Timely initiation of antenatal care and its associated factors among pregnant women in sub-Saharan Africa: A multicountry analysis of Demographic and Health Surveys

Background Timely initiation of antenatal care (ANC) is an important component of ANC services that improve the health of the mother and the newborn. Mothers who begin attending ANC in a timely manner, can fully benefit from preventive and curative services. However, evidence in sub-Saharan Africa (sSA) indicated that the majority of pregnant mothers did not start their first visit timely. As our search concerned, there is no study that incorporates a large number of sub-Saharan Africa countries. Thus, the objective of this study was to assess the prevalence of timely initiation of ANC and its associated factors in 36 sSA countries. Methods The Demographic and Health Survey (DHS) of 36 sSA countries were used for the analysis. The total weighted sample of 233,349 women aged 15–49 years who gave birth in the five years preceding the survey and who had ANC visit for their last child were included. A multilevel logistic regression model was used to examine the individual and community-level factors that influence the timely initiation of ANC. Results were presented using adjusted odds ratio (AOR) with 95% confidence interval (CI). Results In this study, overall timely initiation of ANC visit was 38.0% (95% CI: 37.8–38.2), ranging from 14.5% in Mozambique to 68.6% in Liberia. In the final multilevel logistic regression model:- women with secondary education (AOR = 1.08; 95% CI: 1.06, 1.11), higher education (AOR = 1.43; 95% CI: 1.36, 1.51), women aged 25–34 years (AOR = 1.20; 95% CI: 1.17, 1.23), 35 years (AOR = 1.30; 95% CI: 1.26, 1.35), women from richest household (AOR = 1.19; 95% CI: 1.14, 1.22), women perceiving distance from the health facility as not a big problem (AOR = 1.05; 95%CI: 1.03, 1.07), women exposed to media (AOR = 1.29; 95%CI: 1.26, 1.32), women living in communities with medium percentage of literacy (AOR = 1.51; 95%CI: 1.40, 1.63), and women living in communities with high percentage of literacy (AOR = 1.56; 95%CI: 1.38, 1.76) were more likely to initiate ANC timely. However, women who wanted their pregnancy later (AOR = 0.84; 95%CI: 0.82, 0.86), wanted no more pregnancy (AOR = 0.80; 95%CI: 0.77, 0.83), and women residing in the rural area (AOR = 0.90; 95%CI: 0.87, 0.92) were less likely to initiate ANC timely. Conclusion Even though the WHO recommends all women initiate ANC within 12 weeks of gestation, sSA recorded a low overall prevalence of timely initiation of ANC. Maternal education, pregnancy intention, residence, age, wealth status, media exposure, distance from health facility, and community-level literacy were significantly associated with timely initiation of ANC. Therefore, intervention efforts should focus on the identified factors in order to improve timely initiation of ANC in sSA. This can be done through the providing information and education to the community on the timing and importance of attending antenatal care and family planning to prevent unwanted pregnancy, especially in rural settings. The study used 36 sSA countries’ Demographic and Health Survey (DHS) data which were obtained using a cross-sectional study design. The survey we used were conducted between 2006–07 and 2018 in sSA countries. The data for this study were drawn from recent nationally representative DHS data conducted in 36 countries in sSA. The DHS surveys are routinely collected every five-year period across low- and middle-income countries using structured methodologies and pretested validated quantitative tools. It follows the same standard procedure sampling, questionnaires, data collection, and coding which makes multi-country analysis possible. In order to ensure national representativeness, the DHS survey employs a stratified two-stage sampling technique. In the first stage, clusters/enumeration areas (EAs) that cover the entire country were randomly selected from the sampling frame (i.e. are usually developed from the available latest national census). The second stage is the systematic sampling of households listed in each cluster or EA and interviews are conducted in selected households with target populations (women aged 15–49 and men aged 15–64). In this study, women aged 15–49 years who gave birth in the five years preceding the survey and who had ANC visit for their last child were included. The total sample size from the pooled data analyzed in this study was 233,349 and the sample size ranged from 1,316 in Sao Tome and Principe to 16,543 in Nigeria (Table 1). The outcome variable for this study was timely initiation of first ANC visit which was recorded as: within 12 weeks of gestation “timely” and after 12 weeks of gestation”delayed” [55]. Independent variables were extracted based on literature and the likelihood to influence the outcome of interest from the available DHS [6, 7, 9–11, 14, 19–21, 24, 34–42, 45, 46, 56, 57]. In this study, independent variables included in the analysis are broadly categorized as individual and community-level factors. The individual-level factors include maternal age (categorized as 15–24 years, 25–34 years, and ≥35 years), maternal education (no education, primary, secondary, and higher), marital status (categorized as ever married and never married), household wealth status was derived from a combination of all household variables describing housing and assets and computed using principal component analysis (poorest, poorer, middle, richer, and richest), media exposure (exposed to at least one of radio, magazine/newspaper or television were labeled as ‘yes’ and those who did not were labeled as ‘no’), insurance coverage (yes/no), parity (categorized as primiparous, multiparous, and grand multiparous), ever had a pregnancy terminated (yes/no), pregnancy intention (wanted then, wanted later and wanted no more), perception of distance from the health facility (big problem/not a big problem) and employment status (not employed/employed). Community-level factors were: place of residence (rural/urban), community-level literacy, community-level poverty, and community media exposure. The community-level variables such as community-level literacy, community-level poverty, and community media exposure were obtained by aggregating the individual-level variables into clusters by using the proportion. Community-level literacy is measured as the proportion of women who completed primary and above educational level in the primary sampling unit. It was categorized as low, medium and high if less than 25%, 25%-50% and more than 50% of study population of the cluster had at least eight years of education respectively. Community-level poverty was computed from the household wealth and defined as the proportion of women in the top 3 wealth quantiles (middle, richer and richest) in the clusters. It was categorized as low, medium and high if less than 25%, 25%-50% and more than 50% of study population of the cluster had at least middle quintile respectively. Community media exposure is the proportion of women who had exposure to at least one type of media; radio, newspaper, or television in the primary sampling unit. Similarly, community media exposure was categorized as low, medium, and high. All statistical analysis was carried out with STATA version 14. Since DHS surveys follows the same standard procedure sampling, questionnaires, data collection, and coding, datasets were appended together to explore the timing of ANC and its associated factors among women in sSA. Both descriptive and analytic analysis were carried out after the weighting of data using sample weights to adjust disproportional sampling and non-response as well as to restore the representativeness of the sample so that the total sample looks like the country’s actual population. Frequencies and percentages were used to describe the background characteristics of the study participants. Multilevel logistic regression was employed because our outcome variable (timing of the first ANC visit) was measured as a binary factor and since DHS data are hierarchical, i.e. individuals (level 1) were nested within communities (level 2). To cater for the unexplained variability at the community level, we used clusters as random effect. The log of the probability of the timing of ANC was modeled using a two-level model as follows: Log [Πij /1−Πij] = β0+β1Xij+ B2Zij+ μj+eij Where i and j are the individual (level 1) and community (level 2) units, respectively; X and Z refer to level 1 and (level 2) variables, respectively; πij is the probability of timely initiation of ANC the β’s are fixed coefficients; β0 is the intercept-the effect on the probability of the timing of ANC in the absence of independent variables; μj and eij are random effect (effect of the community on timing of ANC for the jth community) and random errors at the individual levels respectively. In particular, three models were constructed [58]. We first constructed an empty model, which only includes outcome variable and cluster variable to test the random effect between-cluster variability. Then model containing only individual-level variables (model I) was fitted. Finally, in model II, we adjusted for both individual and community-level variables to estimate the association between timely initiation of ANC and the factors. The Intra-class Correlation Coefficient (ICC), the Median Odds Ratio (MOR), and the Proportional Change in Variance (PCV) were computed to assess the clustering effect/variability. ICC shows the variation in timely intiation of ANC for reproductive women due to community characteristics and it was calculated as follows: ICC = VA/ (VA+3.29), where VA is the estimated variance of clusters in each model [59]. The MOR is defined as the median odds ratio between the area at highest risk and the area at the lowest risk when comparing two individuals from two different randomly chosen clusters. It was calculated using the formula: MOR = exp. [√(2 × VA) × 0.6745] ≈ exp(0.95√VA)] Where VA is the cluster level variance in each model [59, 60]. We used PCV to measure total variation attributed to an individual or/and community-level factors at each model. It was calculated as: PCV % = (VA−VB/VA)*100, where VA = variance of the empty model, and VB = variance of the model with more factors [59]. Moreover, deviance information criteria (DIC) was used to compare the candidate model, which was calculated as: deviance = -2log-likelihood ratio. It is always greater or equal than zero, being zero only if the fit is perfect. Therefore, model with the minimum value of deviance was selected for data analysis. First, we fit unadjusted regression models for each explanatory variable to select variables for multivariable analysis, and variables with p-value ≤ 0.20 in the unadjusted regression analysis were included in multivariable analysis. Finally, results for the multivariable analysis have been presented as odds ratios (OR), with their corresponding 95% confidence intervals (CI), and p-value <0.05 were considered to be significant factors associated with the timely initiation of ANC. Ethical approval for this study was not required since this study used existing public domain survey data sets, which are freely available online with all identifier information removed. But to access and use the data we sought permission and approval from Measure DHS through the online request.