Selection criteria

The inclusion criteria were kept deliberately broad in order to identify all relevant reviews. The inclusion criteria were conceptualised using the PECOS (Population, Exposure, Comparison, Outcome, Study Design) framework, as follows:

Population: General populations, including demographic sub-populations. All countries.

Exposure: Advantaged SES, according to the following indicators: education, income, occupation/employment, and/or measures of area-level deprivation/poverty (e.g., the English Indices of Multiple Deprivation (IMD)).

Comparison: Disadvantaged SES, according to the following indicators: education, income, occupation/employment, and/or measures of area-level deprivation/poverty.

Outcome: Variation in the rate (uptake), or proportion of a target population (coverage), that have been vaccinated. Eligible vaccines were those labelled by the World Health Organization (WHO) as universally recommended routine vaccinations [26]: BCG (Tuberculosis), Hepatitis B (Hep B), Polio (IPV/OPV), DTP-containing (Diphtheria, Tetanus and Pertussis) vaccine, Haemophilus influenzae type b (Hib), Pneumococcal (conjugate) (PCV), Rotavirus, Measles, Rubella, and HPV. Studies focusing on influenza and Coronavirus were also eligible for inclusion to account for reviews published in response to the Coronavirus pandemic. Eligible measures of uptake or coverage were: initiation and/or completion of multi-dose vaccines or vaccination schedules as a whole (where uptake or coverage is measured by the initiation/completion of several different vaccines, some with and without multiple doses).

Study Design: Only systematic reviews that synthesised quantitative or qualitative studies were included. The quantitative reviews did not have to include a meta-analysis. A systematic review was classified as such if it met four of the following criteria, as outlined by the Database of Abstracts of Reviews of Effects (DARE) [27]:

1. Were inclusion/exclusion criteria reported?
2. Was the search adequate?
3. Were the included studies synthesised?
4. Was the quality of the included studies assessed?
5. Are sufficient details about the individual included studies presented?

There were no language restrictions, any potentially relevant abstracts and titles were translated using online tools. A publication date of 2011 –present day was applied. The WHO’s 2011 “Global Vaccine Action Plan 2011–2020” report outlined the updated guidance on improving vaccination uptake [28]. The publication of this report may capture vaccination policy changes relevant to the present day, and as the WHO is a global institution, it is relevant to all countries. Further details on the inclusion and exclusion criteria are outlined in the additional material, S2 Appendix.

Search strategy and information sources

The search strategy was developed and piloted in Medline (via Ovid) to test its sensitivity in returning eight previously identified indicator systematic reviews [11–18]. After initial piloting, the following eight databases were searched during the period 03/09/2022-04/09/2022: Medline (via Ovid), Embase (via Ovid), Cumulated Index to Nursing and Allied Health Literature (CINAHL) (via EBSCO), Cochrane CENTRAL, Science Citation Index (SCI) (via Web of Science), Database of Abstract Reviews of Effects (DARE), Scopus (via Elsevier), and Applied Social Sciences Index and Abstracts (ASSIA) (via ProQuest). Grey literature searching was conducted using the WHO repositories and PROSPERO. The full search strategy can be seen in the additional material, S3 Appendix.

Screening, and selection

After the searches, the records were downloaded into Rayyan [29] and deduplicated. Title and abstract screening were conducted by reviewer one (AS), and a 10% sample was checked by reviewer two (KT), in reference to the a priori established eligibility criteria. The full-text screening stage was also performed by reviewer one (AS) and a 10% sample was checked by reviewer two (KT). The process of the double-checking process is to ensure a consistency during screening, thus ensuring all relevant reviews are included and all ineligible reviews are excluded. If an agreement could not be achieved, a third reviewer was consulted (AT) to establish consensus. Both forwards and backwards citation searching was performed using Web of Science to supplement the search.

Data extraction

To aid the data extraction process, an extraction template was designed, and the following information was extracted from each of the systematic reviews: bibliographical details (author, year of publication, title, DOI, abstract), study design (satisfaction of the DARE criteria, method of synthesis, number of included studies); search specificities (databases, date, restrictions); any information relating to PICOS (geographical location, population, vaccine/s, definition of uptake, measures of socioeconomic inequality); the main findings/conclusions relevant to the umbrella reviews’ research questions (uptake percentages (rates), counts, odds ratios etc.); any potential mechanisms or pathways that may help account for the socioeconomic differences in vaccine uptake, as identified in the systematic review. Additionally, the following information was extracted from the primary studies, if provided by the systematic review in which they were included: authors, year of publication, vaccine/s of focus, geographical location, population, measures of socioeconomic inequality, risk of bias/quality verdict, overall uptake of specified vaccine, and the main findings. The data extraction was performed by reviewer one (AS) and checked in full by reviewer two (KT) for accuracy; any disagreements were discussed with reviewer three (AT) who helped establish a consensus.

Dealing with overlap

In umbrella reviews, there is a risk of primary study overlap [23]. To address this, a citation matrix was created, and the corrected coverage area (CCA) was subsequently calculated and reported [30]. A CCA value equal to, or greater than, 15% indicates high overlap of primary studies [30].

Quality appraisal

A MeaSurement Tool to Assess systematic Reviews (AMSTAR-2) was used to assess the quality of the included systematic reviews. This tool was selected because it can be used to assess both randomized control trials and non-randomised studies of interventions [31]. The checklist is comprised of 16 questions, of which seven are considered “critical domains” (see Shea et al. for more information [31]). The responses to each of the questions (see additional material, S4 Appendix) were recorded in a spreadsheet and then inputted into the online AMSTAR-2 tool that produces either a “high”, “moderate”, “low”, or “critically low” verdict [32]. Failure to meet one, or more, of these “critical domains” automatically reduces the overall score to “low” or “critically low”. AMSTAR-2 is not intended to produce an overall score but to identify areas of potential methodological weakness.

Narrative synthesis

This process of narrative synthesis was guided by the Synthesis Without Meta-analysis recommendations [33]. Firstly, a table including the author, year of publication, vaccine/s, definition of uptake, geographical location, population, number of primary studies, number of relevant primary studies, and the measures of socioeconomic inequalities, was created. The reviews were then summarized using descriptive statistics, according to vaccine (childhood/adolescent, HPV, influenza/pneumococcal, or all routine vaccinations), geographical location (high-income countries, or low/middle-income countries, according to the World Bank classification [34]), and by measures of socioeconomic status (education, income, employment, area-level measures, or unspecified).

The key findings of each review were then summarised in a matrix-style table, disaggregated by country economic status. The results were narratively synthesised also according to country economic status to account for differences in healthcare spending. The synthesis approach used a range of terms to explain the findings from the included reviews; definitions of these terms can be seen in Table 1.

Download:

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

Table 1. Classification of findings and their definitions.

https://doi.org/10.1371/journal.pone.0294688.t001

The extracted mechanisms were collated in a table to further support the synthesis. These mechanisms were then mapped onto an adapted version of Levesque et al’s. patient-centred access to healthcare framework [35] (see the additional material S5 Appendix). The framework was adapted for this review to depict the process of vaccination, for more explanation, please refer to the protocol [25]. Mapping the mechanisms onto the framework will provide an understanding of the trickle-down effects of socioeconomic inequalities on vaccine uptake and ascertain which key stages of the vaccination process are impacted. To make this process easier, a table version has been created (Table 2).

Download:

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

Table 2. Patient-centred access to vaccination framework, table version.

https://doi.org/10.1371/journal.pone.0294688.t002

Searching

In total, 14,065 results were retrieved across the databases, and after deduplication, 9,163 results remained. After title and abstract screening 119 full-text reports were assessed for eligibility leaving a total of 26 systematic reviews meeting the inclusion criteria (Fig 1). Exclusion reasons for each of the 119 articles read at full-text stage are shown in the additional material, S6 Appendix. Forwards and backwards citation searching identified an additional 3,282 results, although after title and abstract screening (n = 3,282) and full text checking (n = 8) no further systematic reviews were eligible for inclusion. During this process we did not find any systematic reviews not published in English that were eligible for inclusion.

Download:

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

Fig 1. PRISMA-flow diagram.

A completed PRISMA-flow diagram depicting the searching and screening process.

https://doi.org/10.1371/journal.pone.0294688.g001

A total of 691 primary studies were included across all included reviews. Of the 691, 93 were included in more than one review, meaning there were 598 unique primary studies. The CCA was 13.46%, showing a moderate level of overlap of the primary studies.

Quality appraisal

Of the 26 reviews, all were deemed of “critically low” quality. The verdict for each AMSTAR-2 question can be seen in the additional material, S4 Appendix. Some key areas in which the included reviews frequently scored poorly were as follows:

Question 2—not containing an explicit statement that the review methods were established prior to conducting the review (such as a protocol, or a study registration database) (n = 22).

Question 7—not including a full list of excluded primary studies and their reasoning (n = 26).

Question 10—not providing the funding details of each primary study (n = 24).

Question 13—not accounting for risk of bias in the interpretation/discussion of results (n = 21).

Please refer to Shea et al’s. AMSTAR-2 guidance [31] regarding the implications of scoring low on these four items. Indeed, failing to provide the funding details of included primary studies is a crucial oversight in the context of vaccination. The far-reaching repercussions of this failure were starkly evident in 1998 with the MMR vaccine [36], the effects of which still persist today.

Summary of included studies

Of all the included reviews, a meta-analysis was conducted on 7 occasions [13, 17, 37–41], and a narrative approach was taken in the remaining 18 reviews [11, 12, 15, 18, 42–54]. One of the reviews that conducted a meta-analysis as its main method of synthesis, but the findings in relation to socioeconomic inequalities were only synthesised narratively [38]. The remaining study, by Ali et al., performed both a meta-analysis and narrative synthesis [55].

The included systematic reviews covered a range of contexts. Several countries and geographical groupings were included: n = 14 high-income countries [11–13, 15, 18, 38, 42, 44, 45, 47–49, 52, 56], n = 7 low/middle-income countries [37, 39, 41, 51, 54, 55], and n = 5 a mix of high/low/middle-income countries [17, 40, 46, 50, 53].

The included vaccines were as follows: childhood/adolescent routine schedule, in whole or in part (n = 10), HPV (n = 10), influenza and/or pneumococcal (n = 5), and all routine vaccinations (childhood and adult) (n = 1). Moreover, uptake was referred to, and measured in, a variety of different ways. Eight reviews did not define how they measured uptake [11, 38, 40, 41, 43, 47, 50, 55], although five of these explored influenza vaccination which has no set schedule [38, 40, 41, 47, 50]. One review reported vaccine initiation [52], and two reported vaccine schedule completion [51, 54]. The remaining 15 reviews measured both vaccine initiation and completion [12, 13, 15, 17, 18, 37, 39, 42, 44–46, 48, 49, 53, 56].

The populations across the reviews varied, due to the differing vaccines and their respective target groups. Amongst the publications that focused on childhood vaccines, n = 3 were of children under 2 years of age [37, 39, 53], n = 3 under 5 years [11, 43, 51], n = 1 under 7 years [56], n = 2 under 12 years [17, 18], and n = 1 unspecified [54]. For the reviews that examined HPV vaccination, n = 5 focused on females [12, 13, 44, 45, 49], n = 1 on males [52], and n = 4 on both females and males [15, 42, 46, 48]. Moreover, for the reviews that examined influenza vaccination, n = 2 explored adults aged 65 years and under [40, 50], while n = 4 reviews had no population restrictions [38, 41, 47, 55].

The findings relating to SES were reported in one of two ways; either according to each measure, or under a subheading of “socioeconomic status” incorporating all measures. Five reviews used one measure of SES (n = 1 area-level deprivation [42], n = 4 education [17, 39, 41, 46]); while 21 reviews utilised two, three, or four measures (SES, education, income/wealth, various area-level, occupation/employment). These measures were operationalised in vastly different ways, although it was common for reviews focusing on children or adolescents to refer only to maternal education, rather than a combined parental measure. A ‘Characteristics of Included Studies’ table can be seen in the additional material, S7 Appendix, which details the specific vaccines, geographical areas, populations, and measures of SES.

Socioeconomic inequalities and vaccine uptake

The association between socioeconomic status and vaccination uptake were summarised in a matrix-style table, (Table 3). All 26 reviews reported an association between SES and vaccination uptake. Evidence for inequalities (advantaged SES, higher vaccination uptake, OR disadvantaged SES, lower vaccination uptake) were found on all but two occasions where only inverse associations were identified (n = 24). However, in over half of these reviews (n = 15) the overall conclusions were that of mixed findings as support for inverse associations were also identified. In total, 17 reviews found evidence for inverse associations: either lower vaccination uptake for advantaged SES (n = 6) higher uptake for disadvantaged SES groups (n = 7), or both (n = 4).

Download:

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

Table 3. Summary of findings.

https://doi.org/10.1371/journal.pone.0294688.t003

Overall, the differing measures of SES employed by the reviews did not seem to explain their varying conclusions. Income, education, occupation/employment, and area-level deprivation were neither more, nor less, frequently cited as being associated with vaccination uptake. Additionally, mixed findings were equally prevalent across all measures; evidence for inequalities and/or inverse associations were not restricted to a specific aspect of SES. This result is also applicable across different vaccinations. Across the 26 reviews, there was no evidence to suggest that a particular vaccine, or group of vaccines, were more, or less, prone to inequalities in uptake by SES. Mixed results (evidence of inequalities and inverse associations) were equally common for all vaccinations.

Country context

Overall, the association between SES and vaccination uptake was more complex among high-income countries than in low/middle-income countries. In high-income countries evidence for inverse associations and inequalities were both identified frequently.

Low/Middle-income countries

Seven reviews focused on low/middle-income countries [37, 39, 41, 43, 51, 54, 55]. Five of these reviews explored the childhood/adolescent vaccination schedule [37, 39, 43, 51, 54]; two found evidence for consistent inequalities [39, 43], another two inconsistent inequalities [37, 54], and one had mixed findings [51]. In a review analysing influenza vaccination, a conclusion of consistent inequalities was made [41]. The seventh review, by Ali et al., explored the uptake for all routine vaccinations (childhood/adolescent and adulthood) and conducted both a narrative synthesis and meta-analysis [55]. The narrative synthesis showed consistent, mixed results, whereas the meta-analysis demonstrated consistent evidence for inequalities. Overall, in the context of low/middle-income countries, the findings largely suggest there are socioeconomic inequalities in vaccination uptake across the childhood/adolescent schedule, routine, and influenza vaccinations.

High-income countries

The majority (n = 14) reviews focused exclusively on high-income countries [11–13, 15, 18, 38, 42, 44, 45, 47–49, 52, 56]; four of which analysed the childhood/adolescent vaccination schedule, finding mixed and inconsistent associations with SES in three [11, 18, 56], and one inconsistent evidence for inequalities [15]. A further eight reviews explored HPV vaccination [12, 13, 42, 44, 45, 48, 49, 52]; one found consistent evidence for inverse associations with SES [48], one inconsistent inequalities [13], and one consistent mixed [42]. The results of the remaining five reviews that explored HPV vaccination were inconsistent and mixed [12, 44, 45, 49, 52]. One review, analysing influenza and pneumococcal vaccination uptake, identified some consistent support for inverse associations with SES [38]. Analysing influenza vaccination, Lucyk et al. found inconsistent and mixed results [47]. Broadly, these results suggest that, among high-income countries, vaccination uptake does vary across different socioeconomic groups, but the results are often mixed and inconsistent. These findings appear to be applicable to all vaccines (the childhood/adolescent schedule, HPV, influenza, and pneumococcal) routinely given across high-income countries.

High/middle/low-income countries

Five reviews explored a mixture of high, middle, and low-income country contexts [17, 40, 46, 50, 53]. Two reviews focused on the childhood/adolescent schedule [17, 53]; one identified inconsistent support for SES inequalities in uptake [17], whereas the other found consistently mixed associations [53]. Another review found consistent and mixed results for SES and HPV vaccination [46]. Okoli et al. and Nagata et al. analysed the uptake of influenza vaccination [40, 50]; the former identified inconsistent support for SES inequalities [40], whereas the latter showed inconsistent evidence for mixed associations [50]. Overall, reviews conducted in the context of high/middle/low-income countries identified mixed and inconsistent differences in vaccination uptake across different socioeconomic groups across all vaccines analysed (the childhood/adolescent schedule, HPV, influenza, and pneumococcal.

Mechanisms

The majority (n = 16) of included reviews described potential mechanisms that could explain the differences in uptake according to SES. The following mechanisms were hypothesised had not been tested by the review authors: vaccine cost (n = 2) [18, 48]; access to transport (n = 3) [18, 37, 43]; time costs (n = 1) [18]; the extent of maternal control over household resources (n = 1) [43]; confidence (in vaccination in general, or in oneself to make decisions about uptake) (n = 6) [18, 37, 46, 48, 50, 56]; commitment to health-seeking behaviour (n = 3) [18, 43, 46]; vaccination knowledge (access to relevant information and/or ability to understand information) (n = 8) [12, 18, 37, 41, 43, 50, 56]; attitudes or beliefs about vaccination (n = 3) [43, 48, 56]; trust in healthcare or vaccination providers (n = 3) [46, 50, 56]; ease of access (based on the type of healthcare system) (n = 3) [17, 40, 55]; the vaccine delivery strategy (facility versus school-based) (n = 2) [15, 49]; funding of the vaccination programme (n = 2) [47, 53]. The identified mechanisms have been mapped on to the adapted patient-centred access to vaccination framework (Table 4).

Download:

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

Table 4. Extracted mechanisms mapped onto the patient-centred access to vaccination framework.

https://doi.org/10.1371/journal.pone.0294688.t004

Implications of the findings

This umbrella review has summarised and synthesised a large body of complex literature. It provides a concise description of the association between vaccination uptake and SES globally. The existing literature has acknowledged this complexity, particularly around the different directions of association between socioeconomic status and vaccine uptake [19]. However, what other literature fails to do is adequately portray the extent of inverse associations. This is especially relevant to high-income countries. By collating the evidence from several systematic reviews, a wider perspective reveals that higher uptake for disadvantaged SES groups, and lower uptake for advantaged SES, can occur. Before designing interventions to increase uptake, policy makers must be aware of the nature of the association between SES and specific vaccines within their population. Interventions must target all individuals with low uptake, which may include advantaged SES groups as well as disadvantaged SES groups.

Additionally, the exploration of the mechanisms linking SES and vaccination uptake raises questions about the legitimacy of linking consistently relating low uptake to education, as this neglects the structural barriers that are present. Our review highlights the need for more research into the legitimacy of these claims.

Limitations

There are three main limitations to this umbrella review. Three minor amendments have been made to the methodology since the publication of the protocol [25]. Firstly, the study design inclusion criteria were broadened to allow for studies that analysed secondary data to be eligible. Many identified systematic reviews synthesised studies that analysed secondary data from national or regional vaccination registries. Excluding reviews that did would significantly reduce the number eligible. Secondly, the inclusion criteria were reframed using PECOS, in replacement of PICOS, as the former more accurately reflects what we aimed to explore. Thirdly, a pilot of the narrative synthesis indicated that an additional stage was required to adequately answer the primary research question. Namely, to ascertain whether certain countries demonstrated more evidence for differences in vaccination uptake according to SES. The results were organised by high-income, low/middle-income, or mixed, to address this concern.

Secondly, each level of synthesis (primary study to systematic review, systematic review to umbrella review) requires additional interpretation. Subsequently, there may be discrepancies in the conclusions made by the primary study authors and the systematic review authors. These potential discrepancies have then been collated for the purpose of this umbrella review. This issue is even more pertinent when taking into account that all included systematic reviews were rated “critically low” by the AMSTAR-2 checklist. However, it is important to note that AMSTAR-2 is a blunt instrument, with reviews undergoing an automatic downgrade if they do not satisfy one of the “critical domains” [32], and this does not reflect the quality of the primary studies included in these reviews.

Thirdly, there were some overlaps of primary studies, where they had been synthesised in more than one included systematic review. The CCA (13.46%) was reported, showing a moderate degree of overlap. Although the potential overlap was investigated, no action was taken to address the issue.

Suggestions for future research

Future research on vaccine uptake must be more explicit in detailing their PICO criteria. At minimum, the vaccine should be stated, the number of doses (including the number required for full immunization), and the target age of administration–this is especially relevant when comparing multiple countries as routine schedules are likely to vary. Furthermore, it is important for authors conducting systematic reviews to give more careful consideration to the assessment tools used to appraise their work, such as AMSTAR-2. In doing so, they will have a greater awareness of the criterion which they should satisfy in order to be awarded a higher rating.

Moreover, the legitimacy of the mechanisms could be explored. This could be done through data analysis (if appropriate data are available), or via qualitative methods of inquiry. For instance, in socioeconomically deprived areas where vaccination uptake is low, the lived experience of these inequalities could be explored, both from the perspective of healthcare providers and service users.