Data source and extraction of data
The study used a cross-sectional design. Data for the study was taken from the female files of the Nigeria Demographic and Health Survey (NDHS) waves from 2008, 2013, and 2018. Since our outcome variable and certain important factors used in this study were not measured in the earlier surveys, we decided not to include them. The DHS Program, financed by the United States Agency for International Development (USAID), gives financial support and technical help for population and health surveys. The National Population Commission carried out the surveys while ICF provided technical assistance. These surveys gathered data on a variety of health-related topics, including disability and smoking, as well as fertility, knowledge of and use of family planning methods, breastfeeding practices, nutritional status of women and children, maternal and child health, adult and child mortality, women's empowerment, domestic violence, female genital cutting, the prevalence of malaria, HIV and AIDS awareness and behavior, and other STIs (36).
The same methodology underlies these survey waves. First, the surveys used the Federal Republic of Nigeria's Population and Housing Census's sampling frame (NPHC). Nigeria is separated into states administratively. The local government areas (LGAs) that make up each state are further divided into wards. Every locale is divided into practical regions known as census enumeration areas (EAs). The multistage stratified sample selection was applied and EAs are chosen in the first stage with a probability inversely correlated with EA size. All chosen EAs undergo a household listing procedure, and the lists of homes that emerge serve as a sampling frame for choosing households in the subsequent stage. A set number of households are chosen in each cluster in the second stage of selection using equal probability systematic sampling. In total, 9,085 women between the ages of 15 and 49 who had full information on the factors that were analysed for the study were included in the pooled sample. The 2018 NDHS report includes a full description of the sampling process (36).
Derivation of the outcome variable
To prevent malaria in pregnancy (MiP), it is advised that expectant women take IPTp-SP at least three times (37,38). The best method for reducing MiP in malaria endemic countries like Nigeria is to use insecticide-treated mosquito nets, treat malaria quickly and effectively, and treat it intermittently with sulphadoxine-pyrimethamine (IPTp-SP) while pregnant (39). An optimum intake of IPTp-SP was therefore the study's main outcome variable. Receiving three or more doses of IPTp-SP during pregnancy was the operational definition of this. Women were questioned in the surveys about whether they used SP/Fansidar to prevent malaria when they were pregnant. Women were additionally questioned about how often they took SP/Fansidar in order to specifically determine how frequently they took it. This was posed as “How many times did you take SP/Fansidar during this pregnancy?”. To be consistent with the literature (37,38,40), women who reported that they took less than three doses of SP/Fansiar were classified as “low intake of IPTp-SP” while those that took three or more doses of SP/Fansidar were classified as “optimal intake of IPTp-SP. We recoded “low intake of IPTp-SP” as “0” and “optimal intake of IPTp-SP” as “1”.
Derivation of the explanatory variable
Recommended ANC visits served as the study's main explanatory variable. To conceptualise recommended ANC visits in accordance with prior studies' conceptualization of recommended ANC visits and the WHO recommendation on minimum ANC visits for expecting mothers (41–43), our definition of recommended ANC visits included four or more ANC visits. The NDHS asks women about their past pregnancies as well as other vital examinations that are important for maternal and child health, like ANC visits. Therefore, during the survey, women who gave birth two years before the data collection were asked the number of times they attended ANC, "How many times did you receive antenatal care during this pregnancy?”. To be consistent with the literature (41,42), all women who mentioned that they had less than four ANC visits were classified as “not recommended” while those who made four or more visits were classified as “recommended”. We recoded “not recommended” as “0” and recommended as “1”.
Derivation of covariates
Age, wealth position, education, marital status, parity, occupation, religion, access to mass media, sex of household head, region, autonomy in making health decisions, ability to pay for necessary transportation, and distance to a health center were the fourteen covariates considered in this study. These factors weren't predetermined, but their theoretical significance to IPTp-SP intake led to their selection (40). Some of these variables were recoded in order to make our results easier to interpret. Age was recoded into “15-19=1”, “20-34=2” and “35 and above=3”, wealth status recoded into “poor=1”, “middle=2” and “rich=3”, and education recoded as “no formal education=1” and “formal education=2”.
Considering the Nigerian fertility rate (which is Nigeria is 5.3 children per woman) (36), parity was recoded into “one birth=1”, “two births=2”, “three births=3”, “four births=4” and “five or more births=5”, employment recoded into “none working class=1” and “working class=2”, religion recoded into “Christians=1”, “Muslims=2”, “Others=3”, partner’s education recoded into “no formal education=1” and “formal education=2”, health decision-making autonomy recoded into “respondent alone=1”, “respondent with others=2” and “others=3”, money needed for transport recoded into “problematic=1” and “unproblematic=2” and distance to travel was recoded into “problematic=1” and “unproblematic=2”. Following Appiah et al. (44) derivation of mass media, we computed access to mass media from three variables: frequency of reading newspaper/magazine, frequency of listening to the radio, and frequency of watching television, which was asked during the surveys. Each of these variables had three responses: ‘not at all’, ‘less than once a week’, and ‘at least once a week’. A composite variable was created whereby all ‘less than once a week’ and ‘at least once a week’ responses were classified as having access to mass media whilst ‘not at all’ was considered as not having access to mass media.
Statistical analysis
The analysis proceeded with steps. Firstly, we applied the weighting factor inherent in the dataset (v005/100,000) to cater for sampling biases and over-generalization. Further, we calculated the proportion of rural women who had optimal intake of or otherwise, and the results were left in proportion and percentage. Additionally, proportionality calculations for ANC visits and the rest of the covariates were done to summarise the general background characteristics of the rural women. Next, we computed the distribution of IPTp-SP intake across ANC visits and other covariates. Then, we applied a chi-square test of independence between our main outcome variable and the explanatory variables to assess the association between them. At a cut-off point of 0.05, any explanatory variable that was not associated with the main outcome variable was not entered into our multivariate model.
Further, we conducted multicollinearity test to confirm whether the explanatory variables correlated with each other using variance inflation factor (VIF). The results showed no evidence of multicollinearity between the explanatory variables (Mean VIF=1.43, Maximum VIF=1.95, Minimum VIF=1.02) (see Appendix 1). Two logistic regression models were built at 95% two-tailed confidence intervals (95% CI). Model I only considered our main outcome variable and key explanatory variable, and the results were expressed in odds ratio (OR). In Model II, we accounted for the influence of other covariates on the outcome variable, and the results were presented in adjusted odds ratio (aOR). Our results were declared as having a higher likelihood to the outcome variable when the odds were above 1 and a lesser likelihood to the outcome variable when the odds were below 1. We assessed the model fit using ‘linktest’ and ‘goodness-of-fit’ commands, and the results indicated that our model was well-specified. All the analyses were done using Stata version 16.0.
Ethical considerations
This study relied on a secondary data source. As such, the authors were not directly involved in the data gathering, fieldwork, or any activity connected with the data generation. Therefore, ethical issues applicable to the conduct of the survey did not apply to this study. However, permission to use the dataset was sought from Measure DHS. The datasets were downloaded after Measure DHS gave authors clearance to use the data. Meanwhile, Measure DHS anonymised the dataset before making it available to the public. Also, they reported that all ethical standards applicable to human participation in a survey were followed (45).