Determinants of vaccination willingness among people experiencing homelessness: a cross-sectional study in Warsaw, Poland

Abstract

Background

People experiencing homelessness (PEH) have an increased risk of all-cause mortality, higher rates of vaccine-preventable diseases (VPDs), and poorer related health outcomes compared to the general population. At the same time, delivering vaccinations within this vulnerable group is complex, and there is limited evidence to inform practice in this field. This study aims to explore the key determinants influencing vaccination willingness among PEH.

Methods

Cross-sectional survey of PEH at shelters, night shelters, and free-meal sites in Warsaw (Sep 2024–Aug 2025; convenience sampling). Analyses used ETHOS-weighted descriptives (95%CIs) and logistic regression comparing reluctant (unsure or unwilling) vs. willing (secondary: unwilling vs. willing), adjusted for prespecified sociodemographic and homelessness characteristics.

Results

Of 275 PEH surveyed, 215 (78.2%) were male, and 60 (21.8%) were female. Based on FEANTSA (European Federation of National Organisations Working with the Homeless) categories, 162 (58.9%) were classified as Roofless, 85 (30.9%) as Houseless, and 11 (4.0%) as Insecure. Weighted mean age was 52.6 years (95% CI: 50.9–54.2). The majority of the surveyed individuals have completed secondary education (37.1%), followed by vocational (31.6%), primary (20.7%), and higher (9.8%); two participants (0.7%) had not completed primary school.

Overall vaccination willingness was 69.9% (95%CI 63.1–76.0). Unwillingness and uncertainty were 24.0% (95%CI 18.5–30.5) and 6.1% (95%CI 3.4–10.6), respectively. Lower odds of reluctance were observed for secondary/higher education (vs. primary; OR 0.40, 95%CI 0.22–0.76), Roofless individuals (OR 0.46, 95% CI 0.24–0.89), and those experiencing homelessness for a prolonged period of time (per 1-SD, OR 0.59, 95%CI 0.40–0.86). Sex, age, and chronic illness were not significant. Across attitude items, lower perceived effectiveness and lower personal benefit showed the strongest associations with reluctance; safety concerns were modest. As compared to the full sample, reluctant respondents were about two times more likely to endorse low-benefit/low-effectiveness statements, while safety concerns were common in both groups.

Conclusions

About two-thirds of respondents were willing to vaccinate. Reluctance was most strongly associated with lower perceived vaccine effectiveness and personal benefit. Interventions should address these beliefs and prospectively evaluate whether changes to them translate into greater willingness and, ultimately, higher uptake.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12889-026-26291-x.

Background

Homelessness is a significant public health crisis. Currently, in Poland, more than 30 thousand people are experiencing homelessness [1]. People experiencing homelessness (PEH) face many health consequences, ranging from infectious diseases (including vaccine-preventable diseases, VPDs), chronic illnesses, substance abuse disorders, to mental health issues [2], which all lead to higher all-cause mortality [3]. They have a more than twofold increased risk of all-cause mortality compared to the general population, and almost 4 times higher risk of premature death among both men and women [3]. While in the general population the most common causes of death are cardiovascular events and cancer, the PEH mortality profile is influenced by external, preventable factors. Those include substance-abuse disorders (21-times higher risk), drug overdose (11-times higher risk), accidental injuries (13times higher risk), and alcohol-related causes (6-times higher risk) [3].

The physical health of the homeless population is impacted by factors that can be categorized into three domains: communicable diseases, non-communicable diseases, and external factors associated with inadequate housing (such as low temperatures, a wet environment, a lack of appropriate clothing and footwear). Living conditions associated with homelessness, such as staying in congregate settings, poor sanitation, lack of access to hygiene facilities, and higher prevalence of injectable drug use, increase the risks of infectious diseases, e.g., respiratory infections, bloodborne viral infections, or sexually transmitted diseases [4]. According to a meta-analysis, prevalence among PEH ranged from 0.2% to 7.7% for tuberculosis, 3.9% to 36.2% for hepatitis C virus infection, and 0.3% to 21.1% for HIV infection [5]. Even though the prevalence of chronic conditions may not differ significantly from the general population, it is often unrecognized, untreated, or poorly managed among PEH [6]. This leads to more severe complications, higher rates of emergency department use, and preventable hospitalisations [7]. Untreated chronic illness, malnutrition (often found among PEH) [8], and unfavourable external factors have a profound effect on the immune system, which can lead to disproportionately high burden of vaccine-preventable diseases (VPDs) and significantly worse health outcomes upon infection.

For PEH, access to healthcare is insufficient. This can be attributed to barriers such as bureaucratic difficulties (lack of ID documents, no permanent address, and no access to previous medical history) [9], financial constraints and lack of insurance, social stigma from medical professionals [10], and the inflexibility of the healthcare system [11]. In addition, PEH may disregard basic health recommendations, as their needs for food and shelter are more pressing [12]. Homelessness is also linked with distrust towards the healthcare system and healthcare professionals, for example, due to negative past experiences [13]. All of those factors result in overdependence on emergency departments, as it is often the only way to seek medical advice. Drug-, alcohol-, and injury-related presentations dominate the reasons for emergency department visits by PEH [14]. It was also found that PEH are more likely to be discharged without treatment and to be re-admitted within 7 days [15]. The homeless population is often described as “hard-to-reach”, but in reality, the systemic barriers in accessing healthcare make them “hard-to-reach”.

Although the United Nations’ 17 Sustainable Development Goals (SDGs) aim to address the world’s greatest challenges by “leaving no one behind” [16], the issue of homelessness is not explicitly addressed in any of them [17]. However, homelessness is interconnected with many of the SDGs, including eradicating poverty (SDG 1), reducing inequalities (SDG 10), and ensuring good health (SDG 3) [18]. The relationship between homelessness and poor health is often described as complex, varied, and bidirectional [19], as well as a self-perpetuating cycle, where illness leads to homelessness, and homelessness decimates health. However, it is essential to emphasize that many diseases and deaths are preventable, many of them through the use of vaccination. Unfortunately, the homeless population faces barriers to accessing both healthcare and vaccination, which leads to lower vaccination rates. A meta-analysis has shown that the COVID-19 vaccination rate among PEH was 43% for the first dose and even lower for the second and third doses [20]. In comparison, the uptake of the first dose of the COVID-19 vaccine among the general population was estimated at around 70% [21]. However, under-vaccination is not a phenomenon unique to the COVID-19 pandemic. A study has shown that only 23.7% of eligible homeless individuals received their influenza vaccine, compared to 53.2% of the eligible general population [22]. In addition, studies show that vaccine hesitancy, defined as “delay in acceptance or refusal of vaccination despite availability of vaccination services” [23], among PEH is higher than in the general population [24].

Methods

Aim of study

Before implementing strategies to improve vaccine uptake in a given population, it is crucial to thoroughly assess the population and identify areas that require targeted interventions. The study aimed to explore the key determinants influencing vaccination willingness among PEH, in order to inform the design and implementation of effective, evidence-based vaccination strategies tailored to this underserved population.

Study design and participants

This cross-sectional questionnaire study was conducted in Warsaw, Poland, from September 2024 to August 2025 among individuals experiencing homelessness.

Inclusion criteria were as follows: (1) age ≥ 18 years; (2) capacity to provide informed consent; (3) currently experiencing homelessness per FEANTSA/ETHOS working definition (European Typology of Homelessness and Housing Exclusion, developed by FEANTSA – the European Federation of National Organisations Working with the Homeless); and (4) ability to complete the questionnaire independently or with assistance from a trained interviewer. Exclusion criteria were: severe cognitive impairment precluding consent, substantial language or communication barriers, refusal to participate, or inability to complete the survey even with assistance.

Data collection

Participants were recruited via convenience sampling at homeless shelters, night shelters, and free meal services. The survey was administered in Polish. Trained interviewers were instructed to provide only the minimum assistance necessary, limited to reading items verbatim and offering neutral clarifications upon request.

The questionnaire was developed by the research team, drawing on existing literature, prior PEH health surveys, and input from practitioners. The questionnaire comprised five sections:

1. Socio-demographic and homelessness-related variables (based on the Nationwide Study on the Number of Homeless People in Poland [1]): age, gender, citizenship, education, marital status, duration of homelessness, usual sleeping place in the last month: shelter, night shelter, street, other, used to derive ETHOS category: Rooflessness (individuals without a roof over their heads, sleeping on the streets or in nonresidential areas), Houselessness (individuals temporarily residing in shelters, dormitories, or other institutions providing short-term accommodation), Insecure housing (individuals in an unstable housing situation, at risk of eviction, or compelled to leave their residence due to domestic violence), Inadequate housing (persons living in conditions that do not meet housing standards, such as illegal or unsafe premises and severely overcrowded spaces) [25];

2. Health-related factors: self-rated health (5-point scale), chronic conditions (yes/no/unsure), health insurance status, and an open-ended prompt about recent use of healthcare services;

3. Attitudes toward vaccination: thirteen Likert-type statements on effectiveness, safety, access, importance for self and others, benefits of routine vaccination, perceived risk of newer vaccines, trust in information from healthcare workers, following provider recommendations, fear of serious adverse effects, and perceived need for vaccines against rare diseases, including ten questions derived from the Polish version of the 10-item adult Vaccine Hesitancy Scale (aVHS) [26]. This section also included an open-ended question assessing attitudes towards vaccination (‘Do you believe vaccination is necessary? Explain why?’).

4. Vaccination willingness (primary outcome): an open-ended item asked: “Would you get vaccinated if you were offered a free and easily accessible vaccine against a disease that is dangerous for you?” Responses were recorded as free text and coded during data entry as willing, unwilling, or unsure. Analyses were then dichotomized to distinguish between willing and unwilling, with non-responses treated as missing. Optional free-text justifications were retained.

5. Perceived access to vaccination: a question on whether obtaining vaccination is easy or difficult, with space to explain.

The complete questionnaire can be found in the Supplementary Materials.

Outcome

Willingness to vaccinate was coded into three mutually exclusive categories: willing, unsure, and unwilling, based on a single-item measure.

Data analysis

We summarized participant characteristics using counts and ETHOS-weighted percentages or means with 95% CIs. Post-stratification weights matched the FEANTSA/ETHOS distribution from the 2024 count.

Group differences in categorical variables were tested using χ² or Fisher’s exact tests (as appropriate). For comparisons of two groups, we reported differences in weighted proportions with linearisation-based SEs and z-tests. Continuous variables were compared with Welch’s t-test and Mann–Whitney U as robustness checks. Correlations used Pearson or Spearman as appropriate.

We modeled two binary outcomes using survey-weighted logistic regression: (1) unwilling vs. willing (excluding “unsure”) and (2) reluctant—unsure or unwilling vs. willing. Models used post-stratification weights by ETHOS category and HC3 heteroskedasticity-consistent standard errors. Covariates were pre-specified: age, sex, education, duration of homelessness, ETHOS housing category, and chronic illness. Numeric predictors (age, duration) were z-standardized (effects per 1 SD). Education was collapsed into three groups (Primary, Vocational, Secondary/Higher) due to small cell counts. For all categorical predictors, infrequent levels were collapsed (n < 8) to reduce separation; the reference categories were Primary for education and Houseless for the ETHOS category.

Analyses used complete cases with positive weights. We report odds ratios (OR) with 95% CIs and two-sided p-values.

All tests were two-tailed, with p < 0.05 considered statistically significant (for the attitude-item family, q < 0.05 after FDR). Analyses were conducted using Python (pandas, numpy, statsmodels, SciPy, seaborn, and matplotlib for plots); the code is available upon request.

Responses to the open-ended question “Do you believe vaccination is necessary? Explain why?” were analysed using thematic analysis, focusing on perceived barriers and facilitators to vaccination among people experiencing homelessness.

Ethics considerations

Participation was voluntary, and all participants received a small non-cash token (a pack of coffee). Participants were informed about the scope and objective of the study. Verbal informed consent was obtained from all participants prior to their enrolment in this study. The study was approved by the Bioethics Committee of the Medical University of Warsaw (Decision No. AKBE/216/2024, issued on September 9, 2024).

Results

Participant characteristics

Of 275 individuals surveyed, 215 (78.2%) were male, and 60 (21.8%) were female. Almost all respondents (264, 96.0%) reported Polish citizenship, while 11 (4.0%) reported other citizenships (multiple responses allowed). Based on FEANTSA categories, 162 (58.9%) were classified as Roofless, 85 (30.9%) as Houseless, and 11 (4.0%) as Insecure. The remaining 17 (6.2%) either did not report their sleeping place (n = 10, 3.6%) or could not be classified (n = 7, 2.5%). Weighted mean age was 52.6 years (95% CI: 50.9–54.2). Houseless respondents were significantly older (mean: 56.0 years, 95% CI: 53.6–58.4) than Roofless respondents (mean: 48.5 years, 95% CI: 46.5–50.6; p < 0.001). The most common education level completed was secondary (102, 37.1%), followed by vocational (87, 31.6%), primary (57, 20.7%), and higher (27, 9.8%); 2 (0.7%) had not completed primary school. Most respondents were never married (146, 53.1%), followed by those who were divorced (91, 33.1%), widowed (21, 7.6%), and married (17, 6.2%). Participant characteristics are shown in Table 1.

Table 1.

Sociodemographic characteristics of participants (unweighted)

Characteristic	n (%)
Gender	275
Male	215 (78.2%)
Female	60 (21.8%)
Citizenship¹	275
Polish	264 (96.0%)
Other	11 (4.0%)
FEANTSA/ETHOS category	275
Roofless	162 (58.9%)
Houseless	85 (30.9%)
Insecure	11 (4.0%)
Missing/unclassified	17 (6.3%)
Age, years	272
mean ± SD	50.9 ± 12.6
Education	275
Incomplete primary	2 (0.7%)
Primary	57 (20.7%)
Vocational	87 (31.6%)
Secondary	102 (37.1%)
Higher	27 (9.8%)
Marital status	275
Never married	146 (53.1%)
Divorced	91 (33.1%)
Widowed	21 (7.6%)
Married	17 (6.2%)
Duration of homelessness, years	261
Median (IQR)	4 (1–10)
Mean ± SD	7.9 ± 9.9
Self-rated health	275
Good/very good	127 (46.2%)
Poor/very poor	54 (19.6%)
Chronic condition (self-reported)	130 (47.3%)
Health insurance	275
Insured	167 (60.7%)
Uninsured	97 (35.3%)
Unsure	11 (4.0%)

¹Multiple responses allowed; percentages are with respect to the number of respondents; sums may exceed 100%

Length of homelessness was reported by 261 (94.9%) respondents. Due to substantial right skew (skewness = 1.95), we report both median and mean values. Median duration was 4 years (IQR: 1–10), while the weighted mean was 8.3 years (95% CI: 6.9–10.3; range: 0–50 years). Duration did not differ significantly between ETHOS categories (p > 0.05).

Weighted self-rated health responses show that 54.2% of respondents consider their health to be good or very good, while 30.4% report poor or very poor health. Self-rated health was strongly associated with age (r = − 0.353, p < 0.001) but not with housing category after age adjustment (p > 0.05). The self-reported prevalence of chronic health conditions was significantly higher among Houseless individuals (64.7%, 95% CI: 54.0–74.1%) compared to Roofless individuals (39.5%, 95% CI: 32.3–47.2%; p < 0.001). This association remained significant after adjusting for age (aaOR = 2.71, 95% CI: 1.54–4.78, p < 0.001).

The majority (167, 60.7%) reported having medical insurance, 97 (35.3%) were uninsured, and 11 (4.0%) were unsure. Insurance coverage differed significantly by housing status: 85.9% of Houseless vs. 49.4% of Roofless individuals were insured (p < 0.001). ETHOS-weighted insurance coverage was 69.4% (95% CI: 63.3–74.8).

Perceived access to vaccination

Regarding perceived access to vaccination, over half of the respondents (n = 146, 53.1%) found it difficult to get vaccinated or had difficulty answering this question. Among the 29 participants who specified reasons, the most common barriers were: limited access to healthcare facilities/preventive healthcare services (n = 9, 31.0%), financial constraints (n = 7, 24.1%), absence of health insurance (n = 5, 17.2%), low vaccine literacy (n = 5, 17.2%), and a lack of information specifically targeted at PEH (n = 3, 10.3%).

Willingness to vaccinate and its determinants

Of 275 respondents, 257 (93.5%) answered the willingness question. ETHOS-weighted prevalence was: willing 69.9% (95% CI: 63.1–76.0); unwilling 24.0% (95% CI: 18.5–30.5); and unsure 6.1% (95% CI: 3.4–10.6).

Three predictors showed consistent associations across both models (Fig. 1). Individuals with Secondary or Higher education had substantially lower odds of unwillingness than those with Primary education (unwilling model: OR = 0.38, 95% CI: 0.19–0.74, p = 0.004; reluctant model: OR = 0.40, 95% CI: 0.22–0.76, p = 0.005). Roofless individuals had lower odds of unwillingness than Houseless (unwilling: OR = 0.48, 95% CI: 0.24–0.97, p = 0.042; reluctant: OR = 0.46, 95% CI: 0.24–0.89, p = 0.021). Longer duration of homelessness was associated with lower reluctance (per SD: unwilling OR = 0.61, 95% CI: 0.41–0.92, p = 0.019; reluctant OR = 0.59, 95% CI: 0.40–0.86, p = 0.006). Vocational education showed no difference from Primary education in either model. Age, sex, and chronic illness were not significant in either model.

Fig. 1.

Participant characteristics associated with vaccination unwillingness

Responses to vaccination-attitude items were strongly predictive of unwillingness to vaccinate on the single-item outcome (Fig. 2). Lower perceived personal protection from vaccines (OR = 3.25, 95% CI: 2.31–4.57), lower perceived vaccine effectiveness (OR = 2.70, 95% CI: 1.99–3.67), and lower perceived benefits of routine vaccination (OR = 2.53, 95% CI: 1.89–3.38) showed the strongest associations with stating unwillingness (all p < 0.001). Concerns about serious adverse effects showed only a modest association (OR = 1.45, 95% CI: 1.16–1.82, p < 0.05), while the belief that newer vaccines are riskier showed no significant association.

Fig. 2.

Attitude items associated with being unwilling or unsure

Consistent with these adjusted associations, reluctant respondents endorsed benefit- or efficacy-skeptical items roughly twice as often as the full sample; e.g., ‘not important for my health’ 25.5% vs. 11.3%, ‘not important for others’ health’ 23.9% vs. 10.7%, and ‘vaccines are not effective’ 28.2% vs. 13.4%. By contrast, safety concerns were common in both groups (fear of serious side effects 62.6% vs. 52.3%), and ‘new vaccines carry more risks’ showed little difference (36.4% vs 40.0%), as did reported access barriers (18.1% vs 15.7%) (Fig. 3).

Fig. 3.

Vaccination attitudes by willingness status: overall sample vs. the unwilling or unsure subgroup

Perceived barriers and facilitators to vaccination

One hundred ninety-eight out of 275 participants (72.0%) answered the open-ended question “Do you believe vaccination is necessary? Explain why?”. This analysis of respondents’ insights led to the identification of three themes, which were subsequently categorized as either vaccine facilitators (factors promoting vaccination) or vaccine barriers (factors hindering vaccination).

Theme 1. Importance of being vaccinated regarding personal health and the health of others in the community

Participants often recognized the importance of vaccines in maintaining herd protection and protecting others in the community. “Vaccination is necessary to prevent diseases and protect me and people around me”. Some PEH cited specific health conditions as their primary reason for vaccination (“Vaccines are necessary in my case, especially considering my conditions”). Other highlighted benefits, such as avoiding illness to maintain employment (“I care about my health, because when I am not sick, I can work without taking sick leave”), often without mentioning benefits to the community. As a barrier, participants with high self-rated health expressed the belief that they had sufficient disease protection (“I don’t need any vaccinations because I never get sick. I have excellent immunity”). As a facilitator, participants who perceived their health as poor viewed vaccination as a necessary protective measure (“I have a weak body, without vaccination I would have a problem”).

Theme 2. Perceived vaccination effectiveness and safety

Vaccine-positive participants expressed confidence in their effectiveness (“Vaccination is important - it reduces the risk of a disease and I don’t really have space in my life to get sick”). In contrast, vaccine-hesitant participants demonstrated distrust, with views ranging from mild skepticism to belief that vaccines are detrimental to health, for example, by weakening immunity (“I believe that vaccines, to put it bluntly, weaken the body’s natural immunity”). Concerns about the safety of new vaccine technologies, such as mRNA vaccines, were identified as a barrier to vaccination. Some answers mirror online misinformation, including statements such as “They can implant a microchip through vaccination”. For hesitant participants, fear of side effects was a primary barrier, with statements such as “vaccines have more negative than positive effects” or that they “fear the consequences”. However, for pro-vaccine participants, the fear of consequences of not vaccinating was a facilitator - “Without vaccination, I would probably be dead or disabled right now”.

Theme 3. Vaccine literacy and trust in the healthcare system

Although some participants acknowledged the need for adult vaccinations and booster shots (most commonly citing the flu vaccine), a significant number believed that vaccinations are necessary only for children, assuming that this protection persists throughout adulthood - “I don’t need any vaccines because I was vaccinated as a child and I’m definitely fully protected”. Distrust towards the healthcare system, its providers, and the information they share about vaccines was a significant barrier to vaccination for PEH. The intensity of statements varied from statements about medical practice, “Without a correct diagnosis, doctors should not vaccinate me,” to concerns about global health organisations, “WHO is a political corporation”.

Discussion

In this sample of PEH in Warsaw, 69.1% reported willingness to be vaccinated, 24.3% were unwilling, and 6.6% were uncertain (ETHOS-weighted). Our findings suggest that, despite structural and social barriers, a substantial proportion of PEH are receptive to vaccination, indicating an opportunity for targeted public health interventions. Effective strategies to enhance vaccination uptake among PEH include, among others, delivering vaccinations at convenient and accessible locations; administering vaccines at the first point of contact, irrespective of an individual’s vaccination history or serological status (if clinically safe); extending service operating hours; providing staff with training on working with PEH; implementing promotional campaigns for vaccination clinics; incorporating health education initiatives, reminders, incentives, and co-interventions; ensuring no out-of-pocket costs; and fostering collaborative approaches with relevant stakeholders, including PEH themselves [27].

Willingness was higher among individuals classified as Roofless compared with Houseless, those with secondary or higher education (compared with primary), and those with a longer duration of homelessness. By contrast, lower perceived personal protection and lower perceived vaccine effectiveness were strongly associated with unwillingness, whereas concerns about serious side effects showed only a modest association.

Surveys conducted by Kotronia et al. from March 2021 to April 2022 indicated that concerns regarding side effects and the safety of COVID-19 vaccines were the most frequently cited reasons for hesitancy among the Polish general population; however, these concerns diminished over time [28]. Demographic associations (age, sex, employment) were present but shifted across waves of COVID-19. Our study differs in both population and measurement: we assessed willingness to vaccinate with a single item and examined its associations with structured attitude statements, rather than eliciting self-reported reasons. In addition, the study by Kotronia et al. focused solely on the COVID-19 vaccine, whereas our study was not focused on a single type of vaccine. These design differences, combined with our later data collection window, plausibly explain the differences in identified correlates.

Our study found that the primary factors differentiating willing to vaccinate from unwilling to vaccinate respondents were beliefs regarding vaccine effectiveness and personal protection, while Kotronia et al. [28], and Nguyen et al.’s [24] studies identified safety concerns as primary barriers to vaccinate. Ahillan et al.’s scoping review of early-pandemic studies reported hesitancy among PEH commonly in the 35.7–48.0% range, driven by concerns about trial robustness, side effects, and mistrust of government, with evidence concentrated in Europe and North America [29]. Nguyen et al. estimated pooled COVID-19 vaccine acceptance at 66%, increasing to 78% (95% CI 65%-86%) in studies conducted after June 2021, with higher acceptance among men, older adults, and people with chronic conditions. In our ETHOS-weighted, adjusted models, secondary/higher education, Roofless status, and longer duration of homelessness were associated with greater willingness; age, sex, chronic illness, and vocational education were not. Cross-study comparability is limited by heterogeneity in endpoints (uptake vs. intention/willingness) and sampling frames; together with our non-COVID focus and single-item measure, these issues likely account for differences across studies and are also highlighted by Ahillan et al. [29].

Our study has some limitations. First, selection and nonresponse may bias estimates. Recruitment at shelters, night shelters, and meal distribution sites likely leads to underrepresentation of people sleeping rough who avoid services; non-participants may also differ systematically in vaccine attitudes, so willingness could be overstated. Second, weights calibrated to ETHOS housing categories improve representativeness but do not correct for all imbalances (e.g., age/sex or unobserved factors), and the study was conducted only in Warsaw. Findings may not be generalizable to rough sleepers with minimal service contact or to other regions in Poland. Third, willingness to vaccinate was measured with a single, hypothetical item rather than observed vaccine uptake. Single-item measures are sensitive to wording and context and may misclassify nuanced positions; the “unsure” group was small, limiting three-category modeling and motivating the pooled (unsure + unwilling) contrast in primary analyses. Fourth, data were self-reported. Self-rated health and chronic illness can be misclassified - particularly where diagnoses are incomplete among PEH. Some participants of this study required interviewer assistance. Although interviewers were instructed to read items verbatim and offer only neutral clarifications, interviewer effects and social desirability bias cannot be excluded. Furthermore, attitudes towards vaccination may have been influenced by the contemporaneous information environment (e.g., heightened public debate about vaccine safety, including COVID-19). We did not measure disease-specific attitudes, so it is unclear to what extent such concerns generalize to routine adult vaccination. Finally, the cross-sectional design precludes causal inference. Associations between housing situation, duration of homelessness, sex, and willingness may reflect unmeasured confounding or selection processes rather than causal effects; longitudinal studies are needed to disentangle these pathways.

Programmes aiming to increase vaccination rates among PEH in Warsaw should prioritize low-barrier, on-site delivery co-located with services used by PEH (e.g., mobile teams; pop-up clinics at shelters and meal sites). These recommendations are directly supported by our findings on the perceived barriers to getting vaccinated (e.g., financial constraints, lack of health insurance). Furthermore, such approaches have been associated with successful delivery and uptake in PEH programmes and public health practice [27, 30]. Campaigns should clearly convey the expected protection for the individual and their close contacts and address the common belief that adult vaccination is unnecessary. Because these associations are correlational, interventions should be accompanied by prospective monitoring and evaluation to determine whether changes in beliefs lead to increased willingness and, ultimately, higher vaccine uptake. We recommend a brief, repeatable measurement module (the single-item willingness outcome plus a small set of belief items), ETHOS-calibrated weighting, and disaggregation by housing context to track equity. Although we used an open-ended question in this study to capture response depth and conditional willingness, we acknowledge that a standardized multiple-choice/Likert-style question would allow cross-study comparisons. Given that demographic gradients were small after adjustment, there is limited justification for targeting demographics alone; emphasis should remain on addressing modifiable beliefs and verifying their impact over time.

Future work should measure proximal determinants of willingness and vaccine uptake (e.g., healthcare access, perceived discrimination in healthcare, health literacy, digital exclusion) and model their mediation pathways as well as moderation by ETHOS category. It should clearly distinguish among vaccine attitudes, willingness to vaccinate, and observed uptake, and quantify their relationships over time. Second, following Nguyen et al., staff attitudes should be assessed to test whether they affect client willingness and uptake. The influence of site-level norms and practices on client willingness and uptake should also be examined. A validated, PEH-appropriate measurement module should be developed and used consistently in research and practice (including intervention evaluation) to enable comparisons across populations and programmes. Given our design, this could center on the single-item willingness outcome supplemented by a small set of belief items (e.g., effectiveness, personal benefit) to track change.

Conclusions

We studied willingness to vaccinate among people experiencing homelessness in Warsaw using a single-item measure. In adjusted, ETHOS-weighted analyses, lower perceived effectiveness and lower personal benefit were the strongest correlates of reluctance, while safety concerns were only modestly related. Secondary/higher education (vs. primary), Roofless (vs. Houseless) status, and longer homelessness were associated with greater willingness. Sex, age, and chronic illness were not significant.

Health programmes aimed at increasing vaccination uptake among PEH should prioritize communication that addresses effectiveness and personal benefit, and pair these efforts with prospective monitoring to test whether shifts in beliefs translate into higher willingness and, ultimately, uptake. A brief, consistent measurement module (comprising the willingness item and a few belief items) will enable tracking over time and across settings.

Abbreviations

PEH

People experiencing homelessness

VPDs

vaccine-preventable diseases

ETHOS

European Typology of Homelessness and Housing Exclusion

FEANTSA

European Federation of National Organisations Working with the Homeless

95% CIs

95% confidence intervals

SEs

standard errors

HIV

human immunodeficiency virus

χ²

chi-square

SD

Standard deviation

OR

Odds ratio

FDR

false discovery rate

Authors’ contributions

Concept and design: ZM, PS, ZW, KLAcquisition, analysis, or interpretation of data: ZM, PS, ZWDrafting of the manuscript: ZM, PS, ZW, KLCritical review of the manuscript for important intellectual content: ZM, PS, ZW, KLSupervision: KL.

Funding

This work was conducted with financial support from the Medical University of Warsaw under grant “Attitudes, Knowledge, Behaviours, and Access to Vaccinations among People Experiencing Homelessness: A Case Study of the Warsaw Population” (Reference Number: 1M33/2/Z/MG/N/24 and 1M33/2/Z/MG/N/25) Scientific leader: Paulina Maria Szot.

Data availability

The datasets used and/or analyzed during the current study are available from the corresponding author on reasonable request.

Declarations

Ethics approval and consent to participate

This study was conducted in accordance with the Declaration of Helsinki and approved by the Bioethics Committee of the Medical University of Warsaw (Decision No. AKBE/216/2024, issued on September 9, 2024). Participants were informed about the scope and objective of the study. Verbal informed consent was obtained from all participants prior to their enrolment in this study. All methods were carried out in accordance with relevant guidelines and regulations.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.