COVID-19 vaccine hesitancy among undocumented migrants during the early phase of the vaccination campaign: a multicentric cross-sectional study

Abstract

Study objectives

The marginalisation of undocumented migrants raises concerns about equitable access to COVID-19 vaccination. This study aims to describe migrants’ hesitancy about the COVID-19 vaccination during the early phase of the vaccination campaign.

Setting

This multicentric cross-sectional survey was conducted in health facilities providing care to undocumented migrants in the USA, Switzerland, Italy and France in February–May 2021.

Participants

Eligibility criteria included age >16 years, being of foreign origin and living without valid residency permit in the country of recruitment. A convenience sample of minimum 100 patients per study site was targeted.

Primary and secondary outcome measures

Data were collected using an anonymous structured questionnaire. The main outcomes were perceived access to the local COVID-19 vaccination programme and demand for vaccination.

Results

Altogether, 812 undocumented migrants participated (54.3% Geneva, 17.5% Baltimore, 15.5% Milano and 12.7% Paris). Most (60.9%) were women. The median age was 39 years (interquartile range 1). Participants originated from the Americas (55.9%), Africa (12.7%), Western Pacific (11.2%) Eastern Mediterranean (7.9%), Europe (7.6%) and South-East Asia (4.7%). Overall, 14.1% and 26.2% of participants, respectively, reported prior COVID-19 infection and fear of developing severe COVID-19 infection. Risk factors for severe infection were frequently reported (29.5%). Self-perceived accessibility of COVID-19 vaccination was high (86.4%), yet demand was low (41.1%) correlating with age, comorbidity and views on vaccination which were better for vaccination in general (77.3%) than vaccination against COVID-19 (56.5%). Participants mainly searched for information about vaccination in the traditional and social media.

Conclusions

We found a mismatch between perceived accessibility and demand for the COVID-19 vaccination. Public health interventions using different communication modes should build on trust about vaccination in general to tackle undocumented migrants’ hesitancy for COVID-19 vaccination with a specific attention to men, younger migrants and those at low clinical risk of severe infection.

Strengths and limitations of this study.

• The study included undocumented migrants, a hard-to-reach population, in four countries.

• Efforts were made to overcome language, trust and literacy barriers to participation.

• The number of participants differed in every study site.

Introduction

It is estimated that between 3.9 and 4.8 million undocumented migrants live in Europe and 10.5 million in the USA.1–3 Economic opportunities, integration policies, and the rights and benefits afforded to undocumented migrants vary by host country. However, challenges, including language barriers, fear of deportation, poverty, housing precariousness, and limited access to healthcare and workplace protections, are common experiences for most undocumented migrants.

Although undocumented migrants represent less than 1% of Europe’s and 3.2% of the US total population, emerging evidence points to the devastating impact of COVID-19 in this group. In high-income countries, migrants have high risk of COVID-19 infection, morbidity and mortality.4 Although COVID-19 outcomes by specific immigration status are rarely available, surrogate markers (eg, language, country of origin, housing status, health insurance eligibility and demographics) suggest that undocumented migrants are at particularly high risk.4–13 Community and health facility-based studies in Europe and the USA showed exceptionally high SARS-CoV-2 positivity rates among foreign-born or limited English proficiency patients.7 8 14 15 In the USA, COVID-19 case rates were highest in counties with large immigrant communities, and the correlation was stronger in areas with more Central Americans, a group with high poverty levels and irregular migrant status.1 12 16 In addition, there is evidence of poor outcomes due to delayed presentation to care among undocumented migrants.4 6 10 17 18 Mortality data by migrant status are limited, but what is available shows that compared with native-born citizens, migrants to Europe and the USA, particularly those from low/middle-income countries, have higher excess all-cause and COVID-19 mortality.19–23

Undocumented migrants play an essential role in the global economy but rely heavily on informal and low-wage labour with limited occupational protections. Mitigation strategies to reduce the social, economic and health impact of the COVID-19 pandemic frequently exclude undocumented migrants. Without a social safety net, many continued to work at the peak of the pandemic in high-risk essential jobs, such as logistics, manufacturing, domestic and care activities, construction and the food processing industry.11 24 25 Several European countries provided food assistance to migrants during lockdown, and a few further extended benefits. For example, Ireland implemented a system to pay unemployment benefits to undocumented migrants who lost their jobs, and Portugal granted temporary citizenship rights to migrants.26 The suspension of exclusionary immigrant policies, however, was not uniform and there were many unmet needs and many vulnerable undocumented migrants fell into extreme poverty.26 A survey conducted in Switzerland in April 2020 showed that almost one in six migrants had experienced hunger during the first lockdown.27

Furthermore, long-standing anti-immigrant policies and mistrust of governmental institutions have not been eased during the pandemic, and pre-existing legal, socioeconomic, and linguistic barriers to social and health services have exacerbated the impact of COVID-19 among undocumented migrants.28 29 Although countries deployed health services for COVID-19 without eligibility restrictions based on migration status, no specific measure has been implemented to facilitate access for undocumented migrants who already tended to underuse social and health services even before the pandemic.30 31 As a result, pre-existing barriers to accessing health and social services are exacerbated by the pandemic and likely lead to delaying life-saving care for many.6 10 11 27

The rapid development of effective COVID-19 vaccines was an unprecedented scientific achievement, but equitable vaccine distribution is a major challenge worldwide. Undocumented migrants and other socially disadvantaged populations have faced significant hurdles to get vaccinated, including digital, transportation and health system navigation barriers. The European Centre for Disease Prevention and Control (ECDC) and the Council of Europe have called for tailored vaccination programmes for undocumented migrants that are free from immigration control enforcement activities,32 but only a few national immunisation plans explicitly include provisions for undocumented migrants, or address potential barriers, such as language proficiency or identification requirements.33 34 In addition, the willingness and hesitancy of individuals, including undocumented migrants, to get immunised depends on a variety of factors, such as self-perceived risks and severity of illness; confidence in the safety and effectiveness of the vaccine; trust in medical, governmental, or pharmaceutical institutions; and behavioural and social processes (eg, awareness, information, education, social norms, networks and media). The objective of this multicentric study conducted in the early phase of COVID-19 immunisation programmes was to explore undocumented migrants’ hesitancy about COVID-19 vaccine.

Methods

Design

This multicentric cross-sectional survey was conducted from mid-February to late May 2021 in four facilities providing medical care to undocumented migrants in Switzerland, the USA, Italy and France during the early phase of the vaccination campaign (February–May 2021).

Setting

The four study sites are part of an informal network of health institutions providing care to undocumented migrants which started to share experiences and good practices during the early phase of the COVID-19 pandemic.

Geneva, Switzerland

Geneva (population 500 000) hosts an estimated 10 000–15 000 undocumented migrants, predominantly women from Latin America, the Philippines and South-Eastern Europe who are active in the domestic and care industry.35 While potentially eligible to purchasing the mandatory health insurance to access medical care, less than 10% are actually insured because of financial and administrative barriers. The Geneva University Hospital acts as the main port of entry into the healthcare system for undocumented migrants and other underserved groups of population, providing the full range of preventive, curative and rehabilitation health services.36 While the Swiss Federal Government has decided on the universal access to COVID-19 vaccination to all residents irrespective of their legal status in early 2021, the policy implementation has been delayed at Canton level and Geneva was the first Canton to officially integrate undocumented migrants into the vaccination programme in May 2021.37 At the beginning of the study, the COVID-19 incidence and mortality in Canton Geneva were at their lowest since October 2020. There was then a mild resurgence of new cases not associated with increased mortality that peaked in April before coming back to its baseline in May. The vaccination campaign started on 28 December 2020. Two vaccines were available, BNT162b2 mRNA (Pfizer/BioNTech) and mRNA-1273 (Moderna). In the first 2 months, vaccination was limited to high-risk groups and it became available to all adults in early March 2021. By the end of the study, 37% of the population had received at least one dose. No additional public restrictions were imposed during the study period.

Milan, Italy

According to available estimates, there are currently 517 000 undocumented migrants in Italy.38 Disaggregated estimates at city level including for Milan are not readily available. However, Milan is the economic centre and the most populous region in Italy, hence likely to host a large population of undocumented migrants. In principle, the National Health Service (NHS) system is based on a universalistic model providing healthcare free of charge at the point-of-use against payment of standard flat fees with waivers based on socioeconomic criteria and is decentralised at regional level for both policy and service delivery aspects. Access to the NHS requires a valid health card, which is issued based on residency status. As a result, undocumented migrants do not have access to the NHS. To address this fundamental legal and administrative barrier, the NHS provides a temporary access code, which allows access to emergency care and essential services including maternity and vaccination services. In practice, undocumented migrants face barriers even to obtain a temporary access code and rely on charities for accessing healthcare. Among them, ‘Opera San Francesco per i Poveri’ is a faith-based charity operating a large-sized health clinic in Milan providing free-of-charge outpatient healthcare including consultations, diagnostics and therapy for socially disadvantaged population groups including undocumented migrants. For COVID-19 vaccination, the NHS procures and distributes vaccines and consumables, while the regional health system administers them through a client-initiated online booking system requiring a valid health card. As of 25 June 2021, the Lombardy Region, with Milan as the chief-lieu, granted eligibility for online booking to undocumented migrants with a temporary access code. Charities have mobilised to provide individual support to facilitate administrative, linguistic and practicality challenges. At study inception, COVID-19 incidence and mortality were persistently elevated in Italy. The Lombardy Region, with Milan as its chief-lieu, continued to account for the highest toll in-country. Restrictions including lockdown continued to be implemented in a modular way according to local epidemiology. The national immunisation campaign kicked off officially just before the end of 2020, targeting the health workforce and the elderly in hospices; however, it struggled to pick up pace until summer 2021 and only 1.2% of total target population was fully immunised at study inception. Initially, the campaign used BNT162b2 mRNA, then mRNA-1273, ChAdOx1 nCoV-19 AZD1222 (AstraZeneca), and finally added JNJ-78436735 (Johnson & Johnson) vaccines, the latter having been prioritised for hard-to-reach population groups including undocumented migrants.

Baltimore, USA

Baltimore City is an emergent destination for migrants from Latin America.39 An estimated 20 000 foreign-born Latin Americans live in the city and approximately 13 500 (67%) are not citizens. Migrants from Mexico and Central America have higher non-citizen status (>80%), low educational attainment (50% with less than high school education) and high rates (70%) of limited English proficiency.40 In the USA, the COVID-19 vaccine is freely available to all, regardless of immigration or insurance status, and the Department of Homeland Security has explicitly stated that immigration enforcement activities will not be conducted at vaccination site.41 In the early stages of the COVID-19 immunisation programme, the state of Maryland implemented a phased distribution plan and the vaccine was not available to the general population until 27 April, after data collection for this study was completed. The Access Program, Johns Hopkins Medicine in Baltimore, Maryland (TAP) acts as the main port of entry into the Johns Hopkins Health System. Patients are enrolled in TAP if they have low income (<200% federal poverty line) and are ineligible to enrol in Medicaid or subsidised health insurance because of their irregular immigration status. In Baltimore City, cases of COVID-19 in February of 2021 were the lowest since October 2020, but by March 2021, a fourth wave of COVID-19 emerged which peaked on 10 April 2021. COVID-19 vaccine administration began on 14 December 2020 in a phased approach which sequentially prioritised first responders, the elderly and those with underlying health conditions. The vaccine became available to the general population on 27 April 2021. Three COVID-19 vaccines authorised in the USA for Emergency Use or approved by the US Food and Drug Administration were available for vaccination programmes: BNT162b2 mRNA, mRNA-1273 and JNJ-78436735.

Paris, France

Avicenne University Hospital is located in the Department of Seine Saint Denis in the North-East of Paris. The department is historically a place where migrants use to be provided social lodging after the Second World War (mainly sub-Saharan Africa and North African communities). It is estimated that more than 30% of the population is constituted of immigrants, with recently an additional wave of migrants from South Asia. Moreover, the majority of undocumented migrants in metropolitan France (around 400.000) tend to be concentrated in this department. Undocumented migrants in France have access to health via State Medical Aid, an insurance coverage for individuals with no right to National Health Insurance. Those without any coverage may access healthcare via specific units created for uninsured persons (Permanence d’accès aux soins), located in hospitals principally. Avicenne University Hospital receives uninsured persons via this unit on a daily basis. In France, all eligible persons are entitled to COVID-19 vaccination, as per government declaration. In Paris region, incidence of COVID-19 mid-February 2021 was already high at 237/100 000 inhabitants, and quickly increased further. A third lockdown was ordered on 18 March when incidence was at 426/100 000. The incidence peaked at the end of April, at 682/100 000, and slowly decreased. The survey hence took place about 1 month before the lockdown when virus circulation was already quite high, with a regional curfew in place since mid-January. The rate of study site enrolment was further affected by the lockdown and the increased police controls. COVID-19 vaccine national campaign began on 27 December 2020 in a phased approach which first prioritised the elderly, and those with underlying health conditions. The vaccine became available to the general population on 18 January 2021, while its uptake was very slow during the first weeks. The four COVID-19 vaccines authorised in France were BNT162b2 mRNA, mRNA-1273, ChAdOx1 nCoV-19 AZD1222 and JNJ-78436735.

Participants

Eligibility criteria were age equal or above 16 years and living as a foreigner without valid residency permit (undocumented) in the country of recruitment. Participants were recruited upon spontaneous presentation (walk-in) to one of the participating health facilities.

We used several strategies to reduce the risk of recruitment and measurement bias by addressing the main barriers limiting undocumented migrants’ participations in health programmes such as fear of personal data misuse and sociocultural factors. All consecutive patients consulting at the four health facilities were informed about the study orally and with written material in different languages. We explained that the questionnaire was anonymous, and that no identifying information was collected considering the frequent fear of undocumented migrants to disclose personal information. The questionnaire was translated in French, Spanish, Italian, Portuguese, Arabic, English, Tagalog, Albanian, Ukrainian and Russian to match with the main languages spoken by migrants visiting the participating health facilities. Participants were proposed the support of research assistants competent in various languages to fill the questionnaire to overcome potential difficulties in reading and understanding the questions.

Data source and variables

We designed a 15-item questionnaire (online supplemental material) based on UNICEF and WHO guidance toolkit for COVID-19 vaccination demand,42 43 and an ECDC document exploring vaccine hesitancy.44 Our main outcome of interest was COVID-19 vaccine hesitancy explored through two main perspectives, perception about vaccination accessibility and the drivers and barriers for demands. Accessibility was investigated using the question: ‘Do you believe that migrants in your (legal) situation will have access to the COVID-19 vaccination?’ with ‘yes’, ‘no’ and ‘I don’t know’ as possible responses; we dichotomised ‘yes’ and ‘I don’t know’ versus ‘no’ in order to determine the proportion of participants perceiving that the vaccination would not be inaccessible. We further investigated the type of barrier in those responding ‘no’. Demand was investigated using the question: ‘If the vaccine was offered to you, would you like to get immunised against COVID-19?’. Responses to the latter question included ‘yes no doubt’, ‘probably yes’, ‘probably no’, ‘no’, ‘I don’t know yet’. In the analysis, we dichotomised ‘yes no doubt’ versus all other response to determine the proportion of vaccine-hesitant respondents, based on the definition of vaccine hesitance as the reluctance or refusal to vaccinate despite the availability of vaccines along a continuum with a broad spectrum of attitudes and intentions from active demand to passive acceptance, vaccine hesitancy and refusal of all vaccines.44 We explored enabling and barriers factors for vaccine accessibility and demand such as demographic characteristics, self-reported clinical risk factors for severe SARS-CoV-2 infection, previous infection with SARS-CoV-2 (self and/or household), self-perceived health risks with COVID-19, views about vaccination in general and COVID-19 vaccination in terms of safety and efficacy (both dichotomised as positive vs negative), desirable place of vaccination and finally the main sources of information about COVID-19 vaccine (traditional media, social media and community networks). The questionnaire was pretested in 10 participants before being implemented in all study sites.

Study size

In absence of pre-existing hypothesis regarding the distribution of responses to the two main outcomes, considering the difference in the number of monthly visits in each site and the uncertainties about migrants’ willingness to engage into the study in the different sites, we pragmatically set a minimal sample size of 100 participants per study site to be reached within the predefined study period.

Patient and public involvement

This study was informed by patients expressing interest and concerns to healthcare workers about COVID-19 vaccine accessibility and safety in the four study sites.

Statistical analysis

Categorical data are presented as proportions with percentages and non-normally distributed continuous variable as median with IQR. We compared the distribution of variables in the four study sites using the Kruskal-Wallis test for non-normally distributed variables and the Χ² test or the Fisher’s exact test, as appropriate. The significance level was set at 0.05.

We performed both univariate and multivariate logistic regression analyses to identify factors associated with the two main outcomes. ORs were estimated through multivariate logistic regression models, which were mutually adjusted with all covariates in the models. Missing values, which ranged from 0.2% to 3.6% of the total study size, were imputed by using a multiple (n=100) imputation approach. Briefly, multiple imputation is a Bayesian method that allows to take into account incomplete cases (ie, observations with any missing data) with a two-step approach. First, this method creates multiple imputed datasets, in which missing values are replaced by imputed values. These are sampled from their predictive distribution based on the observed data. The imputation procedure fully accounts for the uncertainty in predicting the missing values by conferring appropriate variability into the multiple imputed values. Second, standard statistical methods are used to fit the model of interest to each of the imputed datasets. Estimates associated to each of the imputed datasets differ because of the variation introduced in the imputation of the missing values (stage 1), and they are, then, average together to give overall estimated associations. Valid inferences are obtained because they are based on the average of the distribution of the missing data given the observed data, and results were reported as ORs along with their 95% CIs. All analyses were performed using SAS V.9.4.

Role of the funding source

The funders had no role in study design, or in data collection, analysis or interpretation.

Results

Participants’ characteristics

A total of 812 individuals completed the survey: 441 (54.3%) in Geneva, 142 (17.5%) in Baltimore, 126 (15.5%) in Milan and 103 (12.7%) in Paris. The median age was 40.1 years (range 17–76) with a predominance of female respondents (60.9%), but gender distribution varied by city and, notably, 69.9% of participants in Paris were male (table 1). They mainly originated from the Americas (55.9%), Africa (12.7%) and the Western Pacific regions (11.2%). Participants born in the Americas accounted for all the respondents in Baltimore, over half in Geneva and Milan, but only 1.9% in Paris, which had the largest representation of African migrants.

Table 1.

Sociodemographic characteristics of the participants (n=812)

	Total N=812, n (%) or median (IQR)	Geneva N=441, n (%) or median (IQR)	Baltimore N=142, n (%) or median (IQR)	Milan N=126, n (%) or median (IQR)	Paris N=103, n (%) or median (IQR)	P value
Female gender	492 (60.9)	279 (63.4)	98 (70.0)	84 (67.2)	31 (30.1)	<0.001
Missing values	4	1	2	1	0
Age	39 (16)	39 (17)	40 (13)	41 (20)	35 (16)	0.001
Missing values	2	1	0		1
Region of origin						0.001
Africa	103 (12.7)	52 (11.8)	0 (0)	8 (6.4)	43 (41.8)
Americas	454 (55.9)	227 (51.5)	142 (100)	83 (65.9)	2 (1.9)
Eastern Mediterranean	64 (7.9)	28 (6.4)	0 (0)	7 (5.6)	29 (28.2)
Europe	62 (7.6)	39 (8.8)	0 (0)	21 (16.7)	2 (1.9)
Asia	38 (4.7)	7 (1.6)	0 (0)	6 (4.8)	25 (24.3)
Western Pacific	91 (11.2)	88 (20.0)	0 (0)	1 (0.8)	2 (1.9)
Missing values	0	0	0	0	0

Accessibility and demand for vaccination and risk factors for severe infection

The vast majority (86.4%) of participants perceived that the COVID-19 vaccination would be accessible to undocumented migrants, but a lower proportion (41.2%) reported they would get vaccinated against COVID-19 (table 2). Approximately one-third (29.5%) of participants reported at least one chronic comorbidity that could predispose to severe COVID-19 infection, 14.1% reported prior COVID-19 infection and 26.2% worried about developing severe COVID-19 (table 2). In all cities, perceptions about vaccination in general were more favourable than about COVID-19 vaccination overall, more than three-quarters (77.3%) of respondents had positive views on vaccination in general, compared with (56.5%) about COVID-19 vaccination. Traditional media was the most common source of information about COVID-19 vaccination, followed by social media. Community networks were a common source of information among participants in Paris (72.8%), but less so among participants in other cities.

Table 2.

Undocumented migrants’ perceived accessibility to and demand for COVID-19 vaccine with related enabling and barrier factors

	Total N=812, n (%)	Geneva N=441, n (%)	Baltimore N=142, n (%)	Milan N=126, n (%)	Paris N=103, n (%)	P value
Access to COVID-19 vaccination	697 (86.4)	377 (86.1)	116 (82.3)	110 (88.0)	94 (91.3)	0.219
Missing values	5	3	1	1	0
Demand for COVID-19 vaccination	327 (41.2)	168 (39.0)	79 (59.0)	65 (52.0)	15 (14.6)	<0.001
	19	10	8	1	0
COVID-19 exposure
COVID-19 infection (self)	114 (14.1)	62 (14.1)	32 (22.5)	11 (8.7)	9 (8.8)	0.003
Missing	3	2	0	0	1
COVID-19 infection (household)	129 (16.1)	74 (17.0)	35 (25.2)	17 (13.5)	3 (2.9)	<0.001
Missing values	9	6	3	0	0
Clinical risk factors for severe COVID-19 infection
Cardiovascular disease	109 (13.7)	46 (10.8)	14 (10.1)	34 (27.0)	15 (14.6)	<0.001
Diabetes	85 (10.7)	21 (4.9)	27 (19.4)	13 (10.3)	24 (23.3)	<0.001
Weight excess	79 (9.9)	29 (6.8)	22 (15.8)	16 (12.7)	12 (11.7)	0.010
Chronic lung disease	40 (5.0)	24 (5.6)	1 (0.7)	11 (8.7)	4 (3.9)	0.022
Chronic kidney disease	29 (3.7)	15 (3.5)	8 (5.8)	5 (4.0)	1 (1.0)	0.272
≥1 comorbidity	234 (29.5)	96 (22.5)	52 (37.4)	57 (45.2)	29 (28.2)	<0.001
Missing values	18	15	3	0	0
Views on COVID-19 risks and vaccination
High self-perceived risk of severe COVID-19 infection	208 (26.2)	95 (22.0)	35 (25.7)	42 (33.9)	36 (35.0)	0.008
Missing values	18	10	6	2	0
Positive views on vaccination in general	605 (77.3)	300 (70.6)	126 (94.0)	98 (79.0)	81 (81.0)	<0.001
Missing values	29	16	8	2	3
Positive views on COVID-19 vaccination	445 (56.5)	218 (51.1)	104 (77.6)	79 (63.7)	44 (42.7)	<0.001
Missing values	24	14	8	2	0
Sources of information about COVID-19 vaccines
Traditional media (TV, radio, web)	626 (79.3)	329 (76.9)	109 (82.0)	104 (83.2)	84 (81.6)	0.309
Social media	361 (45.8)	189 (44.2)	36 (27.1)	56 (44.8)	80 (77.7)	<0.001
Community networks	214 (27.1)	99 (23.1)	6 (4.5)	34 (27.2)	75 (72.8)	<0.001
Other	33 (4.2)	25 (5.8)	0 (0)	7 (5.6)	1 (1.0)	0.007
Missing values	23	13	9	1	0

Barriers to and preferred place for vaccination

Although perceptions about accessibility did not vary by city, demand ranged widely and was lowest (14.6%) among participants living in Paris. Respondents who did not believe that COVID-19 vaccination would be available to undocumented migrants reported lack of health insurance or card as the main barrier to access. Overall, most participants who intended to get vaccinated preferred to do so at a hospital (73.5%) (tables 3 and 4).

Table 3.

Perceived barriers to accessing COVID-19 vaccination in participants mentioning vaccination being not accessible

	Total N=110, n (%)	Geneva N=61, n (%)	Baltimore N=25, n (%)	Milan N=15, n (%)	Paris N=9, n (%)
Lack of insurance/health card (National Health System)	57 (51.8)	32 (52.5)	14 (56.0)	9 (60.0)	2 (22.2)
High cost	25 (22.7)	17 (27.9)	2 (8.0)	3 (20.0)	3 (33.3)
Lack of eligibility to enrol in vaccination programme	18 (16.4)	8 (13.1)	1 (4.0)	5 (33.3)	4 (44.4)
Not knowing where to go	27 (24.5)	13 (21.3)	9 (36.0)	3 (20.0)	2 (22.2)
Other reasons	13 (11.8)	6 (9.8)	0 (0)	5 (33.3)	2 (22.2)
Missing values	0	0	0	0	0

Table 4.

Preferred place for COVID-19 vaccination

	Total N=327, n (%)	Geneva N=168, n (%)	Baltimore N=79, n (%)	Milan N=65, n (%)	Paris N=15, n (%)
Hospital	236 (73.5)	144 (87.8)	40 (50.6)	39 (60.9)	13 (92.9)
Public health/community clinic	65 (20.2)	31 (18.9)	17 (21.5)	16 (25.0)	1 (7.1)
Private physician	20 (6.2)	4 (2.4)	3 (3.8)	11 (17.2)	2 (14.3)
Pharmacy	37 (11.5)	17 (10.4)	6 (7.6)	9 (14.1)	5 (35.7)
Charity	65 (20.2)	22 (13.4)	16 (20.3)	19 (29.7)	8 (57.1)
Other	10 (3.19)	4 (2.4)	2 (2.5)	4 (6.3)	0 (0)
Missing values	6	4	0	1	1

Factors associated with perceived accessibility of COVID-19 vaccination

In univariate and multivariate analyses, female gender was the only factor positively associated with self-perceived accessibility to COVID-19 vaccination overall, while participants originating from the Americas or recruited in Baltimore tended to be more confident about accessibility (table 5).

Table 5.

Factors associated with perceived accessibility of COVID-19 vaccination in regression analysis

	Univariate analysis		Multivariate analysis
	OR (95% CI)	P value	aOR (95% CI)	P value
Study site Geneva	Reference		Reference
Baltimore	0.75 (0.45 to 1.25)	0.276	0.56 (0.30 to 1.03)	0.063
Milan	1.20 (0.65 to 2.19)	0.562	1.07 (0.56 to 2.06)	0.838
Paris	1.70 (0.81 to 3.54)	0.160	2.24 (0.86 to 5.83)	0.100
Gender female	1.57 (1.04 to 2.35)	0.030	1.62 (1.03 to 2.56)	0.038
Age (per additional year)	1.01 (0.99 to 1.03)	0.272	1.01 (0.99 to 1.03)	0.511
Region of origin Europe	Reference		Reference
Africa	1.82 (0.78 to 4.23)	0.165	1.64 (0.66 to 4.05)	0.286
Americas	1.77 (0.90 to 3.46)	0.095	1.97 (0.93 to 4.16)	0.075
Eastern Mediterranean	2.56 (0.91 to 7.25)	0.225	2.13 (0.71 to 6.36)	0.175
South-East Asia	1.12 (0.40 to 3.13)	0.827	0.84 (0.25 to 2.79)	0.773
Western Pacific	1.72 (0.72 to 4.06)	0.220	1.39 (0.55 to 3.48)	0.484
≥1 clinical risk factors	1.24 (0.79 to 1.97)	0.352	1.18 (0.70 to 2.00)	0.533
High self-perceived risk of severe COVID-19	0.89 (0.55 to 1.42)	0.615	0.90 (0.54 to 1.49)	0.681
COVID-19 infection (self)	1.06 (0.60 to 1.88)	0.841	1.01 (0.52 to 1.99)	0.968
COVID-19 infection (household)	0.88 (0.51 to 1.50)	0.637	0.90 (0.47 to 1.70)	0.737
Positive views on vaccination in general	1.39 (0.88 to 2.20)	0.158	1.33 (0.74 to 2.39)	0.336
Positive views on COVID-19 vaccination	1.14 (0.76 to 1.72)	0.518	1.18 (0.71 to 1.98)	0.519
Information through traditional media (TV, radio, web)	1.19 (0.73 to 1.93)	0.494	1.20 (0.69 to 2.11)	0.515
Information through social media	1.29 (0.85 to 1.94)	0.234	1.21 (0.75 to 1.96)	0.427
Information through community network	1.22 (0.76 to 1.97)	0.409	1.00 (0.58 to 1.74)	0.998
Information through other source	2.39 (0.57 to 10.11)	0.236	3.13 (0.70 to 14.08)	0.137

aOR, adjusted OR.

When the analysis was conducted at study site level, the strength of association with covariates associated with perceived availability was different in each location (online supplemental appendix). For instance, Latin American origin in Geneva and information through social media or community network in Paris showed statistically significant associations.

Factors associated with demand for COVID-19 vaccination

Overall, demand for vaccination was associated with a variety of factors (table 6). Before adjustment, living in the USA and Italy, female gender, older age, comorbidity, perception of being at risk of severe COVID-19, positive views on vaccination including COVID-19 and mentioning traditional media as the main source of information were all associated with more chance to demand for the vaccination. On the other hand, living in France and using social media and community networks as the preferred sources of information were negatively associated with demand. After adjustment, increasing age, the presence of comorbidities, and positive views about vaccination in general and COVID-19 in particular were all significantly associated with increased demand for vaccination, while living in France and relying on community network to get informed were associated with lower demand. Of note, the preference for social media lost its significant negative association with demand after adjustment. Although not statistically significant, there was a trend toward more demand among African migrants.

Table 6.

Factors associated with demand for COVID-19 vaccination in regression analysis

	Univariate analysis		Multivariate analysis
	OR (95% CI)	P value	aOR (95% CI)	P value
Study site Geneva	Reference		Reference
Baltimore	2.24 (1.51 to 3.33)	<0.001	0.97 (0.56 to 1.68)	0.920
Milan	1.70 (1.14 to 2.54)	0.009	1.18 (0.66 to 2.09)	0.578
Paris	0.26 (0.15 to 0.47)	<0.001	0.15 (0.06 to 0.38)	<0.001
Gender female	1.43 (1.07 to 1.92)	0.016	1.23 (0.80 to 1.88)	0.344
Age (per additional year)	1.04 (1.02 to 1.05)	<0.001	1.02 (1.00 to 1.04)	0.019
Region of origin Europe	Reference		Reference
Africa	0.75 (0.38 to 1.46)	0.396	2.73 (0.93 to 8.02)	0.069
Americas	1.62 (0.94 to 2.80)	0.085	0.85 (0.36 to 1.96)	0.695
Eastern Mediterranean	0.93 (0.45 to 1.93)	0.852	1.93 (0.63 to 5.86)	0.247
South-East Asia	0.38 (0.15 to 1.01)	0.052	0.45 (0.12 to 1.65)	0.231
Western Pacific	0.90 (0.46 to 1.78)	0.769	0.69 (0.26 to 1.87)	0.467
≥1 comorbidity	1.91 (1.40 to 2.61)	<0.001	1.77 (1.10 to 2.84)	0.018
High self-perceived risk of severe COVID-19	1.46 (1.06 to 2.01)	0.019	1.26 (0.81 to 1.96)	0.315
COVID-19 infection (self)	1.37 (0.92 to 2.05)	0.124	1.23 (0.66 to 2.27)	0.514
COVID-19 infection (household)	1.23 (0.84 to 1.79)	0.292	0.84 (0.48 to 1.49)	0.557
Positive views on vaccination (general)	32.5 (14.2 to 74.4)	<0.001	12.9 (5.17 to 32.22)	<0.001
Positive views on vaccination (COVID-19)	16.70 (11.2 to 24.8)	<0.001	9.70 (6.08 to 15.47)	<0.001
Information through traditional media (TV, radio, web)	2.25 (1.53 to 3.29)	<0.001	1.28 (0.75 to 2.18)	0.360
Information through social media	0.47 (0.35 to 0.62)	<0.001	0.84 (0.55 to 1.28)	0.410
Information through community network	0.47 (0.33 to 0.65)	<0.001	0.61 (0.38 to 1.00)	0.049
Information through other source	0.30 (0.12 to 0.73)	0.008	0.44 (0.13 to 1.43)	0.170
Self-perceived accessibility to COVID-19 vaccination	1.19 (0.78 to 1.81)	0.421	1.08 (0.61 to 1.92)	0.799

aOR, adjusted OR.

In Geneva and Baltimore, positive views about vaccines were strongly associated with demand (online supplemental appendix). In Paris and Milan, the main predictors were the sources of information. Both social media in Milan and community networks in Paris were negatively associated with demand.

Discussion

This study shows that during the early phase of the COVID-19 immunisation programme in four cities in Europe and the USA, most undocumented migrants believed the COVID-19 vaccine would be available to them, but fewer intended to get vaccinated. During this period, participants listed traditional media as the most common source of information, followed by social media and community networks. Although perceptions about vaccination in general were positive, they were much lower for COVID-19 vaccination. We found that factors associated with perceived availability of and demand for COVID-19 vaccination diverged across study sites, reflecting differences in samples, local health policies and cultural preferences. This highlights the importance of collecting data at local level in order to tailor responses. These findings provide insights about the factors underlying vaccine hesitancy among undocumented migrants during the initial phase of the vaccination programme and can help strengthen it as currently ongoing as well as inform the early response for future initiatives. Traditional media appears to play an important role at the early stage and positive views about general immunisation programmes should be leveraged through community engagement and messaging in various languages to address issues of particular concern to undocumented migrants, such as safety of the COVID-19 vaccines, confidentiality and implications on immigration status.

The high confidence in COVID-19 vaccination access among undocumented migrants is telling given their frequent exclusion from many public health benefits. This is reassuring given the legitimate concern that access to vaccination would be limited for this population. Early in the vaccination roll-out, qualitative research among primarily female migrant farmworkers in the USA and migrants with precarious immigration status in the UK showed that misinformation and lack of awareness about entitlements, including access to COVID-19 vaccines, could present substantial barriers to immunisation programmes.45 46 In our study, women were more likely to endorse access than men. This could be related to increased familiarity with the vaccination programmes and overall health system through the use of reproductive health services and as traditional caregivers for children.47 Participants thinking vaccine would not be available to them mentioned the lack of registration within the healthcare system as the predominant reason, more than financial, eligibility or practical issues. This may reflect how migrants in precarious legal situation internalise structural barriers restricting their agency to satisfy their essential needs.48 Of interest, most participants reported hospitals as their preferred place for vaccination. This may reflect concern about vaccine safety requiring specialised care and surveillance and the perception that public hospitals are more accessible and secure regarding the management of personal data than private clinics. Previous studies have indeed shown how migrants used camouflage to avoid detection by immigration authorities and the importance of safe places.49 The gap between accessibility and demand is concerning. One possible explanation might pertain to the timing of the survey. Indeed, in all study locations, the COVID-19 incidence and death toll had sharply dropped by the beginning of the study which may have lessen the feeling of urgency for vaccination. Additionally, at the same time in all four countries, there were widespread public debates about the mRNA-based vaccines’ short-term and long-term safety that may have fuelled hesitancy. Indeed, this may contribute to explain the discrepancy between reported confidence in vaccines in general as compared with COVID-19 vaccines in particular. In future studies, longer period of observation may help identify fluctuation on the perception of the risks and therefore of hesitancy associated with epidemiological fluctuations and the adoption by the population of scientific and lay information about new vaccine technologies. In our study, there was regional variability, with the lowest demand among participants from Paris. Information from community networks tended also to be associated with low demand for vaccination and was more common in Paris, highlighting the need for targeted approaches for different communities. In Paris, the level of literacy (though not measured) may have been lower, given that most respondents could not fill in the questionnaire themselves but had to be helped. This would impact on the potential source of information: information through community networks is more easily accessible in case of language barriers. Also, the second most common source of information was social media, in which content is uncontrolled, opening the debate on how to use social media to harness vaccine hesitancy. Higher demand for vaccination among older people and those with comorbidities is consistent with global trends and may reflect the risk–benefit calculus for people at higher risk of hospitalisation and death from COVID-19. In all four sites, only one-quarter to one-third of participants reported concern about the risk of a severe infection. These low proportions may be related to the overall young age of participants and likely to the comparable proportion of those reporting suffering multiple chronic infections. Interestingly, high self-perceived risk of COVID-19 or prior COVID-19 infection was not associated with demand for vaccination, perhaps because this includes mild cases of the disease.

Intention to get vaccinated against COVID-19 has evolved over time. The successful implementation of large-scale immunisation programmes has encouraged many previously hesitant individuals to get vaccinated, but misinformation and fake news continue to fuel mistrust and slow progress in terms of immunisation coverage in many settings. In our study, only two in five individuals reported they would get vaccinated if the COVID-19 vaccine was offered to them. Although comparison with other groups is difficult due to heterogeneity of methods and timing, hesitancy appears to be higher in our sample compared with the general adult population in the countries studied. For example, in a survey conducted in Italy in December 2020, 82% of adults reported willingness to get vaccinated compared with 52% of our study participants from Milan.50 Similarly, in a survey conducted in France in June 2020, 71.8% of participants reported they would accept vaccination compared with only 14.6% of our Paris participants.51 An international cross-sectional survey conducted between September 2020 and January 2021, however, showed lower intention to get vaccinated among participants from France (49.2%).52 Of note, all these surveys were conducted online, with likely bias towards higher educational and socioeconomic status. Specific data on undocumented migrants are very limited, but in a survey conducted in the USA in late April 2021, 68% of respondents classified as potentially undocumented reported that they had either been vaccinated or planned to get vaccinated.53

This study has several limitations. Participant recruitment was non-random and occurred in health facilities serving undocumented migrants, thereby involving a non-representative sample population of neither the health facilities’ clients nor undocumented migrants at large, and therefore limiting the generalisability of our findings. Specifically, recruitment in healthcare setting may have biased the perception about vaccine accessibility by selecting people with better ability to navigate the healthcare system. Studies conducted in the community would bring important complementary information to our findings. Moreover, differences in sampling strategies and participants’ sociodemographic characteristics imply limitations in comparability among locations. Furthermore, the questionnaire was translated in eight languages and translators were not systematically available during questionnaire administration, hence it is possible that participants speaking a different language had a limited understanding about the questionnaire, thus introducing an information bias and limiting response accuracy. Confidence about access to the COVID-19 vaccine and desire to be vaccinated may differ for undocumented migrants who have not interacted with the health system in their country of residence. Nonetheless, approximately half of respondents in our sample identified lack of health insurance/health card as a major barrier to COVID-19 vaccination. Although concerns about immigration have been shown to dampen healthcare utilisation for COVID-19 services among undocumented migrants,54 we did not specifically ask whether worries about immigration repercussions impacted demand. In our study, public hospitals or clinics were identified as preferred sites for vaccination among those intending to get vaccinated, but we did not collect information about trust in public institutions among vaccine-hesitant participants. Finally, for efficiency purpose, we build the questionnaire using a stringent selection of items previously shown to influence vaccine hesitancy but we cannot claim to cover all areas underlying participants’ assessment of the risk–benefit balance for COVID-19 vaccination.

In summary, our study showed a substantial gap between undocumented migrants’ perceptions about access to COVID-19 vaccines and demand for vaccination. The WHO, UNICEF, the United Nations High Commissioner for Refugees, the ECDC and the Council of Europe have issued recommendations urging access to COVID-19 vaccination to all vulnerable populations, including low-income countries, undocumented migrants and refugees.33 Our results show that building trust and confidence in COVID-19 vaccination is as important as promoting access to tackle hesitancy in this group. Information and promotion of vaccination should particularly focus on men, younger migrants and those with low clinical risks highlighting both individual and collective benefits and reassuring about vaccine safety. Given the marginalisation and criminalisation of undocumented migrants, this may not be simple and requires tailored local solutions.55 Women should be seen as potential key partners in trust-building initiatives promoting vaccination. Our data suggest that during the first phase of a new vaccination programme as for COVID-19, traditional media is an important source of information and communities need to be engaged to leverage existing confidence in general vaccination programmes to reduce hesitancy. Social media plays an important role on how migrants balance risks and benefits and could represent an avenue for disseminating objective information and resources. Community engagement is also important to adequately inform and guide community networks, which can be influential but may undermine vaccination efforts unless equipped with official and verified information. Innovative strategies to foster trust in the equitable access to vaccine for everyone and to ensure a high uptake in all groups through multipronged tailored intervention may help better control the ongoing COVID-19 pandemic. Future research should include the monitoring of hesitancy in this group over longer periods in order to adapt communication strategies and the impact of health promotion interventions using different channels of communication such as social media and community interventions.

Data availability statement

Data are available in a public, open access repository. The dataset that includes all participants’ responses to the questionnaire is available at https://doi.org/10.5281/zenodo.5769319.

Ethics statements

Patient consent for publication

Not required.

Ethics approval

The John Hopkins University (IRB00252774), Geneva Canton (CCER 2021-0246) and the University of Milan-Bicocca (138AQ-38183) ethical boards provided clearance for this survey. In France, the INSERM review board (IRB00003888) considered this study to be exempted of ethical clearance given the nature of the survey. The study was registered with the Office of the Data Protection (DPO) of Sorbonne Paris Nord University. All participants gave oral informed consent to participate.