in access to EmOC. The objective of this study was to develop a simple questionnaire to measure equity in utilization of EmOC at Wolisso Hospital, Ethiopia and compare the wealth status of EmOC users with women in the general population. Methods. Women in the Ethiopia 2005 Demographic and Health Survey (DHS) constituted our reference population. We cross-tabulated DHS wealth variables against wealth quintiles. Five variables that differentiated well across quintiles were selected to create a questionnaire that was administered to women at discharge from the maternity from January to August 2010. This was used to identify inequities in utilization of EmOC by comparison with the reference population. Results: 760 women were surveyed. An a posteriori comparison of these 2010 data to the 2011 DHS dataset, indicated that women using EmOC were wealthier and more likely to be urban dwellers. On a scale from 0 (poorest) to 15 (wealthiest), 31% of women in the 2011 DHS sample scored less than 1 compared with 0.7% in the study population. 70% of women accessing EmOC belonged to the richest quintile with only 4% belonging to the poorest two quintiles. Transportation costs seem to play an important role. Conclusions: We found inequity in utilization of EmOC in favour of the wealthiest. Assessing and monitoring equitable utilization of maternity services is feasible using this simple tool. © 2013 Wilunda et al.; licensee BioMed Central Ltd. The study population consisted of women utilizing EmOC at Wolisso Hospital. Wolisso Hospital is a referral, private, non-profit facility located in Wolisso town (central Ethiopia, 115 km south-west of Addis Ababa), which is the capital of the Southwest Shoa Zone (SWSZ) in the Oromiya region. SWSZ has a population of about 1.1 million inhabitants and is served by 81 health facilities (one hospital, ten HCs, 53 HPs and 17 private clinics). Wolisso Hospital is the only facility that provides comprehensive EmOC in the SWSZ. To promote equity in access to Wolisso Hospital, the price of health services are highly subsidized. The Oromiya Regional Health Bureau provides financial support to the hospital within the framework of a public-private partnership [19]. The hospital is currently an implementer of the National Health Service Extension Program by supporting the activities of health extension workers in the region. In conducting this study, we primarily followed the methodology proposed by Pitchforth et al. in 2007 to develop a proxy wealth index for EmOC in Bangladesh [11]. This consisted of selecting proxy wealth variables, scoring the variables, validating the scores, and comparing the wealth status of women in the study sample with that of women in the general population. The study design has three components: 1) development of the proxy wealth index; 2) using the wealth index variables to construct the survey questionnaire for women attending Wolisso Hospital for EmOC; and 3) comparing women attending EmOC in Wolisso with the general population of the region. For the development of the proxy wealth index, we used data from the Ethiopia 2005 Demographic and Health Survey (DHS). The DHS programme represents the largest worldwide effort to obtain demographic and health data from nationally representative household surveys in developing countries (http://www.measuredhs.com). As the DHS use standardized questionnaires and methods, they are often considered the best available source of data for many health and demographic indicators in developing countries. The surveys are usually conducted periodically every five years. As our study population was women in the region of Oromiya, we selected a dataset of women from the national DHS aged 15–49 years with a previous birth and who usually resided in Oromiya. This constituted the reference population against which we compared the women accessing the hospital. All proxy wealth variables available in the DHS were cross tabulated against the wealth quintiles in the DHS dataset (See Additional file 1 for list of all variables). Five variables that differentiated well across all quintiles and showed proportional progression from the lowest to the highest quintile were selected for our proxy wealth index (Additional file 2). These variables were: roof material, ownership of a table, type of toilet facility, ownership of a radio and educational attainment (Table 1). Crude and weighted wealth scores of selected wealth variables The second step involved assigning wealth scores. Two categories of wealth scores were generated: crude scores and weighted scores using our assigned weights. Crude scores were assigned to show the relative importance of response options within each variable in reflecting the wealth status. For instance, for educational attainment, no education was scored as 1, incomplete primary education as 2, and so forth. Crude scores were then rescaled to take values of between 0 and 1 (Table 1). This was done to ensure that the final score was not influenced by the number of response options a variable had. Some variables may be more influential in differentiating between the poorest and the wealthiest quintiles. We introduced weighting to reflect this relative significance. These weights ranged from 1 to 5, for each variable based on factor loadings obtained from factor analysis of the five selected variables. Factor analysis is a data-reduction technique used to identify a small number of factors that explain most of the variance observed in a larger number of variables, whereas factor loadings show how much each variable correlates with each factor; higher loadings imply greater correlations [20]. These weights were then multiplied by the rescaled crude scores to obtain weighted scores for each variable response option (Table 1). The sum of these scores provided an overall weighted score for each individual. Scores could range from 0 to 15. Lastly, we assessed the validity and reliability of the weighted scores of the five variables. Factor analysis was performed for all DHS wealth variables (including education) to generate factor scores, and the first factor was taken to represent the wealth status [21]. This formed the “gold standard” upon which the scores of the five variables were compared. Validity and reliability were assessed using correlation and kappa analysis respectively [22]. Because scores of the five variables were not normally distributed, Spearman’s correlation was used to assess the validity. The questionnaire was developed on the basis of the 2005 DHS and the data for our study were collected in 2010. We however seized the opportunity of using updated DHS 2011 data that became available soon after our data collection. We re-checked the validity and reliability of the weighted scores of the 5 variables using this 2011 DHS data by following the same process as described above. In order to assess the socio-economic status of the women attending EmOC in Wolisso Hospital, we prepared a questionnaire to be administered to the women. The questionnaire needed to be simple, easy to administer and pose limited inconvenience to the women, hence the need for a limited set of variables. The five variables selected by the process explained above were used to construct the questionnaire which also included information on parity, place of residence: woreda (district) and kebele (village). Kebeles located in big and small towns or trading centers were classified as urban, and the rest were classified as rural. We also collected information on the means and cost of transportation to the hospital. The questionnaire is presented in Additional file 3. A midwife working from 8.00 to 17.00 was responsible for administering the questionnaire to all women who were being discharged from the maternity ward during these hours and had given informed consent. A piloting of the questionnaire had shown that it was easy to use and well understood by the respondents. The midwife received a brief orientation to the questionnaire. Data collection took place from January 18 to August 11, 2010. To assess who was utilizing EmOC, data of EmOC users from the questionnaire were compared with that of parous women in the Oromiya 2011 DHS dataset based on parity, area of residence and the five proxy wealth variables. Wealth variables would be equally distributed in the two groups if there were no inequity. Stata commands that account for the complex sampling design and weighting of DHS data were used. The reported p values are corrected for the study design [23]. Additionally, the assigned weights in Table 1 were used to develop weighted scores for the women accessing maternity services at Wolisso Hospital. In the 2005 DHS dataset, cutoff points of quintiles generated, based on weighted scores i.e. <0.60, 0.60-1.95, 1.20-3.15, 3.20-5.70 and >5.70 for quintiles 1, 2, 3, 4 and 5, respectively, were used to divide the women in our study dataset into five groups, thus grouping users of EmOC into the quintiles that they would have belonged to in the general population based on the 2005 DHS. All analyses, except the comparison with women in the general population; which was done using Stata version 11, were done using SPSS version 16. The response rate of EmOC users was calculated as the proportion of women who were discharged from the maternity ward during the study period and who were interviewed. To assess the extent to which this sub-population of interviewed women is representative or not of the overall population of maternity users at the hospital in the study period, characteristics of interviewed and non-interviewed women were compared on the basis of routine data notified in the hospital registers using their in-patient numbers. The two groups of women were compared for age, district of residence, length of hospital stay and mode of delivery.