Vaccine hesitancy in war-torn Ukraine: A machine learning framework for resilient immunization

Vaccine hesitancy in war-torn Ukraine: A machine learning framework for resilient immunization | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Research Article Vaccine hesitancy in war-torn Ukraine: A machine learning framework for resilient immunization Moeen Hamid Bukhari, Iuliia Pavlova, Fedir Lapii, Ubydul Haque This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-7406595/v1 This work is licensed under a CC BY 4.0 License Status: Posted Version 1 posted You are reading this latest preprint version Abstract Background Vaccination prevents millions of deaths, yet conflict magnifies inequities and erodes trust. Ukraine’s war has jeopardized immunization, exposing urgent gaps in vaccine confidence. We conducted a nationwide survey and applied traditional statistics and interpretable machine-learning models to identify drivers of hesitancy. Our approach uncovers complex behavioral determinants and offers actionable, context-specific insights to strengthen vaccination strategies in Ukraine and other crisis settings. Method We conducted a cross-sectional household survey (November–December 2024) across 557 purposively selected hromadas in Ukraine, excluding occupied or frontline areas. Eligible respondents were parents (≥ 18 years) responsible for vaccination decisions for children aged 0–17. Of 4,972 households approached, 2,526 met the inclusion criteria and completed a structured questionnaire administered online. The 135-item instrument integrated validated scales (WHO Vaccine Confidence Scale, Parent Attitudes about Childhood Vaccines) and Ukraine-specific items. Data were analyzed using descriptive statistics, χ² tests, and machine learning models with cross-validation to identify determinants of vaccine confidence and uptake. Results Among 2,526 surveyed Ukrainian parents, 24% perceived vaccines as risky, 44% reported low trust in vaccine information, and only 34.5% of children were fully vaccinated in the past year. Attitudes and uptake were strongly influenced by cultural norms, perceived safety, social media, and healthcare trust. Using machine learning, the BART model identified novel, high-impact predictors of vaccine hesitancy, including perceived risk, lack of safety information, institutional distrust, and urban–rural access gaps. Interaction analyses revealed that these factors synergistically shape parental beliefs, attitudes, and vaccination practices, offering unprecedented insights into determinants of vaccine uptake in conflict-affected settings. Conclusion Our findings reveal that parental vaccine confidence and uptake are shaped by perceived risks, cultural norms, trust in healthcare professionals, and access disparities. Machine learning identified these factors as key predictors, highlighting targets for interventions to address hesitancy and improve equitable childhood immunization in Ukraine. Health Policy Health Economics & Outcomes Research Vaccine Development Health Diseases Russian invasion Public health vaccine hesitancy Ukraine War Civilian Quality of life Figures Figure 1 Figure 2 Introduction Vaccination remains one of the most powerful public-health interventions responsible for saving millions of lives annually. A landmark analysis marking 50 years of the WHO’s expanded programme on immunization estimates that vaccines have averted ~ 154 million deaths, predominantly among infants and young children 1 . These gains, however, are fragile: the COVID-19 era exposed persistent inequities in access and uptake and rekindled hesitancy, showing that scientific advances do not automatically translate into population protection 2 . In conflict settings these gaps widen as infrastructure and supply chains are damaged, populations are displaced, and institutional trust erodes, elevating vaccine confidence from a demand-side concern to a public-health security imperative 3,4 . Vaccine confidence is a decisive determinant of coverage when supply and access exist. Vaccine hesitancy, identified by the World Health Organization as one of the ten threats to global health in 2019, can erode these gains 5 . The WHO SAGE Working Group defines vaccine hesitancy as delayed acceptance or refusal of vaccines despite availability, shaped by the “3Cs” of confidence, complacency, and convenience 6,7 . During emergencies, this confidence gap typically widens as misinformation circulates and polarization grows 7 . The consequences are not confined within borders: measles resurgence and importations repeatedly demonstrate how under-immunized pockets can ignite cross-border outbreaks and strain health services 8–10 . Ukraine is a key case where conflict and displacement undermine vaccination. Before 2022, the country experienced repeated measles crises, with 53,218 cases in 2018 alone, amid fluctuating childhood coverage and low public trust 11 , also a cVDPV2 outbreak was detected in October 2021 12 . On the eve of the invasion, only about 35% of Ukrainians were fully vaccinated against COVID-19, reflecting pre-existing hesitancy and access barriers 13 . War-time destruction, service interruptions, and mass displacement (≈ 6.9 million refugees and ≈ 3.7 million internally displaced as of early 2025) further jeopardized routine and pandemic immunization efforts 14 . Consistent with broader patterns in fragile and conflict-affected settings, fragile and conflict-affected countries have shown persistently lower routine coverage and slower recovery, sustaining regional outbreak risks 3,15 . Despite several studies charting COVID-19 uptake or demographics in Ukraine and among refugees 13,16,17 , a holistic examination spanning vaccine-related knowledge, beliefs, attitudes, practices, and barriers across both adult and childhood immunization during the current war is still missing. The interaction among sociocultural factors, political alignment, and access constraints remains underexplored, and the use of predictive modeling to isolate the strongest drivers of hesitancy in an active conflict is sparse 4,18 . To address this complexity, we pair a large, nationwide survey of Ukrainian adults across war-affected and less-affected territories with complementary machine-learning models to characterize the beliefs, attitudes, practices, and barriers shaping vaccine confidence in wartime. We emphasize interpretable Bayesian Additive Regression Trees alongside Support Vector Machines and Light Gradient Boosting Machines to stress-test predictive performance and rank drivers of hesitancy 19–21 . Unlike traditional regression approaches, modern ML models can uncover non-linear relationships and complex interactions among variables; they provide a bridge between statistical sophistication and practical utility for health policymakers, enabling targeted, transparent, and context-specific interventions 22–24 . While machine learning has been applied to vaccine-confidence prediction in peacetime contexts, its deployment to guide crisis immunization strategy in an ongoing, high-intensity war is, to our knowledge, novel 25,26 . By combining model accuracy with explainability, we aim to generate immediately actionable insights for targeted, context-specific insights for interventions in Ukraine and an approach that can be applied in other crises. Methods Study area Ukraine, in Eastern Europe, is the continent’s second-largest country by land area and seventh most populous, with 38.98 million residents. Average life expectancy is 70 years for men and 79.5 years for women, below the European average. 27,28 The country comprises 1,469 hromadas, 409 urban, 435 suburban/settlement, and 625 rural, with 175 currently under Russian occupation. 29 For this study, 557 hromadas outside front-line and occupied areas will be selected for household recruitment. The war in Ukraine offers a unique lens to explore health and behavioral challenges amid conflict. Since 2022, over 1,000 healthcare facilities have been damaged or destroyed. 30 Coupled with mass population displacement and sharp drops in childhood vaccination, these disruptions have fueled increased epidemic risk for measles, polio, and diphtheria. A total of 557 hromadas were purposively chosen from the Ministry of Health and the State Statistics Service records to maximize geographic coverage while omitting areas affected by occupation or ongoing conflict. The sample reflects differences in population scale, urban–rural distribution, and health system resources, ensuring adequate power and representativeness for nationwide analyses. By January 2024, 79.2% of Ukraine’s population had internet access. 31 Mobile connectivity was broader, with nearly 90% of households estimated to have access to mobile services in the same period. 32 Study design and participants A cross-sectional household survey was conducted to evaluate vaccine hesitancy, acceptance, and coverage across Ukraine. The study included 557 purposively chosen hromadas, ensuring representation across all accessible oblasts while excluding conflict-affected areas. In each hromada, households with children aged 0–17 were identified, and the parent primarily responsible for vaccination decisions was invited to respond, enabling assessment of both behavioral attitudes and practical vaccination barriers. We randomly approached 4,972 parents from distinct households to ensure representation across different population sizes, settlement types (urban, suburban, rural), and health system capacities. Participants were required to reside in the selected hromada during the survey period, be at least 18 years old, and have made a vaccination decision for at least one child in the preceding year. Survey administration and data collection Between November and December 2024, each respondent completed a structured, interviewer-administered questionnaire. TGM Research, which manages a nationally representative survey panel, facilitated data collection. Participants were approached through phone calls, text messages, and emails, and received a pretested questionnaire in Ukrainian. Following informed consent, respondents accessed the survey via a secure hyperlink. By integrating direct interviews with online survey completion, this approach maximized engagement across diverse regions and conflict-affected contexts while minimizing potential sampling bias. Sampling and representativeness Participants were recruited from all 24 oblasts of Ukraine using 2022 census data to ensure national representativeness. Quotas were applied to reflect key demographics (age, sex, and urban/rural distribution) and geographic coverage. Within each hromada, households were selected using probability proportional to size (PPS) sampling. This integrated approach, combining census-informed quotas, spatial diversity, and structured recruitment, enabled the study to represent population-level differences in vaccination attitudes, practices, and access challenges. Outcomes and measures The study’s primary outcomes assessed parental vaccine confidence and trust in health services, utilizing established instruments such as the WHO Vaccine Confidence Scale and the PACV survey, enhanced with context-specific items for Ukraine. 33,34 Secondary outcomes captured patterns of childhood vaccination, including uptake, delays, refusals, perceived obstacles to access, and exposure to misinformation about vaccines. Participants Among the 4,972 households approached, 2,526 parents successfully completed the survey, providing robust representation across age, sex, settlement type, and region, including areas impacted by armed conflict. Eligibility required participants to be 18 years or older, resident in Ukraine during the survey period, and able to respond to survey questions. All participants provided informed consent, ensuring adherence to ethical standards. This design facilitated the collection of reliable, geographically inclusive, and representative data, providing a foundation to investigate how behavioral, social, and systemic factors shape vaccination decisions in war-affected populations. Data collection Data were collected using a structured questionnaire covering the following domains: (i) Demographics and socioeconomic status, which included age, gender, education, employment status, and religious influence; (ii) Vaccine knowledge and perceptions, which included understanding of efficacy and safety, perceived risk, awareness of immunization programs, and exposure to misinformation; (iii) Vaccination history and side effects, which included self-reported COVID-19 doses, experienced side effects, and history of vaccine-preventable infections; (iv) Attitudes and beliefs, which included perceived safety and efficacy, autonomy, historical influences, and social determinants of vaccine decision-making; (v) External and environmental Influences, which included impact of media, religion, politics, socioeconomic status, and conflict exposure; and (vi) Childhood immunization data, which included vaccine coverage by type and age, side effects, and infection history. Survey instrument and measurement of vaccine confidence We administered a 135-item survey designed to assess demographic, psychological, socio-cultural, behavioral, political, and conflict-related factors influencing vaccination choices. Among these, 116 questions used a 6-point Likert scale to quantify key aspects of vaccine confidence—perceived safety and effectiveness, trust in the health system, and social influence. This approach allowed a detailed assessment of vaccine hesitancy, capturing gradations of attitudes beyond simple yes/no measures. Vaccine confidence was quantified using validated constructs customized for the Ukrainian setting, integrating the WHO Vaccine Confidence Scale and PACV survey. 33,34 Responses were aggregated into continuous and categorical scores, facilitating evaluation of both overall confidence and the behavioral drivers underlying vaccination choices. Education, employment, region, vaccine brand, and reported side effects were measured using ordinal and nominal scales, with vaccination status and conflict exposure recorded as binary variables. This multidimensional dataset facilitates a nuanced analysis of the interactions between psychological predispositions, social influence, access constraints, and contextual determinants of vaccine uptake. Statistical analysis Descriptive statistics were calculated for all variables, with continuous variables reported as means ± standard deviation (SD) and categorical variables as frequencies and percentages. Chi-square tests of independence (χ²), Fisher’s exact test, or the Cochran-Armitage trend test, as appropriate, were used to assess differences between groups. Cramér's V was applied to measure the strength of association, categorized as 0–0.1 (weak), >0.1–0.3 (moderate), >0.3–0.5 (strong), and >0.5 (very strong) 35 . For selected variables, post-hoc analyses were conducted to compare groups and identify categories contributing to the χ² results. Machine learning models Machine learning (ML) analysis was performed to model predictors of vaccine hesitancy across five constructs: knowledge, beliefs, attitudes, practices, and structural barriers. Eight ML models were applied, namely Bayesian Additive Regression Tree (BART), Adaptive Boosting (Adboost), Lasso Logistic Regression (LLR), Decision Tree (DT), Artificial Neural Network (ANN), Random Forest (RF), Support Vector Machine (SVM), and Light Gradient Boosting (LGBM). Model selection and development The dataset was categorized into 135 independent and dependent (target) variables for effective modeling. Feature engineering was applied to identify predictors of vaccine hesitancy, with outcomes simplified into binary values for classification. Variables were omitted if they had i) more than 5% missing data, ii) correlations above 0.90, or iii) were marked as unreliable. The final dataset included 86 variables out of 135. We used an 80:20 train-test split, allocating 80% of the data for training and 20% for testing. To mitigate overfitting and obtain more reliable performance estimates, 10-fold cross-validation was applied by dividing the training data into 10 subsets, iteratively training on nine and validating on the remaining one. Model optimization Hyperparameters were optimized using Randomized Search with stratified 5-fold cross-validation for each machine learning model, a robust technique for parameter tuning that helps avoid local optima and identify configurations yielding superior generalization performance. Performance metrics To ensure a balanced assessment of predictive accuracy and reliability, we evaluated model performance using a comprehensive set of metrics. These included the Area Under the Curve (AUC), which measures the ability to rank positive instances above negative ones, with higher values indicating greater discriminative power; accuracy, the proportion of correctly classified instances; sensitivity (true positive rate), the proportion of actual positives correctly identified; specificity (true negative rate), the proportion of actual negatives correctly identified; positive predictive value (PPV), the proportion of true positives among all predicted positives; negative predictive value (NPV), the proportion of true negatives among all predicted negatives; the F1-score, defined as the harmonic mean of precision and recall; and balanced accuracy, the average sensitivity across classes, particularly useful for imbalanced datasets. We also used the Brier score, defined as the mean squared difference between predicted probabilities and actual outcomes, where lower values indicate better performance (0 = perfect accuracy, 1 = poorest predictive performance). Feature importance Based on the selected model, which showed reliable and optimal performance metrics, we assessed feature importance to identify the strongest determinants within each construct. In addition, three variable importance rules were applied: (i) Local threshold: a predictor is included if its inclusion proportion exceeds the 1-α quantile of its null distribution; (ii) Global max threshold: a predictor is included if its inclusion proportion exceeds the 1-α quantile of the distribution of the maximum null variable inclusion proportions across permutations of the response; and (iii) Global SE threshold: using a global multiplier shared by all predictors, a predictor is included if its inclusion proportion exceeds a threshold based on the mean and standard deviation of its null distribution. All analyses were conducted in R version 4.5.1. Results Participant Characteristics A total of 2526 parents responded to the survey, with a mean age of 44.6 ± 13 years. Of these, 46.1% of respondents belong to the 18-44 years age group, 48.3% were 45-64 years, 5.46% were 65-79 years, and only 0.07% were over 80 years old. 58.8% were female compared to males (41.2%). Overall, 59.4% of respondents completed their university degree, and 64.8% were employed. Only 5.1% reported religious influence on vaccination decisions, but 48.1% indicated religion-related vaccine hesitancy in Ukraine (Table 1). Overall, 24.26% respondents perceived vaccines as risky, including 5.62% respondents who understood vaccine efficiency and safety, 18% exposed to vaccine misinformation, 22.3% valued individual autonomy in healthcare decisions, and 16.7% concerned about vaccine access. Overall, 44.22% had a lack of trust in vaccine information, of whom 35.1% understood vaccine efficiency, 29.1% were exposed to misinformation, 41.2% valued individual autonomy, and 37.5% had concerns about vaccine access. Pandemic-related influence on vaccine hesitancy was reported by 31.20%, including 29.9% exposed to misinformation and 26.5% concerned about equitable access (Table S1). Regarding vaccine doses, 14.0% reported side effects after their second COVID-19 dose, while 17.8% respondents experienced post-vaccination COVID-19 infection after two doses (Table S2). The most common vaccination side effects were site discomfort (4.9% for Pfizer) and muscle pain (6.6% for Pfizer) (Table S3). Regarding childhood vaccinations, 34.5% of children had received age-appropriate vaccines since the previous year, with high uptake for birth Hepatitis B (HepB) (20.5%), Diphtheria, tetanus, and pertussis (DTP) at 2 months (21.9%), and Measles, mumps, rubella (MMR) at 12 months (20.5%). All vaccine-specific coverage rates were statistically significant (p<0.0001) (Table 2). Overall, 5.54% of children with a complete vaccination history experienced side effects, while 5.82% reported side effects from vaccinations received in the past year (Table S4). Overall, 50.3% of participants expressed positive attitudes toward vaccination, 8.9% negative, and 40.8% neutral. Positive attitudes were strongly associated with beliefs in vaccine safety (p < 0.0001, Cramer’s V = 0.293) and infection prevention (p < 0.0001, Cramer’s V = 0.313), while 34.8% reported concerns about serious side effects. Negative attitudes were linked to fears of harmful ingredients and a preference for natural immunity. External influences included social media (11.6% positive, 46.7% negative; p < 0.0001, Cramer’s V = 0.098) and religious leaders (2.5% positive, 45.8% negative; p < 0.0001, Cramer’s V = 0.211) (Table 3). Concerns about safety and risks were reported by 7.21% of those with negative vaccine attitudes (Table 8). Broader concerns about vaccine side effects were associated with political ideology, perception of vaccines (63.1%), safety concerns (72.8%), social influence (20.7%), cultural norms (33.6%), and trust in healthcare workers (23.9%) or doctors (28.5%) (Table S5). Reluctance to receive future vaccines was linked to personal health beliefs influencing vaccine perception (21.8%) and to lack of local access or affordability (23.2%). Low trust in healthcare professionals’ recommendations was significantly associated with personal vaccine beliefs (34.8%) and inadequate access (43.9%) (Table S6). The respondents who perceived vaccine hesitancy in the community related to the factors such as urban–rural disparities in vaccine access (16.0%), lack of access to a trustworthy place (25.6%), influence of perceptions about vaccines (27.2%), local accessibility issues (34.1%), and cultural and societal influences (17.2%) (Table S7). Cultural and societal beliefs influencing vaccine hesitancy were related to vaccination history (72.2%), the impact of personal health beliefs (44.0%), and urban–rural disparities (20.0%) (Table S8). Environmental factors influencing vaccine hesitancy included regional gaps in vaccine availability (p < 0.0001, Cramer’s V = 0.201) and challenges in accessing healthcare facilities (p < 0.0001, Cramer’s V = 0.140) (Table S9). High trust in the Ukrainian healthcare system was associated with greater confidence in vaccine policies, equitable vaccine access, and willingness to be vaccinated in the future. In contrast, residence in conflict-affected areas was modestly associated with lower trust (p = 0.043) (Table S10). Machine Learning Analysis The performance of all machine learning models was evaluated using accuracy, sensitivity, specificity, PPV, NPV, F1 score, balanced accuracy, and Brier score. For the vaccine knowledge model, BART model has the highest accuracy (0.78), specificity (0.75), PPV (0.75), NPV (0.80), F1 score (0.78), balanced accuracy (0.78), and the lowest Brier score (0.16), while LLR had the highest sensitivity (0.87) (Figure S1A). For the belief model, SVM, LGBM, and BART all reached the highest accuracy (0.73), with BART showing the highest sensitivity (0.73) and LGBM the highest specificity (0.76) (Figure S1B). The highest F1 scores were observed in adaptive boosting, RF, and SVM, while the lowest Brier scores were reported for all models except decision tree, neural network, and random forest. For the attitude, practice, and barrier models, the highest model accuracy was obtained with BART at 0.74, 0.75, and 0.71, respectively, with corresponding low Brier scores of 0.18, 0.16, and 0.19 (Figure S1C–E). Overall, validation results indicate that the BART model consistently demonstrated superior performance across all models (Figure S1). The ROC curves with AUC values and corresponding 95% bootstrap confidence intervals are presented in Figure 1. The BART model achieved higher AUC values than other machine learning approaches across all five settings (Figure 1A–E). The inclusion proportion for a predictor represents the fraction of times it is selected as a splitting rule relative to the total number of splitting rules across posterior draws in the sum-of-trees model. In the knowledge model, the top five predictors were Perceived Risk of Vaccines, awareness of general immunization vaccines, pandemic influence, concerns about equitable vaccine access, and historical influences on vaccination attitudes (Figure S2, 1A). For the belief model, the most important variables were lack of information on vaccine safety, concerns about serious vaccine side effects, perceptions of dangerous ingredients in vaccines, preference to delay COVID-19 vaccination, and vaccine safety perception (Figure S2, 2A). The most important features for the attitude model were cultural norms and vaccination perceptions, lack of institutional trust issues, lack of trusted sources for vaccine information, impact of social media on vaccine choice, and religious influence on vaccination decisions (Figure S2, 3A). For the practice model, the key predictors were willingness for future vaccination, impact of previous vaccination status, sharing knowledge on COVID-19 variants, influence of health beliefs on vaccines, and previous adverse vaccine experiences (Figure S2, 4A). In the barrier model, the most important predictors were lack of trust in healthcare professionals, lack of healthcare quality and transparency, vaccine hesitancy in the community, financial constraints, and urban–rural vaccine access gaps (Figure S2, 5A). A similar set of important predictors was identified using the local procedure and the Global Max and Threshold Procedures. In the knowledge model, perceived risk of vaccines, lack of awareness of general vaccines immunization, concerns about equitable vaccine access, and pandemic influence exceeded the green line threshold (Figure S2, 1B). In the belief model, key predictors included lack of information on vaccine safety, concerns about serious vaccine side effects, and perceptions of dangerous ingredients in vaccines (Figure S2, 2B). For the attitude model, cultural norms, institutional trust issues, social media impact on vaccine decisions, safety concerns, religious leaders’ opposition to COVID-19 vaccination, and economic factors were the most important features (Figure S2, 3B). In the practice model, the top predictors were lack of trust in healthcare professionals, COVID-19 variant knowledge sharing, past vaccination status, and impact of health beliefs on vaccines (Figure S2, 4B). For the barrier model, the most influential predictors were lack of trust in healthcare professionals, lack of healthcare quality and transparency, vaccine hesitancy in the community, and limited urban–rural vaccine access (Figure S2, 5B). Global Max and Threshold procedures identified the strongest predictors across models. Perceived risk of vaccines was the top predictor in the knowledge model (Figure S2, 1C); lack of information on vaccine safety in the belief model (Figure S2, 2C); and cultural norms in the attitude model (Figure S2, 3C). In the practice and barrier analyses, Lack of trust in healthcare professionals, Lack of healthcare quality and transparency, and vaccine hesitancy in the community were the only variables to exceed both the Global SE and Global Max thresholds, indicating they are the most important predictors of vaccine hesitancy (Figure S2, 4C-5C). The average relative importance of all significant predictors with 95% confidence intervals is shown in Figure 2. In the knowledge model, the top three interaction effects were between perceived risk of vaccines and awareness of general immunization vaccines, perceived risk of vaccines and pandemic influence on vaccine hesitancy, and autonomy’s influence in healthcare and concerns about equitable vaccine access (Figure 2A). For the belief model, key interactions included perceptions of dangerous ingredients in vaccines and lack of information on vaccine safety , concerns about serious vaccine side effects and lack of information on vaccine safety, and perceptions of dangerous ingredients in vaccines and concerns about serious vaccine side effects (Figure 2B). In the attitude model, the most important interactions were social media impact on vaccine decisions and safety concerns, institutional trust issues and economic factors, and safety concerns and political ideology (Figure 2C). For the practice model, the strongest interactions involved impact of past vaccination status and lack of trust in healthcare professionals, impact of COVID-19 variant knowledge sharing and lack of trust in healthcare professionals, and lack of trust in healthcare professionals and negative impact of health beliefs on vaccines (Figure 2D). In the barrier model, the leading interactions were lack of healthcare quality and transparency and vaccine hesitancy in the community, lack of healthcare quality and transparency and financial constraints, and lack of trust in healthcare professionals and financial constraints (Figure 2E). Discussion This study offers one of the most comprehensive wartime portraits of vaccine attitudes, beliefs, and behaviors in Ukraine, integrating a large behavioral survey with interpretable machine learning. Just over half of respondents held positive views of vaccination (50.3%), while 40.8% were neutral and 8.9% negative. Across multiple predictive domains, Bayesian Additive Regression Trees (BART) outperformed other algorithms and illuminated high-impact predictors and their interactions, sharpening a picture of the social, structural, and informational landscape that shapes vaccination decisions 19,22 . Our findings underscore how prolonged conflict aggravates vaccine hesitancy by both disrupting services and eroding public trust. In this wartime survey of Ukraine, lack of trust in healthcare professionals and institutions emerged as one of the most influential barriers to vaccination, alongside misinformation and concerns about vaccine safety, and these patterns consistent with broader literature linking trust and misinformation to lower vaccine acceptance 36 . This aligns with evidence from other conflict settings where war amplifies immunization gaps and fuels hesitancy 4,18 . Our findings provide Ukraine-specific confirmation of broader guidance that confidence and demand-side factors are as critical as supply in crises. Restoring vaccine confidence requires transparency, consistent risk communication, and community engagement in line with WHO SAGE recommendations on the “3Cs” of confidence, complacency, and convenience 7 . Cultural norms (e.g., preferences for “natural immunity”), religious influence, and political identity exerted indirect effects that often interacted with perceived safety and transparency. These patterns align with regional and global work showing that socio-political identity and norms condition vaccine confidence and demand 37 . Information pathways also mattered: nearly half of respondents citing social media reported negative influence (46.7%). This accords with evidence that exposure to vaccine misinformation depresses intent and uptake, particularly when messages appear “scientific” or circulate in polarized networks 38 . Misinformation-related beliefs, such as fears of dangerous ingredients and doubts about vaccine safety, were among the most important predictors in our belief model, emphasizing the urgent need for tailored counter-messaging strategies. Hesitancy was highest where people perceived poor care quality, high out-of-pocket costs, and limited nearby vaccination services, showing that structural barriers and demand factors are closely linked. Guidance for humanitarian settings recommends keeping services continuous, close to communities, affordable, and culturally appropriate to sustain coverage during crises 39 Ukraine’s program showed resilience despite war: by late 2024, over 80% of children were vaccinated against most infections, a return toward pre-war levels 40 . Coverage still fell short of the ~ 95% threshold for herd protection, and by early 2025 nine measles outbreaks were recorded. While early-life vaccines (for example, birth HepB and MMR at 12 months) had relatively high uptake, only 34.5% of children were fully up to date in the prior year. Nearly half of participants reported community-level hesitancy driven by misinformation, low trust, and convenience barriers - pressures that often worsen during war. These results suggest recovery is robust but fragile, depending on both reliable delivery and renewed confidence and equity across regions 40 . A key implication is the need for targeted, context-aware strategies to bolster vaccine confidence in conflict settings. Our models highlighted belief/attitude factors (e.g., fear of side effects, influence of social networks, and some cultural norms) that predict hesitancy. any of these are modifiable through public health interventions, for instance, community education to counteract misinformation, engagement with local leaders to allay cultural or religious concerns, and visible endorsement of vaccines by trusted healthcare providers 7,41 . Notably, we found that residence in high-conflict areas was associated with slightly lower trust in the health system (p = 0.043), suggesting the war’s direct toll on institutional confidence. This finding aligns with reports that sustained conflict can weaken trust even in previously strong health systems, emphasizing that rebuilding trust should be a priority in Ukraine’s health response 7,18 . Methodologically, this study demonstrates the value of interpretable machine learning in crisis settings. BART offered strong accuracy and calibration across domains, and helped to uncover non-linear relationships and interactions that standard models often miss 22,24 . High risk perception deterred acceptance most when people had low baseline vaccine knowledge. Fears about severe side effects combined with worries about ingredients were more discouraging than either concern alone. Low trust in clinicians plus a prior negative vaccination experience also strongly predicted reluctance to return. Together, these patterns support precision public health that targets clusters of co-occurring risks 25 . These results suggest a multi-level strategy: address structural barriers (urban–rural access gaps, affordability, service quality/transparency) with mobile brigades, extended hours, and robust cold-chain; execute myth-specific communications elevating trusted messengers (clinicians, community leaders); and target high-risk clusters identified by ML (e.g., low trust and poor access) for co-delivered services and counseling 7,41 . Strengths of the study include a large diverse wartime sample; integration of behavioral, cultural, and structural determinants; and interpretable ML that translates prediction into policy-relevant insight. However, limitations must be acknowledged. Limitations include cross-sectional design (no causal inference), potential self-report bias, and further geographic stratification could enhance understanding of regional variations, especially in frontline versus non-frontline areas. Additionally, although our ML models identified strong associations, real-world interventions will require careful testing to confirm the impact of targeting the identified predictors. Given the ongoing conflict and shifting vaccine attitudes, future work should prioritize longitudinal panels to track changes in hesitancy and evaluate intervention impact. Integrating real-time social media listening into hesitancy surveillance can enable rapid counter-messaging as new myths emerge, while experimental designs, including randomized controlled trials, can test interventions tailored to machine-learning–identified risk clusters. Extending this combined survey and machine-learning approach to other fragile or conflict-affected settings will test generalizability and refine precision public health strategies, helping advance goals to reach zero-dose and under-immunized children in FCV contexts. Conclusion Our comprehensive analysis of 2,526 Ukrainian parents highlights the complex determinants of vaccine confidence and uptake amid a conflict-affected context. While 50% of respondents expressed positive vaccination attitudes, nearly one-quarter perceived vaccines as risky, and 44% reported low trust in vaccine information. Childhood vaccine coverage was moderate, with 34.5% of children receiving age-appropriate immunizations in the prior year, and reported side effects were generally mild. Attitudes were shaped by beliefs in vaccine safety and infection prevention, while negative perceptions were linked to fears of harmful ingredients, preference for natural immunity, and external influences such as social media and religious guidance. Machine learning models, particularly BART, identified perceived vaccine risk, lack of safety information, cultural norms, and trust in healthcare professionals as the most predictive factors for vaccine knowledge, beliefs, attitudes, practices, and barriers. Interactions between misinformation, institutional trust, social media influence, and access disparities further exacerbated hesitancy. These findings underscore the need for multifaceted interventions targeting knowledge gaps, societal norms, healthcare trust, and equitable access to improve vaccine acceptance in vulnerable populations. Abbreviations KNOWQ1: Perceived Risk of Vaccines; KNOWQ3: Awareness of General Immunization Vaccines; KNOWQ8: Pandemic Influence; KNOWQ11: Concerns About Equitable Vaccine Access; KNOWQ9: Historical Influences on Vaccination Attitudes; KNOWQ10: Importance of Autonomy in Healthcare Decisions; KNOWQ6: Exposure to Vaccine Misinformation.ATTQ4: Lack of Information on Vaccine Safety; ATTQ3: Concerns About Serious Vaccine Side Effects; ATTQ2: Perceptions of Dangerous Ingredients in Vaccines; ATTQ5: Preference to Delay COVID-19 Vaccination; ATTQ1: Vaccine Safety Perception; ATTQ8: Belief in Natural Immunity over Vaccination; ATTQ6: Perceived Protection from Vaccines; ATTQ7: Belief in Alternatives to Vaccination; ATTQ16: Cultural Norms and Vaccination Perceptions; ATTQ18: Lack of Institutional Trust Issues; ATTQ19: Lack of Trusted Sources for Vaccine Information; ATTQ9: Impact of Social Media on Vaccine Choice; ATTQ15: Religious Influence on Vaccination Decisions; ATTQ11: Religious Leaders’ Stance Against COVID-19 Vaccination; ATTQ21: Influence of Political Ideology; ATTQ23: Financial Factors Influence Vaccination Decision; ATTQ20: Religious Factors Influence on Vaccination Opinions; ATTQ24: Other Factors in Vaccination; ATTQ10: Influence of Media Reports on COVID-19 Vaccination; ATTQ22: Influence of Social Influence; ATTQ17: Familial and Peer Impact on Vaccination. PRACQ9: Willingness for Future Vaccination; PRACQ8: Impact of Previous Vaccination Status; PRACQ3: Sharing Knowledge on COVID-19 Variants; PRACQ6: Influence of Health Beliefs on Vaccines; PRACQ5: Previous Adverse Vaccine Experiences; PRACQ2: Travel Abroad During COVID-19; PRACQ1: Avoidance of Crowded Places During COVID-19; PRACQ4: Comfort in Sharing COVID-19 Knowledge; PRACQ7: Impact of Socioeconomic Status on Vaccination; PRACQ12: Lack of Trust in Healthcare Professionals; PRACQ13: Lack of Healthcare Quality and Transparency; PRACQ00: Vaccine Hesitancy in the Community; PRACQ15: Financial Constraints; PRACQ17: Urban–Rural Disparities in Vaccine Access; PRACQ19: Environmental Impact on Vaccine Perceptions; PRACQ14: Influence of Government Vaccine Policies. Declarations This study was approved by Poltava Medical University, Ukraine (No. 227 dated 5/23/2024), and Rutgers University (#Pro2024001126). Written informed consent was obtained from all participants, in addition to which all data collection was performed in accordance with relevant guidelines and regulations in Ukraine and the USA. Acknowledgement : We sincerely appreciate TGM Research for facilitating data collection for this study. Author contributions : All authors made equal contributions to this study. ; Declarations of conflict : None declared. Data availability : Deidentified, delinked datasets can be obtained from the corresponding author upon request. References Shattock, A. J. et al. Contribution of vaccination to improved survival and health: modelling 50 years of the Expanded Programme on Immunization. The Lancet 403 , 2307-2316 (2024). WHO. Global immunization efforts have saved at least 154 million lives over the past 50 years. World Health Organization. [2024, April 24] , (2024). Jones, C. E. Routine vaccination coverage—worldwide, 2023. MMWR. Morbidity and mortality weekly report 73 (2024). Ciccacci, F. et al. Between war and pestilence: the impact of armed conflicts on vaccination efforts: a review of literature. Frontiers in Public Health 13 , 1604288 (2025). WHO. Ten threats to global health in 2019. World Health Organization [2019, January 10] (2019). MacDonald, N. E. Vaccine hesitancy: Definition, scope and determinants. Vaccine 33 , 4161-4164 (2015). SAGE, W. WHO SAGE Working Group on Vaccine Hesitancy. Report of the SAGE Working Group on Vaccine Hesitancy. (2014, November 12). (2014). Patel, M. et al. National update on measles cases and outbreaks—United States, January 1–October 1, 2019. American Journal of Transplantation 20 , 311-314 (2020). Zucker, J. R. et al. Consequences of undervaccination—measles outbreak, New York City, 2018–2019. New England Journal of Medicine 382 , 1009-1017 (2020). CDC/WHO. Centers for Disease Control and Prevention, & World Health Organization. (2023, November 16). Global measles threat continues to grow as another year passes with millions of children unvaccinated [Joint news release]. (2023). WHO. World Health Organization. Measles – European Region [Disease outbreak news]. (2019, May 6) , (2019). WHO/Europe. World Health Organization, Regional Office for Europe. (2023, September 21). Polio outbreak in Ukraine closed – A success story for public health despite extreme challenges of war [News release]. , (2023). Costantino, V. & MacIntyre, C. R. Impact of vaccine coverage and disruption to health services on COVID-19 in Ukraine. Scientific Reports 14 , 14729 (2024). UNHCR, U. f. Ukraine refugee crisis: Aid, statistics and news. Retrieved August 17, 2025, from (2025). UNICEF. Measles cases across Europe continue to surge, putting millions of children at risk [Press release]. ( May 28, 2024) , (2024). Korolchuk, O., Vasiuk, N., Klymkova, I., Shvets, D. & Piddubnyi, O. COVID-19 vaccination under conditions of war in Ukraine. Asian bioethics review 15 , 259-281 (2023). Ganczak, M. et al. “Why would we?” A qualitative study on COVID-19 vaccination decision making among Ukrainian economic female migrants in Poland. Frontiers in Public Health 12 , 1380627 (2024). Dwyer, H., Enria, L., Palmer, J., Ayub, S. & Beckmann, N. Trust, Information and Vaccine Aonfidence in Crisis Settings: A Scoping Review. Public Health Challenges 4 , e70073 (2025). Chipman, H. A., George, E. I. & McCulloch, R. E. BART: Bayesian additive regression trees. (2010). Cortes, C. & Vapnik, V. Support-vector networks. Machine learning 20 , 273-297 (1995). Ke, G. et al. Lightgbm: A highly efficient gradient boosting decision tree. Advances in neural information processing systems 30 (2017). Tan, Y. V. & Roy, J. Bayesian additive regression trees and the General BART model. Statistics in medicine 38 , 5048-5069 (2019). Chen, T. & Guestrin, C. in Proceedings of the 22nd acm sigkdd international conference on knowledge discovery and data mining. 785-794. Lundberg, S. M. & Lee, S.-I. A unified approach to interpreting model predictions. Advances in neural information processing systems 30 (2017). Lincoln, T. M. et al. Taking a machine learning approach to optimize prediction of vaccine hesitancy in high income countries. Scientific reports 12 , 2055 (2022). Carrieri, V., Lagravinese, R. & Resce, G. Predicting vaccine hesitancy from area‐level indicators: A machine learning approach. Health Economics 30 , 3248-3256 (2021). Life expectancy at birth, total (years) - Ukraine. (accessed April 8 2022). ( Number of Present Population of Ukraine, as of January 1. 2021. , ( Institute for the Study of War 2024. Fact Sheet: The Kremlin's Occupation Playbook: Coerced Russification and Ethnic Cleansing in Occupied Ukraine. , <Available at [https://www.understandingwar.org/sites/default/files/Fact%20Sheet%20%20--%20The%20Kremlin%27s%20Occupation%20Playbook-Coerced%20Russification%20and%20Ethnic%20Cleansing%20in%20Occupied%20Ukraine%202.pdf], last accessed 08.22.2024.> ( Haque, U. et al. A Comparison of Ukrainian Hospital Services and Functions Before and During the Russia-Ukraine War. Jama-Health Forum 5 , e240901, doi:10.1001/jamahealthforum.2024.0901 (2024). DataReportal. Digital 2024: Ukraine. , (2024). International Telecommunication Union (ITU) and Freedom House. Freedom on the Net 2024: Ukraine. Accessed August 17, 2025. , ( Larson, H. J. et al. Measuring vaccine hesitancy: The development of a survey tool. Vaccine 33 , 4165-4175, doi:10.1016/j.vaccine.2015.04.037 (2015). Opel, D. J. et al. The relationship between parent attitudes about childhood vaccines survey scores and future child immunization status: a validation study. JAMA Pediatr 167 , 1065-1071, doi:10.1001/jamapediatrics.2013.2483 (2013). Scheffé, H. H. cramér, mathematical methods of statistics. (1947). Adhikari, B., Cheah, P. Y. & von Seidlein, L. Trust is the common denominator for COVID-19 vaccine acceptance: a literature review. Vaccine: X 12 , 100213 (2022). Karakulak, A. et al. Trust in government moderates the association between fear of COVID-19 as well as empathic concern and preventive behaviour. Communications psychology 1 , 43 (2023). Pierri, F. et al. Online misinformation is linked to early COVID-19 vaccination hesitancy and refusal. Scientific reports 12 , 5966 (2022). WHO. Vaccination in humanitarian emergencies: Implementation guide (WHO/IVB/17.13). Geneva, Switzerland: World Health Organization., (2017). UNICEF/Ukraine. Ministry of Health of Ukraine; WHO/Ukraine. (2025, April 24). Vaccination is both personal protection and a contribution to collective immunity (Joint statement)., (2025). OECD. Enhancing public trust in COVID-19 vaccination: The role of governments. , (2021). Tables Table 1: : Demographic characteristics of Participants Characteristics Total n =2526 Age, year a 44.6 ± 13 Age group 18–44 years 1165 (46.1) 45–64 years 1221 (48.3) 65–79 years 138 (5.46) ≥80 years 8 (0.07) Gender Male 1041 (41.2) Female 1485 (58.8) Education attainment Junior high or middle school Professional (vocational) education 750 (29.7) Grade school Complete general secondary education 245 (9.7) Post-college (graduate school) Academic degree 30 (1.2) University degree 1501 (59.4) Employment Status Employed 1638 (64.8) Un-Employed 228 (9.0) Student 54 (2.1) Retired 299 (11.8) Houseworker or Other 307 (12.15) Religious Influence on Vaccination Yes 130 (5.1) No 2292 (90.7) Unsure 104 (4.1) Religious Vaccine Hesitancy in Ukraine Yes 1226 (48.1) No 471 (18.6) Unsure 829 (32.8) a mean ±standard deviation Table 2: Age‑Specific Vaccination in children by Vaccine Type Age Vaccines Total vaccinated Since last year 870 (34.5%) Counts (%) p-value p-value Birth Hepatitis B (HepB) 365 (20.51) <0.0001 a <0.0001 c 2 months Hepatitis B (HepB) 310 (17.42) <0.0001 a <0.0001 c Diphtheria, tetanus, pertussis (DTP) 390 (21.91) <0.0001 a Haemophilus influenzae type b (Hib) 221 (12.42) <0.0001 Inactivated poliovirus (IPV) 250 (14.04) <0.0001 a 4 months Diphtheria, tetanus, pertussis (DTP) 365 (20.51) <0.0001 a <0.0001 c Haemophilus influenzae type b (Hib) 203 (11.4) <0.0001 b Inactivated poliovirus (IPV) 233 (13.09) <0.0001 a 6 months Diphtheria, tetanus, pertussis (DTP) 335 (18.82) <0.0001 a <0.0001 c Haemophilus influenzae type b (Hib) 198 (11.12) <0.0001 b Live poliovirus (bOPV) 220 (12.36) <0.0001 b Hepatitis B (HepB) 286 (16.07) <0.0001 a 12 months Measles, mumps, rubella (MMR) 365 (20.51) <0.0001 a <0.0001 c Haemophilus influenzae type b (Hib) 197 (11.07) <0.0001 b 18 months Diphtheria, tetanus, pertussis (DTP) 305 (17.13) <0.0001 b <0.0001 c 18 months Live poliovirus (bOPV) 201 (11.29) <0.0001 b 6 years Diphtheria, tetanus (DT) 239 (13.43) <0.0001 a <0.0001 c Live poliovirus (bOPV) 162 (9.1) <0.0001 Measles, mumps, rubella (MMR) 273 (15.34) <0.0001 a 14 years Live poliovirus (bOPV) 121 (6.8) <0.0001 b <0.0001 c 16 years Diphtheria, tetanus (Td) 155 (8.71) <0.0001 b <0.0001 c Counts, percentages a Fisher exact test . b Pearson’s chi-square test, c Cochran-Armitage trend test. Table 3: Attitudes Across Vaccine Safety, Efficacy, Hesitancy, and Influence Overall Attitude about vaccine Positive 1271(50.3) Negative 225(8.9) Neutral 1030(40.8) p-value a Cramer's V b Vaccine Safety Beliefs Vaccines are safe 536(42.2) 195(86.7) 379(36.8) <0.0001 0.293 Vaccines contain harmful ingredients 164(12.9) 44(19.6) 379(36.8) <0.0001 0.295 Information & Novelty Hesitancy Concern about serious side effects 442(34.8) 28(12.4) 193(18.7) <0.0001 0.204 Insufficient vaccine information 322(25.3) 77(34.2) 282(27.4) <0.0001 0.133 Vaccine Efficacy Perceptions Prefer to wait on new vaccine 268(21.1) 54(24.0) 290(28.2) <0.0001 0.226 Vaccines prevent infection 801(63.0) 159(70.7) 312(30.3) <0.0001 0.313 Preference for Alternative Preventive Measures 99(7.8) 36(16.0) 398(38.6) <0.0001 0.248 Prefer immunity from infection 163(12.8) 80(35.6) 359(34.9) <0.0001 0.182 External Influences on Vaccination Decisions Influenced by social media reports 148(11.6) 105(46.7) 353(34.3) <0.0001 0.131 Influenced by mainstream media 203(16.0) 106(47.1) 352(34.2) <0.0001 0.098 Influenced by religious leaders 32(2.5) 103(45.8) 342(33.2) <0.0001 0.211 Counts, percentages, a chi-square test, b estimated Cramer’s V Additional Declarations The authors declare no competing interests. Supplementary Files 4.Supplementary.docx supplementary file Cite Share Download PDF Status: Posted Version 1 posted You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-7406595","acceptedTermsAndConditions":true,"allowDirectSubmit":true,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":502441791,"identity":"526cb8d6-6426-474f-9313-c159918b8912","order_by":0,"name":"Moeen Hamid Bukhari","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAABEElEQVRIiWNgGAWjYLACxgYGBjaGHAbGRhCDHUgwGFiQooXnAEiLBGEtDDAtDBIJIB5uLbrtzQcfF+6wi+Zjzz0mOXOHXT6f5POrG34USDDwt3cnYNNiduZYsvHMM8m5bTzv0iQ3nkm2bJPOKbvZA3SYxJmzG7BquZFjJs3bxpzbJpFjJvmwjdmATTon7QYPUIuBRC4OLfnff/O21cO01BuwSZ5Ju/kHr5YcNmbetsMQLRvbDhuwSbAfu43XljPHjIEOOw7yS7LlzLbjBmw8OWy3ZQwkeHD65Xjzw8+8bdW589tzD97sbas2kG8//uzmmz82cvztvVi1YAM8BmCSWOUgwP6AFNWjYBSMglEw/AEAqndja88uFgQAAAAASUVORK5CYII=","orcid":"","institution":"Department of Statistics, Quaid-i-Azam University, Islamabad, Pakistan","correspondingAuthor":true,"prefix":"","firstName":"Moeen","middleName":"Hamid","lastName":"Bukhari","suffix":""},{"id":502441792,"identity":"7ed897f0-e116-4812-aa25-44bd4c9bbc8c","order_by":1,"name":"Iuliia Pavlova","email":"","orcid":"","institution":"Theory and Methods of Physical Culture Department, Lviv State University of Physical Culture, Lviv, Ukraine","correspondingAuthor":false,"prefix":"","firstName":"Iuliia","middleName":"","lastName":"Pavlova","suffix":""},{"id":502441793,"identity":"999ed529-2919-4428-8ed0-a0ccc8e76a0b","order_by":2,"name":"Fedir Lapii","email":"","orcid":"","institution":"Department of Pediatrics, Immunology, Infectious and Rare Diseases, European Medical School, International European University, Kyiv, Ukraine","correspondingAuthor":false,"prefix":"","firstName":"Fedir","middleName":"","lastName":"Lapii","suffix":""},{"id":502441794,"identity":"aa19c1c5-962f-4446-9c67-4f8d6afe8f98","order_by":3,"name":"Ubydul Haque","email":"","orcid":"","institution":"Rutgers Global Health Institute, Rutgers University, New Brunswick, NJ","correspondingAuthor":false,"prefix":"","firstName":"Ubydul","middleName":"","lastName":"Haque","suffix":""}],"badges":[],"createdAt":"2025-08-19 08:58:34","currentVersionCode":1,"declarations":{"humanSubjects":true,"vertebrateSubjects":false,"conflictsOfInterestStatement":false,"humanSubjectEthicalGuidelines":true,"humanSubjectConsent":true,"humanSubjectClinicalTrial":false,"humanSubjectCaseReport":false,"vertebrateSubjectEthicalGuidelines":false},"doi":"10.21203/rs.3.rs-7406595/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-7406595/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":89642563,"identity":"f205c440-b8e2-425c-9b6c-a177a8fba5ed","added_by":"auto","created_at":"2025-08-22 08:21:13","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":836461,"visible":true,"origin":"","legend":"\u003cp\u003eA: Receiver operating characteristic, area under the curves and 95% confidence intervals in machine learning algorithms based on the test set. (1.knowledge)\u003c/p\u003e\n\u003cp\u003eB: Receiver operating characteristic, area under the curves and 95% confidence intervals in machine learning algorithms based on the test set. (2.belief)\u003c/p\u003e\n\u003cp\u003eC: Receiver operating characteristic, area under the curves and 95% confidence intervals in machine learning algorithms based on the test set. \u0026nbsp;(3.Attitude)\u003c/p\u003e\n\u003cp\u003eD: Receiver operating characteristic, area under the curves and 95% confidence intervals in machine learning algorithms based on the test set. ((4.Practice)\u003c/p\u003e\n\u003cp\u003eE: Receiver operating characteristic, area under the curves and 95% confidence intervals in machine learning algorithms based on the test set. (5.Access, Policy \u0026amp; Structural Barriers)\u003c/p\u003e\n\u003cp\u003e\u003cbr\u003e\u003c/p\u003e","description":"","filename":"1.png","url":"https://assets-eu.researchsquare.com/files/rs-7406595/v1/191aa30bd28d7353b78bf041.png"},{"id":89642126,"identity":"563d4c24-b614-465e-8c89-2fed14d3228b","added_by":"auto","created_at":"2025-08-22 08:13:13","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":248037,"visible":true,"origin":"","legend":"\u003cp\u003eA: The top 10 interaction effects based on relative importance. . (A.knowledge)\u003c/p\u003e\n\u003cp\u003eB: The top 10 interaction effects based on relative importance. .(B.belief)\u003c/p\u003e\n\u003cp\u003eC: The top 10 interaction effects based on relative importance. . \u0026nbsp;(C.Attitude)\u003c/p\u003e\n\u003cp\u003eD: The top 10 interaction effects based on relative importance. . (D.Practice)\u003c/p\u003e\n\u003cp\u003eE: The top 10 interaction effects based on relative importance. . \u0026nbsp;(E.Access, Policy \u0026amp; Structural Barriers)\u003c/p\u003e","description":"","filename":"2.png","url":"https://assets-eu.researchsquare.com/files/rs-7406595/v1/6693a1a4c0f28c8ba4af8926.png"},{"id":89643300,"identity":"e071d8f4-11fa-4345-96c7-6fc332781abb","added_by":"auto","created_at":"2025-08-22 08:29:14","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":2011945,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-7406595/v1/3e667885-108e-4ef7-932f-15e0c4640d91.pdf"},{"id":89642131,"identity":"6ace7776-dcc0-4cec-916b-a17cfc6616ba","added_by":"auto","created_at":"2025-08-22 08:13:13","extension":"docx","order_by":1,"title":"","display":"","copyAsset":false,"role":"supplement","size":1060198,"visible":true,"origin":"","legend":"\u003cp\u003esupplementary file\u003c/p\u003e","description":"","filename":"4.Supplementary.docx","url":"https://assets-eu.researchsquare.com/files/rs-7406595/v1/3c33552882a8694bfd7494fd.docx"}],"financialInterests":"The authors declare no competing interests.","formattedTitle":"\u003cp\u003eVaccine hesitancy in war-torn Ukraine: A machine learning framework for resilient immunization\u003c/p\u003e","fulltext":[{"header":"Introduction","content":"\u003cp\u003eVaccination remains one of the most powerful public-health interventions responsible for saving millions of lives annually. A landmark analysis marking 50 years of the WHO\u0026rsquo;s expanded programme on immunization estimates that vaccines have averted\u0026thinsp;~\u0026thinsp;154\u0026nbsp;million deaths, predominantly among infants and young children \u003csup\u003e1\u003c/sup\u003e. These gains, however, are fragile: the COVID-19 era exposed persistent inequities in access and uptake and rekindled hesitancy, showing that scientific advances do not automatically translate into population protection \u003csup\u003e2\u003c/sup\u003e. In conflict settings these gaps widen as infrastructure and supply chains are damaged, populations are displaced, and institutional trust erodes, elevating vaccine confidence from a demand-side concern to a public-health security imperative \u003csup\u003e3,4\u003c/sup\u003e.\u003c/p\u003e\u003cp\u003eVaccine confidence is a decisive determinant of coverage when supply and access exist. Vaccine hesitancy, identified by the World Health Organization as one of the ten threats to global health in 2019, can erode these gains \u003csup\u003e5\u003c/sup\u003e. The WHO SAGE Working Group defines vaccine hesitancy as delayed acceptance or refusal of vaccines despite availability, shaped by the \u0026ldquo;3Cs\u0026rdquo; of confidence, complacency, and convenience\u003csup\u003e6,7\u003c/sup\u003e. During emergencies, this confidence gap typically widens as misinformation circulates and polarization grows\u003csup\u003e7\u003c/sup\u003e. The consequences are not confined within borders: measles resurgence and importations repeatedly demonstrate how under-immunized pockets can ignite cross-border outbreaks and strain health services \u003csup\u003e8\u0026ndash;10\u003c/sup\u003e.\u003c/p\u003e\u003cp\u003eUkraine is a key case where conflict and displacement undermine vaccination. Before 2022, the country experienced repeated measles crises, with 53,218 cases in 2018 alone, amid fluctuating childhood coverage and low public trust \u003csup\u003e11\u003c/sup\u003e, also a cVDPV2 outbreak was detected in October 2021\u003csup\u003e12\u003c/sup\u003e. On the eve of the invasion, only about 35% of Ukrainians were fully vaccinated against COVID-19, reflecting pre-existing hesitancy and access barriers\u003csup\u003e13\u003c/sup\u003e. War-time destruction, service interruptions, and mass displacement (\u0026asymp;\u0026thinsp;6.9\u0026nbsp;million refugees and \u0026asymp;\u0026thinsp;3.7\u0026nbsp;million internally displaced as of early 2025) further jeopardized routine and pandemic immunization efforts \u003csup\u003e14\u003c/sup\u003e. Consistent with broader patterns in fragile and conflict-affected settings, fragile and conflict-affected countries have shown persistently lower routine coverage and slower recovery, sustaining regional outbreak risks\u003csup\u003e3,15\u003c/sup\u003e. Despite several studies charting COVID-19 uptake or demographics in Ukraine and among refugees\u003csup\u003e13,16,17\u003c/sup\u003e, a holistic examination spanning vaccine-related knowledge, beliefs, attitudes, practices, and barriers across both adult and childhood immunization during the current war is still missing. The interaction among sociocultural factors, political alignment, and access constraints remains underexplored, and the use of predictive modeling to isolate the strongest drivers of hesitancy in an active conflict is sparse\u003csup\u003e4,18\u003c/sup\u003e.\u003c/p\u003e\u003cp\u003eTo address this complexity, we pair a large, nationwide survey of Ukrainian adults across war-affected and less-affected territories with complementary machine-learning models to characterize the beliefs, attitudes, practices, and barriers shaping vaccine confidence in wartime. We emphasize interpretable Bayesian Additive Regression Trees alongside Support Vector Machines and Light Gradient Boosting Machines to stress-test predictive performance and rank drivers of hesitancy \u003csup\u003e19\u0026ndash;21\u003c/sup\u003e. Unlike traditional regression approaches, modern ML models can uncover non-linear relationships and complex interactions among variables; they provide a bridge between statistical sophistication and practical utility for health policymakers, enabling targeted, transparent, and context-specific interventions \u003csup\u003e22\u0026ndash;24\u003c/sup\u003e. While machine learning has been applied to vaccine-confidence prediction in peacetime contexts, its deployment to guide crisis immunization strategy in an ongoing, high-intensity war is, to our knowledge, novel\u003csup\u003e25,26\u003c/sup\u003e. By combining model accuracy with explainability, we aim to generate immediately actionable insights for targeted, context-specific insights for interventions in Ukraine and an approach that can be applied in other crises.\u003c/p\u003e"},{"header":"Methods","content":"\u003cp\u003e\u003cstrong\u003e\u003cem\u003eStudy area\u003c/em\u003e\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eUkraine, in Eastern Europe, is the continent\u0026rsquo;s second-largest country by land area and seventh most populous, with 38.98 million residents. Average life expectancy is 70 years for men and 79.5 years for women, below the European average.\u003csup\u003e27,28\u003c/sup\u003e The country comprises 1,469 hromadas, 409 urban, 435 suburban/settlement, and 625 rural, with 175 currently under Russian occupation.\u003csup\u003e29\u003c/sup\u003e For this study, 557 hromadas outside front-line and occupied areas will be selected for household recruitment.\u003c/p\u003e\n\u003cp\u003eThe war in Ukraine offers a unique lens to explore health and behavioral challenges amid conflict. Since 2022, over 1,000 healthcare facilities have been damaged or destroyed.\u003csup\u003e30\u003c/sup\u003e Coupled with mass population displacement and sharp drops in childhood vaccination, these disruptions have fueled increased epidemic risk for measles, polio, and diphtheria.\u003c/p\u003e\n\u003cp\u003eA total of 557 hromadas were purposively chosen from the Ministry of Health and the State Statistics Service records to maximize geographic coverage while omitting areas affected by occupation or ongoing conflict. The sample reflects differences in population scale, urban\u0026ndash;rural distribution, and health system resources, ensuring adequate power and representativeness for nationwide analyses.\u003c/p\u003e\n\u003cp\u003eBy January 2024, 79.2% of Ukraine\u0026rsquo;s population had internet access.\u003csup\u003e31\u003c/sup\u003e Mobile connectivity was broader, with nearly 90% of households estimated to have access to mobile services in the same period.\u003csup\u003e32\u003c/sup\u003e\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e\u003cem\u003eStudy design and participants\u003c/em\u003e\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eA cross-sectional household survey was conducted to evaluate vaccine hesitancy, acceptance, and coverage across Ukraine. The study included 557 purposively chosen hromadas, ensuring representation across all accessible oblasts while excluding conflict-affected areas. In each hromada, households with children aged 0\u0026ndash;17 were identified, and the parent primarily responsible for vaccination decisions was invited to respond, enabling assessment of both behavioral attitudes and practical vaccination barriers.\u003c/p\u003e\n\u003cp\u003eWe randomly approached 4,972 parents from distinct households to ensure representation across different population sizes, settlement types (urban, suburban, rural), and health system capacities. Participants were required to reside in the selected hromada during the survey period, be at least 18 years old, and have made a vaccination decision for at least one child in the preceding year.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e\u003cem\u003eSurvey administration and data collection\u003c/em\u003e\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eBetween November and December 2024, each respondent completed a structured, interviewer-administered questionnaire. TGM Research, which manages a nationally representative survey panel, facilitated data collection. Participants were approached through phone calls, text messages, and emails, and received a pretested questionnaire in Ukrainian. Following informed consent, respondents accessed the survey via a secure hyperlink. By integrating direct interviews with online survey completion, this approach maximized engagement across diverse regions and conflict-affected contexts while minimizing potential sampling bias.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e\u003cem\u003eSampling and representativeness\u003c/em\u003e\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eParticipants were recruited from all 24 oblasts of Ukraine using 2022 census data to ensure national representativeness. Quotas were applied to reflect key demographics (age, sex, and urban/rural distribution) and geographic coverage. Within each hromada, households were selected using probability proportional to size (PPS) sampling. This integrated approach, combining census-informed quotas, spatial diversity, and structured recruitment, enabled the study to represent population-level differences in vaccination attitudes, practices, and access challenges.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e\u003cem\u003eOutcomes and measures\u003c/em\u003e\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe study\u0026rsquo;s primary outcomes assessed parental vaccine confidence and trust in health services, utilizing established instruments such as the WHO Vaccine Confidence Scale and the PACV survey, enhanced with context-specific items for Ukraine.\u003csup\u003e33,34\u003c/sup\u003e Secondary outcomes captured patterns of childhood vaccination, including uptake, delays, refusals, perceived obstacles to access, and exposure to misinformation about vaccines.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e\u003cem\u003eParticipants\u003c/em\u003e\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eAmong the 4,972 households approached, 2,526 parents successfully completed the survey, providing robust representation across age, sex, settlement type, and region, including areas impacted by armed conflict. Eligibility required participants to be 18 years or older, resident in Ukraine during the survey period, and able to respond to survey questions. All participants provided informed consent, ensuring adherence to ethical standards. This design facilitated the collection of reliable, geographically inclusive, and representative data, providing a foundation to investigate how behavioral, social, and systemic factors shape vaccination decisions in war-affected populations.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e\u003cem\u003eData collection\u003c/em\u003e\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eData were collected using a structured questionnaire covering the following domains: (i) Demographics and socioeconomic status, which included age, gender, education, employment status, and religious influence; (ii) Vaccine knowledge and perceptions, which included understanding of efficacy and safety, perceived risk, awareness of immunization programs, and exposure to misinformation; (iii) Vaccination history and side effects, which included self-reported COVID-19 doses, experienced side effects, and history of vaccine-preventable infections; (iv) Attitudes and beliefs, which included perceived safety and efficacy, autonomy, historical influences, and social determinants of vaccine decision-making; (v) External and environmental Influences, which included impact of media, religion, politics, socioeconomic status, and conflict exposure; and (vi) Childhood immunization data, which included vaccine coverage by type and age, side effects, and infection history.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e\u003cem\u003eSurvey instrument and measurement of vaccine confidence\u003c/em\u003e\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eWe administered a 135-item survey designed to assess demographic, psychological, socio-cultural, behavioral, political, and conflict-related factors influencing vaccination choices. Among these, 116 questions used a 6-point Likert scale to quantify key aspects of vaccine confidence\u0026mdash;perceived safety and effectiveness, trust in the health system, and social influence. This approach allowed a detailed assessment of vaccine hesitancy, capturing gradations of attitudes beyond simple yes/no measures.\u003c/p\u003e\n\u003cp\u003eVaccine confidence was quantified using validated constructs customized for the Ukrainian setting, integrating the WHO Vaccine Confidence Scale and PACV survey.\u003csup\u003e33,34\u003c/sup\u003e Responses were aggregated into continuous and categorical scores, facilitating evaluation of both overall confidence and the behavioral drivers underlying vaccination choices.\u003c/p\u003e\n\u003cp\u003eEducation, employment, region, vaccine brand, and reported side effects were measured using ordinal and nominal scales, with vaccination status and conflict exposure recorded as binary variables. This multidimensional dataset facilitates a nuanced analysis of the interactions between psychological predispositions, social influence, access constraints, and contextual determinants of vaccine uptake.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e\u003cem\u003eStatistical analysis\u003c/em\u003e\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eDescriptive statistics were calculated for all variables, with continuous variables reported as means \u0026plusmn; standard deviation (SD) and categorical variables as frequencies and percentages. Chi-square tests of independence (\u0026chi;\u0026sup2;), Fisher\u0026rsquo;s exact test, or the Cochran-Armitage trend test, as appropriate, were used to assess differences between groups. Cram\u0026eacute;r\u0026apos;s V was applied to measure the strength of association, categorized as 0\u0026ndash;0.1 (weak), \u0026gt;0.1\u0026ndash;0.3 (moderate), \u0026gt;0.3\u0026ndash;0.5 (strong), and \u0026gt;0.5 (very strong)\u003csup\u003e35\u003c/sup\u003e. For selected variables, post-hoc analyses were conducted to compare groups and identify categories contributing to the \u0026chi;\u0026sup2; results.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e\u003cem\u003eMachine learning models\u0026nbsp;\u003c/em\u003e\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eMachine learning (ML) analysis was performed to model predictors of vaccine hesitancy across five constructs: knowledge, beliefs, attitudes, practices, and structural barriers. Eight ML models were applied, namely Bayesian Additive Regression Tree (BART), Adaptive Boosting (Adboost), Lasso Logistic Regression (LLR), Decision Tree (DT), Artificial Neural Network (ANN), Random Forest (RF), Support Vector Machine (SVM), and Light Gradient Boosting (LGBM).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e\u003cem\u003eModel selection and development\u003c/em\u003e\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe dataset was categorized into 135 independent and dependent (target) variables for effective modeling. Feature engineering was applied to identify predictors of vaccine hesitancy, with outcomes simplified into binary values for classification. Variables were omitted if they had i) more than 5% missing data, ii) correlations above 0.90, or iii) were marked as unreliable. The final dataset included 86 variables out of 135.\u003c/p\u003e\n\u003cp\u003eWe used an 80:20 train-test split, allocating 80% of the data for training and 20% for testing. To mitigate overfitting and obtain more reliable performance estimates, 10-fold cross-validation was applied by dividing the training data into 10 subsets, iteratively training on nine and validating on the remaining one.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e\u003cem\u003eModel optimization\u003c/em\u003e\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eHyperparameters were optimized using Randomized Search with stratified 5-fold cross-validation for each machine learning model, a robust technique for parameter tuning that helps avoid local optima and identify configurations yielding superior generalization performance.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e\u003cem\u003ePerformance metrics\u003c/em\u003e\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eTo ensure a balanced assessment of predictive accuracy and reliability, we evaluated model performance using a comprehensive set of metrics. These included the Area Under the Curve (AUC), which measures the ability to rank positive instances above negative ones, with higher values indicating greater discriminative power; accuracy, the proportion of correctly classified instances; sensitivity (true positive rate), the proportion of actual positives correctly identified; specificity (true negative rate), the proportion of actual negatives correctly identified; positive predictive value (PPV), the proportion of true positives among all predicted positives; negative predictive value (NPV), the proportion of true negatives among all predicted negatives; the F1-score, defined as the harmonic mean of precision and recall; and balanced accuracy, the average sensitivity across classes, particularly useful for imbalanced datasets. We also used the Brier score, defined as the mean squared difference between predicted probabilities and actual outcomes, where lower values indicate better performance (0 = perfect accuracy, 1 = poorest predictive performance).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e\u003cem\u003eFeature importance\u003c/em\u003e\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eBased on the selected model, which showed reliable and optimal performance metrics, we assessed feature importance to identify the strongest determinants within each construct. In addition, three variable importance rules were applied: (i) Local threshold: a predictor is included if its inclusion proportion exceeds the 1-\u0026alpha; quantile of its null distribution; (ii) Global max threshold: a predictor is included if its inclusion proportion exceeds the 1-\u0026alpha; quantile of the distribution of the maximum null variable inclusion proportions across permutations of the response; and (iii) Global SE threshold: using a global multiplier shared by all predictors, a predictor is included if its inclusion proportion exceeds a threshold based on the mean and standard deviation of its null distribution. All analyses were conducted in R version 4.5.1.\u003c/p\u003e"},{"header":"Results","content":"\u003cp\u003e\u003cstrong\u003eParticipant Characteristics\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eA total of 2526 parents responded to the survey, with a mean age of 44.6 \u0026plusmn; 13 years. Of these, 46.1% of respondents belong to the 18-44 years age group, 48.3% were 45-64 years, 5.46% were 65-79 years, and only 0.07% were over 80 years old. 58.8% were female compared to males (41.2%). Overall, 59.4% of respondents completed their university degree, and 64.8% were employed. Only 5.1% reported religious influence on vaccination decisions, but 48.1% indicated religion-related vaccine hesitancy in Ukraine (Table 1).\u003c/p\u003e\n\u003cp\u003eOverall, 24.26% respondents perceived vaccines as risky, including 5.62% respondents who understood vaccine efficiency and safety, 18% exposed to vaccine misinformation, 22.3% valued individual autonomy in healthcare decisions, and 16.7% concerned about vaccine access. Overall, 44.22% had a lack of trust in vaccine information, of whom 35.1% understood vaccine efficiency, 29.1% were exposed to misinformation, 41.2% valued individual autonomy, and 37.5% had concerns about vaccine access. Pandemic-related influence on vaccine hesitancy was reported by 31.20%, including 29.9% exposed to misinformation and 26.5% concerned about equitable access (Table S1). Regarding vaccine doses, 14.0% reported side effects after their second COVID-19 dose, while 17.8% respondents experienced post-vaccination COVID-19 infection after two doses (Table S2). \u0026nbsp;The most common vaccination side effects were site discomfort (4.9% for Pfizer) and muscle pain (6.6% for Pfizer) (Table S3).\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eRegarding childhood vaccinations, 34.5% of children had received age-appropriate vaccines since the previous year, with high uptake for birth Hepatitis B (HepB) (20.5%), Diphtheria, tetanus, and pertussis (DTP) at 2 months (21.9%), and Measles, mumps, rubella (MMR) at 12 months (20.5%). All vaccine-specific coverage rates were statistically significant (p\u0026lt;0.0001) (Table 2). Overall, 5.54% of children with a complete vaccination history experienced side effects, while 5.82% reported side effects from vaccinations received in the past year (Table S4).\u003c/p\u003e\n\u003cp\u003eOverall, 50.3% of participants expressed positive attitudes toward vaccination, 8.9% negative, and 40.8% neutral. Positive attitudes were strongly associated with beliefs in vaccine safety (p \u0026lt; 0.0001, Cramer\u0026rsquo;s V = 0.293) and infection prevention (p \u0026lt; 0.0001, Cramer\u0026rsquo;s V = 0.313), while 34.8% reported concerns about serious side effects. Negative attitudes were linked to fears of harmful ingredients and a preference for natural immunity. External influences included social media (11.6% positive, 46.7% negative; p \u0026lt; 0.0001, Cramer\u0026rsquo;s V = 0.098) and religious leaders (2.5% positive, 45.8% negative; p \u0026lt; 0.0001, Cramer\u0026rsquo;s V = 0.211) (Table 3).\u003c/p\u003e\n\u003cp\u003eConcerns about safety and risks were reported by 7.21% of those with negative vaccine attitudes (Table 8). Broader concerns about vaccine side effects were associated with political ideology, perception of vaccines (63.1%), safety concerns (72.8%), social influence (20.7%), cultural norms (33.6%), and trust in healthcare workers (23.9%) or doctors (28.5%) (Table S5).\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eReluctance to receive future vaccines was linked to personal health beliefs influencing vaccine perception (21.8%) and to lack of local access or affordability (23.2%). Low trust in healthcare professionals\u0026rsquo; recommendations was significantly associated with personal vaccine beliefs (34.8%) and inadequate access (43.9%) (Table S6).\u003c/p\u003e\n\u003cp\u003eThe respondents who perceived vaccine hesitancy in the community related to the factors such as urban\u0026ndash;rural disparities in vaccine access (16.0%), lack of access to a trustworthy place (25.6%), influence of perceptions about vaccines (27.2%), local accessibility issues (34.1%), and cultural and societal influences (17.2%) (Table S7).\u003c/p\u003e\n\u003cp\u003eCultural and societal beliefs influencing vaccine hesitancy were related to vaccination history (72.2%), the impact of personal health beliefs (44.0%), and urban\u0026ndash;rural disparities (20.0%) (Table S8). Environmental factors influencing vaccine hesitancy included regional gaps in vaccine availability (p \u0026lt; 0.0001, Cramer\u0026rsquo;s V = 0.201) and challenges in accessing healthcare facilities (p \u0026lt; 0.0001, Cramer\u0026rsquo;s V = 0.140) (Table S9).\u003c/p\u003e\n\u003cp\u003eHigh trust in the Ukrainian healthcare system was associated with greater confidence in vaccine policies, equitable vaccine access, and willingness to be vaccinated in the future. In contrast, residence in conflict-affected areas was modestly associated with lower trust (p = 0.043) (Table S10).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eMachine Learning Analysis\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe performance of all machine learning models was evaluated using accuracy, sensitivity, specificity, PPV, NPV, F1 score, balanced accuracy, and Brier score. For the vaccine knowledge model, BART model has the highest accuracy (0.78), specificity (0.75), PPV (0.75), NPV (0.80), F1 score (0.78), balanced accuracy (0.78), and the lowest Brier score (0.16), while LLR had the highest sensitivity (0.87) (Figure S1A). For the belief model, SVM, LGBM, and BART all reached the highest accuracy (0.73), with BART showing the highest sensitivity (0.73) and LGBM the highest specificity (0.76) (Figure S1B). The highest F1 scores were observed in adaptive boosting, RF, and SVM, while the lowest Brier scores were reported for all models except decision tree, neural network, and random forest. For the attitude, practice, and barrier models, the highest model accuracy was obtained with BART at 0.74, 0.75, and 0.71, respectively, with corresponding low Brier scores of 0.18, 0.16, and 0.19 (Figure S1C\u0026ndash;E). Overall, validation results indicate that the BART model consistently demonstrated superior performance across all models (Figure S1).\u003c/p\u003e\n\u003cp\u003eThe ROC curves with AUC values and corresponding 95% bootstrap confidence intervals are presented in Figure 1. The BART model achieved higher AUC values than other machine learning approaches across all five settings (Figure 1A\u0026ndash;E).\u003c/p\u003e\n\u003cp\u003eThe inclusion proportion for a predictor represents the fraction of times it is selected as a splitting rule relative to the total number of splitting rules across posterior draws in the sum-of-trees model. In the knowledge model, the top five predictors were Perceived Risk of Vaccines, awareness of general immunization vaccines, pandemic influence, concerns about equitable vaccine access, and historical influences on vaccination attitudes (Figure S2, 1A). For the belief model, the most important variables were lack of information on vaccine safety, concerns about serious vaccine side effects, perceptions of dangerous ingredients in vaccines, preference to delay COVID-19 vaccination, and vaccine safety perception \u0026nbsp;(Figure S2, 2A). The most important features for the attitude model were cultural norms and vaccination perceptions, lack of institutional trust issues, lack of trusted sources for vaccine information, impact of social media on vaccine choice, and religious influence on vaccination decisions \u0026nbsp; (Figure S2, 3A). For the practice model, the key predictors were willingness for future vaccination, impact of previous vaccination status, sharing knowledge on COVID-19 variants, influence of health beliefs on vaccines, and previous adverse vaccine experiences (Figure S2, 4A). In the barrier model, the most important predictors were lack of trust in healthcare professionals, lack of healthcare quality and transparency, vaccine hesitancy in the community, financial constraints, and urban\u0026ndash;rural vaccine access gaps \u0026nbsp;(Figure S2, 5A).\u003c/p\u003e\n\u003cp\u003eA similar set of important predictors was identified using the local procedure and the Global Max and Threshold Procedures. In the knowledge model, perceived risk of vaccines, lack of awareness of general vaccines immunization, concerns about equitable vaccine access, and pandemic influence exceeded the green line threshold (Figure S2, 1B). In the belief model, key predictors included lack of information on vaccine safety, concerns about serious vaccine side effects, and perceptions of dangerous ingredients in vaccines (Figure S2, 2B). For the attitude model, cultural norms, institutional trust issues, social media impact on vaccine decisions, safety concerns, religious leaders\u0026rsquo; opposition to COVID-19 vaccination, and economic factors were the most important features (Figure S2, 3B). In the practice model, the top predictors were lack of trust in healthcare professionals, COVID-19 variant knowledge sharing, past vaccination status, and impact of health beliefs on vaccines \u0026nbsp;(Figure S2, 4B). For the barrier model, the most influential predictors were lack of trust in healthcare professionals, lack of healthcare quality and transparency, vaccine hesitancy in the community, and limited urban\u0026ndash;rural vaccine access \u0026nbsp;(Figure S2, 5B).\u003c/p\u003e\n\u003cp\u003eGlobal Max and Threshold procedures identified the strongest predictors across models. Perceived risk of vaccines was the top predictor in the knowledge model (Figure S2, 1C); lack of information on vaccine safety in the belief model (Figure S2, 2C); and cultural norms in the attitude model (Figure S2, 3C). In the practice and barrier analyses, Lack of trust in healthcare professionals, Lack of healthcare quality and transparency, and vaccine hesitancy in the community were the only variables to exceed both the Global SE and Global Max thresholds, indicating they are the most important predictors of vaccine hesitancy (Figure S2, 4C-5C).\u003c/p\u003e\n\u003cp\u003eThe average relative importance of all significant predictors with 95% confidence intervals is shown in Figure 2. In the knowledge model, the top three interaction effects were between perceived risk of vaccines and awareness of general immunization vaccines, perceived risk of vaccines and pandemic influence on vaccine hesitancy, and autonomy\u0026rsquo;s influence in healthcare \u0026nbsp;and concerns about equitable vaccine access \u0026nbsp;(Figure 2A). For the belief model, key interactions included perceptions of dangerous ingredients in vaccines and lack of information on vaccine safety\u003cs\u003e,\u0026nbsp;\u003c/s\u003econcerns about serious vaccine side effects and lack of information on vaccine safety, and perceptions of dangerous ingredients in vaccines and concerns about serious vaccine side effects \u0026nbsp;(Figure 2B). In the attitude model, the most important interactions were social media impact on vaccine decisions and safety concerns, institutional trust issues and economic factors, and safety concerns and political ideology (Figure 2C). For the practice model, the strongest interactions involved impact of past vaccination status and lack of trust in healthcare professionals, impact of COVID-19 variant knowledge sharing and lack of trust in healthcare professionals, and lack of trust in healthcare professionals and negative impact of health beliefs on vaccines (Figure 2D). In the barrier model, the leading interactions were lack of healthcare quality and transparency and vaccine hesitancy in the community, lack of healthcare quality and transparency and financial constraints, and lack of trust in healthcare professionals and financial constraints (Figure 2E).\u003c/p\u003e"},{"header":"Discussion","content":"\u003cp\u003eThis study offers one of the most comprehensive wartime portraits of vaccine attitudes, beliefs, and behaviors in Ukraine, integrating a large behavioral survey with interpretable machine learning. Just over half of respondents held positive views of vaccination (50.3%), while 40.8% were neutral and 8.9% negative. Across multiple predictive domains, Bayesian Additive Regression Trees (BART) outperformed other algorithms and illuminated high-impact predictors and their interactions, sharpening a picture of the social, structural, and informational landscape that shapes vaccination decisions\u003csup\u003e19,22\u003c/sup\u003e. Our findings underscore how prolonged conflict aggravates vaccine hesitancy by both disrupting services and eroding public trust. In this wartime survey of Ukraine, lack of trust in healthcare professionals and institutions emerged as one of the most influential barriers to vaccination, alongside misinformation and concerns about vaccine safety, and these patterns consistent with broader literature linking trust and misinformation to lower vaccine acceptance\u003csup\u003e36\u003c/sup\u003e. This aligns with evidence from other conflict settings where war amplifies immunization gaps and fuels hesitancy\u003csup\u003e4,18\u003c/sup\u003e. Our findings provide Ukraine-specific confirmation of broader guidance that confidence and demand-side factors are as critical as supply in crises. Restoring vaccine confidence requires transparency, consistent risk communication, and community engagement in line with WHO SAGE recommendations on the \u0026ldquo;3Cs\u0026rdquo; of confidence, complacency, and convenience\u003csup\u003e7\u003c/sup\u003e.\u003c/p\u003e\u003cp\u003eCultural norms (e.g., preferences for \u0026ldquo;natural immunity\u0026rdquo;), religious influence, and political identity exerted indirect effects that often interacted with perceived safety and transparency. These patterns align with regional and global work showing that socio-political identity and norms condition vaccine confidence and demand \u003csup\u003e37\u003c/sup\u003e. Information pathways also mattered: nearly half of respondents citing social media reported negative influence (46.7%). This accords with evidence that exposure to vaccine misinformation depresses intent and uptake, particularly when messages appear \u0026ldquo;scientific\u0026rdquo; or circulate in polarized networks \u003csup\u003e38\u003c/sup\u003e. Misinformation-related beliefs, such as fears of dangerous ingredients and doubts about vaccine safety, were among the most important predictors in our belief model, emphasizing the urgent need for tailored counter-messaging strategies.\u003c/p\u003e\u003cp\u003eHesitancy was highest where people perceived poor care quality, high out-of-pocket costs, and limited nearby vaccination services, showing that structural barriers and demand factors are closely linked. Guidance for humanitarian settings recommends keeping services continuous, close to communities, affordable, and culturally appropriate to sustain coverage during crises \u003csup\u003e39\u003c/sup\u003e\u003c/p\u003e\u003cp\u003eUkraine\u0026rsquo;s program showed resilience despite war: by late 2024, over 80% of children were vaccinated against most infections, a return toward pre-war levels\u003csup\u003e40\u003c/sup\u003e. Coverage still fell short of the ~\u0026thinsp;95% threshold for herd protection, and by early 2025 nine measles outbreaks were recorded. While early-life vaccines (for example, birth HepB and MMR at 12 months) had relatively high uptake, only 34.5% of children were fully up to date in the prior year. Nearly half of participants reported community-level hesitancy driven by misinformation, low trust, and convenience barriers - pressures that often worsen during war. These results suggest recovery is robust but fragile, depending on both reliable delivery and renewed confidence and equity across regions\u003csup\u003e40\u003c/sup\u003e.\u003c/p\u003e\u003cp\u003eA key implication is the need for targeted, context-aware strategies to bolster vaccine confidence in conflict settings. Our models highlighted belief/attitude factors (e.g., fear of side effects, influence of social networks, and some cultural norms) that predict hesitancy. any of these are modifiable through public health interventions, for instance, community education to counteract misinformation, engagement with local leaders to allay cultural or religious concerns, and visible endorsement of vaccines by trusted healthcare providers\u003csup\u003e7,41\u003c/sup\u003e. Notably, we found that residence in high-conflict areas was associated with slightly lower trust in the health system (p\u0026thinsp;=\u0026thinsp;0.043), suggesting the war\u0026rsquo;s direct toll on institutional confidence. This finding aligns with reports that sustained conflict can weaken trust even in previously strong health systems, emphasizing that rebuilding trust should be a priority in Ukraine\u0026rsquo;s health response\u003csup\u003e7,18\u003c/sup\u003e.\u003c/p\u003e\u003cp\u003eMethodologically, this study demonstrates the value of interpretable machine learning in crisis settings. BART offered strong accuracy and calibration across domains, and helped to uncover non-linear relationships and interactions that standard models often miss\u003csup\u003e22,24\u003c/sup\u003e. High risk perception deterred acceptance most when people had low baseline vaccine knowledge. Fears about severe side effects combined with worries about ingredients were more discouraging than either concern alone. Low trust in clinicians plus a prior negative vaccination experience also strongly predicted reluctance to return. Together, these patterns support precision public health that targets clusters of co-occurring risks\u003csup\u003e25\u003c/sup\u003e.\u003c/p\u003e\u003cp\u003eThese results suggest a multi-level strategy: address structural barriers (urban\u0026ndash;rural access gaps, affordability, service quality/transparency) with mobile brigades, extended hours, and robust cold-chain; execute myth-specific communications elevating trusted messengers (clinicians, community leaders); and target high-risk clusters identified by ML (e.g., low trust and poor access) for co-delivered services and counseling\u003csup\u003e7,41\u003c/sup\u003e.\u003c/p\u003e\u003cp\u003eStrengths of the study include a large diverse wartime sample; integration of behavioral, cultural, and structural determinants; and interpretable ML that translates prediction into policy-relevant insight. However, limitations must be acknowledged. Limitations include cross-sectional design (no causal inference), potential self-report bias, and further geographic stratification could enhance understanding of regional variations, especially in frontline versus non-frontline areas. Additionally, although our ML models identified strong associations, real-world interventions will require careful testing to confirm the impact of targeting the identified predictors.\u003c/p\u003e\u003cp\u003eGiven the ongoing conflict and shifting vaccine attitudes, future work should prioritize longitudinal panels to track changes in hesitancy and evaluate intervention impact. Integrating real-time social media listening into hesitancy surveillance can enable rapid counter-messaging as new myths emerge, while experimental designs, including randomized controlled trials, can test interventions tailored to machine-learning\u0026ndash;identified risk clusters. Extending this combined survey and machine-learning approach to other fragile or conflict-affected settings will test generalizability and refine precision public health strategies, helping advance goals to reach zero-dose and under-immunized children in FCV contexts.\u003c/p\u003e"},{"header":"Conclusion","content":"\u003cp\u003eOur comprehensive analysis of 2,526 Ukrainian parents highlights the complex determinants of vaccine confidence and uptake amid a conflict-affected context. While 50% of respondents expressed positive vaccination attitudes, nearly one-quarter perceived vaccines as risky, and 44% reported low trust in vaccine information. Childhood vaccine coverage was moderate, with 34.5% of children receiving age-appropriate immunizations in the prior year, and reported side effects were generally mild. Attitudes were shaped by beliefs in vaccine safety and infection prevention, while negative perceptions were linked to fears of harmful ingredients, preference for natural immunity, and external influences such as social media and religious guidance. Machine learning models, particularly BART, identified perceived vaccine risk, lack of safety information, cultural norms, and trust in healthcare professionals as the most predictive factors for vaccine knowledge, beliefs, attitudes, practices, and barriers. Interactions between misinformation, institutional trust, social media influence, and access disparities further exacerbated hesitancy. These findings underscore the need for multifaceted interventions targeting knowledge gaps, societal norms, healthcare trust, and equitable access to improve vaccine acceptance in vulnerable populations.\u003c/p\u003e"},{"header":"Abbreviations","content":"\u003cp\u003eKNOWQ1: Perceived Risk of Vaccines; KNOWQ3: Awareness of General Immunization Vaccines; KNOWQ8: Pandemic Influence; KNOWQ11: Concerns About Equitable Vaccine Access; KNOWQ9: Historical Influences on Vaccination Attitudes; KNOWQ10: Importance of Autonomy in Healthcare Decisions; KNOWQ6: Exposure to Vaccine Misinformation.ATTQ4: Lack of Information on Vaccine Safety; ATTQ3: Concerns About Serious Vaccine Side Effects; ATTQ2: Perceptions of Dangerous Ingredients in Vaccines; ATTQ5: Preference to Delay COVID-19 Vaccination; ATTQ1: Vaccine Safety Perception; ATTQ8: Belief in Natural Immunity over Vaccination; ATTQ6: Perceived Protection from Vaccines; ATTQ7: Belief in Alternatives to Vaccination; ATTQ16: Cultural Norms and Vaccination Perceptions; ATTQ18: Lack of Institutional Trust Issues; ATTQ19: Lack of Trusted Sources for Vaccine Information; ATTQ9: Impact of Social Media on Vaccine Choice; ATTQ15: Religious Influence on Vaccination Decisions; ATTQ11: Religious Leaders\u0026rsquo; Stance Against COVID-19 Vaccination; ATTQ21: Influence of Political Ideology; ATTQ23: Financial Factors Influence Vaccination Decision; ATTQ20: Religious Factors Influence on Vaccination Opinions; ATTQ24: Other Factors in Vaccination; ATTQ10: Influence of Media Reports on COVID-19 Vaccination; ATTQ22: Influence of Social Influence; ATTQ17: Familial and Peer Impact on Vaccination. PRACQ9: Willingness for Future Vaccination; PRACQ8: Impact of Previous Vaccination Status; PRACQ3: Sharing Knowledge on COVID-19 Variants; PRACQ6: Influence of Health Beliefs on Vaccines; PRACQ5: Previous Adverse Vaccine Experiences; PRACQ2: Travel Abroad During COVID-19; PRACQ1: Avoidance of Crowded Places During COVID-19; PRACQ4: Comfort in Sharing COVID-19 Knowledge; PRACQ7: Impact of Socioeconomic Status on Vaccination; PRACQ12: Lack of Trust in Healthcare Professionals; PRACQ13: Lack of Healthcare Quality and Transparency; PRACQ00: Vaccine Hesitancy in the Community; PRACQ15: Financial Constraints; PRACQ17: Urban\u0026ndash;Rural Disparities in Vaccine Access; PRACQ19: Environmental Impact on Vaccine Perceptions; PRACQ14: Influence of Government Vaccine Policies.\u003c/p\u003e\n\u003cp\u003e\u003cbr\u003e\u003c/p\u003e"},{"header":"Declarations","content":"\u003cp\u003e\u003cspan\u003eThis study was approved by Poltava Medical University, Ukraine (No. 227 dated 5/23/2024), and Rutgers University (#Pro2024001126). Written informed consent was obtained from all participants, in addition to which all data collection was performed in accordance with relevant guidelines and regulations in Ukraine and the USA.\u003c/span\u003e\u003c/p\u003e\u003cp\u003e\u003cstrong\u003eAcknowledgement\u003c/strong\u003e: We sincerely appreciate TGM Research for facilitating data collection for this study.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAuthor contributions\u003c/strong\u003e: All authors made equal contributions to this study. ;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eDeclarations of conflict\u003c/strong\u003e: None declared.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eData availability\u003c/strong\u003e: Deidentified, delinked datasets can be obtained from the corresponding author upon request.\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\n\u003cli\u003eShattock, A. J.\u003cem\u003e et al.\u003c/em\u003e Contribution of vaccination to improved survival and health: modelling 50 years of the Expanded Programme on Immunization. \u003cem\u003eThe Lancet\u003c/em\u003e \u003cstrong\u003e403\u003c/strong\u003e, 2307-2316 (2024).\u003c/li\u003e\n\u003cli\u003eWHO. \u003cem\u003eGlobal immunization efforts have saved at least 154 million lives over the past 50 years. World Health Organization. [2024, April 24]\u003c/em\u003e, \u0026lt; https://www.who.int/news/item/24-04-2024-global-immunization-efforts-have-saved-at-least-154-million-lives-over-the-past-50-years\u0026gt; (2024).\u003c/li\u003e\n\u003cli\u003eJones, C. E. Routine vaccination coverage\u0026mdash;worldwide, 2023. \u003cem\u003eMMWR. Morbidity and mortality weekly report\u003c/em\u003e \u003cstrong\u003e73\u003c/strong\u003e (2024).\u003c/li\u003e\n\u003cli\u003eCiccacci, F.\u003cem\u003e et al.\u003c/em\u003e Between war and pestilence: the impact of armed conflicts on vaccination efforts: a review of literature. \u003cem\u003eFrontiers in Public Health\u003c/em\u003e \u003cstrong\u003e13\u003c/strong\u003e, 1604288 (2025).\u003c/li\u003e\n\u003cli\u003eWHO. \u003cem\u003eTen threats to global health in 2019. World Health Organization [2019, January 10] \u003c/em\u003e\u0026lt;https://www.who.int/news-room/spotlight/ten-threats-to-global-health-in-2019\u0026gt; (2019).\u003c/li\u003e\n\u003cli\u003eMacDonald, N. E. Vaccine hesitancy: Definition, scope and determinants. \u003cem\u003eVaccine\u003c/em\u003e \u003cstrong\u003e33\u003c/strong\u003e, 4161-4164 (2015).\u003c/li\u003e\n\u003cli\u003eSAGE, W. \u003cem\u003eWHO SAGE Working Group on Vaccine Hesitancy. Report of the SAGE Working Group on Vaccine Hesitancy. (2014, November 12).\u003c/em\u003e \u0026lt; https://www.asset-scienceinsociety.eu/sites/default/files/sage_working_group_revised_report_vaccine_hesitancy.pdf\u0026gt; (2014).\u003c/li\u003e\n\u003cli\u003ePatel, M.\u003cem\u003e et al.\u003c/em\u003e National update on measles cases and outbreaks\u0026mdash;United States, January 1\u0026ndash;October 1, 2019. \u003cem\u003eAmerican Journal of Transplantation\u003c/em\u003e \u003cstrong\u003e20\u003c/strong\u003e, 311-314 (2020).\u003c/li\u003e\n\u003cli\u003eZucker, J. R.\u003cem\u003e et al.\u003c/em\u003e Consequences of undervaccination\u0026mdash;measles outbreak, New York City, 2018\u0026ndash;2019. \u003cem\u003eNew England Journal of Medicine\u003c/em\u003e \u003cstrong\u003e382\u003c/strong\u003e, 1009-1017 (2020).\u003c/li\u003e\n\u003cli\u003eCDC/WHO. \u003cem\u003eCenters for Disease Control and Prevention, \u0026amp; World Health Organization. (2023, November 16). Global measles threat continues to grow as another year passes with millions of children unvaccinated [Joint news release].\u003c/em\u003e \u0026lt;https://www.cdc.gov/media/releases/2023/p1116-global-measles.html (CDC)\u0026gt; (2023).\u003c/li\u003e\n\u003cli\u003eWHO. \u003cem\u003eWorld Health Organization. Measles \u0026ndash; European Region [Disease outbreak news]. (2019, May 6)\u003c/em\u003e, \u0026lt;https://www.who.int/emergencies/disease-outbreak-news/item/2019-DON140 \u0026gt; (2019).\u003c/li\u003e\n\u003cli\u003eWHO/Europe. \u003cem\u003eWorld Health Organization, Regional Office for Europe. (2023, September 21). Polio outbreak in Ukraine closed \u0026ndash; A success story for public health despite extreme challenges of war [News release]. \u003c/em\u003e, \u0026lt;https://www.who.int/europe/news/item/21-09-2023-polio-outbreak-in-ukraine-closed-a-success-story-for-public-health-despite-extreme-challenges-of-war\u0026gt; (2023).\u003c/li\u003e\n\u003cli\u003eCostantino, V. \u0026amp; MacIntyre, C. R. Impact of vaccine coverage and disruption to health services on COVID-19 in Ukraine. \u003cem\u003eScientific Reports\u003c/em\u003e \u003cstrong\u003e14\u003c/strong\u003e, 14729 (2024).\u003c/li\u003e\n\u003cli\u003eUNHCR, U. f. \u003cem\u003eUkraine refugee crisis: Aid, statistics and news. Retrieved August 17, 2025, from \u003c/em\u003e\u0026lt;https://www.unrefugees.org/emergencies/ukraine/\u0026gt; (2025).\u003c/li\u003e\n\u003cli\u003eUNICEF. \u003cem\u003eMeasles cases across Europe continue to surge, putting millions of children at risk [Press release]. ( May 28, 2024)\u003c/em\u003e, \u0026lt;https://www.unicef.org/eca/press-releases/measles-cases-across-europe-continue-surge-putting-millions-children-risk\u0026gt; (2024).\u003c/li\u003e\n\u003cli\u003eKorolchuk, O., Vasiuk, N., Klymkova, I., Shvets, D. \u0026amp; Piddubnyi, O. COVID-19 vaccination under conditions of war in Ukraine. \u003cem\u003eAsian bioethics review\u003c/em\u003e \u003cstrong\u003e15\u003c/strong\u003e, 259-281 (2023).\u003c/li\u003e\n\u003cli\u003eGanczak, M.\u003cem\u003e et al.\u003c/em\u003e \u0026ldquo;Why would we?\u0026rdquo; A qualitative study on COVID-19 vaccination decision making among Ukrainian economic female migrants in Poland. \u003cem\u003eFrontiers in Public Health\u003c/em\u003e \u003cstrong\u003e12\u003c/strong\u003e, 1380627 (2024).\u003c/li\u003e\n\u003cli\u003eDwyer, H., Enria, L., Palmer, J., Ayub, S. \u0026amp; Beckmann, N. Trust, Information and Vaccine Aonfidence in Crisis Settings: A Scoping Review. \u003cem\u003ePublic Health Challenges\u003c/em\u003e \u003cstrong\u003e4\u003c/strong\u003e, e70073 (2025).\u003c/li\u003e\n\u003cli\u003eChipman, H. A., George, E. I. \u0026amp; McCulloch, R. E. BART: Bayesian additive regression trees. (2010).\u003c/li\u003e\n\u003cli\u003eCortes, C. \u0026amp; Vapnik, V. Support-vector networks. \u003cem\u003eMachine learning\u003c/em\u003e \u003cstrong\u003e20\u003c/strong\u003e, 273-297 (1995).\u003c/li\u003e\n\u003cli\u003eKe, G.\u003cem\u003e et al.\u003c/em\u003e Lightgbm: A highly efficient gradient boosting decision tree. \u003cem\u003eAdvances in neural information processing systems\u003c/em\u003e \u003cstrong\u003e30\u003c/strong\u003e (2017).\u003c/li\u003e\n\u003cli\u003eTan, Y. V. \u0026amp; Roy, J. Bayesian additive regression trees and the General BART model. \u003cem\u003eStatistics in medicine\u003c/em\u003e \u003cstrong\u003e38\u003c/strong\u003e, 5048-5069 (2019).\u003c/li\u003e\n\u003cli\u003eChen, T. \u0026amp; Guestrin, C. in \u003cem\u003eProceedings of the 22nd acm sigkdd international conference on knowledge discovery and data mining.\u003c/em\u003e 785-794.\u003c/li\u003e\n\u003cli\u003eLundberg, S. M. \u0026amp; Lee, S.-I. A unified approach to interpreting model predictions. \u003cem\u003eAdvances in neural information processing systems\u003c/em\u003e \u003cstrong\u003e30\u003c/strong\u003e (2017).\u003c/li\u003e\n\u003cli\u003eLincoln, T. M.\u003cem\u003e et al.\u003c/em\u003e Taking a machine learning approach to optimize prediction of vaccine hesitancy in high income countries. \u003cem\u003eScientific reports\u003c/em\u003e \u003cstrong\u003e12\u003c/strong\u003e, 2055 (2022).\u003c/li\u003e\n\u003cli\u003eCarrieri, V., Lagravinese, R. \u0026amp; Resce, G. Predicting vaccine hesitancy from area‐level indicators: A machine learning approach. \u003cem\u003eHealth Economics\u003c/em\u003e \u003cstrong\u003e30\u003c/strong\u003e, 3248-3256 (2021).\u003c/li\u003e\n\u003cli\u003e\u003cem\u003eLife expectancy at birth, total (years) - Ukraine. (accessed April 8 2022).\u003c/em\u003e \u0026lt; https://data.worldbank.org/indicator/SP.DYN.LE00.IN?locations=UA\u0026gt; (\u003c/li\u003e\n\u003cli\u003e\u003cem\u003eNumber of Present Population of Ukraine, as of January 1. 2021. \u003c/em\u003e, \u0026lt;http://www.ukrstat.gov.ua/druk/publicat/kat_u/2021/zb/05/zb_chuselnist%202021.pdf (accessed April 8 2022).\u0026gt; (\u003c/li\u003e\n\u003cli\u003e\u003cem\u003eInstitute for the Study of War 2024. Fact Sheet: The Kremlin\u0026apos;s Occupation Playbook: Coerced Russification and Ethnic Cleansing in Occupied Ukraine. \u003c/em\u003e, \u0026lt;Available at [https://www.understandingwar.org/sites/default/files/Fact%20Sheet%20%20--%20The%20Kremlin%27s%20Occupation%20Playbook-Coerced%20Russification%20and%20Ethnic%20Cleansing%20in%20Occupied%20Ukraine%202.pdf], last accessed 08.22.2024.\u0026gt; (\u003c/li\u003e\n\u003cli\u003eHaque, U.\u003cem\u003e et al.\u003c/em\u003e A Comparison of Ukrainian Hospital Services and Functions Before and During the Russia-Ukraine War. \u003cem\u003eJama-Health Forum\u003c/em\u003e \u003cstrong\u003e5\u003c/strong\u003e, e240901, doi:10.1001/jamahealthforum.2024.0901 (2024).\u003c/li\u003e\n\u003cli\u003eDataReportal.\u003cem\u003e Digital 2024: Ukraine. \u003c/em\u003e, \u0026lt;Retrieved from https://datareportal.com/reports/digital-2024-ukraine.\u0026gt; (2024).\u003c/li\u003e\n\u003cli\u003e\u003cem\u003eInternational Telecommunication Union (ITU) and Freedom House. Freedom on the Net 2024: Ukraine. Accessed August 17, 2025.\u003c/em\u003e, \u0026lt; https://freedomhouse.org/country/ukraine/freedom-net/2024.\u0026gt; (\u003c/li\u003e\n\u003cli\u003eLarson, H. J.\u003cem\u003e et al.\u003c/em\u003e Measuring vaccine hesitancy: The development of a survey tool. \u003cem\u003eVaccine\u003c/em\u003e \u003cstrong\u003e33\u003c/strong\u003e, 4165-4175, doi:10.1016/j.vaccine.2015.04.037 (2015).\u003c/li\u003e\n\u003cli\u003eOpel, D. J.\u003cem\u003e et al.\u003c/em\u003e The relationship between parent attitudes about childhood vaccines survey scores and future child immunization status: a validation study. \u003cem\u003eJAMA Pediatr\u003c/em\u003e \u003cstrong\u003e167\u003c/strong\u003e, 1065-1071, doi:10.1001/jamapediatrics.2013.2483 (2013).\u003c/li\u003e\n\u003cli\u003eScheff\u0026eacute;, H. H. cram\u0026eacute;r, mathematical methods of statistics. (1947).\u003c/li\u003e\n\u003cli\u003eAdhikari, B., Cheah, P. Y. \u0026amp; von Seidlein, L. Trust is the common denominator for COVID-19 vaccine acceptance: a literature review. \u003cem\u003eVaccine: X\u003c/em\u003e \u003cstrong\u003e12\u003c/strong\u003e, 100213 (2022).\u003c/li\u003e\n\u003cli\u003eKarakulak, A.\u003cem\u003e et al.\u003c/em\u003e Trust in government moderates the association between fear of COVID-19 as well as empathic concern and preventive behaviour. \u003cem\u003eCommunications psychology\u003c/em\u003e \u003cstrong\u003e1\u003c/strong\u003e, 43 (2023).\u003c/li\u003e\n\u003cli\u003ePierri, F.\u003cem\u003e et al.\u003c/em\u003e Online misinformation is linked to early COVID-19 vaccination hesitancy and refusal. \u003cem\u003eScientific reports\u003c/em\u003e \u003cstrong\u003e12\u003c/strong\u003e, 5966 (2022).\u003c/li\u003e\n\u003cli\u003eWHO. \u003cem\u003eVaccination in humanitarian emergencies: Implementation guide (WHO/IVB/17.13). Geneva, Switzerland: World Health Organization., \u0026lt;https://www.who.int/publications/i/item/WHO-IVB-17.13\u0026gt; (2017).\u003c/em\u003e\u003c/li\u003e\n\u003cli\u003eUNICEF/Ukraine. \u003cem\u003eMinistry of Health of Ukraine; WHO/Ukraine. (2025, April 24). Vaccination is both personal protection and a contribution to collective immunity (Joint statement)., \u0026lt;https://www.unicef.org/ukraine/en/press-releases/vaccination-is-both\u0026gt; (2025).\u003c/em\u003e\u003c/li\u003e\n\u003cli\u003eOECD. \u003cem\u003eEnhancing public trust in COVID-19 vaccination: The role of governments.\u003c/em\u003e, \u0026lt;https://www.oecd.org/content/dam/oecd/en/publications/reports/2021/05/enhancing-public-trust-in-covid-19-vaccination-the-role-of-governments_36749555/eae0ec5a-en.pdf\u0026gt; (2021).\u003c/li\u003e\n\u003c/ol\u003e"},{"header":"Tables","content":"\u003cp\u003eTable 1: : Demographic characteristics of Participants\u0026nbsp;\u003c/p\u003e\n\u003ctable border=\"1\" cellspacing=\"0\" cellpadding=\"0\" width=\"564\"\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 332px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eCharacteristics\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 232px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eTotal n =2526\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 332px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eAge, year \u003csup\u003ea\u003c/sup\u003e\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 232px;\"\u003e\n \u003cp\u003e44.6 \u0026plusmn; 13\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 332px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eAge group\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 232px;\"\u003e\u003cbr\u003e\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 332px;\"\u003e\n \u003cp\u003e18\u0026ndash;44 years\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 232px;\"\u003e\n \u003cp\u003e1165 (46.1)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 332px;\"\u003e\n \u003cp\u003e45\u0026ndash;64 years\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 232px;\"\u003e\n \u003cp\u003e1221 (48.3)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 332px;\"\u003e\n \u003cp\u003e65\u0026ndash;79 years\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 232px;\"\u003e\n \u003cp\u003e138 (5.46)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 332px;\"\u003e\n \u003cp\u003e\u0026ge;80 years\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 232px;\"\u003e\n \u003cp\u003e8 (0.07)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 332px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eGender\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 232px;\"\u003e\u003cbr\u003e\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 332px;\"\u003e\n \u003cp\u003eMale\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 232px;\"\u003e\n \u003cp\u003e1041 (41.2)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 332px;\"\u003e\n \u003cp\u003eFemale\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 232px;\"\u003e\n \u003cp\u003e1485 (58.8)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 332px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eEducation attainment\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 232px;\"\u003e\n \u003cp\u003e\u003cstrong\u003e\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 332px;\"\u003e\n \u003cp\u003eJunior high or middle school Professional (vocational) education\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 232px;\"\u003e\n \u003cp\u003e750 (29.7)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 332px;\"\u003e\n \u003cp\u003eGrade school Complete general secondary education\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 232px;\"\u003e\n \u003cp\u003e245 (9.7)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 332px;\"\u003e\n \u003cp\u003ePost-college (graduate school) Academic degree\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 232px;\"\u003e\n \u003cp\u003e30 (1.2)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 332px;\"\u003e\n \u003cp\u003eUniversity degree\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 232px;\"\u003e\n \u003cp\u003e1501 (59.4)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 332px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eEmployment Status\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 232px;\"\u003e\n \u003cp\u003e\u003cstrong\u003e\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 332px;\"\u003e\n \u003cp\u003eEmployed\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 232px;\"\u003e\n \u003cp\u003e1638 (64.8)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 332px;\"\u003e\n \u003cp\u003eUn-Employed\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 232px;\"\u003e\n \u003cp\u003e228 (9.0)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 332px;\"\u003e\n \u003cp\u003eStudent\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 232px;\"\u003e\n \u003cp\u003e54 (2.1)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 332px;\"\u003e\n \u003cp\u003eRetired\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 232px;\"\u003e\n \u003cp\u003e299 (11.8)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 332px;\"\u003e\n \u003cp\u003eHouseworker or Other\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 232px;\"\u003e\n \u003cp\u003e307 (12.15)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 332px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eReligious Influence on Vaccination\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 232px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 332px;\"\u003e\n \u003cp\u003eYes\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 232px;\"\u003e\n \u003cp\u003e130 (5.1)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 332px;\"\u003e\n \u003cp\u003eNo\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 232px;\"\u003e\n \u003cp\u003e2292 (90.7)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 332px;\"\u003e\n \u003cp\u003eUnsure\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 232px;\"\u003e\n \u003cp\u003e104 (4.1)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 332px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eReligious Vaccine Hesitancy in Ukraine\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 232px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 332px;\"\u003e\n \u003cp\u003eYes\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 232px;\"\u003e\n \u003cp\u003e\u0026nbsp;1226 (48.1)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 332px;\"\u003e\n \u003cp\u003eNo\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 232px;\"\u003e\n \u003cp\u003e471 (18.6)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 332px;\"\u003e\n \u003cp\u003eUnsure\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 232px;\"\u003e\n \u003cp\u003e829 (32.8)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n\u003c/table\u003e\n\u003cp\u003e\u003csup\u003ea\u003c/sup\u003emean \u0026plusmn;standard deviation\u003c/p\u003e\n\u003cp\u003eTable 2: Age‑Specific Vaccination in children by Vaccine Type\u0026nbsp;\u003c/p\u003e\n\u003ctable border=\"1\" cellspacing=\"0\" cellpadding=\"0\" width=\"666\"\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 94px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eAge\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 205px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eVaccines\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd colspan=\"2\" valign=\"top\" style=\"width: 213px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eTotal vaccinated Since last year\u003c/strong\u003e\u003c/p\u003e\n \u003cp\u003e\u003cstrong\u003e870 (34.5%)\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 154px;\"\u003e\n \u003cp\u003e\u003cstrong\u003e\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 94px;\"\u003e\n \u003cp\u003e\u003cstrong\u003e\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 205px;\"\u003e\n \u003cp\u003e\u003cstrong\u003e\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 106px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eCounts (%)\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 106px;\"\u003e\n \u003cp\u003e\u003cstrong\u003ep-value\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 154px;\"\u003e\n \u003cp\u003e\u003cstrong\u003ep-value\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 94px;\"\u003e\n \u003cp\u003eBirth\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 205px;\"\u003e\n \u003cp\u003eHepatitis B (HepB) \u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 106px;\"\u003e\n \u003cp\u003e365 (20.51) \u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 106px;\"\u003e\n \u003cp\u003e\u0026lt;0.0001\u003csup\u003ea\u003c/sup\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 154px;\"\u003e\n \u003cp\u003e\u0026lt;0.0001\u003csup\u003e\u0026nbsp;c\u003c/sup\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd rowspan=\"4\" valign=\"top\" style=\"width: 94px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003cp\u003e2 months\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 205px;\"\u003e\n \u003cp\u003eHepatitis B (HepB) \u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 106px;\"\u003e\n \u003cp\u003e310 (17.42) \u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 106px;\"\u003e\n \u003cp\u003e\u0026lt;0.0001\u003csup\u003ea\u003c/sup\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd rowspan=\"4\" valign=\"top\" style=\"width: 154px;\"\u003e\n \u003cp\u003e\u0026lt;0.0001\u003csup\u003e\u0026nbsp;c\u003c/sup\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 205px;\"\u003e\n \u003cp\u003eDiphtheria, tetanus, pertussis (DTP) \u0026nbsp; \u0026nbsp;\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 106px;\"\u003e\n \u003cp\u003e390 (21.91) \u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 106px;\"\u003e\n \u003cp\u003e\u0026lt;0.0001\u003csup\u003e\u0026nbsp;a\u003c/sup\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 205px;\"\u003e\n \u003cp\u003eHaemophilus influenzae type b (Hib) \u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 106px;\"\u003e\n \u003cp\u003e221 (12.42) \u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 106px;\"\u003e\n \u003cp\u003e\u0026lt;0.0001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 205px;\"\u003e\n \u003cp\u003eInactivated poliovirus (IPV) \u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 106px;\"\u003e\n \u003cp\u003e250 (14.04) \u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 106px;\"\u003e\n \u003cp\u003e\u0026lt;0.0001\u003csup\u003e\u0026nbsp;a\u003c/sup\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd rowspan=\"3\" valign=\"top\" style=\"width: 94px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003cp\u003e4 months\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 205px;\"\u003e\n \u003cp\u003eDiphtheria, tetanus, pertussis (DTP) \u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 106px;\"\u003e\n \u003cp\u003e365 (20.51) \u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 106px;\"\u003e\n \u003cp\u003e\u0026lt;0.0001\u003csup\u003e\u0026nbsp;a\u003c/sup\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd rowspan=\"3\" valign=\"top\" style=\"width: 154px;\"\u003e\n \u003cp\u003e\u0026lt;0.0001\u003csup\u003e\u0026nbsp;c\u003c/sup\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 205px;\"\u003e\n \u003cp\u003eHaemophilus influenzae type b (Hib) \u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 106px;\"\u003e\n \u003cp\u003e203 (11.4) \u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 106px;\"\u003e\n \u003cp\u003e\u0026lt;0.0001\u003csup\u003e\u0026nbsp;b\u003c/sup\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 205px;\"\u003e\n \u003cp\u003eInactivated poliovirus (IPV) \u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 106px;\"\u003e\n \u003cp\u003e233 (13.09) \u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 106px;\"\u003e\n \u003cp\u003e\u0026lt;0.0001\u003csup\u003e\u0026nbsp;a\u003c/sup\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd rowspan=\"4\" valign=\"top\" style=\"width: 94px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003cp\u003e6 months\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 205px;\"\u003e\n \u003cp\u003eDiphtheria, tetanus, pertussis (DTP) \u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 106px;\"\u003e\n \u003cp\u003e335 (18.82) \u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 106px;\"\u003e\n \u003cp\u003e\u0026lt;0.0001\u003csup\u003e\u0026nbsp;a\u003c/sup\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd rowspan=\"4\" valign=\"top\" style=\"width: 154px;\"\u003e\n \u003cp\u003e\u0026lt;0.0001\u003csup\u003e\u0026nbsp;c\u003c/sup\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 205px;\"\u003e\n \u003cp\u003eHaemophilus influenzae type b (Hib) \u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 106px;\"\u003e\n \u003cp\u003e198 (11.12) \u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 106px;\"\u003e\n \u003cp\u003e\u0026lt;0.0001\u003csup\u003e\u0026nbsp;b\u003c/sup\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 205px;\"\u003e\n \u003cp\u003eLive poliovirus (bOPV) \u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 106px;\"\u003e\n \u003cp\u003e220 (12.36) \u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 106px;\"\u003e\n \u003cp\u003e\u0026lt;0.0001\u003csup\u003e\u0026nbsp;b\u003c/sup\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 205px;\"\u003e\n \u003cp\u003eHepatitis B (HepB) \u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 106px;\"\u003e\n \u003cp\u003e286 (16.07) \u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 106px;\"\u003e\n \u003cp\u003e\u0026lt;0.0001\u003csup\u003e\u0026nbsp;a\u003c/sup\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd rowspan=\"2\" valign=\"top\" style=\"width: 94px;\"\u003e\n \u003cp\u003e12 months\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 205px;\"\u003e\n \u003cp\u003eMeasles, mumps, rubella (MMR) \u0026nbsp; \u0026nbsp;\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 106px;\"\u003e\n \u003cp\u003e365 (20.51) \u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 106px;\"\u003e\n \u003cp\u003e\u0026lt;0.0001\u003csup\u003e\u0026nbsp;a\u003c/sup\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd rowspan=\"2\" valign=\"top\" style=\"width: 154px;\"\u003e\n \u003cp\u003e\u0026lt;0.0001\u003csup\u003e\u0026nbsp;c\u003c/sup\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 205px;\"\u003e\n \u003cp\u003eHaemophilus influenzae type b (Hib) \u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 106px;\"\u003e\n \u003cp\u003e197 (11.07) \u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 106px;\"\u003e\n \u003cp\u003e\u0026lt;0.0001\u003csup\u003e\u0026nbsp;b\u003c/sup\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 94px;\"\u003e\n \u003cp\u003e18 months\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 205px;\"\u003e\n \u003cp\u003eDiphtheria, tetanus, pertussis (DTP) \u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 106px;\"\u003e\n \u003cp\u003e305 (17.13) \u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 106px;\"\u003e\n \u003cp\u003e\u0026lt;0.0001\u003csup\u003e\u0026nbsp;b\u003c/sup\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd rowspan=\"2\" valign=\"top\" style=\"width: 154px;\"\u003e\n \u003cp\u003e\u0026lt;0.0001\u003csup\u003e\u0026nbsp;c\u003c/sup\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 94px;\"\u003e\n \u003cp\u003e18 months\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 205px;\"\u003e\n \u003cp\u003eLive poliovirus (bOPV) \u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 106px;\"\u003e\n \u003cp\u003e201 (11.29) \u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 106px;\"\u003e\n \u003cp\u003e\u0026lt;0.0001\u003csup\u003e\u0026nbsp;b\u003c/sup\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd rowspan=\"3\" valign=\"top\" style=\"width: 94px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003cp\u003e6 years\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 205px;\"\u003e\n \u003cp\u003eDiphtheria, tetanus (DT) \u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 106px;\"\u003e\n \u003cp\u003e239 (13.43) \u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 106px;\"\u003e\n \u003cp\u003e\u0026lt;0.0001\u003csup\u003e\u0026nbsp;a\u003c/sup\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd rowspan=\"3\" valign=\"top\" style=\"width: 154px;\"\u003e\n \u003cp\u003e\u0026lt;0.0001\u003csup\u003e\u0026nbsp;c\u003c/sup\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 205px;\"\u003e\n \u003cp\u003eLive poliovirus (bOPV) \u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 106px;\"\u003e\n \u003cp\u003e162 (9.1) \u0026nbsp; \u0026nbsp;\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 106px;\"\u003e\n \u003cp\u003e\u0026lt;0.0001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 205px;\"\u003e\n \u003cp\u003eMeasles, mumps, rubella (MMR) \u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 106px;\"\u003e\n \u003cp\u003e273 (15.34) \u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 106px;\"\u003e\n \u003cp\u003e\u0026lt;0.0001\u003csup\u003e\u0026nbsp;a\u003c/sup\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 94px;\"\u003e\n \u003cp\u003e14 years\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 205px;\"\u003e\n \u003cp\u003eLive poliovirus (bOPV) \u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 106px;\"\u003e\n \u003cp\u003e121 (6.8) \u0026nbsp; \u0026nbsp;\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 106px;\"\u003e\n \u003cp\u003e\u0026lt;0.0001\u003csup\u003e\u0026nbsp;b\u003c/sup\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 154px;\"\u003e\n \u003cp\u003e\u0026lt;0.0001\u003csup\u003e\u0026nbsp;c\u003c/sup\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 94px;\"\u003e\n \u003cp\u003e16 years\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 205px;\"\u003e\n \u003cp\u003eDiphtheria, tetanus (Td)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 106px;\"\u003e\n \u003cp\u003e155 (8.71)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 106px;\"\u003e\n \u003cp\u003e\u0026lt;0.0001\u003csup\u003e\u0026nbsp;b\u003c/sup\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 154px;\"\u003e\n \u003cp\u003e\u0026lt;0.0001\u003csup\u003ec\u003c/sup\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd colspan=\"5\" valign=\"top\" style=\"width: 666px;\"\u003e\n \u003cp\u003eCounts, percentages \u003csup\u003ea\u0026nbsp;\u003c/sup\u003eFisher exact test . \u003csup\u003eb\u003c/sup\u003e Pearson\u0026rsquo;s chi-square test, \u003csup\u003ec\u003c/sup\u003e Cochran-Armitage trend test.\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n\u003c/table\u003e\n\u003cp\u003e\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eTable 3: Attitudes Across Vaccine Safety, Efficacy, Hesitancy, and Influence\u003c/p\u003e\n\u003ctable border=\"1\" cellspacing=\"0\" cellpadding=\"0\" width=\"617\"\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 212px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd colspan=\"3\" valign=\"top\" style=\"width: 258px;\"\u003e\n \u003cp\u003eOverall Attitude about vaccine\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 74px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 74px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 212px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 85px;\"\u003e\n \u003cp\u003ePositive\u003c/p\u003e\n \u003cp\u003e1271(50.3)\u003c/p\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 88px;\"\u003e\n \u003cp\u003eNegative\u003c/p\u003e\n \u003cp\u003e225(8.9)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 85px;\"\u003e\n \u003cp\u003eNeutral\u003c/p\u003e\n \u003cp\u003e1030(40.8) \u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 74px;\"\u003e\n \u003cp\u003ep-value\u003csup\u003ea\u003c/sup\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 74px;\"\u003e\n \u003cp\u003eCramer\u0026apos;s V\u003csup\u003eb\u003c/sup\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd colspan=\"4\" style=\"width: 470px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eVaccine Safety Beliefs\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 74px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 74px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 212px;\"\u003e\n \u003cp\u003eVaccines are safe\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 85px;\"\u003e\n \u003cp\u003e536(42.2)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 88px;\"\u003e\n \u003cp\u003e195(86.7)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 85px;\"\u003e\n \u003cp\u003e379(36.8)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 74px;\"\u003e\n \u003cp\u003e\u0026lt;0.0001\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 74px;\"\u003e\n \u003cp\u003e0.293\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 212px;\"\u003e\n \u003cp\u003eVaccines contain harmful ingredients\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 85px;\"\u003e\n \u003cp\u003e164(12.9)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 88px;\"\u003e\n \u003cp\u003e44(19.6)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 85px;\"\u003e\n \u003cp\u003e379(36.8)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 74px;\"\u003e\n \u003cp\u003e\u0026lt;0.0001\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 74px;\"\u003e\n \u003cp\u003e0.295\u003c/p\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd colspan=\"4\" style=\"width: 470px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eInformation \u0026amp; Novelty Hesitancy\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 74px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 74px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 212px;\"\u003e\n \u003cp\u003eConcern about serious side effects\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 85px;\"\u003e\n \u003cp\u003e442(34.8)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 88px;\"\u003e\n \u003cp\u003e28(12.4)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 85px;\"\u003e\n \u003cp\u003e193(18.7)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 74px;\"\u003e\n \u003cp\u003e\u0026lt;0.0001\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 74px;\"\u003e\n \u003cp\u003e0.204\u003c/p\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 212px;\"\u003e\n \u003cp\u003eInsufficient vaccine information\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 85px;\"\u003e\n \u003cp\u003e322(25.3)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 88px;\"\u003e\n \u003cp\u003e77(34.2)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 85px;\"\u003e\n \u003cp\u003e282(27.4)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 74px;\"\u003e\n \u003cp\u003e\u0026lt;0.0001\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 74px;\"\u003e\n \u003cp\u003e0.133\u003c/p\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd colspan=\"4\" style=\"width: 470px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eVaccine Efficacy Perceptions\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 74px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 74px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 212px;\"\u003e\n \u003cp\u003ePrefer to wait on new vaccine\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 85px;\"\u003e\n \u003cp\u003e268(21.1)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 88px;\"\u003e\n \u003cp\u003e54(24.0)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 85px;\"\u003e\n \u003cp\u003e290(28.2)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 74px;\"\u003e\n \u003cp\u003e\u0026lt;0.0001\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 74px;\"\u003e\n \u003cp\u003e0.226\u003c/p\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 212px;\"\u003e\n \u003cp\u003eVaccines prevent infection\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 85px;\"\u003e\n \u003cp\u003e801(63.0)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 88px;\"\u003e\n \u003cp\u003e159(70.7)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 85px;\"\u003e\n \u003cp\u003e312(30.3)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 74px;\"\u003e\n \u003cp\u003e\u0026lt;0.0001\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 74px;\"\u003e\n \u003cp\u003e0.313\u003c/p\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 212px;\"\u003e\n \u003cp\u003ePreference for Alternative Preventive Measures\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 85px;\"\u003e\n \u003cp\u003e99(7.8)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 88px;\"\u003e\n \u003cp\u003e36(16.0)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 85px;\"\u003e\n \u003cp\u003e398(38.6)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 74px;\"\u003e\n \u003cp\u003e\u0026lt;0.0001\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 74px;\"\u003e\n \u003cp\u003e0.248\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 212px;\"\u003e\n \u003cp\u003ePrefer immunity from infection\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 85px;\"\u003e\n \u003cp\u003e163(12.8)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 88px;\"\u003e\n \u003cp\u003e80(35.6)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 85px;\"\u003e\n \u003cp\u003e359(34.9)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 74px;\"\u003e\n \u003cp\u003e\u0026lt;0.0001\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 74px;\"\u003e\n \u003cp\u003e0.182\u003c/p\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd colspan=\"3\" style=\"width: 385px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eExternal Influences on Vaccination Decisions\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 85px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 74px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 74px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 212px;\"\u003e\n \u003cp\u003eInfluenced by social media reports\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 85px;\"\u003e\n \u003cp\u003e148(11.6)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 88px;\"\u003e\n \u003cp\u003e105(46.7)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 85px;\"\u003e\n \u003cp\u003e353(34.3)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 74px;\"\u003e\n \u003cp\u003e\u0026lt;0.0001\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 74px;\"\u003e\n \u003cp\u003e0.131\u003c/p\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 212px;\"\u003e\n \u003cp\u003eInfluenced by mainstream media\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 85px;\"\u003e\n \u003cp\u003e203(16.0)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 88px;\"\u003e\n \u003cp\u003e106(47.1)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 85px;\"\u003e\n \u003cp\u003e352(34.2)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 74px;\"\u003e\n \u003cp\u003e\u0026lt;0.0001\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 74px;\"\u003e\n \u003cp\u003e0.098\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 212px;\"\u003e\n \u003cp\u003eInfluenced by religious leaders\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 85px;\"\u003e\n \u003cp\u003e32(2.5)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 88px;\"\u003e\n \u003cp\u003e103(45.8)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 85px;\"\u003e\n \u003cp\u003e342(33.2)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 74px;\"\u003e\n \u003cp\u003e\u0026lt;0.0001\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 74px;\"\u003e\n \u003cp\u003e0.211\u003c/p\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n\u003c/table\u003e\n\u003cp\u003eCounts, percentages, \u003csup\u003ea\u003c/sup\u003e chi-square test, \u003csup\u003eb\u0026nbsp;\u003c/sup\u003eestimated Cramer\u0026rsquo;s V\u003c/p\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":true,"hideJournal":true,"highlight":"","institution":"N/A","isAcceptedByJournal":false,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true},"keywords":"Health, Diseases, Russian invasion, Public health, vaccine hesitancy, Ukraine, War, Civilian, Quality of life","lastPublishedDoi":"10.21203/rs.3.rs-7406595/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-7406595/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003ch2\u003eBackground\u003c/h2\u003e\u003cp\u003eVaccination prevents millions of deaths, yet conflict magnifies inequities and erodes trust. Ukraine\u0026rsquo;s war has jeopardized immunization, exposing urgent gaps in vaccine confidence. We conducted a nationwide survey and applied traditional statistics and interpretable machine-learning models to identify drivers of hesitancy. Our approach uncovers complex behavioral determinants and offers actionable, context-specific insights to strengthen vaccination strategies in Ukraine and other crisis settings.\u003c/p\u003e\u003ch2\u003eMethod\u003c/h2\u003e\u003cp\u003eWe conducted a cross-sectional household survey (November\u0026ndash;December 2024) across 557 purposively selected hromadas in Ukraine, excluding occupied or frontline areas. Eligible respondents were parents (\u0026ge;\u0026thinsp;18 years) responsible for vaccination decisions for children aged 0\u0026ndash;17. Of 4,972 households approached, 2,526 met the inclusion criteria and completed a structured questionnaire administered online. The 135-item instrument integrated validated scales (WHO Vaccine Confidence Scale, Parent Attitudes about Childhood Vaccines) and Ukraine-specific items. Data were analyzed using descriptive statistics, χ\u0026sup2; tests, and machine learning models with cross-validation to identify determinants of vaccine confidence and uptake.\u003c/p\u003e\u003ch2\u003eResults\u003c/h2\u003e\u003cp\u003eAmong 2,526 surveyed Ukrainian parents, 24% perceived vaccines as risky, 44% reported low trust in vaccine information, and only 34.5% of children were fully vaccinated in the past year. Attitudes and uptake were strongly influenced by cultural norms, perceived safety, social media, and healthcare trust. Using machine learning, the BART model identified novel, high-impact predictors of vaccine hesitancy, including perceived risk, lack of safety information, institutional distrust, and urban\u0026ndash;rural access gaps. Interaction analyses revealed that these factors synergistically shape parental beliefs, attitudes, and vaccination practices, offering unprecedented insights into determinants of vaccine uptake in conflict-affected settings.\u003c/p\u003e\u003ch2\u003eConclusion\u003c/h2\u003e\u003cp\u003eOur findings reveal that parental vaccine confidence and uptake are shaped by perceived risks, cultural norms, trust in healthcare professionals, and access disparities. Machine learning identified these factors as key predictors, highlighting targets for interventions to address hesitancy and improve equitable childhood immunization in Ukraine.\u003c/p\u003e","manuscriptTitle":"Vaccine hesitancy in war-torn Ukraine: A machine learning framework for resilient immunization","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2025-08-22 08:13:09","doi":"10.21203/rs.3.rs-7406595/v1","editorialEvents":[{"type":"communityComments","content":0}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"258258f0-9134-4bc4-87c3-3f616f7dd83c","owner":[],"postedDate":"August 22nd, 2025","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"posted","subjectAreas":[{"id":53544006,"name":"Health Policy"},{"id":53544007,"name":"Health Economics \u0026 Outcomes Research"},{"id":53544008,"name":"Vaccine Development"}],"tags":[],"updatedAt":"2025-08-22T08:13:09+00:00","versionOfRecord":[],"versionCreatedAt":"2025-08-22 08:13:09","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-7406595","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-7406595","identity":"rs-7406595","version":["v1"]},"buildId":"8U1c8b4HqxoKbykW_rLl7","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}