Individual-level and community-level determinants of use of maternal health services in Northwest Ethiopia: a prospective follow-up study

Objective The use of maternal health services is an important indicator of maternal health and socioeconomic development. Evidence on individual-level and community-level determinants of use of maternal health services in Benishangul Gumuz Region was not well known so far. Hence, this study fills this gap. Design A prospective follow-up study Setting This study was conducted from March 2020 to January 2021 in Northwest Ethiopia. Participants A total of 2198 pregnant women participated in the study. Main outcome measures A multistage random sampling technique was used to select study subjects. Data were collected via face-to-face interviews using pretested semistructured questionnaires. Collected data were coded, cleaned and analysed using Stata software. Multilevel regression models were applied to determine individual-level and community-level factors of use of maternal health services. Results This study found that the proportions of women who visited recommended antenatal care (ANC), received skilled delivery care and postnatal care (PNC) were 66.1%, 58.3% and 58.6%, respectively. Being rural (adjusted OR (AOR)=3.82, 95% CI 1.35 to 10.78), having information on maternal health services (AOR=2.13, 95% CI 1.21 to 3.75), with a history of pregnancy-related problems (AOR=1.83, 95% CI 1.15 to 2.92) and women with decision-making power (AOR=1.74, 95% CI 1.14 to 2.68) were more likely to attend fourth ANC visits. Similarly, women who attended tertiary school (AOR=4.12, 95% CI 1.49 to 11.33) and who initiated the first ANC visit within 4-6 months of pregnancy (AOR=0.66, 95% CI 0.45 to 0.96) were determinants of skilled delivery care. Finally, women whose partners attended tertiary education (AOR=3.67, 95% CI 1.40 to 9.58), women with decision-making power (AOR=1.8, 95% CI 1.09 to 2.97), women who attended the fourth ANC visit (AOR=10.8, 95% CI 6.79 to 17.2), women received iron-folic acid during pregnancy (AOR=1.96, 95% CI 1.11 to 3.49) and women who received skilled delivery care (AOR=1.63, 95% CI 1.1 to 2.42) were more likely to get PNC services. Conclusion The proportion of women who attended ANC, received skilled delivery services and PNCs was low. Different individual-level and community-level factors that influenced the use of these services were discovered. Therefore, community-based interventions should target those identified factors to improve maternal health services. This study was conducted in Benishangul Gumuz Regional State. It is 1 of the 11 regions constituting the Federal Democratic Republic of Ethiopia (FDRE), located in Northwest Ethiopia. The capital city of the region is Assosa town, located at 670 km away from Addis Ababa, the capital city of Ethiopia. Administratively, the region has three zones (namely, Assosa zone, Metekel zone and Kamashi zone), three town administrations (namely, Assosa, Gilgel Beles and Kamashi town administrations), one special woreda (namely, Mao-Komo special woreda), and 475 kebeles (439 rural and 36 urban). The region hosts nearly 60 000 refugees. Based on the 2007 national population and household census, the 2018 population projection revealed that the total population of the region was 1 127 001, which covers 1.1% of the national population, the total number of pregnant women in the region, and the selected study districts were 36 754 and 15 368 pregnant women, respectively.33 Health facilities serving these populations were 446 public health facilities (401 health posts, 41 health centres, 4 primary hospitals and two general hospitals); 119 private and non-governmental organisation health institutes (15 medium clinics and 104 primary clinics) and 91 private pharmaceutics (three pharmacies, 50 drug stores and 38 rural drug vendors). A community and health facility-linked prospective follow-up study design was carried out from March 2020 to January 2021. All pregnant women within the study area during the time of the baseline survey were the source population. Randomly selected pregnant women using the sampling technique were study participants. The inclusion criteria were women who were permanent residents (living more than 6 months) in the selected districts and categorised as pregnant women, women whose gestational age is >8 weeks and also fulfil pregnancy screening criteria, whereas the exclusion criteria were pregnant women who have hearing or other disabilities hindering communication; severely ill and mentally disturbed, pregnant women who reported their pregnancy is less than 8 weeks, and pregnant women who completed the fourth ANC visit during the time of baseline survey. The sample size was computed using both single and double population proportion formulas. For the single population proportion, the following assumptions were considered while computing the sample size: the proportion of women who used the whole maternal healthcare service is 60% (p=0.6).34 The margin of error is 5% (d=0.05) with a 95% level of CI (1.96), taking a design effect of 2 and a non-response rate of 10%. Then, the sample size calculated is 812 pregnant women. Similarly, the double population proportion formula was used to compute the sample size for each determinant of use of maternal health services. Among all the factors considered for sample size calculation, women’s age is found to have the maximum sample size. Thus, considering the following assumption for double population formula: the proportion of women who completed the whole maternal health services (ANC, skilled delivery and PNC services), among women whose age is greater than 35 is 48% (p1=0.48) and among women whose age between 20 and 35 years old is 62% (p2=0.62)34; pooled population proportion (p=0.55); r=1:1 ratio of exposure to non-exposure; 5% significant level; 80% power, design effect of 2 and 10 non-response rate. Then, 874 sample sizes were generated through Stata/MP V.13.0 software. As a result, a total of 874 pregnant women were calculated for this study. This study, however, was part of larger research work,35 and the sample size determined for another objective was 2402 pregnant women, which was used as the final sample size for this study. A multistage clustered sampling technique was employed to reach the study participants. In this study, the study area was first stratified into three zones and three town administrations with one special woreda. In the first stage, of these stratified areas, two zones and one town administration were selected using a simple random sampling technique. Then after, seven districts/woredas and two town districts/woredas were randomly selected from two zones and one town administration, respectively, as the second stage. Subsequently, at the third stage, 51 Kebeles/clusters were randomly selected from the selected districts/woredas. A 1-month baseline census was conducted to identify pregnant women using a pregnancy screening criterion to prepare a sampling frame. Then, all pregnant women who resided in the selected kebeles/clusters were included in the study and then followed up for an average of 11 months. Mean time of the house-to-house survey and public health facilities that provide at least basic maternal health services for the community were identified. Then, all eligible public health facilities were recruited and made a candidate for a facility-based survey. Based on these, 46 health facilities (3 hospitals, 12 health centres and 31 health posts) were included in the health facility-based survey. The research questionnaire was prepared in English, which was adapted from Ethiopia Demographic and Health Survey (EDHS) 2011,2 National Technical Guidance for MPDSR 2017,36 MCH Program Indicator Survey 2013,37 survey tools conducted in Jimma Zone, Southwest Ethiopia,38 survey tools conducted in Rural South Ethiopia5 and other relevant different works of literature. After finalising the research instrument preparation, training, pretest, supervision and use of local languages were made to ensure the quality of data. Then, the trained data collectors gather information through face-to-face interviews at comfortable and convenient places. After all, completed questionnaires were reviewed by supervisors on a monthly base for accuracy and consistency. In this study, maternal health service encompasses care during pregnancy, childbirth and after birth within 42 days. Therefore, we have three primary outcomes: attending recommended ANC visits (fourth visits or more), receiving skilled delivery care and attending PNC fourth visits within 42 days. Independent variables were categorised into two levels. Individual-level variables (level 1) included individual-related factors: women’s age, educational level (women and partner), occupational status (women and partner), information on maternal health services, age at first marriage and pregnant, past and present bad obstetric history, women and partner decision-making power in health-seeking behaviours, iron and folic acid (IFA) supplementation during pregnancy and provision of tetanus toxoid (TT) vaccination during pregnancy. Higher-level variables (cluster 2) included community and health facility-related factors such as place of residence, Household Wealth Index, accessibility of health facilities, availability of health facilities within the community and quality of maternal health services. The collected data were coded and entered into Epi. Info V.7.2.2.6. After data entry was completed, it was exported to Stata software V.14.1. Then, data were cleaned, edited and analysed using Stata software. Descriptive statistics and crude OR at 95% CI were computed for all variables to select candidate variables for multivariable analysis (p<0.25). The composite indicator of the Household Wealth Index was computed and categorised into three categories using principal component analysis. Before running the full model, the coefficient of the interaction term at p>0.1 and the multicollinearity effect between independent variables were determined by using variance inflation factors (>10%). Finally, all included variables had no multicollinearity and interaction effect. Even though a multistage clustered sampling method was used in the study, a multilevel regression model was applied by using Stata V.14.1 to identify community and individual-level factors having significant association with use of maternal healthcare (ANC fourth visits or more, skilled delivery care and PNC fourth visits). Kebele/ketena was considered as cluster, and cluster-level variables including a place of residence, access to health posts and Household Wealth Index were taken as higher levels (level 2), whereas individual factors such as sociodemographic, obstetric history, age at first marriage and pregnancy, information on maternal health services, women decision-making power, key services offered during pregnancy and pregnancy-related problems were taken as lower levels (level 1). The goodness of fit of the multilevel model was tested by the log-likelihood ratio test and found to be statistically significant such as data fit the model. In this study, patients or the public were not involved in the design, conduct, report or dissemination plans of our research.