Menu
Menu
Menu
Menu
Objective We assessed whether geographic distance and difference in altitude between home to health facility and household socioeconomic status were associated with utilisation of maternal and child health services in rural Ethiopia. Design Household and health facility surveys were conducted from December 2018 to February 2019. Setting Forty-six districts in the Ethiopian regions: Amhara, Oromia, Tigray and Southern Nations, Nationalities, and Peoples. Participants A total of 11 877 women aged 13-49 years and 5786 children aged 2-59 months were included. Outcome measures The outcomes were four or more antenatal care visits, facility delivery, full child immunisation and utilisation of health services for sick children. A multilevel analysis was carried out with adjustments for potential confounding factors. Results Overall, 39% (95% CI: 35 to 42) women had attended four or more antenatal care visits, and 55% (95% CI: 51 to 58) women delivered at health facilities. One in three (36%, 95% CI: 33 to 39) of children had received full immunisations and 35% (95% CI: 31 to 39) of sick children used health services. A long distance (adjusted OR (AOR)=0.57; 95% CI: 0.34 to 0.96) and larger difference in altitude (AOR=0.34; 95% CI: 0.19 to 0.59) were associated with fewer facility deliveries. Larger difference in altitude was associated with a lower proportion of antenatal care visits (AOR=0.46; 95% CI: 0.29 to 0.74). A higher wealth index was associated with a higher proportion of antenatal care visits (AOR=1.67; 95% CI: 1.02 to 2.75) and health facility deliveries (AOR=2.11; 95% CI: 2.11 to 6.48). There was no association between distance, difference in altitude or wealth index and children being fully immunised or seeking care when they were sick. Conclusion Achieving universal access to maternal and child health services will require not only strategies to increase coverage but also targeted efforts to address the geographic and socioeconomic differentials in care utilisation, especially for maternal health. Trial registration number ISRCTN12040912.
We conducted a cross-sectional study in 46 districts of the four most populous Ethiopian regions, namely, Amhara, Oromia, Tigray, and Southern Nations, Nationalities, and Peoples (figure 1). The regions have a population scattered in dispersed settlements, and the primary source of income is farming. Maps showing (A) Ethiopia, (B) the four study regions (Tigray, Amhara, Oromia, SNNP (South Nation, Nationalities and People) and (C) the 46 study districts in the four regions (graph constructed in ArcGIS V.10.4 software using freely available data from the Ethiopian Central Statistics Agency). In the Ethiopian health delivery system, the primary healthcare unit is comprised of five satellite health posts and one health centre. Each health post serves approximately 5000 people and is staffed by two female health extension workers. The study was conducted from December 2018 to February 2019. The Ethiopian Public Health Institute, in collaboration with the London School of Hygiene and Tropical Medicine, University of Gondar, Mekelle University, Jimma University and Hawassa University, conducted the survey. This study used data from an evaluation of the Optimizing of Health Extension Program intervention that was implemented in four regions of Ethiopia. This intervention aimed to improve care utilisation for sick under 5 children.15 The evaluation included baseline and endline surveys, and this study was a secondary analysis of end-line data. A two-stage stratified cluster sampling was applied in the selected study districts. The first stage used lists of enumeration areas from the 2007 Ethiopian Housing and Population Census as the sampling frame.16 Thereafter, 194 enumeration areas were selected with probability proportional to size (sampling from a finite population in which a size measure is available for each population unit before sampling and where the probability of selecting a unit is proportional to its size). In the second stage, all households within each enumeration area were listed and a sampling interval was calculated. A random start number between 1 and the sampling interval was selected. The households that matched the random start number on the list were chosen as the first household to be included. This process was repeated until the targeted number of 60 households in each cluster was reached. All women aged 13–49 years and children under the age of 5 years, who lived in the selected households, were included in the study. The health posts and the health centres serving the enumeration area were included in the study. We used a standard sample size formula to calculate the sample size. We estimated a design effect of 1.3, 80% power and the assumption of a ratio of children less than 5 years of age per household was considered, based on sampling reported in the Ethiopia Demographic Health Survey 2011.17 We used the assumption to detect 15% improved childcare utilisation in two groups (intervention and comparison) between baseline and end-line surveys. The sample size was estimated to be 6000 households per group (12 000 in total) to have 80% power to detect differences of 15 percentage points for care seeking for 2–59 months children. The protocol for the evaluation study was registered with a trial number ISRCTN12040912 and published.15 We used a modular questionnaire that was based on the demographic and health survey and similar assessment tools. Information on the antenatal care attendance and place of delivery was collected from all women aged 13–49 who had a live birth in the last 12 months before the survey. Caregivers of children aged 2–59 months were asked whether the child had any illness in the previous 2 weeks and if they had sought care from an appropriate provider. Caregivers were invited to show the immunisation cards and asked additional questions on the different vaccinations if the card was not available. Besides, the questionnaire included information on sociodemographic data, assets and geographic coordinates of households and health facilities. The coordinates were measured by geographical positioning system (GPS) dongles (ND-100S). We estimated the overall coverage of four selected maternal and child health services. For women, this included (1) the proportion of women aged 13–49 with a live birth within the last 12 months before the survey who had attended antenatal care four or more times and (2) the proportion of women with a live birth within 12 months before the survey who gave birth in a health facility. For the child health services utilisation, we included: (1) the proportion of children aged 12–23 months who had received full immunisation, which was defined as BCG, three pentavalent vaccinations (diphtheria, tetanus, pertussis, hepatitis B and Haemophilus influenzae), oral polio vaccine and one dose of measles vaccine18 and (2) the proportion of children aged 2–59 months with fever, diarrhoea or suspected pneumonia (cough and difficult or fast breathing) in the last 2 weeks for whom care was sought from an appropriate provider, that is, health posts, health centres, hospitals or private clinics. We used the Euclidian distance and the difference in altitude from the participants’ houses to the nearest health facility and household wealth as explanatory variables. Furthermore, we included other sociodemographic variables such as age, sex, parity, region, caregiver’s education level and number of children in the household to adjust for potential confounding in the analyses. Household data on care utilisation were linked with the GPS information of the 142 health centres and 164 health posts that served the households. The linking was done using cluster identification. The Euclidian distance from participants’ houses to the nearest health facility was calculated using the geographical location of health facilities and households. Thus, we calculated the distance to the nearest health post for child health services (full child immunisation and sick child healthcare utilisation) and distance to the nearest health centre for maternal healthcare services (antenatal care and facility delivery). These distances were divided into tertiles and labelled as short, medium and long. Similarly, the difference in altitude was estimated using the altitude information of both the households and the health facility and classified as small, medium and large. Corrections of the GPS readings were done whenever the reading erroneously was positioned outside the study areas. If that was the case, we adjusted by considering the nearest GPS reading captured. To estimate household wealth, we constructed wealth index using 15 household assets and characteristics. We used principal component analysis to construct the wealth index and divided it into tertiles.19 We first analysed the prevalence of the outcomes variables and graphically displayed the relation between distance, the difference in altitude, household wealth and facility visits for antennal care, facility delivery, child immunisation and sick child services utilisation. After that, we used two-level mixed-effect logistic regression modelling to examine the associations with individual-level variables (wealth, number of children in the household, sex, age, parity and educational status), group-level variables (region, distance and the difference in altitude) and the interactions between the two. We analysed the crude and adjusted ORs with 95% CIs. As individual observations were grouped within the cluster specifically by geography (region, distance to the health facility and their altitude), the regression models we used accounted for confounding and clustering effects. Considering both the distance as well as the difference in altitude would yield a good estimate of the extent of space between the health facilities and the households. The model quality was assessed using Akaike’s information criterion and intracluster correlation coefficient. The significance level was set at a p value of less than 5%. All analyses were done using STATA V.15 (Stata Corporation, College Station, Texas). Patients or the public were not involved in the design or conduct, or reporting, or dissemination plans of this research.
