Introduction

The COVID-19 pandemic has disproportionately affected people with pre-existing socioeconomic and health disadvantages, with wide inequalities in COVID-19 mortality driven by socially patterned risk factors for both SARS-CoV-2 infection and subsequent severe COVID-19 disease [1]. Older adults from minority ethnic groups have been particularly at risk, with COVID-19 mortality rates in England more than 3-fold higher amongst people belonging to Black ethnic groups, and more than 2-fold higher amongst those from Asian ethnic groups, compared to rates amongst those from White ethnic groups [2]. These disproportionate impacts are likely due in part to socioeconomic disadvantages and wide existing health inequalities for those belonging to minority ethnic groups compared to White British adults in England [2,3].

The development and rollout under emergency use legislation of several safe and effective vaccines against COVID-19 has offered the potential to protect individuals and communities from severe COVID-19 [4]. COVID-19 vaccination has been offered free at the point of use through the English National Health Service (NHS) to priority groups, based on age and clinical vulnerability to COVID-19 [5]. However, even for healthcare services provided free at the point of use, disadvantaged populations with the greatest ‘need’ for care tend to have lower uptake of services, formalised as the ‘inverse care law’ [6]. Potential barriers to vaccine uptake include logistical and practical barriers, such as the need to book and travel to vaccination appointments, as well as informational barriers [7]. Personal, family, and cultural views may also influence individuals’ intention to take up vaccination, as may concerns about vaccine safety.

Access to COVID-19 vaccination initially required an appointment booked in advance via the NHS national booking service, a predominantly online booking service available only in English and that requires users to be registered with a general practice (GP) surgery [8]. Many people booked appointments following public announcements indicating that their priority group was eligible to book, and some individuals received letters in the post and/or text messages or telephone calls from their GP inviting them to book an appointment through the national booking service. During the first few months of the COVID-19 vaccine rollout, the majority of vaccination appointments available were at out-of-town mass vaccination centres or hospital hubs. However, as the rollout has progressed, vaccinations have increasingly been provided at local and pop-up sites and without the need for pre-booked appointments, and information is increasingly (though not consistently) available in additional non-English languages.

Existing evidence from England indicates early COVID-19 vaccine uptake has been lower amongst minority ethnic groups in older adults and healthcare workers [9–14]. However, these studies focused on specific population groups, lacked granularity in ethnic group definitions, and did not compare inequalities in uptake to those of other vaccination programmes. Here, we estimate ethnic inequalities in COVID-19 vaccine uptake using a novel electronic health record dataset across a large, ethnically diverse city region in England (Greater Manchester) [15]. We analyse vaccine uptake amongst adults aged 50+ years or with clinical vulnerability to severe COVID-19, and undertake subgroup analysis to identify populations where inequalities are concentrated. To contextualise inequalities in COVID-19 vaccine uptake, we also estimate inequalities in uptake of influenza vaccine in winter 2019/2020, the season immediately prior to the major impacts of the COVID-19 pandemic. We use this comparison because similar groups of adults to those eligible for priority COVID-19 vaccination are routinely encouraged to take up seasonal influenza vaccination [16], and so this comparison allows us to see if the inequalities in vaccination uptake existed before.

Methods

Results

Descriptive statistics

There were 2,543,124 individuals aged 18+ years, resident and registered with a GP surgery in GM, and living in the community included in the GMCR dataset; 1,255,117 individuals were eligible for COVID-19 vaccination in priority cohorts (aged 50+ years or with specified health conditions). Within this group we restricted the sample to 1,099,503 individuals (87.60%) with complete data (Fig A in S1 Appendix).

Most individuals were White British (871,231; 79.24%). The largest minority ethnic groups were Pakistani (50,268; 4.57%), ‘other White background’ (43,195; 3.93%), ‘other ethnic group’ (34,568; 3.14%), and Black African (18,802; 1.71%). The population was more income-deprived than the national average, with 37.59% (413,325) living in the 20% most income-deprived neighbourhoods (Table 1).

Download:

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

Table 1. Baseline characteristics and vaccine uptake.

https://doi.org/10.1371/journal.pmed.1003932.t001

Rapid rollout of vaccination to priority groups occurred over the 138-day analysis period (1 December 2020–18 April 2021) (Fig C in S1 Appendix), with high uptake (>76%) by the end of the period amongst all priority groups (Table A in S1 Appendix). Uptake was highest amongst those aged 75–79 years (89.34%, 95% CI 89.14% to 89.55%), with uptake decreasing monotonically across younger age groups to 76.20% (95% CI 75.99% to 76.41%) in those aged 50–54 years. Amongst those eligible due to health conditions, uptake was higher for those at high clinical risk (87.21%, 95% CI 87.02% to 87.40%) than for those at moderate clinical risk (80.02%, 95% CI 79.87% to 80.17%). COVID-19 vaccine uptake was slightly higher amongst women (83.24%, 95% CI 83.15% to 83.33%) than men (81.70%, 95% CI 81.60% to 81.80%).

Overall, 2019/2020 seasonal influenza vaccine uptake was lower than for COVID-19 vaccination, though patterns across COVID-19 vaccination eligibility groups were similar to those for COVID-19 vaccine uptake (Table A and Fig C in S1 Appendix). Uptake was highest amongst those aged 80+ years (70.18%, 95% CI 69.90% to 70.46%), and lowest amongst those eligible due to moderate clinical risk (36.82%, 95% CI 36.64% to 37.00%). Influenza vaccine uptake was slightly lower amongst women (52.93%, 95% CI 52.78% to 53.08%) compared to men (53.78%, 95% CI 53.62% to 53.94%).

Ethnic inequalities in COVID-19 vaccine uptake

Inequalities in COVID-19 vaccine uptake between the White British group and minority ethnic groups were large across all eligibility groups (Fig 1A; Table B in S1 Appendix). Compared to the White British group, uptake differences tended to be largest for the ethnic groups Black African (HR range: 0.28 [95% CI 0.24 to 0.32] to 0.47 [95% CI 0.45 to 0.49]), Black Caribbean (HR range: 0.36 [95% CI 0.33 to 0.38] to 0.57 [95% CI 0.51 to 0.64]), ‘other Black background’ (HR range: 0.38 [95% CI 0.29 to 0.50] to 0.48 [95% CI 0.40 to 0.57]), and Arab (HR range: 0.38 [95% CI 0.30 to 0.49] to 0.54 [95% CI 0.40 to 0.74]). Patterns of inequalities in uptake across eligibility groups varied between ethnic groups, decreasing with decreasing age for some (6 of 16) groups (e.g., Black African and Indian) but increasing amongst younger age groups for others (e.g., ‘other White background’ and ‘other ethnic group’). For 5 of 16 minority ethnic groups, the widest inequality compared to the White British group occurred within the high clinical risk eligibility group. In line with these results, there were substantial differences in standardised total vaccine uptake at the end of the analysis period, with eligibility-group-standardised uptake at least 20 percentage points lower than that of the White British group for 6 of 16 minority ethnic groups (Table C in S1 Appendix).

Download:

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

Fig 1. Associations between ethnic group and vaccine uptake.

Hazard ratios with 95% confidence intervals from Cox proportional hazards models estimating time to vaccination across ethnic groups stratified by vaccine eligibility group for (A) COVID-19 vaccine and (B) 2019/2020 seasonal influenza vaccine. (C) Hazard ratios with 95% confidence intervals from Cox proportional hazards models estimating time to vaccination across ethnic groups adjusted by vaccine eligibility group amongst individuals eligible for both the COVID-19 and 2019/2020 seasonal influenza vaccines. High clinical = high clinical risk. Mod. clinical = moderate clinical risk.

https://doi.org/10.1371/journal.pmed.1003932.g001

Subgroup analysis by income deprivation

Uptake of COVID-19 and influenza vaccines followed the income deprivation gradient, with uptake of each vaccine approximately 9 percentage points lower in the most, compared to the least, income-deprived areas (Table A in S1 Appendix). Restricting the population eligible for both vaccinations to those living in the least income-deprived areas attenuated the differences in ethnic inequalities in COVID-19 vaccine uptake as compared to those for influenza vaccine uptake, with inequalities greater for COVID-19 vaccine uptake in only 4 of 16 minority ethnic groups (Fig 2A; Table H in S1 Appendix). In contrast, wide ethnic inequalities in COVID-19 vaccine uptake persisted within the most income-deprived areas, exceeding the magnitude of influenza vaccine uptake inequalities for almost all (15 of 16) minority ethnic groups (Fig 2B; Table H in S1 Appendix).

Download:

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

Fig 2. Associations between ethnic group and vaccine uptake by income deprivation.

Hazard ratios with 95% confidence intervals from Cox proportional hazards models estimating time to vaccination across ethnic groups adjusted by vaccine eligibility group amongst individuals eligible for both the COVID-19 and 2019/2020 seasonal influenza vaccines living in (A) the 20% least income-deprived neighbourhoods and (B) the 20% most income-deprived neighbourhoods.

https://doi.org/10.1371/journal.pmed.1003932.g002

Subgroup analysis by prior influenza vaccine uptake

Amongst those eligible for both vaccinations, COVID-19 vaccine uptake was substantially higher amongst those who did previously take the influenza vaccine (95.46%, 95% CI 95.40% to 95.52%) than those who did not (74.64%, 95% CI 74.55% to 74.74%) (Table A in S1 Appendix). Amongst those who did previously take the influenza vaccine, wide ethnic inequalities in COVID-19 vaccine uptake persisted (15 of 16 groups) and were similar in magnitude to the inequalities amongst those not previously vaccinated against influenza for half (8 of 16) of the minority ethnic groups (Fig 3; Table I in S1 Appendix). Where there were differences, these were relatively small in magnitude and varied in directionality.

Download:

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

Fig 3. Associations between ethnic group and vaccine uptake by prior influenza vaccine uptake.

Hazard ratios with 95% confidence intervals from Cox proportional hazards models estimating time to COVID-19 vaccination across ethnic groups adjusted by vaccine eligibility group amongst individuals eligible for both the COVID-19 and 2019/2020 seasonal influenza vaccines who (A) did take up 2019/2020 seasonal influenza vaccination and (B) did not take up 2019/2020 seasonal influenza vaccination.

https://doi.org/10.1371/journal.pmed.1003932.g003

Discussion

We found wide inequalities in time to COVID-19 vaccination for most minority ethnic groups compared to the White British group. Inequalities were particularly wide for all Black or Black British groups and the Arab group. Vaccine uptake inequalities were substantially wider for COVID-19 vaccination than for influenza vaccination. We also found wide ethnic inequalities in COVID-19 vaccine uptake amongst individuals who previously took up influenza vaccination. COVID-19 vaccine inequalities were concentrated amongst those most at risk of severe COVID-19 (older and clinically vulnerable adults) and those living in income-deprived neighbourhoods.

Our finding that COVID-19 vaccine uptake was lower amongst almost all minority ethnic groups, and particularly low amongst people belonging to Black or Black British ethnic groups, is broadly consistent with existing reports [9–14]. However, previous studies have not analysed uptake inequalities for Arab individuals, so these inequalities were not previously recognised [9–14].

To the best of our knowledge, there is no published information about ethnic inequalities in influenza vaccine uptake amongst older or clinically vulnerable adults in the UK. However, several studies assessing intention to receive vaccination against the 2009 pandemic H1N1 influenza (‘swine flu’) found that people from minority ethnic groups were more likely than the White British group to intend to take up vaccination [22]. It has been suggested this could be attributable to the higher incidence of severe influenza amongst the minority ethnic population [22]. However, this factor appears not to apply to COVID-19 vaccination. These findings reinforce the conclusion that the magnitude of ethnic inequality in COVID-19 vaccine uptake is unusual, and far exceeds the inequalities associated with uptake of other vaccines.

Recent prior influenza vaccine uptake suggests individuals have a degree of confidence in vaccination, and have chosen and successfully navigated access to vaccination, perhaps somewhat explaining the higher COVID-19 vaccine uptake amongst this group. However, even within the group with recent influenza vaccine uptake, we found lower uptake of COVID-19 vaccination amongst people from minority ethnic groups relative to the White British group. This suggests that new ethnic health inequalities have been generated, perhaps resulting from additional barriers to vaccine access, lower vaccine confidence, or both.

There were many differences in how individuals were able to access COVID-19 vaccination compared to 2019/2020 influenza vaccination, which may have introduced additional barriers that disproportionately affected minority ethnic groups. For example, there was a strict requirement for booked COVID-19 vaccination appointments, and much delivery was through mass vaccination centres or hospital hubs. As discussed by Razai et al. [7], residential segregation driven by systemic racism may therefore have resulted in additional barriers—in terms of journey time and cost—to accessing centralised vaccination sites for some communities. In contrast, influenza vaccines are available through GPs and community pharmacies, often via drop-in services [16]. Community pharmacies have been shown to be more accessible in areas of high deprivation than in less deprived areas [23]. Participants in PCIE discussion groups also raised concerns that vaccine information was rarely presented in languages other than English, and that public health messaging sometimes failed to address the concerns of target populations. It is unclear whether any type of health equity impact assessment or community engagement was undertaken during the very rapid centralised design of the COVID-19 vaccine rollout, and this may have contributed to the observed exacerbation of inequalities. Further interdisciplinary research is needed to understand the potential impacts of each of these factors on equity of access to COVID-19 vaccination.

Several studies have investigated attitudes towards COVID-19 vaccination across sociodemographic groups, with relatively consistent reports of lower vaccine confidence amongst minority ethnic groups, which likely contributes to observed uptake disparities [24–29]. Concerns about the novelty and potential unknown side effects or long-term effects of COVID-19 vaccines were frequently reported [24,26–28]. These concerns may have been heightened amongst people from minority ethnic groups due to their underrepresentation in COVID-19 vaccine clinical trials [7,26,29]. Erosion of trust in government, mainstream media, and medical authorities during the COVID-19 pandemic has also been reported [7,25–31] and was a key theme raised in PCIE discussions. A combination of factors—such as the disproportionate burden of COVID-19 mortality amongst minority ethnic groups [2], apparent government suppression of a report into COVID-19 disparities [30], and disregard for the impact of restrictions on non-Christian religious festivals—may have exacerbated loss of trust amongst minority ethnic groups [7]. Moreover, these factors likely compounded existing mistrust amongst some minority ethnic communities stemming from racism, experiences of culturally insensitive healthcare, and awareness of previous unethical healthcare research [7,24]. It will be particularly important to understand whether these COVID-19 vaccination findings represent a shift in attitudes towards vaccination per se, as this may lead to wider inequalities in future routine influenza and other vaccination campaigns.

A strength of this study is the comparison of inequalities in uptake of COVID-19 and influenza vaccines. This brings crucial context to observed inequalities, and demonstrates that COVID-19 vaccine uptake reproduces previous health inequalities, but also harbours new inequalities that are particularly concentrated in income-deprived neighbourhoods. Stratification by previous influenza vaccine uptake demonstrated that the COVID-19 vaccine rollout has generated inequalities even amongst individuals previously accessing influenza vaccination.

Another strength is the population-level coverage of electronic health record data across a large, ethnically diverse metropolitan area, which allowed estimation of inequalities for 16 minority ethnic groups. However even with this large sample, we were unable to estimate inequalities in vaccine uptake for the Gypsy or Irish Traveller group due to very low numbers. There are also limitations in the recording of sociodemographic information in electronic health records. For example, there is incomplete granularity in the recording of ethnicity, which leads to relatively large populations in each of the ‘other’ ethnic groupings, a problem highlighted by our community advisory group. Analysis stratified by gender indicated that ethnic inequalities in vaccine uptake were moderately wider for women than for men for many minority ethnic groups. This highlights the interactions between sociodemographic factors that may exacerbate health inequalities amongst some groups. However, we were unable to explore the role of other factors that may intersect with ethnicity and affect vaccination uptake, such as faith, country of birth, or preferred spoken and written languages, as these are rarely recorded in electronic health records.

The high clinical risk group was robustly identified using clinical codes for the ‘extremely clinically vulnerable’ population of individuals who were instructed to ‘shield’ during the first COVID-19 wave in England. However, the moderate clinical risk group was approximated using codes for clinical eligibility for seasonal influenza vaccination, potentially introducing some misclassification. Similarly, health and social care workers were not identified in our dataset but were eligible for priority vaccination. These misclassifications may have differed with respect to ethnicity, but Kaplan–Meier plots indicated reasonable temporal agreement between eligibility and uptake, suggesting this is unlikely to have substantially biased the estimates. Data were highly complete (≥99.9%) for most variables, but 12.31% of the population was missing data for ethnic group, which could have introduced bias. Detailed exploration of the demographic patterns of missing ethnicity data suggested that a large majority of those with missing data were likely to be White British. Sensitivity analysis showed this did not substantially bias our estimates of vaccine uptake inequalities. Another limitation of our analysis is that we used the Breslow method to handle ties in the Cox regression models, as opposed to the Efron method, which has been shown to generally be less biased [32]. Although using the Breslow method may have biased our results towards underestimation of inequalities in vaccine uptake, this method is more computationally feasible with a large dataset and is unlikely to have markedly biased estimates, given our large sample size [32].

Another limitation is that we considered only first-dose COVID-19 vaccine uptake. Reported differential uptake of second doses [14] may already have led to wider inequalities in terms of full COVID-19 vaccination than those reported here. It may be particularly important to understand whether inequalities widen for subsequent doses, given the planned rollout of ‘booster’ COVID-19 vaccine doses [33].

Finally, given the ethnically diverse urban study setting, our results are likely informative for similar regions, but may not be fully generalizable to England as a whole, or applicable across international contexts.

In summary, we identified wide ethnic inequalities in COVID-19 vaccine uptake, and demonstrated that inequalities were concentrated amongst groups at greatest risk of severe COVID-19—older adults, extremely clinically vulnerable individuals, and those living in the most income-deprived neighbourhoods. By contextualising inequalities in relation to previous influenza vaccine uptake, we found a marked exacerbation of ethnic health inequalities in vaccine uptake. Further research and policy action is needed to understand and remove barriers to vaccine uptake, and to build trust and confidence amongst minority ethnic communities.

Supporting information

Acknowledgments

We would like to thank John Ainsworth and all members of the Greater Manchester Population Research Resource team at the University of Manchester for helpful discussion and suggestions that have helped shape this project. We are grateful to Nicky Timmis, Aneela McAvoy, Joanna Ferguson, and Sue Wood for their support in organising and hosting PCIE discussion groups. We would particularly like to thank Nasrine Akhtar, Basma Issa, Nicholas Filer, and Charles Kwaku-Odoi (public contributor advisory group members) for their invaluable insights, as well as all members of the NIHR Applied Research Collaboration for Greater Manchester Public and Community Involvement and Engagement Panel and the Health Innovation Manchester Public Community Involvement and Engagement Forum.

We would also like to recognise the GMCR (a partnership of Greater Manchester Health and Social Care Partnership, Health Innovation Manchester, and Graphnet Health, on behalf of Greater Manchester localities) for the provision of data required to undertake this work. This work uses data provided by patients and collected by the NHS as part of patient care and support. Using patient data is vital to improving health and care for everyone. There is huge potential to make better use of information from people’s patient records, to understand more about disease, develop new treatments, monitor safety, and plan NHS services. Patient data should be kept safe and secure, to protect everyone’s privacy, and it’s important that there are safeguards to make sure that data are stored and used responsibly. Everyone should be able to find out about how patient data is used.