Spatial Patterns and Multilevel Analysis of Factors Associated with Antenatal Care Visits in Nigeria: Insight from the 2018 Nigeria Demographic Health Survey

Despite global progress towards antenatal care (ANC) uptake, ANC utilization in a number of countries in sub-Saharan Africa, such as Nigeria, is low. Although several studies have identified the determinants and factors associated with ANC services utilization in Nigeria, there is a gap in knowledge about the spatial patterns in ANC use. Therefore, this study aims to map the spatial distribution and factors associated with ANC visits in Nigeria. A cross-sectional dataset was obtained from the 2018 Nigeria Demographic and Health Survey. A total of 20,003 women aged 15–49 were considered in this study. Both spatial and multilevel analyses were carried out. The results were presented in spatial maps and adjusted odds ratios (aOR) at a 95% confidence interval (CI). Hot spot areas (high proportion of an incomplete ANC visit) were located in Sokoto, Kebbi, Zamfara, Katsina, Kano, Jigawa, Bauchi, Niger, Borno, Gombe, and Bayelsa. Regional disparities in incomplete ANC visits were found in this study. Maternal age, maternal education, partner’s level of education, working status, ethnicity, parity, religion, exposure to media, place of residence, wealth index, region, and community literacy level were factors associated with incomplete ANC. There is a need to consider these factors in the design and strengthening of existing interventions (e.g., mini-clinics) aimed at increasing ANC visits to help attain maternal health-related Sustainable Development Goals by 2030. The regional disparities in incomplete ANC visits also need to be considered by encouraging pregnant women in hotspot areas to attend ANC visits.


Introduction
Antenatal care (ANC) is an essential element in the continuum of reproductive health care for preventing preventable pregnancy-related morbidity and mortality. It provides a platform for integrated health functions, including health promotion, screening and diagnosis, and disease prevention before and after pregnancy [1]. ANC also improves newborn making [11]. Since Nigeria still lags behind in achieving the recommended World Health Organization ANC visits, investigating the multi-level factors and spatial distribution of ANC use in Nigeria is an important step to closing existing gaps. Therefore, this study aims to map the spatial distribution and factors associated with ANC visits in Nigeria.

Data Source
Data were obtained from the 2018 Nigeria Demographic and Health Survey (NDHS). The NDHS is a cross-sectional survey that gathers data on health information on men, women, and children. The NDHS is nationally representative and collects data from 36 administrative units and the Federal Capital Territory using a two-stage sampling process. The survey data cover various health topics, including antenatal care visits [18]. The NDHS methodology is reported in the National Population Commission (NPC) and International Center for Migration (ICF). We followed the guidelines for enhancing the reporting of observational studies in epidemiology when producing this publication [19]. The dataset can be downloaded from https://dhsprogram.com/data/available-datasets. cfm (accessed on 20 March 2021).

Sample Size Determination and Inclusion Criteria
The primary sampling unit for the survey consisted of samples chosen at random from clusters. A total of 20,003 women aged 15-49 were considered for this study, based on data from the children's recode file. The selection process involved the selection of eligible households from the clusters. All eligible interviewed women of reproductive age between 15 and 45 years were initially considered; however, after the exclusion of women without a complete information on ANC visits and other variables of interest, the initial sample selection then dropped from 34,082 to 20,003 respondents, and this is the sample size included in the study analysis.

Outcome Variable
The outcome variable in this study was the "number of ANC visits." In this study, the number of ANC visits was defined as the percentage of women aged 15-49 who had a live birth in a given time period that received ANC services during pregnancy. This variable was derived from the question, "How many times did you receive ANC during this pregnancy?" The responses ranged from 0 to 20. Following the revised World Health Organization guidelines on the recommended number of ANC visits [8], the responses were categorized as <8 visits or ≥8 visits, where ≥8 visits were considered complete ANC visits and <8 incomplete ANC visits. Similar categorizations have been used in previous studies [20][21][22].

Independent Variables
Based on theoretical and practical significance and the availability of the variables in the dataset, we considered both individual-and household-level factors in our study. The selection of the variables was influenced by their association with the number of ANC visits in previous studies [20][21][22].

Household-Level Factors
The household-level factors were place of residence (urban vs. rural), wealth index (categorized by DHS as poorest, poorer, middle, richer, and richest, and the same categories were used in this study), region (North Central, North East, North West, South East, South-South, South West), sex of household head (male, female), community literacy level (low, medium, high), and community socioeconomic status (low, medium, high).

Statistical Analyses
Both spatial and multi-level analyses were carried out.

Multi-Level Analysis
A two-level multi-level binary logistic regression was built to assess the individualand household-level factors associated with incomplete ANC visits in Nigeria. Women were nested within households in the modeling, and subsequently, households were nested within clusters. Clusters were considered as random effects to account for the unexplained variability at the household level. A total of four models were fitted. Firstly, we fitted an empty model, model 0, which contained no predictors (random intercept). After that, model I contained individual-level variables alone, model II included household-level variables and model III was the complete model that comprised both individual-level and household-level variables. The odds ratio and its corresponding 95% confidence intervals (CIs) were provided for models I-III. These models were fitted by a Stata command "melogit." The log-likelihood ratio (LLR), Akaike Information Criterion (AIC) measure, and Schwarz's Bayesian Information Criterion (BIC) were used for model comparison. The best fit model has the highest log-likelihood and the lowest AIC [23]. We also tested for multicollinearity by using variance inflation factor (VIF), which showed no collinearity among the independent variables (Mean VIF= 1.79, Maximum VIF= 2.99, and Minimum VIF= 1.07). In the individual populations, sample weight (v005/1,000,000) was used in all analyses to account for over-and under-sampling, whereas the "svy" command was used to account for the survey's complex nature, which also helps in the generalizability of the findings. Stata version 16.0 (Stata Corporation, College Station, TX, USA) was used for statistical analyses.

Spatial Analysis Spatial Autocorrelation
Spatial autocorrelation analysis was performed to check whether there is a clustering effect on incomplete ANC visits in Nigeria. This analysis result gives Global Moran's I value, Z-score, and p-value for deciding whether the data are dispersed or random or clustered. A Moran's I value close to positive 1 indicates a clustering effect, close to negative one indicates dispersed, and close to zero random. If the p-value is significant and the I value is close to zero, this means that incomplete ANC visit had a clustering effect.
Hot Spot Analysis (Getis-OrdGi* Statistic) The hot spot analysis tool gives Getis-Ord or Gi* statistics for a cluster in the dataset. Statistical values such as Z-score and p-value are computed to determine the statistical significance of clusters. Results of the analysis with a high GI* value mean hot spot areas (high prevalence of an incomplete ANC visit), and a low GI* value means cold spot areas (low prevalence of an incomplete ANC visit).

Spatial Interpolation or Prediction
Spatial prediction is one of the techniques of estimating unsampled areas based on sampled areas. In the 2018 NDHS, a total of 1400 clusters were selected to take a sample for this area that is believed to be representative of the country. An Ordinary Kriging prediction method was used for this study to predict an incomplete ANC visit in unobserved areas of Nigeria.

Spatial Scan Statistical Analysis
Bernoulli's purely spatial model was applied to identify primary and secondary, and incomplete ANC visit clusters using 1393 enumeration areas. SatTscan software (Havard Medical School, Boston, MA, USA) was used for the analysis. First, the dataset was managed as appropriate for SaTScan software. Women who had no complete ANC visits were taken as cases, and women who had complete ANC visits were taken as controls. The cluster number, longitude, and latitude data were obtained from the GPS dataset. The cluster size of less than 50% of the population was taken as the upper bound. A 999 Monte Carlo replication was used for this study. Based on the above criteria, primary and secondary clusters were identified.

Sociodemographic Characteristics of Respondents
A total of 20,003 women of reproductive age were included in the study. At the individual level, 9665 (48.32%) of the respondents were aged 25 to 34. A total of 9245 (46.22%) had no education, while 9886 (49.42%) of the respondents' partners had secondary education and above. A total of 19,387 (96.92%) of the respondents were currently married, and 12,326 (61.62%) of the women had mass media exposure. At the household level, 12,266 (61.32%) of the study women resided in rural areas, 4464 (22.32%) were from the poorest wealth index households, 7269 (36.34%) were residing in the North West region, 6786 (33.93%) were from a community with a high literacy level, and 11,957 (59.77%) were from a community with low socioeconomic status. All the individual and household-level factors showed significant associations with ANC visits in Nigeria (Table 1).

Spatial Autocorrelation
Spatial autocorrelation analysis was carried out to check whether an incomplete ANC visit was random or not. According to the Global Moran's I value z-score and p-value of p < 0.0001, an incomplete ANC visit was clustered in Nigeria across all regions ( Figure 1).

Spatial Autocorrelation
Spatial autocorrelation analysis was carried out to check whether an incomplete ANC visit was random or not. According to the Global Moran's I value z-score and p-value of p < 0.0001, an incomplete ANC visit was clustered in Nigeria across all regions ( Figure 1).

Hot Spot Analysis
Hot spot analysis was conducted using Getis-Ord GI* analysis to detect hot and cold spot areas. Hot spot areas (high proportion of an incomplete ANC visit) were located in Sokoto, Kebbi, Zamfara, Katsina, Kano, Jigawa, Bauchi, Niger, Borno, Gombe, and Bayelsa. The cold spot areas were located in Lagos, Osun, Ekiti, Kogi, Enugu, Imo, Abia, River, and Oyo ( Figure 2).

Hot Spot Analysis
Hot spot analysis was conducted using Getis-Ord GI* analysis to detect hot and cold spot areas. Hot spot areas (high proportion of an incomplete ANC visit) were located in Sokoto, Kebbi, Zamfara, Katsina, Kano, Jigawa, Bauchi, Niger, Borno, Gombe, and Bayelsa. The cold spot areas were located in Lagos, Osun, Ekiti, Kogi, Enugu, Imo, Abia, River, and Oyo ( Figure 2).

Prediction of an Incomplete ANC Visit
Prediction analysis was performed. The result gives the prevalence of an incomplete ANC visit for the unsampled area of Nigeria based on the sample area. The prediction revealed that areas such as those shown in red are high-risk areas for an incomplete ANC visit in Nigeria (Figure 3).

Prediction of an Incomplete ANC Visit
Prediction analysis was performed. The result gives the prevalence of an incomplete ANC visit for the unsampled area of Nigeria based on the sample area. The prediction revealed that areas such as those shown in red are high-risk areas for an incomplete ANC visit in Nigeria (Figure 3).

Spatial SaTScan Analysis of Stillbirth Bernoulli-Based Model
Most likely (primary) and secondary clusters of an incomplete ANC visit were identified. A total of 336 clusters were identified: 336 were primary, and 8 were secondary clusters. The primary clusters' spatial window was located in the Northern part of Nigeria (12.458115 N, 10.398881 E)/690.22 km, and log-likelihood ratio (LLR) of 1307 and relative risk (RR) 1.79, at p < 0.0001. It showed that women within the spatial window had 1.79 times higher risk of an incomplete ANC visit than women outside the window. Likewise, the secondary clusters were centered at (6.224563 N, 7.971626 E)/16.54 km radius, LLR of 11.90, and relative risk (RR) 1.28 at p-value 0.0062. This showed that women within the spatial window had 1.28 times higher risk of an incomplete ANC visit than children outside the window (Table 2, Figure 4).

Multi-Level Fixed Effects (Measures of Associations) Results
The factors associated with an incomplete ANC visit in Nigeria at the individual level include maternal age, maternal education, partner's level of education, working status, ethnicity, parity, religion, and exposure to media, while the place of residence, wealth index, region, and community literacy level were factors associated with incomplete ANC at the household level (Table 3).

Cluster
Enumeration Area

Multi-Level Fixed Effects (Measures of Associations) Results
The factors associated with an incomplete ANC visit in Nigeria at the individual level include maternal age, maternal education, partner's level of education, working status, ethnicity, parity, religion, and exposure to media, while the place of residence, wealth index, region, and community literacy level were factors associated with incomplete ANC at the household level (Table 3).

Random Effects (Measures of Variations) Results
The empty model (model 0), as shown below in Table 3, depicted a substantial variation in the likelihood of incomplete ANC visits in Nigeria across the Primary Sampling Unit (PSU) clustering (σ2 = 5.33; 95% (CI = 4.73-6.02)). Model 0 indicated that 62% of the variation in incomplete ANC visits in Nigeria was attributed to Intra-Class Correlation variation, i.e., (ICC = 0.62). The between-cluster variation decreased to 29% (0.29) in model I (individual-level variables only). In the household-level variables (model II), the ICC decreased to 26% (ICC = 0.26). In comparison, the ICC declined further to 24% (ICC = 0.24) in the complete model with both the individual and household-level factors (model III). This further reiterates that the variations in the likelihood of incomplete ANC visits in Nigeria are attributed to the clustering variation in PSUs. The Akaike Information Criterion (AIC) and Schwarz's Bayesian Information Criterion (BIC) values showed a successive reduction, which means a substantial improvement in each of the models over the previous model, and affirmed the goodness of model III developed in the analysis. Therefore, model III, the complete model with both the selected individual and household-level factors, was selected to predict the likelihood of incomplete ANC visits in Nigeria.

Discussion
This study examined the spatial pattern and multi-level factors associated with ANC visits in Nigeria using the recent NDHS data collected in 2018. It was observed that areas with high levels of incomplete ANC visits were located in Sokoto, Kebbi, Zamfara, Katsina, Kano, Jigawa, Bauchi, Niger, Borno, Gombe, and Bayelsa. The probable explanation for this finding could be the differences in the level of development across the various regions in the country as well as rural-urban differences, as explained in a comparative study on factors associated with the under-utilization of ANC services in Nigeria by Adewuyi et al. [24], which is also linked to the availability of maternal health services in some of these regions [25,26]. Another possible explanation is the role of sociocultural factors in maternal health services utilization in Nigeria [27,28]. Our results showed statistically significant associations between individual factors, household factors, and ANC visits in Nigeria. At the individual level, the factors associated with incomplete ANC visits included maternal age, maternal education, partner's level of education, working status, ethnicity, parity, religion, and exposure to media, while the place of residence, wealth index, region, and community literacy level were factors associated with incomplete ANC visits at the household level. On the other hand, marital status, distance to the health facility, sex of household head, and community SES showed no statistically significant associations with incomplete ANC visits.
Among the individual-level factors examined, our findings showed that the likelihood of having an incomplete ANC visit in Nigeria was high among women of reproductive age who had three children and those practicing traditional and other religious beliefs compared to women with one child and those practicing Christianity. This can be explained by the fact that first-time mothers are mostly less experienced and more mindful about the health of their forthcoming baby as compared to multiparous women, hence the differences in ANC visits; this argument is supported by a cross-sectional study on the utilization of antenatal care among rural-to-urban migrant women in Shanghai [29]. Our result is consistent with other studies being carried out across the world, which suggests that parity is inversely related to ANC [29][30][31]. Moreover, completing ANC visits among first-time mothers could be because of time management, more resources in the family, and positive perceptions about the benefits of ANC, which is similar to the findings of Simkhada et al. [32]. Hence, parity influences ANC visits. As parity increases, the experience of timely initiation of ANC decreases as these women might give less value to ANC services than women with one child [30]. Likewise, a woman's religion was found to be statistically associated with the number of ANC visits. Christian women with one child were less likely to report incomplete ANC visits than women with three children with traditional and other religious backgrounds.
Additionally, our study showed that respondents from Yoruba and Igbo ethnic groups and those exposed to mass media were less likely to report incomplete ANC visits than those from Hausa ethnicity and those not exposed to mass media. This reflects the important role ethnicity, religion, and mass media play in ANC attendance. Comparable to some previous reports, our findings are consistent with the existing literature [24,33,34]. Ethnicity and religion can affect norms and values placed on the use of ANC services.
The contribution of media exposure to ANC visits cannot be underestimated. This was highlighted in our findings. We found a significant positive association between exposure to media and ANC visits. Women exposed to mass media were less likely to report incomplete ANC visits compared to women who were not exposed to mass media. This result is consistent with the findings of Basha [35] and Odesanya et al. [36] in their cross-sectional study on factors affecting the utilization of antenatal care services in Nigeria.
They indicated that media access has the power to enlighten women positively about utilizing maternal health care, including ANC services.
Moreover, we found that women of reproductive age who were aged 35 years and above were less likely to report incomplete ANC visits than women of reproductive age between ages 15 and 24. This pattern may be due to the stigmatization of teenage and young mothers. Many teenage and young pregnant women avoid public engagements until after delivery for fear of the social stigma attached to such pregnancy at an immature stage of life. As a result of this stigma, very few ANC contacts are made by such women, and in the extreme, ANC contacts are avoided throughout the pregnancy. This finding is consistent with previous cross-sectional studies in Nigeria by Fagbamigbe et al. and Akanbiemu et al. [33,37].
Again, we found that women with secondary education and above and women whose partners had secondary education and above were less likely to report incomplete ANC visits compared to those with primary and no education. A study in Bangladesh found education to be positively associated with eight or more ANC contacts during pregnancy [38]. Other studies that examined ANC visits in different settings, including Nigeria, have consistently reported that women with higher levels of education have a higher prevalence and greater odds of ANC visits than those with no education [24,39,40]. Thus, maternal education is associated with the use of basic maternal health services, including ANC [40,41], and lack of formal education has been associated with underutilization or non-use of ANC [24,42]. Maternal education improves health care literacy and makes women appreciate the advantages of using health care services, including ANC (Basha 2019; Mumtaz 2019). In addition, maternal education increases decision-making power and enhances women's socioeconomic opportunities and status [35,[43][44][45] and also their confidence to take action about their health [46,47]. In addition, maternal education and spousal educational attainment were associated with a higher number of ANC visits. This finding emphasizes the importance of education in achieving the recommended ANC visits by the WHO. In Bangladesh, parents' higher educational attainment was positively associated with having eight or more ANC contacts [38]. A higher level of education among pregnant women and a higher level of education among their spouses may positively influence joint decision-making among couples regarding health-seeking during pregnancy. At the household/community level, women of reproductive age currently residing in rural areas and those currently residing in the North East were more likely to report incomplete ANC visits than those residing in urban areas and the North Central area. Similar studies on ANC visits in Nigeria have equally reported significant differences in rural-urban utilization of maternal health services, including ANC [35,[43][44][45]. The inequities can explain this finding in the distribution of accessible health resources between rural and urban areas [4,24,35,41,48]. In situations where facilities exist in rural areas, they may be inaccessible due to poor road networks, inefficient transport systems, or very far distances, making ANC visits difficult [24,35]. In addition to the aforementioned points, rural areas are inadequately financed. It is difficult to attract and retain health workers in such places, which is consistent with the cross-sectional study of Fagbamigbe and Idemudia in Nigeria [4], making the utilization of maternal health services difficult.
Additionally, lower socioeconomic status and the influence of cultural beliefs and social norms discourage rural women from using ANC compared to their counterparts in urban areas [24,28,49]. These population dynamics could be a plausible explanation for the observed regional differences between women in rural areas and those in the North-Eastern part of Nigeria. Therefore, these findings imply that regional differences in ANC utilization are a possible reflection of educational differences and other community-level factors across the regions.

Strengths and Limitations
This study has various strengths that are worth mentioning. First, we relied on the use of a nationally representative dataset, and hence the findings are applicable to all pregnant women of their reproductive age in Nigeria. Additionally, the use of the Geographical Information System (GIS) in the analysis of the spatial distribution enabled us to identify the hotspots of incomplete ANC visits. However, it is important to interpret the findings against these limitations. The cross-sectional design cannot allow causal interpretation of the findings, but only associations can be made. Recall and social desirability biases cannot be overruled from this study as the number of ANC visits were self-reported. Again, DHS samples are generally not large enough to provide estimates for small geographic areas, which often are needed for monitoring and evaluating decentralized programs. DHS surveys in a specific country are typically conducted every three to six years so that annual estimates of key indicators are not available from the surveys. Finally, the data analysis was restricted to only variables that were in the DHS, and hence some variables might have been excluded.

Policy and Public Health Implications
The findings of this study are relevant to policy and public health practice. The differences in the completion of optimal ANC visits based on spatial analysis emphasize the need to pay attention to the hot spot areas to increase the number of ANC visits for women to benefit from the importance of optimal ANC visits. In addition to this, it is prudent to pay attention to the rural-urban variations and the sociocultural differences associated with ANC uptake. These parameters will need urgent consideration as far as education on the importance of ANC visits by health professionals is concerned. The government could adopt a short-term policy that will make ANC visits compulsory and free for all women within the poor wealth quintile, whilst the long-term policy could be to incentivize ANC visits from eight visits and above.

Conclusions
It was observed that the hot spot areas were located in Sokoto, Kebbi, Zamfara, Katsina, Kano, Jigawa, Bauchi, Niger, Borno, Gombe, and Bayelsa. The factors associated with incomplete ANC attendance were maternal age, maternal education, partner's education, working status, ethnicity, parity, religion, exposure to media, place of residence, wealth index, region, and community literacy level. There is a need to consider these factors in designing and strengthening existing interventions aimed at increasing ANC visits to help attain maternal health-related SDGs by 2030. Funding: This study received no external funding. However, we sincerely thank Bielefeld University, Germany, for providing financial support through the Open Access Publication Fund for the article processing charge.

Institutional Review Board Statement: Not applicable.
Informed Consent Statement: Not applicable.

Data Availability Statement:
The datasets utilized in this study can be accessed at https://dhsprogram. com/data/available-datasets.cfm (accessed on 20 March 2021).

Ethical Approval:
Since the authors of this manuscript did not collect the data, we sought permission from the MEASURE DHS website and access to the data was provided after our intent for the request was assessed and approved on the 20th of March 2021. The DHS surveys are ethically accepted by the ORC Macro Inc. Ethics Committee and the Ethics Boards of partner organizations in different countries, such as the Ministries of Health. The women who were interviewed gave either written or verbal consent during each of the surveys.