Introduction

While there has been great success in increasing the coverage of new childhood vaccines globally, expanding routine immunization services to reliably reach all children and communities has proven more challenging [1–4]. Certain populations remain un- or under-vaccinated in many LMICs [2, 5, 6] and infectious diseases still account for a large proportion of deaths among children [2, 7, 8]. Lack of accessibility to health care, poverty and overcrowding, as well as civil and political unrest, are all considered major contributors to the transmission of vaccine preventable diseases, and present significant challenges in reaching global coverage targets [9]. In 2019, 13.6 million children failed to receive even a first dose of the diphtheria-tetanus-pertussis vaccine (DTP1), rising significantly to 17.1 million in 2020, due to disruptions to health services amidst the ongoing coronavirus pandemic [10]. Moreover, the percentage of children receiving three doses of the diphtheria-tetanus-pertussis vaccine (DTP3) declined by 5 percent from 2019 to 2021, reflecting a sustained decline in DTP3 coverage [11].

Considering the substantial burden of vaccine-preventable diseases and the frequent occurrence of outbreaks among disadvantaged populations [2, 12] there have been continuous efforts in the health policy environment to design vaccination strategies and interventions that prioritize and target these population groups [3, 13, 14]. In 2017, senior leaders in the immunization sector formed the Equity Reference Group for Immunization (ERG), identifying remote-rural, conflict-affected and the urban poor as likely groups to be un- or under-vaccinated; also recognizing gender-related barriers as a cross-cutting issue [15, 16]. Gavi, the Vaccine Alliance, has adopted a new five-year strategy (“Gavi 5.0”) for 2021–2025, with a vision of "Leaving no one behind with immunization" by increasing equitable and sustainable use of vaccines [17]. Further, the World Health Organization’s Immunization Agenda 2030 sets a global vision and strategy for vaccines and immunization, committing to an ambitious target of reducing the number of zero-dose children by 50% by 2030 [18].

Despite these developments, estimates of how many children living in remote-rural, urban slum or conflict-affected settings, their vaccination rates and relative distributions within countries remain scarce, limiting robust resource allocation and strategy design [19]. Previous studies have revealed significant geographic inequalities in routine vaccination coverage across many LMICs, through analysis at the district level, helping to identify areas of low coverage and vaccine delivery system vulnerabilities [5, 8, 20–22]. However, to date, there has been no global scale analysis at these finer spatial resolutions focusing on the numbers of unvaccinated children in these key geographically marginalised populations [6, 9, 18]. Achieving equitable coverage, which has been identified as a core priority moving forward [17, 18, 23] needs to be guided by the most current and detailed evidence regarding the size, distribution, and characteristics of these key at risk groups [20, 21, 24–26], if current global immunization targets are to be met [17, 18].

Here, we present outputs from the integration of a range of geospatial datasets to estimate the number and distribution of zero-dose and under immunised children, across remote-rural, conflict-affected, and urban settings. We estimate the relative proportions and numbers of these key populations, to assess variabilities both between and within countries, to help guide global, national, and regional strategies and interventions, and to highlight those areas where children may remain unvaccinated. Given available data and to reflect key immunization program performance indicators, our study focuses on receipt of DTP1, DTP3, and the first dose of the measles containing vaccine (MCV1) among children under the age of one in 2019.

Results

We found substantial heterogeneities in the spatial distribution and setting (or context) of unvaccinated U1, both among and within countries, with a large proportion of them found to be concentrated within a small number of countries. Whilst many of those unvaccinated were found to be in remote-rural (11.8%), conflict-affected (5.9% to 14.7%), and urban locations (20.3%), a large proportion were found in settings other than these (59.7%), with nearly 40% found to be in rural (non-remote) locations within 1-hour of the nearest town or city (not including urban and peri-urban areas) (DTP1).

Geographical distributions of unvaccinated U1

We first examined the geographical distributions of unvaccinated U1. Fig 1 shows the estimated proportion of U1 who did not receive the first dose of the DTP vaccine in 2019 at ADM2 and highlights the substantial differences in vaccination coverage both among and within countries. S1 and S2 Figs, show the same for DTP3 and MCV1 and display broadly similar spatial patterns. The largest numbers of unvaccinated U1 for all three doses were in India and Nigeria, whereas the largest proportions were in Somalia, Papua New Guinea, Guinea, Chad, and Angola (S3 Fig).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1.

Proportion of children under 1 year of age estimated to have not received the first dose of the DTP vaccine in 2019 at GADM administrative level 2 [31, 37] for Latin America (A), Africa (B), and Asia (C).

https://doi.org/10.1371/journal.pgph.0001126.g001

Breakdown of unvaccinated U1 by setting

Next, we examined the numbers and proportions of unvaccinated U1 residing in remote-rural, conflict-affected (broad and narrow), urban and peri-urban, and rural (non-remote) locations at the global, national, and subnational levels to further characterise those unvaccinated.

Table 1 shows the distribution of unvaccinated U1 broken down by geographical setting, vaccine dose, and conflict for all countries in the analysis. We found that more than 19% lived in urban areas, with a smaller proportion in peri-urban areas, whilst more than 11% lived in remote-rural areas more than 3 hours travel-time from the nearest town or city. Around 60% of those unvaccinated lived in other rural areas, not classed as remote, urban, or peri-urban (rural non-remote). Table 1 also shows the percentage of unvaccinated U1 considered to be conflict-affected, which was estimated to be between 6% and 18%, depending on the definition of conflict used (narrow and broad).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Global distribution of unvaccinated children under 1 year of age (percentage and number) in 2019 broken down by geographical setting, vaccine dose, and conflict for all countries in the analysis (Note: Conflict can overlap with other settings in the table).

https://doi.org/10.1371/journal.pgph.0001126.t001

Unvaccinated U1 in remote-rural locations

Fig 2 shows where the unvaccinated U1 in remote-rural areas were concentrated, by mapping the proportions of U1 unvaccinated for DTP1 (of the total unvaccinated) at ADM2. The map highlights countries where significant proportions of U1 were estimated to be unvaccinated, including the Central African Republic, Republic of Congo, and Mauritania, as identified above. However, it additionally reveals regions not previously identified at the national level, for example large areas of northern Chad, central DRC, and South-eastern Angola/Eastern Botswana. The distributions of unvaccinated U1 in remote-rural areas were similar for all three vaccine indicators, though generally appeared slightly lower for MCV1 (S4 and S5 Figs).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2.

Proportion of children under 1 year of age in remote-rural areas estimated to have not received the first dose of the DTP vaccine in 2019 at GADM administrative level 2 [31, 37] for Latin America (A), Africa (B), and Asia (C).

https://doi.org/10.1371/journal.pgph.0001126.g002

S6–S8 Figs, show the top 20 ranked countries, by the estimated proportion (of the total unvaccinated) and corresponding total number of U1 not receiving DTP1/DTP3/MCV1 in remote-rural areas. In terms of proportion, Madagascar, Mauritania, Central African Republic, Papua New Guinea, and Republic of Congo were in the top 5 ranked countries for both DTP1 and DTP3, where around 60% to 70% of unvaccinated U1 were estimated to be in remote-rural locations. These figures are similar for MCV1, where around 60 to 66% of U1 unvaccinated for MCV1 were found to be in remote-rural locations, though Eritrea additionally ranks in the top five countries, rather than the Republic of Congo. However, when considering absolute numbers DR Congo was estimated to have the greatest number of unvaccinated U1 in remote-rural locations for all vaccine doses (on the order of 202,025 children, for DTP1), though Chad and Madagascar remain high in the rankings.

Unvaccinated U1 in conflict-affected locations

Fig 3 shows where the unvaccinated U1 (DTP1) in conflict-affected locations were concentrated according to the broad definition of conflict (at least 30 fatalities per million people) at ADM2 (S24 and S25 Figs). The map highlights conflict-affected regions where large proportions of U1 were unvaccinated, including much of Somalia (52%), Chad (46%), northeast Nigeria (34%), large parts of Central African Republic (26%), and Afghanistan (17%).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 3.

Proportion of children under 1 years of age in conflict-affected areas (broad definition) estimated to have not received the first dose of the DTP vaccine in 2019 at GADM administrative level 2 [31, 37] for Africa and Asia, with close ups shown for northern sub-Saharan Africa (A), and the Middle East/West Asia region (B).

https://doi.org/10.1371/journal.pgph.0001126.g003

S26–S28 Figs, show the top 20 countries, by estimated proportion and numbers of U1 not receiving DTP1/DTP3/MCV1 in conflict-affected areas according to the broad definition. In terms of the proportions of U1 unvaccinated, Syria, South Sudan and Afghanistan had the highest prevalence of unvaccinated U1, with over 80% of U1 not receiving DTP1 found to be in conflict-affected areas. However, when considering absolute numbers, Nigeria had the largest numbers of unvaccinated U1 in conflict -affected areas, though Somalia, South Sudan, and DRC also ranked high. Note that conflict-affected unvaccinated U1 are disbursed across all other geographic settings (urban, peri-urban, remote-rural, and rural non-remote) and included all those unvaccinated within a conflict affected ADM2 unit.

Unvaccinated U1 across all settings

Fig 4 shows the proportion of U1 not receiving DTP1, broken down by country and urban/rural setting. It highlights the variations that exist between countries, with some countries (e.g. Egypt, South Africa) found to have the largest proportion of U1 not receiving DTP1 in urban areas, and other countries (e.g. Mauritania, Madagascar) found to have substantial numbers in remote-rural areas. Similar proportions for urban/rural settings were observed for the different vaccines (S29 and S30 Figs).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 4. Plot showing the estimated breakdown of children under 1 year of age not receiving DTP1 in 2019 by urban/peri-urban/rural (non-remote)/rural (remote) characteristics, for all countries in the study area in Latin America (AM), Africa (AF), and Asia (AS), and also regional averages.

https://doi.org/10.1371/journal.pgph.0001126.g004

Fig 5 maps the setting where the greatest number of unvaccinated U1 (DTP1) were found, at ADM2, for urban, peri-urban, rural (non-remote), rural (remote), and conflict areas. Again, large heterogeneities were observed both among and within countries, with different characteristics of those unvaccinated observed at both national and sub-national levels. For example, in India, the majority of unvaccinated U1 were found to be in rural (non-remote) areas. However, within Africa, most of those unvaccinated were found to be in conflict-affected, rural (remote) locations, and rural (non-remote) locations. In Latin America, we observed significant proportions of unvaccinated U1 in rural (non-remote) and rural (remote) areas, however in Brazil, higher numbers were estimated to be found in urban and peri-urban areas (S31 and S32 Figs).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 5.

Map highlighting the geographical setting with the estimated greatest number of children under 1 years of age in 2019 not receiving DTP1 at GADM administrative level 2 [31, 37] for Latin America (A), Africa (B), and Asia (C).

https://doi.org/10.1371/journal.pgph.0001126.g005

Discussion

If global targets to reduce the number of zero-dose children are to be met [18], then immunisation programmes need to identify and reach those children [6, 17, 18, 23, 24, 38, 39]. This requires the use of data driven actionable evidence, regarding the location and characteristics of those unvaccinated, and the design of tailored vaccination strategies appropriate for a given context [9, 18, 20, 21, 24, 25]. This study integrates a range of geospatial datasets [5, 27, 29, 30] to estimate the numbers of unvaccinated children living within key at-risk settings, including urban, remote-rural, and conflict affected locations, at the sub-national level. It is complementary to country-specific work, which identifies community and other household and individual level factors that characterise zero-dose populations and quantifies their relationship to vaccination coverage [26].

We find that coverage varies widely among and within countries, with substantial heterogeneities existing in the context and spatial distributions of those unvaccinated. Considering the settings identified by the ERG, we estimate that over 11% of zero-dose children are living in remote-rural areas, more than 26% in urban and peri-urban areas, and between 6 and 18% in conflict-affected locations. Altogether, this means that nearly 40% of zero-dose children are living in remote-rural and urban settings, and nearly 20% in conflict settings. Yet, whilst many of those unvaccinated are found to be living in these key at-risk settings, a significant proportion are located within areas other than these, in areas with lower population and/or built-up density than an urban or peri-urban setting. These findings highlight the variability of immunization coverage, both geographically and in terms of the characteristics of those populations, demonstrating the need for both country and regional specific strategies for immunization, that consider country size, available resources, and the context in which people live [17, 18], as well as further identification and characterisation of those most at risk of missing out on immunisation [17, 18].

There is a need for broader ranging yet tailored interventions, particularly while there are ongoing immunisation system vulnerabilities due to the COVID-19 pandemic [24]. One of the key principles guiding Universal Health Coverage is community designed and driven implementation of services, and there has been previous success in strategies that prioritise community engagement to identify and reach vulnerable populations [9, 40], as well improved public-private partnerships and microplanning to help identify and reach vulnerable populations [9, 40]. Additionally, the ERG recommends a range of targeted data-driven interventions that can help to address vulnerable populations in difficult to identify, hard to reach, and dangerous and fragile settings [40]. Which emphasise the need for timely and actionable data on vulnerable populations, through the use and implementation of electronic immunisation registries, dashboards, and visualisations, that allow current data to be collected and aggregated to the appropriate level for informative decision making [41–43].

Whilst we aim to provide the most detailed and up to date analysis of un- and under- vaccinated children characterised as most at risk, the analysis is limited by the choice and definition of those settings, data quality limitations, and both the availability of and ability to link together data from a variety of disparate sources. Regarding the choice of the settings, we chose to focus on key at risk populations, as identified by the ERG [40] to begin to characterise those most at risk from un- and under-vaccination. However, as highlighted by the results, nearly 60% are found to be living outside of these settings. There is, therefore, scope to expand on and further characterise the key geographical settings in which the most disadvantaged populations are located, as well as the consideration of non-geographic dimensions, such as gender related barriers. Further work could look to examine the numbers of zero-dose children living outside of urban, remote-rural, and conflict-affected settings, to explore other potentially relevant variables that may help to explain barriers to immunisation [18, 44]. For example, when measuring remoteness, using travel-time to the nearest health facility [45], as opposed to the nearest town or city, which may be a more relevant measure in assessing the effect of geographical accessibility on vaccination coverage [46, 47].

Additionally, defining what is urban or rural can be challenging, with governments typically establishing their own definitions, making it difficult to meaningfully compare indicators across national borders [48]. For this study, we use the most current and recent globally available standard definitions of urban and rural [30]. However, the datasets feeding into these decisions are typically continuous in nature and so the choice of threshold or data to use naturally affects the estimates produced [49], and such simple classifications may unduly lead to the concealment of smaller areas of geographical inequality [50]. There is additionally the challenge of defining informal urban settlements, without access to globally available comparable data. An attempt was made to use accessibility to utility services [51] as a proxy for informal settlements, though comparison to alternative slum data proved validation difficult.

Furthermore, there is scope for exploring the sensitivity of these definitions, such as evaluating the adjustment of threshold values and the comparison of alternative data sets, to examine how much of an effect each has on the estimated numbers of unvaccinated children found to be living within these settings [52]. More recently, there have been collaborative efforts to address these issues, such as the development of a new global definition of urban and rural areas, using the ‘Degree of Urbanisation,’ to facilitate international comparisons. This method considers the urban-rural continuum, using three classes of ‘cities,’ ‘towns & semi-dense areas,’ and ‘rural areas,’ instead of two, as well as population size and density, and considers the functional urban area around each city [53]. Such data may provide more insight into the numbers of zero-dose children found to be living outside urban areas, though within 1 hour to the nearest town or city.

The analysis links together data from a variety of disparate sources, each with their own limitations of use that need to be considered. Whilst we aim to assemble to most recent and spatially detailed data available, certain compromises need to be made to facilitate comparability between the data sets and global coverage. For example, whilst more recent and higher resolution population estimates are available elsewhere [27], the vaccination coverage estimates are available at 5km, and it was deemed sufficient to conduct the analysis using the 1km population estimates within the same period of the vaccination coverage estimates. Additionally, whilst the accuracy and quality of the vaccination coverage estimates have been assessed elsewhere [5], uncertainties remain as to whether survey samples have been conducted in areas of conflict, for instance. Additionally, while we consider remoteness as travel-time to the nearest town/or city [28], one could also consider travel-time to the nearest health facility or hospital, or distance to major roads, which may be of interest in further analysis. Another key factor influencing the estimated outputs is the choice of administrative boundaries used to summarise the numbers of unvaccinated children, as naturally, the use of different or updated boundaries inherently affects how these numbers are aggregated, as demonstrated by the Modifiable Areal Unit Problem (MAUP), which describes how spatial summary measures are inherently influenced by the boundaries that they are reported at [54]. Finally, whilst we attempt to capture conflict using both broad and narrow definitions, the data available [29] largely relies on media coverage, and whilst we aim to consider two years prior to the vaccination coverage estimates, areas of conflict or insecurity can change significantly over time.

Despite the inherent uncertainties observed in the data quality, definitions and thresholds applied, we present here the first available sub-national estimates, across 99 low- and middle-income countries, on the distributions and numbers of children not receiving vaccine doses in key at-risk settings. These baseline estimates provide significant insights into the inequalities in vaccination coverage, that can assist policy makers in the development of effective immunisation strategies that identify and reach all zero-dose children and missed communities [17]. We demonstrate additionally the need for further inquiry and characterisation of those un- and under immunised, the thresholds used to define these, and for more country-specific and targeted approaches to immunisation.

Supporting information

Acknowledgments

We would like to acknowledge the support, data provision and insights provided by John Mosser at the Institute for Health Metrics and Evaluation (IHME), University of Washington, and Francesco Checchi at the London School of Hygiene & Tropical Medicine (LSHTM).