Coverage and determinants of COVID-19 vaccination among 36 555 Nigerian women of childbearing age: a mixed-effects analysis of nationally representative data

Abstract

Introduction

Even though COVID-19 has spread globally, with millions of deaths, vaccination remains low in resource-constrained settings. Uptake of COVID-19 vaccine has remained suboptimal despite its potential for protection against the disease. The objective of this study was to assess the factors associated with higher COVID-19 vaccination uptake among Nigerian women of reproductive age.

Methods

A sample of 36 555 women aged 15–49 years was extracted from the 2024 Nigeria Demographic and Health Survey data. The outcome variable was COVID-19 vaccination. Percentage and χ² tests were used in univariate and bivariate analyses. In addition, self-reported COVID-19 vaccination uptake was examined using multilevel ordinal logistic regression analysis.

Results

Approximately 73.9% of Nigerian women received no dose of COVID-19 vaccination. Those who received one dose were 13.4%, two doses were 10.5% and >2 doses were 2.1%, respectively. The full model (model V) was selected based on model fitness selection criteria. Older women, educated, Muslims, those covered by health insurance, formerly married, employed, exposed to media, from higher household wealth, have community-level willingness for COVID-19 vaccine uptake, native of the community and living in North East/North Central/South East/South South/South West had higher odds of being in a higher COVID-19 vaccination category when compared with women aged 15–19 years, with no formal education, Christians, uninsured, single, unemployed, not exposed to media, from poorest household, no community-level willingness for COVID-19 uptake, not a native of the community and from North West, respectively. On the other hand, women having a household size >6 were less likely to be in a higher COVID-19 vaccination category when compared with those having a household size of 1–2 members.

Conclusions

COVID-19 vaccination was approximately one-quarter. This finding is grossly suboptimal for herd immunity. Demographic, socioeconomic and environmental factors were associated with higher COVID-19 vaccination uptake among Nigerian women of reproductive age.

WHAT IS ALREADY KNOWN ON THIS TOPIC

• National uptake of COVID-19 vaccines in Nigeria remained well below programme targets and showed marked subnational variation, meaning many states lagged considerably behind national averages.

• Acceptance and uptake of COVID-19 vaccine varied by population groups; including healthcare workers’ acceptance, individual perceived risk of COVID-19 vaccination, trust in health regulators, education and exposure to correct information.

WHAT THIS STUDY ADDS

• This study provides a foremost report of national COVID-19 vaccination uptake among women of reproductive age in Nigeria.

• It investigates contextual factors associated with higher COVID-19 vaccination uptake among Nigerian women of reproductive age.

• The findings set a roadmap for the design and implementation of women-centred improved uptake strategies for COVID-19 vaccination, informing tailored interventions and policies for moderate-to-high transmission settings.

HOW THIS STUDY MIGHT AFFECT RESEARCH, PRACTICE OR POLICY

• Health programmes can use the findings of this study to design and implement context-specific service-delivery changes (integrating vaccination into maternal and child health contacts, mobile outreach, community-led risk communication) to increase uptake among reproductive-age women.

• Evidence on subnational disparities and population groups at highest risk of non-uptake would support policymakers to prioritise funding, cold-chain and outreach resources, and to adopt gender-sensitive policy measures to improve equitable vaccine coverage and strengthen pandemic preparedness.

Introduction

The WHO designated the COVID-19 as a global public health emergency on 30 January 2020.¹ The declaration marked the beginning of a new normal, where countries had to adopt various control measures, including travel restrictions, phased lockdowns, bans on public gatherings and mandatory mask-wearing. The pandemic had a far-reaching impact on health, food security and economy.² Since the advent of COVID-19, there have been millions of reported cases and deaths globally. These staggering figures necessitated the development of vaccines among other preventive measures to halt the spread. By the end of 2024, more than 13.64 billion doses of COVID-19 vaccines had been administered worldwide.³

Several pharmaceutical laboratories have developed COVID-19 vaccines to fight the spread of COVID-19.^4,5 Although WHO approved 12 vaccines for emergency use globally,⁶ the Nigerian Centre for Disease Control, however, recommended seven of these vaccines. These include Vaxveria by Oxford/AstraZeneca Pharma, Covishield by Serum Institute of India, Comirnaty by Pfizer/BioNTech, Jcovden by Johnson and Johnson, Spikevax by Moderna, Sputnik V by Gamaleya and Covilo by Sinopharm. The Africa Centre for Disease Control recently reported 51.8% full vaccination coverage, indicating a lower-than-expected vaccination coverage across the continent.⁷ In Nigeria, only 59.3% of the 115 983 921 eligible population had been fully vaccinated.⁷ This suboptimal vaccination uptake will defeat herd immunity.

COVID-19 vaccination is administered according to standardised dosing schedules designed to optimise immunogenicity and population-level protection. Most primary vaccination series consist of one or two doses, depending on the vaccine platform (mRNA, viral vector or inactivated vaccines), administered intramuscularly at manufacturer-recommended intervals.⁸ The initial (primary) series aims to elicit a robust immune response by stimulating both humoral and cellular immunity against SARS-CoV-2.⁹ Booster doses are subsequently recommended for eligible populations to restore waning immunity and enhance protection against severe disease, hospitalisation and death, particularly in the context of emerging variants.¹⁰ Dosing schedules may vary by age group, risk status and vaccine product, with lower antigen doses typically used for children and adolescents. Public health guidelines emphasise adherence to recommended intervals to ensure optimal effectiveness and safety. Monitoring of vaccine dosing and completion is essential for assessing coverage, evaluating programme performance and supporting the achievement of population immunity through widespread and timely vaccination uptake.

Safety concerns,^11–13 myths and misconceptions, misapprehensions, religious beliefs and hesitancy^12,14–16 have been attributed to low uptake of COVID-19 vaccination. In addition, demographic, socioeconomic and environmental factors have also been identified to influence vaccination uptake.^12,16–21 Though evidence-based studies have been conducted to understand the determinants of COVID-19 vaccination uptake,^{4,12,14,16–18} however, there is a dearth of data on the factors associated with COVID-19 vaccination uptake among Nigerian women of reproductive age. Understanding the uptake of COVID-19 vaccination among a key population will be useful in designing effective interventions for preventing infectious diseases in Nigeria being a multiethnic populous nation with diverse contexts in healthcare system. The objective of this study was to assess factors associated with higher COVID-19 vaccination uptake among Nigerian women of reproductive age.

Methods

Data source

A cross-sectional study was conducted for 36 555 women of reproductive age (15–49 years) using the 2024 Nigeria Demographic and Health Survey (NDHS) data. The National Population Commission (NPC) acted on behalf of the Federal Ministry of Health and Social Welfare (FMoHSW) to support the data collection process. Data collection took place between 1 December 2023 and 7 May 2024. The US Agency for International Development’s funds Demographic and Health Surveys Program and offers technical support for population and health surveys across the globe. In addition, ICF, the WHO, the United Nations Population Fund, UNICEF and the Global Fund to Fight AIDS, Tuberculosis and Malaria (Global Fund) were among the other organisations and agencies that provided financial or technical assistance to enable the survey’s successful implementation. The data analysed for this study can be accessed here.²²

Exclusion criteria

Institutional populations, including those in hotels, barracks and prisons, were not included in the study.

Sample design

The upgraded cartographical frame created for Nigeria’s anticipated first fully digital Population and Housing Census served as the sample frame for the 2024 NDHS. Nigeria is separated into six administrative zones: North Central, North East, North West, South East, South South and South West. The Federal Capital Territory (FCT) and 36 states make up the zones, which result in 37 subnational entities. Local government areas (LGAs) make up each state, while communities make up each LGA. During the delineation process for the impending census, each locality is further divided into suitable areas known as enumeration areas (EAs). These EAs served as the basis for primary sampling units (PSUs), also known as clusters in the 2024 NDHS. The 2024 NDHS sample was created to provide representative data for all six zones, the 36 states and the FCT, urban and rural areas separately and the nation as a whole.

The 2024 NDHS used a two-stage stratified sample design. By dividing each of the 36 states and the FCT into urban and rural areas, stratification was accomplished. There were 74 sampling strata in all. In each stratum, samples were chosen separately using a two-stage selection process. The initial step was to choose sample locations or clusters made up of EAs. Within each sampling stratum, EAs were selected with a probability that was proportionate to their size. There were 701 urban and 699 rural clusters out of the 1400 total that were chosen. A systematic sample of households was used in the second step. All of the chosen clusters underwent a household listing procedure and a systematic equal probability selection process was used to choose a set number of 30 households per cluster for a total sample size of about 42 000 households. At the time of listing and during interviews, Global Positioning System data were gathered for every household. The report of the sampling design, training of field assistants and data collection process has been published previously by NDHS (https://www.dhsprogram.com/data/dataset/Nigeria_Standard-DHS_2024.cfm?flag=1).²³

Selection and measurements of variables

Outcome variable

The response variable was COVID-19 vaccination and was determined from the question: ‘S1112R: Number of doses of COVID-19 vaccination received’, with responses as: ‘0, 1, 2, >2’. COVID-19 vaccination data was self-reported by the women based on the number of doses received during and after the global COVID-19 epidemic.

Explanatory variables

Individual-level factors

Age in 5-year groups between 15 and 49; education: none, primary, secondary, higher; religion: Christianity, Islam, traditional/no religion; covered by health insurance scheme: yes versus no; marital status: single, currently married, formerly married; employment status: not working versus working; cultural norms about wife-beating were created by aggregating responses from women in each community.^24,25 Here, we used the items: ‘beating justified if wife goes out without telling husband’, ‘beating justified if wife neglects the children’, ‘beating justified if wife argues with husband’, ‘beating justified if wife refuses to have sex with husband’ and ‘beating justified if wife burns the food’. A binary variable was created by aggregating responses from women for acceptance of wife beating, where ‘no’ indicates empowered women and ‘yes’ to wife beating justified indicates not empowered women; socioeconomically disadvantaged status: neighbourhood socioeconomic disadvantaged level used items such as rural residence, poorest household wealth, no formal education and not working. Using principal component analysis (PCA), the standardised z-score was used to disentangle the overall assigned scores to low, medium or high; exposed to media was created using standard indicators including, reading newspaper, listening to radio, watching TV, use of internet): respondents who reported not using any of the media (no) were coded as ‘0’, while those who used were coded ‘1’—yes versus no.

Household-level factors

Sex of household head: male versus female; household size: 1–2, 3–4, 5–6, 7+; household wealth: wealth indicator weights were assigned using PCA. The wealth indicator variables (bicycle, motorcycle/scooter, car/truck, main floor material, main wall material, main roof material, sanitation facilities, water source, radio, television, electricity, refrigerator, cooking fuel, furniture and number of people per room) were standardised and scores were assigned. Factor loadings, or factor coefficient scores, and z-scores were computed. To determine each household’s wealth index value, the indicator values were multiplied by the loadings and added together. The overall scores were divided into the poorest/poorer/middle/richer/richest groups using the standardised z-score.

Community-level factors

Since the Demographic and Health Survey (DHS) did not gather aggregate-level data at the community-level, we used EAs to represent communities. Therefore, the analysis’s community-level factors were centred on women’s characteristics, especially those that affect the coverage of COVID-19 vaccinations. Community-level women’s empowerment was created from the individual woman cultural norms about wife-beating justification by aggregating responses from women in each community; hence, community-level empowerment was categorised as high if the proportion of women who reported wife beating justified in a given community was 79%–100% and low if the proportion was 0%–78%. Community-level willingness for COVID-19 vaccine uptake was categorised as high if the proportion of women who reported willingness to COVID-19 vaccine uptake in a given community was 47%–100% and low if the proportion was 0%–46%. Community-level nativity was categorised as high if the proportion of women who had lived 5 years or more in a given community was 88%–100% and low if the proportion was 0%–87%. Community-level urbanicity was categorised as high if the proportion of women who reside in urban areas was 50%–100% and low if the proportion was 0%–49%. Community-level illiteracy was categorised as high if the proportion of women who cannot read at all in a given community was 40%–100% and as low if the proportion was 0%–39%. Community-level media use was categorised as high if the proportion of women who are exposed to media use in a given community was 66%–100% and as low if the proportion was 0%–65%. Community-level poverty was categorised as high if the proportion of women from the two lowest wealth quintiles in a given community was 34%–100% and low if the proportion was 0%–33%. Geographical region: North West, North East, North Central, South East, South South, South West.

Rationale for creating community-level factors

In this study, community-level variables were derived from individual-level DHS responses using an aggregation approach consistent with established multilevel modelling procedures in demographic and public health research. The DHSs employ a hierarchical sampling design in which women are nested within households and households are nested within clusters or PSUs, which are often used as proxies for communities. Because individuals living within the same cluster are likely to share similar social norms, socioeconomic conditions, cultural practices and access to health information and services, aggregating individual characteristics at the cluster level provides an appropriate measure of contextual or community influences beyond individual-level effects. This use of cluster-level aggregates of individual responses to characterise the social context is an established, if necessarily approximate, strategy in multilevel social-epidemiological research.^26–29 Because such derived measures cannot, by themselves, fully separate contextual from compositional effects, we entered each individual-level characteristic into the model simultaneously with its community-level aggregate, which is the recommended approach for distinguishing, as far as cross-sectional data allow, the contextual contribution from the compositional one.³⁰

The construction of community-level variables involved calculating the proportion of women within each DHS cluster who possessed a specific characteristic or behaviour of interest. For example, community-level women’s empowerment was generated by aggregating women’s responses regarding justification of wife beating within each cluster. Similarly, community-level willingness for COVID-19 vaccine uptake, nativity, urbanicity, illiteracy, media use and poverty were created by computing the cluster-level proportions of women exhibiting those attributes. This aggregation strategy is statistically justified because it transforms individual responses into contextual indicators reflecting the prevailing social environment within a community.

The use of aggregated proportions in multilevel analysis is supported by contextual theory, which posits that community norms and collective behaviours may independently influence health outcomes over and above individual characteristics. For instance, women residing in communities with high acceptance of wife beating may experience shared sociocultural constraints that influence health-seeking behaviours and vaccine uptake irrespective of their own personal beliefs. Likewise, living in communities with high media exposure or low poverty may facilitate greater diffusion of health information and improved access to healthcare services.

Following aggregation, the resulting community-level variables were categorised into ‘high’ and ‘low’ groups using empirically derived cut-offs based on the distribution of cluster-level proportions across the study population. Specifically, the cut-points corresponded approximately to the median or distributional thresholds of the aggregated variables, a commonly applied strategy in DHS-based multilevel studies when no universally established thresholds exist. These cut-points were defined a priori from the empirical distribution of each variable and were not selected by reference to the outcome; this distinction matters because it is outcome-optimised cut-points, rather than distribution-based ones, that introduce the serious bias described in the methodological literature.³¹ We nonetheless recognise that any dichotomisation discards information, can reduce power, may introduce some misclassification and residual confounding, and limits comparison with studies using different thresholds; we therefore report all cut-points explicitly and, as a robustness check, also model these variables on their continuous scale (see Results and online supplemental table S2).

This approach is methodologically appropriate because multilevel ordinal regression models are specifically designed to simultaneously estimate both fixed effects of individual-level and community-level characteristics and random effects attributable to clustering. By incorporating aggregated community variables, the analysis accounts for contextual heterogeneity and, because the outcome is measured and modelled at the individual level, avoids the ecological (cross-level) inferential fallacy that arises in purely aggregate designs,³⁰ while preserving the multilevel structure of the DHS data; it does not, however, eliminate the possibility of residual cross-level confounding. Furthermore, inclusion of these contextual variables improves estimation of between-community variation and provides insights into how shared environmental and sociocultural conditions shape women’s health behaviours and outcomes. Overall, the derivation of community-level variables from aggregated DHS cluster responses represents a robust and widely accepted statistical strategy for examining contextual determinants in multilevel population health research. This approach was used in a previous study.^32–34

Analytical approach

Stata software, V.17.0 (Stata), was used to analyse the data. The survey module’s (‘svy’) function was used to account for sampling design (weighting, clustering and stratification) since the study employed multistage stratified cluster sample design data. Multicollinearity was assessed using the variance inflation factor, which is known to seriously impair logit models.³⁵ Percentage and chi-square test were computed in the univariate and bivariate analyses. There were no missing data on the variables included in the analysis; all 36 555 women had complete information on the outcome and explanatory variables, and no imputation was therefore required. Furthermore, multilevel ordinal logistic regression models with random intercepts at household and community levels were fitted to examine the factors associated with higher COVID-19 vaccination uptake. Specifically, a cumulative (proportional-odds) specification was used, in which the four ordered categories (0, 1, 2 and >2 doses) share a common set of slope coefficients separated by three thresholds, so that each OR represents the odds of being in a higher rather than a lower vaccination category. The proportional-odds (parallel-lines) assumption was formally assessed using the Brant test.³⁶ As a sensitivity analysis, we fitted a multilevel ordinal logistic regression model using COVID-19 vaccination status categorised as 0 dose, 1 dose, 2 doses and booster dose to assess whether factors associated with vaccine uptake remained consistent across increasing vaccination intensity. In addition, to test the robustness of the community-level findings to the dichotomisation of the aggregated variables, we re-fitted the model with the six community-level variables specified on their continuous (cluster-proportion) scale, with community-clustered robust SEs.

Fixed and random effects

A model for self-reported COVID-19 vaccination was developed at three levels. Households were nested within communities and individual women were nested within households. Five models were constructed in this study. In the empty or unconditional model, the community-level and household-level variances were first calculated without any explanatory variables. The purpose of the null or empty model was to assess the degree to which factors at the household and community levels explained the observed effects. In order to examine the factors associated with COVID-19 vaccination, the second model included individual-level factors; the third model included household-level factors; and the fourth model included community-level factors. Lastly, the fifth model (complete model) accounted for individual, household and community-level factors associated with COVID-19 vaccination. Fixed effects, or measures of association, of the factors associated with COVID-19 vaccination were reported using ORs and their 95% CI. The log-likelihood, Akaike information criterion (AIC) or Bayesian information criterion (BIC) with the lowest value indicated the best fit model among the five models.³⁷ The intracluster correlation coefficient (ICC) was employed to measure the probable contextual effects. We evaluated the degree of similarity among women residing in the same community and household using ICC. Using the Snijders and Bosker formula,³⁸ the linear threshold was utilised to calculate ICC.³⁹

Results

Table 1 shows the uptake of 0, 1, 2, >2 doses of COVID-19 vaccine across women’s characteristics. The uptake of 1, 2, >2 doses of COVID-19 vaccine increased steadily by older age groups, higher educational level, health insurance coverage, working status, exposure to media. The converse was, however, true for women who received no dose of COVID-19 vaccine.

Table 1. Distribution of COVID-19 vaccination uptake by individual-level factors among Nigerian women of reproductive age, n=36 555.

Variable	n (%)	Doses of COVID-19 vaccine received (%, 95% CI)				P value
		0	1	2	>2
Age in 5-year groups						<0.001*
15–19	7377 (20.2)	80.8 (78.9 to 82.6)	12.5 (11.3 to 13.8)	6.0 (5.1 to 7.0)	0.7 (0.5 to 1.0)
20–24	6263 (17.1)	74.7 (72.8 to 76.6)	14.3 (13.1 to 15.7)	9.7 (8.7 to 10.8)	1.2 (0.9 to 1.6)
25–29	5956 (16.3)	73.4 (71.6 to 75.2)	13.6 (12.4 to 15.0)	11.2 (10.1 to 12.3)	1.8 (1.4 to 2.2)
30–34	5262 (14.4)	69.1 (66.9 to 71.2)	15.3 (14.0 to 16.7)	13.1 (11.8 to 14.5)	2.5 (1.9 to 3.2)
35–39	4749 (13.0)	68.3 (66.3 to 70.3)	15.5 (14.2 to 16.9)	13.2 (11.8 to 14.6)	3.0 (2.4 to 3.7)
40–44	3926 (10.7)	65.9 (63.8 to 68.0)	14.2 (12.8 to 15.7)	15.8 (14.1 to 17.5)	4.2 (3.4 to 5.0)
45–49	3022 (8.3)	65.7 (63.3 to 68.0)	14.5 (13.0 to 16.1)	15.8 (14.2 to 17.4)	4.1 (3.2 to 5.3)
Highest educational level						<0.001*
No education	10 836 (29.6)	80.2 (77.4 to 82.7)	11.3 (9.9 to 12.9)	7.5 (6.3 to 8.9)	1.0 (0.7 to 1.4)
Primary	4098 (11.2)	71.1 (68.7 to 73.5)	15.5 (14.0 to 17.1)	10.8 (9.5 to 12.3)	2.5 (1.9 to 3.3)
Secondary	16 024 (43.8)	71.2 (69.9 to 72.5)	15.3 (14.5 to 16.2)	11.2 (10.4 to 12.0)	2.2 (2.0 to 2.6)
Higher	5597 (15.3)	59.3 (57.4 to 61.1)	16.1 (14.9 to 17.4)	20.3 (18.9 to 21.7)	4.3 (3.7 to 5.1)
Religion						0.005*
Christianity	18 413 (50.4)	71.3 (70.0 to 72.6)	14.1 (13.3 to 14.9)	11.9 (11.2 to 12.7)	2.7 (2.3 to 3.1)
Islam	17 947 (49.1)	73.3 (71.1 to 75.3)	14.2 (13.0 to 15.5)	10.7 (9.7 to 11.9)	1.8 (1.5 to 2.1)
Traditional/no religion	195 (0.5)	83.9 (76.0 to 89.5)	7.2 (3.6 to 13.6)	7.3 (3.7 to 14.1)	1.6 (0.6 to 4.7)
Covered by health insurance						<0.001*
No	35 306 (96.6)	73.3 (72.0 to 74.6)	13.9 (13.2 to 14.7)	10.8 (10.1 to 11.5)	2.0 (1.8 to 2.3)
Yes	1249 (3.4)	49.4 (45.8 to 53.1)	20.0 (16.9 to 23.4)	24.3 (21.3 to 27.5)	6.4 (4.8 to 8.3)
Marital status						<0.001*
Single	10 726 (29.3)	75.9 (74.5 to 77.3)	14.2 (13.2 to 15.3)	8.7 (8.0 to 9.5)	1.2 (1.0 to 1.5)
Currently married	23 929 (65.5)	71.5 (69.9 to 73.1)	14.0 (13.1 to 14.9)	12.2 (11.3 to 13.1)	2.4 (2.1 to 2.7)
Formerly married	1900 (5.2)	66.3 (63.5 to 69.0)	16.1 (14.1 to 18.3)	13.1 (11.3 to 12.0)	4.5 (3.5 to 5.7)
Employment status						<0.001*
Not working	14 766 (40.4)	79.7 (78.0 to 81.2)	11.5 (10.6 to 12.4)	8.0 (7.1 to 9.1)	0.9 (0.7 to 1.1)
Working	21 789 (59.6)	67.4 (66.0 to 68.7)	16.0 (15.1 to 17.0)	13.5 (12.8 to 14.3)	3.1 (2.8 to 3.5)
Wife beating justified						<0.001*
Low	28 824 (78.9)	71.3 (70.0 to 72.6)	14.3 (13.5 to 15.1)	12.0 (11.2 to 12.7)	2.5 (2.2 to 2.7)
High	7731 (21.1)	76.5 (74.3 to 78.7)	13.6 (12.2 to 15.1)	8.7 (7.5 to 10.1)	1.2 (0.9 to 1.6)
Socioeconomically disadvantaged status						<0.001*
Low	12 208 (33.4)	66.8 (65.2 to 68.3)	14.5 (13.6 to 15.5)	14.7 (13.7 to 15.8)	4.0 (3.5 to 4.5)
Medium	12 182 (33.3)	71.3 (69.3 to 73.2)	15.9 (14.6 to 17.2)	11.1 (9.9 to 12.4)	1.8 (1.5 to 2.2)
High	12 165 (33.3)	78.5 (75.4 to 81.4)	12.3 (10.7 to 14.1)	8.3 (6.9 to 9.9)	0.9 (0.6 to 1.3)
Exposed to media						<0.001*
No	14 727 (34.8)	79.9 (77.8 to 81.8)	12.3 (11.1 to 13.7)	6.8 (5.9 to 7.8)	1.0 (0.8 to 1.4)
Yes	23 828 (65.2)	68.3 (67.0 to 69.6)	15.2 (14.4 to 16.0)	13.7 (12.9 to 14.6)	2.8 (2.5 to 3.1)

P value obtained from χ2 test.

^*Significant at p<0.05.

Table 2 shows the uptake of 0, 1, 2 and >2 doses of COVID-19 vaccine across women’s characteristics. The uptake of 1, 2 and >2 doses of COVID-19 vaccine increased steadily by higher household wealth quintiles, community-level women’s empowerment, community-level willingness for COVID-19 vaccine uptake, community-level nativity, community-level urbanicity, community-level media use, respectively. The converse was however true for women who received no dose of COVID-19 vaccine. In addition, there were differences in the patterns of COVID-19 vaccination uptake across geographical regions.

Table 2. Distribution of COVID-19 vaccination uptake by household-level and community-level factors among Nigerian women of reproductive age, n=36 555.

Variable	n (%)	Doses of COVID-19 vaccine received (%, 95% CI)				P value
		0	1	2	>2
Sex of household head						<0.001*
Male	28 847 (78.9)	73.3 (71.8 to 74.8)	14.0 (13.1 to 14.9)	10.8 (10.0 to 11.6)	1.9 (1.7 to 2.2)
Female	7708 (21.1)	69.1 (67.6 to 70.6)	14.7 (13.7 to 15.8)	13.1 (12.1 to 14.1)	3.1 (2.6 to 3.6)
Household wealth						<0.001*
Poorest	6255 (17.1)	83.1 (80.7 to 85.2)	10.6 (9.1 to 12.3)	5.6 (4.6 to 6.9)	0.6 (0.4 to 1.0)
Poorer	5963 (16.3)	76.5 (73.2 to 79.5)	13.6 (12.0 to 15.5)	8.8 (7.2 to 10.7)	1.1 (0.7 to 1.6)
Middle	7279 (19.9)	70.2 (67.9 to 72.3)	16.6 (15.2 to 18.1)	11.4 (10.1 to 12.8)	1.9 (1.5 to 2.4)
Richer	8430 (23.1)	68.3 (66.6 to 69.9)	15.9 (14.7 to 17.1)	12.9 (11.8 to 14.1)	2.9 (2.5 to 3.4)
Richest	8628 (23.6)	67.8 (66.0 to 69.5)	13.2 (12.1 to 14.3)	15.4 (14.2 to 16.6)	3.6 (3.1 to 4.3)
Household size						0.001*
1–2	3769 (10.3)	71.8 (69.8 to 73.7)	12.5 (11.2 to 13.9)	13.1 (11.9 to 14.5)	2.5 (2.0 to 3.2)
3–4	9263 (25.3)	70.8 (69.2 to 72.4)	13.5 (12.5 to 14.5)	12.9 (11.9 to 13.9)	2.8 (2.4 to 3.3)
5–6	9882 (27.0)	71.2 (69.7 to 72.7)	14.3 (13.3 to 15.4)	11.8 (10.9 to 12.8)	2.7 (2.3 to 3.1)
>6	13 641 (37.3)	74.6 (72.4 to 76.6)	14.8 (13.6 to 16.1)	9.4 (8.3 to 10.6)	1.3 (1.1 to 1.7)
Community-level women empowerment						<0.001*
Low	15 100 (41.3)	77.1 (74.8 to 79.3)	13.3 (12.1 to 14.6)	8.4 (7.2 to 9.7)	1.2 (0.9 to 1.5)
High	21 455 (58.7)	68.9 (67.2 to 70.5)	14.8 (13.8 to 15.8)	13.4 (12.5 to 14.3)	2.9 (2.6 to 3.3)
Community-level willingness for COVID-19 vaccine uptake						<0.001*
Low	18 423 (50.4)	86.6 (85.8 to 87.3)	7.3 (6.8 to 7.8)	5.1 (4.7 to 5.6)	1.0 (0.8 to 1.3)
High	18 132 (49.6)	61.5 (59.6 to 63.3)	19.5 (18.3 to 20.7)	16.0 (14.9 to 17.1)	3.1 (2.7 to 3.5)
Community-level nativity						0.0647
Low	13 599 (37.2)	73.0 (71.0 to 74.9)	13.5 (12.3 to 14.8)	11.4 (10.2 to 12.7)	2.1 (1.8 to 2.5)
High	22 956 (62.8)	72.2 (70.4 to 74.0)	14.5 (13.5 to 15.5)	11.1 (10.2 to 12.1)	2.2 (1.9 to 2.5)
Community-level urbanicity						<0.001*
High	17 937 (49.1)	68.9 (67.5 to 70.3)	14.8 (13.9 to 15.8)	13.2 (12.3 to 14.1)	3.1 (2.7 to 3.5)
Low	18 618 (50.9)	75.9 (73.6 to 78.0)	13.5 (12.3 to 14.8)	9.4 (8.3 to 10.6)	1.3 (1.0 to 1.6)
Community-level illiteracy						<0.001*
Low	21 800 (59.6)	68.6 (67.3 to 70.0)	15.3 (14.4 to 16.2)	13.3 (12.5 to 14.2)	2.8 (2.5 to 3.1)
High	14 755 (40.4)	77.7 (75.1 to 80.1)	12.5 (11.2 to 14.0)	8.4 (7.2 to 9.8)	1.4 (1.0 to 1.8)
Community-level media use						<0.001*
Low	15 009 (41.1)	77.3 (74.7 to 79.7)	13.4 (12.1 to 14.9)	8.2 (7.1 to 9.6)	1.1 (0.8 to 1.4)
High	21 546 (58.9)	68.8 (67.4 to 70.1)	14.7 (13.8 to 15.6)	13.5 (12.7 to 14.4)	3.0 (2.7 to 3.4)
Community-level poverty						<0.001*
Low	22 706 (62.1)	68.7 (67.3 to 70.0)	15.0 (14.2 to 15.9)	13.4 (12.6 to 14.3)	2.9 (2.6 to 3.3)
High	13 849 (37.9)	78.3 (75.5 to 80.8)	12.8 (11.4 to 14.4)	7.9 (6.7 to 9.3)	1.0 (0.7 to 1.4)
Geographical region						<0.001*
North West	8853 (24.2)	78.7 (76.1 to 81.1)	14.0 (12.4 to 15.7)	6.8 (5.6 to 8.3)	0.5 (0.3 to 0.8)
North East	5627 (15.4)	69.5 (65.3 to 73.3)	16.3 (14.4 to 18.4)	12.2 (10.2 to 14.6)	2.0 (1.4 to 2.9)
North Central	7117 (19.5)	71.6 (67.7 to 75.3)	13.3 (11.4 to 15.5)	12.3 (10.5 to 14.4)	2.7 (2.2 to 3.4)
South East	4823 (13.2)	81.9 (79.8 to 83.7)	10.5 (9.4 to 11.7)	7.0 (5.8 to 8.5)	0.7 (0.4 to 1.1)
South South	5383 (14.7)	76.1 (73.7 to 78.2)	13.2 (11.5 to 15.1)	8.5 (7.6 to 9.5)	2.3 (1.8 to 2.8)
South West	4752 (13.0)	54.4 (52.0 to 56.7)	16.2 (14.9 to 17.6)	23.3 (21.5 to 25.1)	6.1 (5.2 to 7.2)

P value obtained from χ2 test.

^*Significant at p<0.05.

Figure 1 shows the proportion of the type of COVID-19 vaccine received by Nigerian women of reproductive age. Over one-third of vaccinated women could not report the type of vaccine received. However, slightly less than one-fifth reported Astrazeneca/Oxford (19.6%), Johnson & Johnson (18.3%) and Pfizer (17.3%), respectively. Only 8.0% reported to have received Moderna.

Figure 1. Proportion of COVID-19 vaccine type received by the vaccinated women (n=9541).

Figure 2 shows that approximately 73.9% of Nigerian women received no dose of COVID-19 vaccination. Those who received one dose were 13.4%, two doses 10.5% and >2 doses 2.1%, respectively.

Figure 2. Proportion of COVID-19 vaccine doses received by Nigerian women (n=36 555).

Measures of variations (random effects) and model fit statistics

Table 3 shows the full model had the lowest Log-likelihood, AIC and BIC values (−25747.51, 51 583.02 and 51 957.31, respectively) and was selected. The variations in the odds of COVID-19 vaccination at community and household levels were estimated (σ²=0.71 and σ²=2.07), respectively. A variance of 0.71 at the community level means there is moderate clustering of the outcome at the community level. In addition, a household-level variance of 2.07 means that household membership explains a large portion of variability in the outcome, indicating strong clustering and strong influence of unmeasured household characteristics. The ICC values show that 12.0% and 46.0% of the variances in the odds of higher COVID-19 vaccination were due to community-level and household-level factors, respectively. The multilevel ordinal logistic regression model estimated three statistically significant threshold parameters: first threshold=5.79 (95% CI 5.36 to 6.23), second threshold=7.21 (95% CI 6.76 to 7.66) and third threshold=9.76 (95% CI 9.28 to 10.24). These thresholds represent the cumulative logit boundaries separating successive categories of the ordinal outcome variable. Their statistical significance indicates that the ordered categories of COVID-19 vaccination were distinctly differentiated within the model. The test of the proportional-odds assumption was statistically significant overall (χ²=530.7, df=78, p<0.001); this is expected given the very large sample (n=36 555), in which the test detects even small departures, and category-specific (generalised ordered logit) estimates were in a consistent direction across all three cumulative cut-points for every principal exposure (online supplemental table S1), indicating that the proportional-odds ORs provide a valid weighted-average summary of the category-specific effects.^36,40 In the sensitivity analysis specifying the community-level variables on their continuous scale, the associations were materially unchanged in direction and significance: community-level willingness for vaccination remained strongly associated with higher uptake (OR per 10 percentage-point increase=1.50, 95% CI 1.47 to 1.53) and community-level nativity likewise (OR 1.04, 95% CI 1.02 to 1.07); dichotomisation was thus conservative rather than a source of spurious association (online supplemental table S2).

Table 3. Random effect estimates and model intercepts of individual, household and community-level factors associated with higher COVID-19 vaccination uptake.

Random-effect	Model I	Model II	Model III	Model IV	Model V
Community-level
Variance (95% CI)	2.08 (1.86 to 2.33)*	1.96 (1.75 to 2.00)*	1.85 (1.65 to 2.06)*	0.64 (0.56 to 0.73)*	0.71 (0.62 to 0.81)*
ICC	0.29	0.27	0.27	0.11	0.12
Household-level
Variance (95% CI)	1.69 (1.47 to 1.95)*	2.08 (1.83 to 2.38)*	1.69 (1.47 to 1.94)*	1.72 (1.50 to 1.97)*	2.07 (1.82 to 2.36)*
ICC	0.53	0.55	0.52	0.42	0.46
Model intercepts
Threshold 1	1.58 (1.48 to 1.67)*	3.57 (3.33 to 3.81)*	2.27 (2.08 to 2.47)*	3.40 (3.12 to 3.69)*	5.79 (5.36 to 6.23)*
Threshold 2	2.88 (2.77 to 2.99)*	4.98 (4.73 to 5.24)*	3.58 (3.37 to 3.79)*	4.71 (4.42 to 5.00)*	7.21 (6.76 to 7.66)*
Threshold 3	5.27 (5.11 to 5.43)*	7.53 (7.24 to 7.83)*	5.97 (1.47 to 1.94)*	7.10 (6.78 to 7.42)*	9.76 (9.28 to 10.24)*
AIC	54 355.80	52 516.16	54 120.49	53 220.94	51 583.02
BIC	54 398.34	52 720.32	54 231.08	53 365.55	51 957.31
Log-likelihood	−27 172.90	−26 234.08	−27 047.25	−26 593.47	−25 747.51
Sample size
Community	1380	1380	1380	1380	1380
Household	26 697	26 697	26 697	26 697	26 697
Individual woman	36 555	36 555	36 555	36 555	36 555

Model I: baseline model with no explanatory variables, or empty null model (unconditional model); Model II: adjusted for individual-level factors; Model III: adjusted for household-level factors; Model IV: adjusted for community-level factors; Model V: full model adjusted for characteristics at the individual, household-level and community-level factors.

^*Significant at p<0.05.

AIC, Akaike information criterion; BIC, Bayesian information criterion; ICC, intracluster correlation coefficient.

Measures of association (fixed effects)

Table 4 shows the factors associated with higher COVID-19 vaccination uptake among Nigerian women of reproductive age (model V). Older women, educated, Muslims, those covered by health insurance, formerly married, employed, exposed to media, from higher household wealth, has community-level willingness for COVID-19 vaccine uptake, native of the community and living in North East/North Central/South East/South South/South West had higher odds of being in a higher COVID-19 vaccination category when compared with women aged 15–19 years, with no formal education, Christians, uninsured, single, unemployed, not exposed to media, from poorest household, no community-level willingness for COVID-19 uptake, not a native of the community, and from North West, respectively. On the other hand, women having a household size >6 were less likely to be in a higher COVID-19 vaccination category when compared with those having a household size of 1–2 members.

Table 4. Fixed effect of individual, household and community-level factors associated with higher COVID-19 vaccination uptake.

Variable	Adjusted OR (95% CI)
	Model I	Model II	Model III	Model IV	Model V
Age in 5-year groups
15–19		1.00			1.00
20–24		1.74 (1.53 to 1.97)*			1.71 (1.51 to 1.93)*
25–29		1.97 (1.71 to 2.27)*			1.97 (1.71 to 2.67)*
30–34		2.54 (2.19 to 2.96)*			2.57 (2.21 to 3.00)*
35–39		2.84 (2.43 to 3.32)*			2.88 (2.46 to 3.37)*
40–44		3.44 (2.92 to 4.05)*			3.49 (2.96 to 4.11)*
45–49		3.65 (3.07 to 4.33)*			3.67 (3.09 to 4.36)*
Highest educational level
No education		1.00			1.00
Primary		1.70 (1.48 to 1.95)*			1.62 (1.41 to 1.85)*
Secondary		2.25 (1.98 to 2.56)*			2.12 (1.87 to 2.41)*
Higher		3.83 (3.29 to 4.47)*			3.55 (3.04 to 4.15)*
Religion
Christianity		1.00			1.00
Islam		1.32 (1.16 to 1.49)*			1.14 (1.01 to 1.29)*
Traditional/no religion		0.82 (0.45 to 1.47)			0.80 (0.45 to 1.41)
Covered by health insurance
No		1.00			1.00
Yes		2.27 (1.91 to 2.69)*			2.24 (1.89 to 2.65)*
Marital status
Single		1.00			1.00
Currently married		1.11 (0.99 to 1.24)			1.11 (0.99 to 1.24)
Formerly married		1.27 (1.07 to 1.50)*			1.21 (1.02 to 1.43)*
Employment status
Not working		1.00			1.00
Working		1.53 (1.41 to 1.65)*			1.51 (1.39 to 1.63)*
Wife beating justified
No		1.00			1.00
Yes		1.06 (0.97 to 1.16)			1.05 (0.96 to 1.15)
Socioeconomically disadvantaged status
Low		1.00			1.00
Medium		0.87 (0.71 to 1.06)			1.02 (0.83 to 1.26)
High		0.55 (0.43 to 0.69)*			1.04 (0.73 to 1.47)
Exposed to media
No		1.00			1.00
Yes		1.32 (1.20 to 1.46)*			1.29 (1.17 to 1.42)*
Sex of household head
Male			1.00		1.00
Female			1.11 (1.00 to 1.21)*		1.07 (0.96 to 1.18)
Household wealth
Poorest			1.00		1.00
Poorer			1.41 (1.21 to 1.64)*		1.21 (1.03 to 1.41)*
Middle			2.03 (1.73 to 2.37)*		1.40 (1.18 to 1.67)*
Richer			2.83 (2.40 to 3.34)*		1.58 (1.30 to 1.93)*
Richest			3.38 (2.84 to 4.02)*		1.38 (1.11 to 1.71)*
Household size
1–2			1.00		1.00
3–4			0.94 (0.83 to 1.06)		0.91 (0.80 to 1.03)
5–6			0.98 (0.86 to 1.11)		0.92 (0.81 to 1.05)
>6			0.84 (0.74 to 0.95)*		0.85 (0.74 to 0.98)*
Community-level women empowerment
Low				1.00	1.00
High				1.14 (0.99 to 1.31)	1.15 (0.99 to 1.34)
Community-level willingness for COVID-19 vaccine uptake
Low				1.00	1.00
High				5.66 (4.99 to 6.42)*	6.05 (5.29 to 6.93)*
Community-level nativity
Low				1.00	1.00
High				1.23 (1.09 to 1.39)*	1.31 (1.15 to 1.49)*
Community-level urbanicity
High				1.00	1.00
Low				0.99 (0.86 to 1.13)	1.08 (0.88 to 1.31)
Community-level illiteracy
Low				1.00	1.00
High				0.77 (0.64 to 0.92)*	1.12 (0.92 to 1.38)
Community-level media use
Low				1.00	1.00
High				1.35 (1.11 to 1.63)*	1.04 (0.85 to 1.29)
Community-level poverty
Low				1.00	1.00
High				0.73 (0.59 to 0.89)*	0.97 (0.76 to 1.23)
Geographical region
North West				1.00	1.00
North East				3.01 (2.47 to 3.66)*	3.07 (2.48 to 3.79)*
North Central				2.50 (2.05 to 3.05)*	2.19 (1.75 to 2.73)*
South East				1.97 (1.55 to 2.50)*	1.61 (1.21 to 2.12)*
South South				1.93 (1.55 to 2.42)*	1.69 (1.30 to 2.20)*
South West				4.94 (3.95 to 6.17)*	4.13 (3.19 to 5.35)*

Model I: baseline model with no explanatory variables or empty null model (unconditional model); Model II: adjusted for individual-level factors; Model III: adjusted for household-level factors; Model IV: adjusted for community-level factors; Model V: full model adjusted for characteristics at the individual, household-level and community-level factors.

^*Significant at p<0.05.

Discussion

This study provides the first comprehensive analysis of the doses of COVID-19 vaccination uptake among Nigerian women. This has revealed several important findings with health policy implications. We used data from 36 555 Nigerian women of reproductive age. The study found 26.1% Nigerian women reported COVID-19 vaccination uptake (one dose was 13.4%, two doses was 10.5% and above two doses was 2.1%). This is similar to the population-level uptake from other African countries. In Tanzania, only 18% out of 3470 community members had received at least one dose of COVID-19 vaccine. From a previous study conducted in Nigeria, only 32.8% out of 2370 respondents were vaccinated against COVID-19.^16,18 From a study conducted across 52 African countries, only nine of the 52 African countries (Tunisia, Cabo Verde, Lesotho, Mozambique, Rwanda, Seychelles, Morocco, Botswana and Mauritius) were on track to achieve full COVID-19 vaccination coverage rates ranging from 72% to 97%; while countries such as Burundi, Chad, the Democratic Republic of the Congo, Niger, Nigeria, Madagascar, South Sudan, Mali and Tanzania had low vaccination coverage, with only 0.1%–4.9% of their populations fully vaccinated after 9 months of vaccine deployment.⁴¹ The low vaccination uptake can be attributed to women’s hesitancy,^5,42,43 due to several reasons including perception, fear of side effects among others.^12,19,21

We identified age, educational attainment, access to health insurance scheme, employment, exposure to media, household wealth, community-level willingness to COVID-19 vaccine uptake, community-level nativity and geographical region as significantly associated with higher odds of being in a higher COVID-19 vaccination category. Our result shows a progressive increase in the odds of being in a higher COVID-19 vaccination category as women get older. When compared with women aged 15–19 years, the older age groups had increase in the odds of being in a higher COVID-19 vaccination category. This implies an age-dependent relationship with increasing vaccination category. This could be due to the experience during the pandemic where older people were more susceptible to COVID-19 symptoms compared with younger people. The fact that ageing is associated with weakening body system and suboptimal organ functions could make older people more prone to infectious diseases. Hence, they will have more tendency to be in higher vaccination category. This finding corroborated with previous reports that older people had higher rate to take COVID-19 vaccine compared with people of younger age.^16,18,19

In addition, socioeconomic factors such as education, employment, health insurance coverage, exposure to media and household wealth were found to be associated with higher COVID-19 vaccination uptake. Educated women, those employed, covered by health insurance, exposed to media, and from higher household wealth quintile had greater odds of being in a higher COVID-19 vaccination category. This finding is not unexpected as education is a known significant social determinant of health literacy and positive health-seeking behaviour. Educated women will have better knowledge of COVID-19 vaccine benefits. This is consistent with previous reports that found education as a predictor for vaccine uptake.^18,44,45 Women with poor literacy or health knowledge are less able to comprehend health promotion messages unless communicated in a manner that they can relate to, such as using local language during campaigns. Notably, education would enhance understanding of the benefits or risks inherent in the uptake of new health interventions such as COVID-19 vaccines. Furthermore, exposure to media was found to have greater odds of being in a higher COVID-19 vaccination category. Exposure to media will enhance the utilisation of health information and improve maternal health services uptake.^46–48

Some caution is required in interpreting the media-exposure finding. Our measure captures whether a woman regularly accessed mass media (newspapers, radio, television or the internet); it does not capture the content, source or accuracy of the messages to which she was exposed. Mass media can disseminate accurate, provaccination information but can equally amplify rumour, conspiracy theories and vaccine misinformation, which depress vaccination intent and confidence.^49,50 The variable therefore indexes access to information channels rather than exposure to any particular message, and the positive association observed at the individual level (adjusted OR 1.29, 95% CI 1.17 to 1.42) most plausibly reflects the greater overall reach of public-health communication during the campaign rather than the effect of media content per se; the attenuation of the community-level media association in the fully adjusted model is consistent with this. Future studies should distinguish the type, source and valence of media exposure, and ideally measure exposure to misinformation directly, to clarify the pathways linking media to vaccination behaviour. The findings from household wealth and employment suggest that socioeconomic well-being plays a significant positive role in being in a higher COVID-19 vaccination category. Though COVID-19 vaccine was commonly administered free of charge, socioeconomic empowerment could improve uptake through capacity to afford transport fare to the vaccination centres, decision-making power or as a requirement by many organisations for staff to be vaccinated. Women who are employed are commonly enrolled in health insurance schemes and usually have better health care-seeking behaviour, which could have contributed to having these women in a higher COVID-19 vaccination category. The impediment of catastrophic out-of-pocket expenditure, which negatively affects health services uptake, could have been mitigated.^51,52 Women who are covered by health insurance are also likely to have more trust in the healthcare system and better access. This is consistent with reports from previous studies.^{5,17,43,44,53}

This study shows that women from households with more than six members were less likely to belong to a higher COVID-19 vaccination category compared with women from smaller households with one to two members. This finding reflects the complex socioeconomic and structural disadvantages commonly experienced by large households in low- and middle-income settings.⁵⁴ Women in large households may experience greater caregiving responsibilities and domestic workloads, limiting the time available to seek vaccination services. The observed association may also reflect underlying socioeconomic deprivation, as larger household size is frequently correlated with poverty, lower educational attainment, overcrowding and limited access to health information. These conditions can reduce exposure to accurate public health messaging and increase susceptibility to misinformation and vaccine hesitancy. It further highlights the importance of considering household-level contextual factors when designing public health vaccination strategies. Interventions aimed at improving COVID-19 vaccine uptake among women may therefore benefit from targeted community outreach, flexible vaccination schedules, mobile vaccination services and household-centred health education strategies that specifically address barriers faced by women in large households.

Women who are native to their community of residence and those from communities with high-level willingness for COVID-19 vaccine uptake had greater odds of being in a higher COVID-19 vaccination category. This shows that community attitude towards vaccination is critical in vaccination uptake. Poor local public health campaigns or outreaches, inadequate or unavailability of vaccination centres especially in rural communities could negatively affect COVID-19 vaccination uptake. Our study found that the odds of being in a higher COVID-19 vaccination category were greater among women who reside in communities with willingness to be vaccinated and had lived at least 5 years in the community. A previous study reported that community-level willingness for vaccination promotes access, provision of infrastructure, peer influence, positive social norms and group mobilisation for vaccination.¹²

Geographical region was significantly associated with higher COVID-19 vaccine uptake. Women from North Central, North East, South West, South South and South East had higher odds of being in a higher COVID-19 vaccination category when compared with their counterparts from North West. Evidence-based studies have reported differences in healthcare services uptake between the Southern and Northern parts of Nigeria.^55–57 The low vaccine uptake observed in the North West region may be attributed to historical trust issues, sociocultural and religious influences and physical barriers and rurality. Past negative experiences such as the widely-remembered Kano/Pfizer episode with adverse reactions may have created a collective memory of harm or exploitation thereby lowering trust in new vaccines or government health campaigns.²⁰ The northern women who are believed to be less educated with less access to modern education, when compared with their southern counterparts, may have fallen for a variety of false information, such as the idea that the virus is a biological weapon from the West or China that wants to lower the global population. Conspiracy theories and misinformation have been known to impede international efforts for widespread COVID-19 vaccination. The rapid spread of false claims about vaccine safety, infertility, microchips or government plots increases perceived side effect risk and may have reduced women’s willingness to vaccination uptake. Worst still, the patriarchal dominance in northern Nigeria, whereby women majorly require permission from husbands or elders to visit healthcare centres and limited female autonomy, could be attributed to the women’s less likelihood of being in a higher COVID-19 vaccination category. A previous study on vaccine hesitancy in Nigeria revealed higher hesitancy in largely Muslim northern regions,⁵⁸ contrary to the findings of this study where the Muslim women had greater odds of being in a higher COVID-19 vaccination category when compared with the Christian women. This suggests the influence of religious beliefs and cultural practices on acceptance of COVID-19 vaccination uptake.

Strengths and limitations

The comprehensive, nationally representative dataset used in this study was gathered by NDHS 2024. The multilevel analytical technique employed in this study accounts for the hierarchical nature of the data and enables the analysis of factors at the individual, household and community levels. However, only an association can be established due to the cross-sectional nature of the data, which prevents the identification of a causal effect. The COVID-19 vaccination report may contain recall bias based on women’s self-reported responses, particularly on the type of vaccination received. Other limitations that are common to studies using self-reported behavioural data apply to this. For instance, social desirability bias may have led some women to misreport how many dosages they received. Residual confounding from unmeasured factors (eg, healthcare access and vaccine hesitancy) and community-level variables being constructed from respondents may have inflated effect estimates. The very large effect size observed for community-level willingness warrants cautious interpretation, as it may reflect contextual clustering or shared measurement rather than an independent effect.

Several further limitations should be noted. The community-level exposures were derived by aggregating individual responses within clusters and then dichotomising the resulting proportions; this approach cannot fully separate contextual from compositional effects, discards information and may introduce some misclassification, although a continuous-specification sensitivity analysis confirmed that the principal findings were robust (online supplemental table S2). The cumulative model assumes proportional odds; although the Brant test was significant in this large sample, category-specific estimates were directionally consistent (online supplemental table S1), so the ORs are interpretable as average effects across the ordered categories. Finally, vaccination status and dose number were self-reported and could not be verified against records; the ordinal outcome (0, 1, 2 and >2 doses) was used to distinguish, as far as the data allow, partial from more complete vaccination.

Conclusions

Approximately one-quarter of Nigerian women of reproductive age reported at least one dose of COVID-19 vaccination uptake. This estimate of COVID-19 vaccination indicates the need for a thorough vaccination information campaign with clear and concise COVID-19 vaccine messaging while taking the benefactors’ language and context into account. In addition, health promotion initiatives should focus on increasing the uptake of the COVID-19 vaccine and greater attention should be directed towards individuals with low socioeconomic status. Opportunities to use social media platforms to disseminate health education and health promotion messaging about the advantages of COVID-19 vaccinations could raise awareness of the virus, spread knowledge about it and enhance vaccination rates. As a result, Nigerian women from low socioeconomic backgrounds, those without health insurance, populations with low COVID-19 vaccine uptake willingness and those from the North western geopolitical zone should be targeted for increased acceptance of the vaccine.

Data availability statement

Data are available in a public, open access repository.