Menu
Menu
Menu
Menu
Background: Studies have observed rural-urban inequity in the use of skilled delivery in Nigeria. A number of studies have explicitly examined associated factors of assistance during delivery in rural areas. However, the studies so far conducted in rural Nigeria have investigated mainly individual-level characteristics with near exclusion of community-level characteristics. Also, most of the studies that have investigated community-level influence on use of maternal healthcare services in Nigeria did not isolate rural areas for specific research attention. The objective of this study was to investigate the individual-level and community-level characteristics associated with assistance during delivery in rural Nigeria. Methods: The study analysed women data of 2013 Nigeria Demographic and Health Survey. A weighted sample size of 12,665 rural women was analysed. The outcome variable was assistance during delivery, dichotomised into ‘skilled assistance’ and ‘unskilled assistance’. The explanatory variables are selected individual-level characteristics (maternal education, parity, age at first birth, religion, healthcare decision, employment status, access to mass media, and means of transportation); and selected community-level characteristics (community literacy level, community childcare burden, proportion of women employed outside agriculture, proportion of women who perceived distance to facility as a big problem, community poverty level, and geographical region). The mixed-effects logistic regression was applied. Results: During the most recent deliveries, 23.0% of rural women utilised skilled assistance compared with 77.0% who utilised unskilled assistance. Maternal education, parity, religion, healthcare decision, access to mass media, and means of transportation were the individual-level characteristics that revealed significant effects on the likelihood of utilising skilled assistance during delivery, while community literacy level, community poverty level, community perception of distance to health facility, and geographic region were the community-level characteristics that revealed significant effects on the odds of using skilled assistance during delivery. Results of Intra-Class Correlation (ICC) supported significant community-level effects on the likelihood of using skilled assistance during delivery. Conclusions: Assistance during delivery is influenced by individual-level and community-level characteristics. Health policies and programmes seeking to reduce rural-urban inequity in skilled delivery should endeavour to identify and address important factors at both the individual and community levels of the social environment.
The study location is Nigeria, the most populous country in Africa [1]. Nigeria has a weak health delivery system that contributes to adverse maternal and child health outcomes [10, 51]. Also, the health delivery system in the country is inadequately funded [52]. Average national indices of maternal and child health, particularly utilisation of crucial maternal healthcare services such as skilled care delivery in the country is among the poorest in sub-Saharan Africa [2]. In the 5 years preceding the 2013 Nigeria Demographic and Health Survey (NDHS), among women who had a live birth, 61% received antenatal care from a skilled health provider; a lower proportion (51%) reported the recommended four or more antenatal care visits; 36% had facility-delivery, and 38% of deliveries were attended by a skilled health provider [9]. These show that utilisation of maternal healthcare services in the country need improvement. The Integrated Maternal, Newborn and Child Health (IMNCH) Strategy currently being implemented in the country seeks to boost utilisation of essential maternal care services across the continuum of maternal healthcare [2]. For instance, some of the IMNCH priority actions include increasing the coverage and quality of the Focused Antenatal Care; increasing demand for facility-based deliveries with skilled birth attendance; and ensuring that all mothers and newborns receive prompt postnatal check within 2 days. Nevertheless, the coverage of the IMNCH interventions is still very low particularly in rural and remote areas of the country [53]. The Federal Government has also made efforts to boost the funding of maternal and child health delivery in the country through The Subsidy Reinvestment and Empowerment Programme [51]. The programme has however recorded only marginal improvement in the situation of maternal, newborn and child health in the country, as drop out from the continuum of maternity care remain high in the country [54] with substantial rural-urban differentials in the use of maternal healthcare services [55]. Some health practitioners have commenced advocacy for free maternal and child health in the country to further boost utilisation of maternal healthcare services [56]. Some other health professionals have also advocated for the implementation of Conditional Cash Transfer (CCT) Scheme in the country, as another means of encouraging maternal healthcare use, particularly among socially disadvantaged women [57]. In spite of this efforts, negative perception of public health facilities may have continue to hinder improved utilisation among women in the country [58]. The Community-based health insurance programme, another initiative designed to enhance access to healthcare in the country is yet to successfully commence in many communities in the country [59]. These have provided compelling need for further investigation of associated factors of maternal healthcare use in the country. This study was based on data collected from women of reproductive age in the 2013 NDHS. The 2013 NDHS is part of the series of cross-sectional Demographic and Health Survey (DHS) conducted across developing countries to provide reliable and internationally comparable information on the current state of fertility, childhood and adult mortality, family planning, and other sexual and reproductive health issues in developing countries. Relevant information about the design and implementation of the survey has been published [9]. In this study, only rural women were analysed. The study however, excluded women who were not currently married, and women who had no live birth in the 5 years preceding the survey. A weighted sample of 12,665 women were analysed in the study. A formal request to analyse the dataset was made to MEASURE DHS (the custodian of the DHS data) through online platform. Authorisation was granted. The outcome variable in the study was assistance during delivery. This was a dichotomous variable with ‘skilled assistance’ and ‘unskilled assistance’ as categories. Skilled assistance in the study refers to delivery assistance provided by a doctor, nurse, midwife, or auxiliary nurse/midwife. This was based on the classification of skilled and unskilled health provider adopted in the 2013 NDHS [9]. But in some other countries, auxiliary nurse/midwife may not be classified as a skilled health provider [60]. Unskilled assistance in the study refers to delivery assistance provided by community extension worker, traditional birth attendants, friends/relatives, and no one. The category of interest in the study was the skilled assistance category. The explanatory variables in the study are individual-level and community-level characteristics. The individual-level characteristics analysed are maternal education, healthcare decision, parity, age at first birth, religion, employment status, access to mass media, and means of transportation. These characteristics were selected for analysis because previous studies have established their associations with utilisation of maternal healthcare services particularly in developing countries [20, 21, 23–26, 61–65]. Some of the variables were re-coded to suit the analytic framework of the study. Healthcare decision was based on whether women were involved in making decision about their own healthcare. Women who solely or jointly with male partner had final say on their healthcare decision were grouped as ‘participation’, while other women who had no say in the decision were grouped as ‘no participation’. Women’s access to the mass media was based on the frequency of listening to radio, watching television, or reading newspaper weekly. Women who did not listen to radio, watch television, or read newspaper during the week were grouped as ‘no access’, while those who listened to radio, watched television, or read newspaper at least once weekly were grouped as ‘moderate’ access. Women who listened to radio, watched television, or read newspaper more than once weekly were grouped as ‘high’ access. Women’s parity was divided into primiparity (one child ever born), multiparity (two to four children ever born), and grand multiparity (five or more children ever born). Six community-level characteristics were analysed in the study. These are community childcare burden (proportion of women who had five or more children), community literacy level (proportion of women who cannot read and write at all), community poverty level (proportion of women in the poorest household quintile), proportion of women employed outside agricultural sector, community perception of distance to health facility (proportion of women who perceived distance to health facility as a big problem), and geographic region. The community-level characteristics were derived by aggregating the selected characteristic at the cluster level and then dividing into suitable categories. This was done because the variables are not directly available in the DHS data. Most studies that have analysed community-level variables using DHS data adopted the method [28, 29, 48]. In addition, to the selected individual-level and community-level variables, three variables were selected for statistical control. These are household wealth quintile, number of antenatal care visits, and timing of first antenatal care visits. The selection of these variables was guided by literature [24–27, 66]. Three levels of statistical analyses were employed. The univariate analysis describes assistance during delivery using the pie chart, while respondents’ characteristics were presented using frequency distributions and percentages. At the bivariate level of analysis, cross tabulation of the research variables were carried out to show percentage of skilled assistance as changes in the categories of the explanatory variable occurs. This analytic level also describes the relationship between the variables using unadjusted binary logistic regression coefficient to reveal whether the relationships are positive or negative. Before the multivariate analysis was carried out, a Variance Inflation Factor (VIF) was calculated to detect extent of multi-collinearity of the independent variables using the mean VIF score. This was needed to determine the suitability of the selected variables for multivariate analysis. As a rule of thumb in regression analysis, a mean VIF score of less than 5 is tolerated, while a mean VIF score of 5–10 suggests that the regression coefficients might be inadequately estimated [67, 68]. At the multivariate level, the mixed-effects logistic regression was used to determine extent of variation in the use of skilled assistance attributable to individual-level and community-level characteristics. The mixed-effects logistic regression model consists of two parts, namely, the fixed effect and the random effect [43]. The model was specified as: Where: πij is the log of odds of skilled assistance. (1 − πij) is the log of unskilled assistance. x and z are the explanatory variables for the likelihood of skilled assistance. x1 to x8 are the individual-level characteristics. z1 to z6 are the community-level characteristics. β0is the overall intercept. β1…β14 are the regression coefficients for the explanatory variables x1 to x8, and z1 to z6. u0j is the community-level random effect (assumed to be normally distributed with mean equal to 0 and variance equal to σμ02). The fixed effects were estimated using odds ratio of adjusted binary logistic regression, while the random effects were estimated using the Intra-Class Correlation (ICC) calculated as: ττ+π23where τ is the estimated community-level variance [43]. The ICC ranges from 0 to 1, with ICC of 1 indicating that women in the community have identical use of skilled assistance during delivery, and with ICC of 0 indicating that women in the community do not have identical use of skilled assistance in the community. Four models were fitted in the study. Model 1 included only individual-level characteristics, while Model 2 was based solely on community-level characteristics. Model 3 included both individual-level and community-level characteristics. Model 4 was the full model that included the explanatory and control variables. The models were fitted using the xtmelogit command of Stata version 12 [69]. Model adequacy was examined using the Wald chi-square which assesses the statistical significance of the model. The 5% alpha level was considered statistically significant. Analyses were performed using Stata version 12.
N/A
