Using three national Demographic and Health surveys, 2003, 2008 and 2013, we quantify the associations between FIC and child, household and community factors in Nigeria. Allowing for child, household and community level factors, in multilevel logistic analysis of a pooled data set covering the three DHS, the odds of being fully immunised in 2008 was triple, and in 2013 it was 4.5-times that in 2003. FIC differed substantially by place of residence, with children in urban formal and slum areas being 69 and 45% respectively, more likely to be immunised than those in rural areas. Overall, and in each of the three place of residence strata separately, we found place of delivery, antenatal care, maternal education, maternal age at child birth, religion, place of residence, media exposure and distance to the health facility to be significantly associated with FIC. The pattern of the adjusted odds in overall pooled and stratified logistic models were similar, although the magnitude of odds varied. The pooled analyses provided quantification of changes over time, as well as reliable quantification of the associations with sociodemographic variables allowing for year and setting. Then stratified analyses allowed quantification of the association, highlighting potential differences by setting in these associations. In line with this adjusted odds, the PAR analysis quantified what could be achieved if these variables are optimised and provided more explanation of the FIC variation like health system related factors being more influential in the rural area compared to urban.
Overall, our results are in line with those from previous studies [6,7,8,9], based on more robust methodology we provided more details such as the association in the different places of residence and the improvement that could be obtained by ensuring all mothers obtain the level of the socio-demographic variable associated with optimum FIC.
Child level factors
Our finding that children of lower birth orders were more likely to be fully immunised than higher birth orders confirm evidence from studies in Nigeria and India [6, 24,25,26,27]. Competing demands for family resources and time with increasing number of children likely explains at least part of this association . In our study FIC was higher among children delivered in a health facility than those delivered at home in all three settings, similar to earlier studies [6,7,8,9], which were criticised on methodological grounds for including the 1999 NDHS data in their analyses . Health facility delivery would likely increase the uptake of birth dose vaccines, BCG, OPV and HBV, and the provision of immunisation health messages, including date and place for subsequent immunisation sessions [9, 27, 30, 31]. Further, non-delivery in a health facility may be suggestive of distrust and lack of confidence in modern medicine and its providers . Similarly, children of mothers who attended antenatal care during their pregnancy independently had higher FIC coverage and significantly greater odds of being fully immunised than children whose mothers did not attend antenatal care. Adding to previously reported evidence [9, 26, 31], this study provided the strength and pattern of the significance of the association and residence disparity between antenatal care and fully immunised odds in Nigeria, with stratified multilevel analysis that showed the FIC adjusted odds of delivery in health facility was highest in the slums. PAR analysis estimated the percentage point increase in FIC that can be achieved if all mothers had attended antenatal care.
Household level factors
Maternal education level was consistently associated with FIC: as the mothers’ level of education increased, the child’s fully immunised odds increased, in line with results of several studies from Nigeria and elsewhere [6, 7, 17,18,19, 24, 25, 28, 32,33,34,35,36,37,38,39,40,41,42,43,44,45]. We additionally explored and quantified the pattern and trend of the association between maternal education and FIC status over the years in the general population and in each of the three settings and found its influence positive. Further analysis with PAR estimated that if maternal education was improved the whole population would benefit but the rural population stand to gain more. Increased maternal education is likely associated with increased health seeking behaviour, improved understanding of immunisation messages, knowledge of available immunisation delivery sites and having more money to cover the transport cost to health facilities [27, 46, 47].
Mother’s religion was also associated with the likelihood of the child to be fully immunised, such that a Christian child was several times more likely to be fully immunised than a Muslim child, with at least 25 point difference in their FIC irrespective of place of residence. This finding corroborated results from previous research in Nigeria and other countries [25, 48, 49]. Renne explained that Muslim leaders in Nigeria felt the vaccines were contaminated with HIV and anti-fertility substances aimed at reducing the Muslim population , while Taylor identified political and socioeconomic factors as reasons for acceptance or refusal of polio vaccine administration . New evidence provided by this study was the lack of significance of the relationship between religion and fully immunised child odds in the urban formal areas. As found in earlier studies, the association between the age of the mother and fully immunised child status was significant [25, 41, 50]. Reasons suggested include lack of child care experience by young mother, experience gained by older mothers on the effectiveness of vaccines, and the effect of treating child illness on family income [6, 9]. Our study population was larger than those in previous studies, and nationally representative, thus increasing reliability of our findings.
Households with regular exposure to media had children with higher FIC than households without regular media exposure. This relationship had been documented as a determinant of childhood immunisation in previous research based on secondary analysis of NDHS datasets [9, 25]. The media provides information on the benefits of immunisation, health activities and location of health facilities, as such can serve as a tool to improve childhood immunisation .
Contrary to the findings of previous studies in Nigeria [9, 25], we found that the association between household wealth and childhood immunisation was not statistically significant, which may be attributed to the recoding of the wealth variable in this study to make it urban or rural relative, lest all rural population were classed as being poor relative to urban populations. Despite the free provision of immunisation services in public owned health facilities, there are still indirect costs that can be barriers, such as transport cost to the health facility that is far from the mothers, and for the low income earner's inability to be excused from work as a result of the considerable time spent for journey to and time spent in the health facility . Also, the lack of money has been reported to hinder appropriate health care seeking behaviour .
Community level factors
Overall, the Northern regions (Northcentral, Northeast and Northwest) had lower FIC adjusted odds than the Southern (Southeast, Southsouth and Southwest) regions, also reported by others [17,18,19, 40, 43]. Reasons suggested include higher education level and higher household wealth in the South than in the North, northerners are mostly Muslims and the southerners predominantly Christian, and the recent Muslim insurgency in the Northeast [9, 25]. However, across the six regions in Nigeria, there was high variability in the likelihood of being fully immunised. The Northwest region which is the most populous had the lowest FIC coverage and least fully immunised child odds compared to the other regions, while the Southsouth region had the highest fully immunised children odds. As expected, children of mothers who felt the distance to the nearest health facility to seek health was a big problem had lower FIC coverage and lower fully immunised child odds than children with mothers who did not see the distance as a big problem. This finding is in line with the results of several studies in Nigeria [9, 25, 51].
Our novel contribution lies in the PAR estimates, which showed that the population FIC in each of the three areas of residence was dependent on key sociodemographic exposure. Simply, the PAR is the additional number of children that would have been vaccinated if the sociodemographic variable had been maximised. Though no particular PAR trend was found across the places of residence, variables such as maternal education and media exposure that are linked to building knowledge about immunisation had the greatest effects in the rural areas, next was the slums and least in the urban formal settings. The PAR estimated highest percentage point increase in FIC can be achieved in all places of residence by giving all mothers higher education: rural (38.15), urban formal (22.88) and slum (23.76). Also, with all households having regular media exposure, the PAR was rural (6.30), urban formal (2.81) and slum (3.56). Hence, PAR analysis makes planning more realistic since target setting can be more correctly done as the number of children to be reached to increase the FIC by a percent as well as the variable to minimise or maximise to achieve it in each residence is known.
The analyses are based on data from retrospective cross-sectional surveys; immunisation data was collected for children aged between 12 and 23 months from the child’s immunisation card, with the mother recalling the information if the card was not available. Maternal report may be subject to recall bias, especially as event could have occurred more than a year before the interview. Similarly, maternal education level has been reported to be associated with recall bias, with the mothers of more educated level being more likely to accurately recall the child’s immunisation history. Thus, adjusting for maternal education in the multilevel logistic models reduced the effect of maternal education on the recall bias . The household wealth variable in NDHS is a proxy based on the presence or otherwise of a number of assets, rather than on direct measurement of household income . However, contrary to earlier studies, we calculated household wealth relative to the area of residence rather than for the overall population, and our wealth variable thus indicates wealth relative to the wealth of others living in the same area. Some of the selected NDHS variables were not perfectly aligned to the research question, for example the variable, “distance to the nearest health facility when seeking health care”, does not specifically refer to child immunisation. As the NDHS data provides information only on urban and rural place of residence, we used UN HABITAT guidelines to recode the urban data into formal and slum households. This may have introduced selection bias, as the DHS sampling process was based on projections from general census held several years earlier, and as such may not be fully representative of the population when the DHS was conducted.