Segmenting Attitudes Toward Vaccination—Behavioral Insights into Influenza Vaccination Refusal in Romania

Abstract

Introduction Vaccine hesitancy remains a significant public health challenge, particularly in rural populations where geographic and socioeconomic barriers exacerbate the issue. This study aimed to examine the factors influencing vaccine hesitancy, focusing on influenza vaccinations in rural and urban communities in Romania. Methods The study was conducted between October 4 and October 30, 2024, across 16 family medicine practices, including seven in rural and nine in urban areas from Romania. A 13-item vaccine hesitancy questionnaire, assessing factors such as fear of adverse effects, distrust in vaccines, and perceived necessity included 272 responses. Latent Class Analysis (LCA) was used to identify distinct subgroups of hesitancy, while Chi-square and odds ratio analyses assessed geographic differences. We performed logistic regression for the most observed root attitudes on influenza vaccination refusal. Results Rural residents were significantly more likely to refuse influenza vaccination compared to urban residents. The LCA revealed three distinct latent classes, characterized by varying levels of agreement with key hesitancy items. Class 1 demonstrated low hesitancy, Class 2 exhibited high hesitancy driven by beliefs in natural immunity and perceived lack of necessity, and Class 3 showed intermediate hesitancy, influenced by distrust in vaccines and past negative experiences. Among the key significant triggers underlying vaccine refusal were fear of adverse effects, parental status, chronic diseases, and previous vaccination experiences, while factors such as age and gender showed limited impact. Conclusions Vaccine hesitancy persists as a complex, multidimensional issue, with rural populations disproportionately affected. Targeted, context-specific interventions addressing key drivers such as distorted risk perception, fear of adverse effects and distrust in vaccines are critical for improving vaccination rates These findings underscore the importance of tailored public health strategies to promote equitable vaccine uptake.

Introduction

Influenza viruses are a major cause of severe respiratory illness worldwide, leading to significant hospitalizations and deaths, particularly among older adults [1,2].

Despite declining mortality rates and disability-adjusted life years over the past three decades, influenza-related lower respiratory tract infections (LRTIs) continue to disproportionately affect males, older adults, and low-sociodemographic index regions, with the highest burden observed in individuals over 85 years old [3].

Over the past century, advancements in science and technology, particularly in vaccinology, have revolutionized public health by preventing numerous infectious diseases. Building on the pioneering work of Benjamin Jesty and Edward Jenner in smallpox control, these innovations culminated in milestones such as the licensing of the first Haemophilus influenzae type b conjugated vaccine in the 1980s [4,5]. Today, the World Health Organization estimates that vaccines save over 3.5 million lives annually by preventing diseases such as diphtheria, tetanus, pertussis, influenza, and measles [4,6].

Despite the critical role of vaccination in preventing influenza-related disease and the rising rates of LRTIs associated with influenza, vaccine hesitancy remains a significant barrier, hindering vaccination efforts and impacting the effectiveness of this preventive measure [7,8]. Vaccine hesitancy, characterized by a delay in accepting or refusal of vaccines despite their availability, has become an issue of increasing concern [9]. The internet and social media platforms have transformed the way information is shared and consumed, reshaping public health communication on a global scale. This shift has had a significant influence on vaccine acceptance, as these platforms serve as both a tool for spreading accurate health information and a channel for misinformation [10,11,12].

Understanding the psychological underpinnings of vaccine hesitancy is crucial for developing effective public health strategies. The JitsuVAX project has identified several key factors contributing to vaccine hesitancy: conspiracist ideation, distrust, unwarranted beliefs, worldview and politics, religious concerns, fear and phobias, distorted risk perception, perceived self-interest, and reactance [13].

Although numerous questionnaires have been developed to address vaccine hesitancy [14,15,16], data specific to Romania remain scarce, with one study that highlighted a notable fear of adverse effects associated with new vaccines compared to those included in the national immunization program [17].

In Romania, influenza vaccination rates are on a declining trend, despite the vaccines being available to specific population groups [18].

In the past years, the vaccines were purchased centrally by the Ministry of Health based on pre-orders provided by the county public health departments in consultation with family doctors. Subsequently, the vaccine was distributed to hospitals for healthcare personnel, social or medical-social assistance units for institutionalized individuals, and primarily to primary healthcare offices to be administered to at-risk groups as follows: medical and auxiliary personnel, pregnant women, children aged between 6 months and 5 years, individuals over

65 years old, and people with certain chronic illnesses. The vaccination system faced challenges, including delays from centralized procurement, uneven distribution across counties, limited urban supply, and inadequate logistics, requiring family doctors to manage transportation. The National Vaccination Program focused on pediatric and at-risk groups, with influenza vaccination being the primary vaccine for adults, limiting their perception of lifelong vaccination needs.

In September 2023, a new vaccine procurement and distribution mechanism was introduced through the electronic prescription system of the National Health Insurance House. This initiative expanded public access to vaccines by including HPV, dTap, pneumococcal, meningococcal, hepatitis B, measles, mumps, and rubella (MMR), and varicella vaccines, alongside the influenza vaccine. The new system lays the foundation for implementing the “lifelong vaccination” principle to prevent vaccine-preventable diseases in the adult population. Also, influenza vaccination reimbursement coverage has been expanded to include children aged 6 months to 18 years, pregnant women, individuals 65 and older, healthcare personnel, and adults aged 19-64 with chronic conditions. Additionally, adults aged 45-64 without chronic diseases receive 50% compensation for the vaccine [19,20].

Influenza vaccination has experienced a noticeable decline in the post-pandemic period, influenced by misinformation and disinformation surrounding COVID-19 and the SARS-CoV-2 vaccine. Additionally, administrative challenges highlighted at both European and national level regarding the measures taken to limit the spread of the disease or vaccination against COVID-19 have contributed to this trend.

In this context, we aimed to identify and categorize the reasons invoked by patients for refusing vaccination in the 2024-2025 influenza vaccination season, providing insights to guide educational programs for vaccinating and to improve public communication strategies by authorities, using a tailored questionnaire distributed across several family medicine practices in both rural and urban areas of Romania.

Methods

Data collection

The study was conducted in Romania between October 4 and October 30, 2024, across 16 family medicine practices, including seven in rural areas and nine in urban areas with national distribution that met the following inclusion criteria (1) they agreed to collect anonymized data from office records on patients who presented in October 2024; (2) they integrated the recommendation for influenza vaccination into routine practice for all patients visiting the office, regardless of the reason; (3) they documented in primary records the reasons for vaccine refusal and collected anamnestic data on vaccination for patients on their own list.

The 16 participating practices had as their main physicians members (11 medical practices) of the Immunization Working Group (IWG) of the Romanian National Society of Family Medicine (SNMF) who agreed to participate in this study. In order to balance the number of patients included in the study by urban or rural area where the activity was carried out, we also called on 5 other family doctors from Romania who agreed to participate and met the inclusion criteria. These offices work in 9 counties (Alba, Arad, Brașov, Dolj, Iași, Ilfov, Timiș, Vaslui and Bucharest) out of the 42.

We chose to conduct the study taking into account all presentations at the practice only in October because it is the most appropriate month to recommend influenza vaccination in accordance with the recommendations of the international authority and the influenza vaccine prescription protocol developed by National Immunization Technical Advisory Groups (NITAG). We did not continue the study in November because the patients presenting to the consultation could have been the same considering the characteristics of presentations in family medicine.

The research focused on patients who declined the recommended influenza vaccination and included a total of 272 responses. A 13-item vaccine refusal questionnaire was utilized, comprising "Yes" or "No" responses to specific statements: (R1) – “I don't think it's necessary”, (R2) – “I am afraid of adverse effects”, (R3) – “I lack sufficient information about the vaccine”, (R4) – “I have had a previous negative experience with a vaccine”, (R5) – “I have never had the flu and do not consider vaccination necessary”, (R6) – “I consider myself to have a strong immune system”, (R7) – “I have chronic illnesses that may worsen”, (R8) – “I do not trust the vaccine”, (R9) – “I refuse vaccination for religious reasons”, (R10) – “I believe the vaccine is made to control people”, (R11) – “I no longer use vaccines as I view the COVID-19 vaccine as a business venture”, (R12) – “It is too expensive”, (R13) – “I am afraid of injections”. The items were then ranked according to the highest number of “Yes”. Those that accumulated the highest numbers were considered relevant for our research.

The questionnaire was designed by a team of experts in vaccinations (Gheorghe Gindrovel Dumitra – GGD, Roxana Surugiu – RS and members of IWG) and included the most important factors influencing vaccine hesitancy, as highlighted in the literature or identified through their clinical practice. Prior to distribution, a brief pilot study including 10 files of patients was conducted to ensure that the questions were well-constructed and easily understood. IWG of SNMF is a special interest group in the main professional society in family medicine in Romania. The group provides expertise in the field of vaccination to family doctors in Romania, authorities and the general public. GGD is the coordinator of this group and a member of NITAG of the Ministry of Health of Romania and the group is formed by 14 family physicians with a special interest in the field of vaccination who provide medical services in 9 counties out of the 42 of the country.

The study received ethical approval from the Ethics Committee of the University of Medicine and Pharmacy of Craiova, Romania (approval number 399/04.10.2024).

Statistical analysis

Descriptive statistics are reported for sample characteristics and hesitancy of a flu vaccine. All statistical tests were conducted using R, and statistical significance was set to 0.05. We conducted a Latent Class Analysis (LCA) using the poLCA [21] package for R (v.1.6.0.1 Linzer and Lewis, 2011). We computed the Akaike Information Criterion (AIC) and the Bayesian Information Criterion (BIC) to evaluate and compare the goodness-of-fit of different models while penalizing for model complexity. In the context of our LCA, these criteria help us determine the optimal number of latent classes that best represent our data on vaccine refusal. Lower AIC values indicate a better model fit, balancing goodness-of-fit with model simplicity. Lower BIC values suggest a better model, with a stronger emphasis on model simplicity compared to AIC.

We performed the logistic regression for the most observed root attitudes on influenza vaccination refusal. We first assessed the multicollinearity and casewise list of residuals, then the odds ratio greater than 1 indicates the odds of the outcome increase as the predictor variable increases.

Results

We collected and used data from participating physicians who selected from the database 272 patient records that expressed a reason for refusing to receive vaccination. We included in the study a number of 16 offices distributed in urban areas (n=9) and rural areas (n=7) which allowed the analysis on a capitation list consisting of 37858 patients distributed approximately equally (19427 in urban areas and 18431 in rural areas). During the time interval taken into the study, more patients presented in urban areas, but the number of refusals recorded was higher in rural areas (

Table 1. Distribution of consultations, patients, and vaccination refusals across urban and rural medical practices.

	Number of offices (physicians praxis)	Number of patients allocated	Number of consultations	Number of patients consulted	Number of refusals recorded
Urban	9	19427	8209	5333 (27.5% of the allocated patients)	118 (2.2% of the consulted patients)
Rural	7	18431	5768	4120 (22.4% of the allocated patients)	154 (3.7% of the consulted patients)
Total	16	37858	13977	9453 (25% of the allocated patients)	272 (2.9% of the consulted patients)

Number of refusals recorded includes only patients who expressed a refusal reason to being vaccinated. This category did not include patients who were vaccinated, as well as those who presented a reason but stated that they would be vaccinated later.

). There was a significant difference in refusal rates between urban and rural settings (χ²(1)=18.81, p<0.001). Rural areas exhibited a higher proportion of refusals relative to their consultations compared to urban areas, as seen in Figure 1.

Odds of vaccine refusal given perceptions characteristics are presented in

Table 2. Characteristics of the respondents.

Characteristics	All (n=272)	Adults (n=245)	Parents (n=27)	p value
Age, years mean±SD median (IQR) range	57.5±17.2 58.5 (44-72) 18-91	59.5±16.8 62 (47-73) 18-91	39.7±6.48 38 (36-45) 28-50	<0.001**
Gender, male	109 (40.1%)	92 (36.2%)	17 (63.0%)	0.011* χ2(1)=6.54
Environment, urban	118 (43.4%)	95 (37.4%)	23 (85.2%)	<0.001** χ2(1)=21.3
Marital status, married (%)	168 (61.8%)	152 (58.8%)	16 (59.3%)	0.778 χ2(1)=0.08
Chronic diseases, yes (%)	187 (68.8%)	186 (73.2%)	1 (3.7%%)	<0.001** χ2(1)=59.0
Having had flu in the past 5 years Yes No I do not know	28 (10.3%) 189 (69.5%) 55 (20.2%)	24 (9.4%) 170 (66.9%) 51 (20.1%)	4 (14.8%) 19 (70.4%) 4 (14.8%)	0.597 χ2(2)=1.03
Past flu vaccination, yes	62 (22.8%)	59 (23.2%)	3 (11.1%)	0.127 χ2(1)=2.32
Past COVID-19 vaccination, yes	96 (35.3%)	96 (37.8%)	0	<0.001** χ2(1)=16.4
Other vaccines, yes	50 (18.4%)	32 (12.6%)	18 (66.7%)	<0.001** χ2(1)=46.6

Data is presented as n (%) unless otherwise stated. *p value <0.05, **p value < 0.001.

and

Table 3. Odds of vaccine refusal given perceptions characteristics.

	Answers with “Yes”	All (n=272)	Adults (n=245)	Parents (n=27)	p value
R1	I don't think it's necessary	84 (31%)	78 (30.7%)	6 (22.2%)	0.299 χ2(1)=1.08
R2	I am afraid of adverse effects	91 (33.6%)	80 (31.5%)	11 (40.7%)	0.406 χ2(1)=0.69
R3	I don't have enough information about the vaccine	2 (0.7%)	2 (0.8%)	0	0.637 χ2(1)=0.222
R4	I have had a previous negative experience with a vaccine	28 (10.3%)	28 (11.0%)	0	0.064 χ2(1)=3.44
R5	I've never had the flu and I don't think I need it	63 (23.2%)	55 (21.7%)	8 (29.6%)	0.401 χ2(1)=0.705
R6	I consider myself to have a strong immune system	31 (11.4%)	28 (11.3%)	3 (11.1%)	0.961 χ2(1)=0.002
R7	I have chronic diseases that can get worse	2 (0.7%)	2 (0.8%)	0	0.637 χ2(1)=0.222
R8	I don't trust the vaccine	51 (18.8%)	45 (17.7%)	6 (22.2%)	0.626 χ2(1)=0.237
R9	I do not vaccinate for religious reasons	6 (2.2%)	4 (1.6%)	2 (7.4%)	0.053 χ2(1)=3.76
R10	The vaccine is made to control people	7 (2.6%)	7 (2.8%)	0	0.374 χ2(1)=0.792
R11	I don't use any vaccine anymore. I believe that the COVID vaccine was a business	25 (9.2%)	25 (9.8%)	0	0.082 χ2(1)=3.03
R12	It's too expensive	3 (1.1%)	3 (1.2%)	0	0.563 χ2(1)=0.334
R13	I am afraid of injections	3 (1.1%)	3 (1.2%)	0	0.563 χ2(1)=0.334

Data is presented as n (%). Perceptions were self-reported yes/no.

.

The 3-class model was found to be better among the three models as seen in

Table 4. Comparison among models with different number of classes.

Number of classes	log-likelihood	AIC	BIC	p-value
2-Class Model	-844.5	1737	1823	0.140
3-Class Model	-805.5	1693	1841	0.080
4-Class Model	-777.6	1671	1880	0.060

, being the most balanced choice considering BIC, interpretability, and parsimony (AIC was best for 4-Class Model, but could be unnecessarily complex, whereas BIC was best for 3-Class Model) – Table 4.

Lower AIC and BIC values indicate better models, with BIC placing a stronger emphasis on model simplicity compared to AIC. In our analysis, we observed that as the number of latent classes increased, AIC values continued to decrease, reflecting improved model fit. However, BIC values continued to increase and reached a minimum at two latent classes, but with the highest p-value. We considered 3-classe model indicating that this model strikes the best balance between goodness-of-fit and parsimony.

To ensure robustness in model selection, we prioritized the BIC values in our final decision because of its tendency to favor simpler models, especially in smaller sample sizes. Additionally, the interpretability of the model with three classes aligned with the theoretical framework of vaccine refusal, further justifying its selection as the optimal model. Models with more classes exhibited smaller gains in AIC/BIC improvements while increasing model complexity and interpretive challenges.

Thus, the final model with three latent classes was selected based on a combination of statistical evidence and practical interpretability.

The probabilities of different responses for the key items (R1, R2, R4, R5, R6, R8, R11) across the three latent classes derived from the Latent Class Analysis (LCA) are shown in Figure 1. The three classes are expected to differ in their response patterns to these seven items (Figure 1).

Class 1 predominantly responds with “No” (blue) for most items, indicating a strong tendency toward this type of response across all items, except for R2 (“I am afraid of adverse effects”), where all members of this class responded with “Yes” (orange).

Class 2 shows greater variability in responses, alternating between “No” (blue) and “Yes” (orange) depending on the item. For certain items (e.g., R1, R5, R6), there is a substantial probability of responding with “Yes” suggesting that members of this class are more likely to choose this response for these specific items. However, for other items, the preference for “No” (blue) remains dominant, albeit less consistent compared to Class 1.

Class 3 also predominantly chooses “No” (blue) for most items, but the probability of responding with “Yes” (orange) is higher for certain items, such as R4 (“I have had a previous negative experience with a vaccine”). For some items, the responses in Class 3 align with those of Class 2, particularly for R2 (“I am afraid of adverse effects”) and R8 (“I don't trust the vaccine”).

Items like R1 (“I don't think it's necessary”), R2 (“I am afraid of adverse effects”), R4 (“I have had a previous negative experience with a vaccine”), and R5 (“I've never had the flu, so I don't think I need it”) appear to distinguish the classes more clearly. For instance, Class 1 demonstrates a much higher likelihood of responding “Yes” to R2, while Class 3 shows a higher likelihood of responding “Yes” to both R2 and R5. Additionally, Class 3 has a markedly higher likelihood of responding “Yes” to R4 compared to the other classes.

The three latent classes represented in Figure 2 represent the three distinct subgroups within our data, each with unique response patterns. Class 1 (Red polygon) has lower probabilities of responding positively across most items compared to the other classes. R2 has a higher probability, but its overall profile suggests more uniform and lower levels of agreement with the items. Class 2 (Green polygon) has the highest probabilities of responding positively (“Yes”) for several items (e.g., R1, R5, and R6). This indicates that individuals in Class 2 are more likely to agree or identify with the statements in these items, distinguishing this group as the most positively inclined class. Class 3 (Blue polygon) falls between Class 1 and Class 2. Class 3 has noticeable peaks for certain items (e.g., R4 and R8), suggesting that individuals in this class are more likely to agree with these items compared to Class 1 but less so than Class 2 (Figure 2).

Figure 2. Item by class. Blue represents the response "No" (Y=1), while orange represents the response "Yes" (Y=2). Taller bars for a given response level indicate a higher likelihood of that response within the respective class. Each subplot corresponds to a specific item from the questionnaire.

Figure 2. Item by class. Blue represents the response "No" (Y=1), while orange represents the response "Yes" (Y=2). Taller bars for a given response level indicate a higher likelihood of that response within the respective class. Each subplot corresponds to a specific item from the questionnaire.

Figure 3. Radar chart representing the three distinct classes in the LCA. Each polygon (red, green, blue) represents the response patterns of a specific class. The average polygon represents the overall pattern across all classes. The values along the axes represent probabilities (from 0 to 1) showing the likelihood of a particular response for that item within each class.

Figure 3. Radar chart representing the three distinct classes in the LCA. Each polygon (red, green, blue) represents the response patterns of a specific class. The average polygon represents the overall pattern across all classes. The values along the axes represent probabilities (from 0 to 1) showing the likelihood of a particular response for that item within each class.

Items like R1, R2, R4, R5 and R6 are the most discriminative, as they show the most variation across classes. These items may be key to distinguishing the underlying traits or attitudes of each class.

We superimposed the patients' answers regarding the reasons for refusing vaccination on the roots of attitudes described in the specialized literature [22]. Statistical analysis shows that the most frequent roots of the identified attitudes are: distorted risk perception (n=128), fear and phobia (n=93), distrust (n=53), perceived selfinterest (n=31) –

Table 5. Root of attitudes for influenza vaccine refusal.

Root of attitude	Reason of refusal	n (%) Total=272
Conspiracist ideation	R10 The vaccine is made to control people	7 (2.6%)
Distrust	R3 I don't have enough information about the vaccine R8 I don't trust the vaccine	53 (19.5%)
Unwarranted beliefs	-	0
Worldview and politics	R11 I don't use any vaccine anymore. I believe that the COVID vaccine was a business R12 It's too expensive	27 (9.9%)
Religious concerns	R9 I do not vaccinate for religious reasons	6 (2.2%)
Moral concerns	-	0
Fear and phobias	R2 I am afraid of adverse effects R13 I am afraid of injections	93 (34.2%)
Distorted risk perception	R1 I don't think it's necessary R5 I've never had the flu and I don't think I need it	128 (47.1%)
Perceived self-interest	R6 I consider myself to have a strong immune system	31 (11.4%)
Epistemic relativism	R4 I have had a previous negative experience with a vaccine R7 I have chronic diseases that can get worse	30 (11.0%)
Reactance	-	0

.

Interpretation of the logistic regression results for distorted risk perception as a predictor of influenza vaccination refusal found that marital status did not have a statistically significant impact on flu vaccination refusal (p>0.05). Age was not statistically significant (p>0.05), although it approached significance (p=0.080). There was a slight increase in the odds of refusal with each additional year of age (OR=1.018). Having chronic conditions was found to be associated with a significant decrease in the odds of flu vaccination refusal by approximately 57.2% (p=0.027). Individuals with a history of influenza vaccination had 76.2% lower odds of refusal flu vaccination due to a distorted risk perception (p<0.001).

Based on the logistic regression analysis examining fear and phobia, significant predictors include gender (females exhibited an 83.3% increase in the odds of refusal, p=0.046), being a parent (associated with a substantial 1445.7% increase in the odds of refusal, p<0.001), and having chronic diseases (associated with a 1393.1% increase in the odds of refusal, p<0.001). Non-significant predictors include age, COVID-19 vaccination status, and a history of past vaccination with other vaccines (p>0.05).

According to the logistic regression analysis, marital status (unmarried or widowed) significantly increased the odds of distrust in flu vaccination by 175.8% (p=0.021). Membership in the category of parents significantly reduced the odds of distrust by approximately 91.3% (p=0.004). Similarly, being vaccinated against COVID-19 decreased the odds of distrust by approximately 84.8% (p=0.001). Conversely, having a history of receiving other vaccines significantly increased the odds of distrust by approximately 604.6% (p<0.001). Age, gender, environmental factors, chronic diseases, and past flu vaccination did not have a statistically significant effect on distrust (p>0.05).

Discussion

This study provides valuable insights into the factors influencing influenza vaccine refusal, within rural and urban populations in Romania, highlighting key sociodemographic and behavioral predictors. The findings underscore the complexity of vaccine hesitancy, which is influenced by diverse factors ranging from marital status and parenting status to prior vaccination history and gender.

Vaccine hesitancy remains a significant issue in Romania, affecting numerous vaccines, including pediatric ones. This has contributed to several measles outbreaks in Romania, resulting in 64 deaths between 2016 and 2020 [23,24], and 21 additional deaths during 2023-2024 [25]. Among adults, the perception of the need for lifelong vaccination is notably low, as reflected in Romania’s poor acceptance rates of vaccination against COVID-19.

Through the 13-item vaccine hesitancy questionnaire, key patterns of hesitancy were identified. The LCA distinguished three distinct subgroups (latent classes) among respondents. These patterns could help us understand underlying behavioral or attitudinal differences related to the items. The three latent classes identified through the LCA align with patterns observed in previous research on vaccine hesitancy, emphasizing the diversity in hesitancy drivers and attitudes across populations. Class 2, representing individuals more open to vaccination, reflects a segment that aligns closely with those described in the literature as “acceptors” or “pro-vaccine” individuals. Studies have shown that this group often exhibits higher trust in health authorities and vaccines, influenced by positive experiences with vaccination or a strong belief in the collective benefits of herd immunity [26]. Interventions for this group may not require extensive effort, as they are likely to respond well to general public health messaging reinforcing the benefits and safety of vaccines.

Class 1, on the other hand, appears to align with those labeled as "vaccine hesitant" or "resisters" in prior research [27]. These individuals often harbor deep-seated concerns about vaccine safety, efficacy, and trust in health systems. Factors such as exposure to misinformation, cultural or religious beliefs, and previous negative healthcare experiences may play a significant role in shaping their attitudes [28]. Addressing this class requires more intensive and personalized strategies, such as community-based education, engagement with trusted community leaders, or transparent communication addressing specific fears and misconceptions.

Class 3, the intermediate group, aligns with descriptions of “fence-sitters” in the literature – individuals who may neither strongly accept nor outright reject vaccines but are influenced by contextual factors such as perceived risk, social norms, or specific concerns like the one highlighted (e.g., distrust in the vaccine itself). Studies have suggested that targeted interventions, such as addressing perceived risks versus benefits and leveraging positive peer influence, can be effective in nudging this group toward vaccine acceptance [29]. Tailored messaging that acknowledges their concerns while providing clear, evidence-based information may be key to converting hesitancy into acceptance.

Understanding these classes not only highlights the multifaceted nature of vaccine hesitancy but also underscores the importance of segmented approaches in designing public health interventions. Such differentiation allows for the deployment of resources more effectively, targeting those who need them most while maintaining the trust of those already inclined to vaccinate.

The logistic regression analysis revealed that marital status is significantly associated with distrust toward influenza vaccination, suggesting that factors such as the lack of familial support networks may contribute to attitudes of distrust. Similar results have been reported in the literature, with a positive association between marital status and influenza vaccine uptake [30,31]. Conversely, being a parent was associated with a significant reduction in the odds of distrust, which may indicate a greater awareness of the role of vaccination in protecting both the individual and their family [32,33].

Secondly, the majority of respondents were middle-aged or older, with a significant proportion reporting chronic illnesses. Despite these risk factors, hesitancy persisted, particularly among those citing fears of adverse effects, distrust in vaccines, or a belief in natural immunity. Similar findings have been reported in the literature, with fear of adverse effects emerging as the most prominent factor [34,35].

We further identified geographic disparities in vaccine refusal, with a significant association between the area of residence (urban/rural) and influenza vaccine refusal, with rural residents being more likely to decline vaccination. This finding is consistent with previous studies showing that rural populations often face unique barriers [36], including limited healthcare access [37], lower health literacy [38], and deeply ingrained cultural or religious beliefs [39], or other socio-demographic factors [40].

Our study highlights the need for comprehensive interventions to address the fear of post-vaccination adverse effects, including the implementation of social prescribing measures, special in rural areas, to facilitate greater acceptance of vaccination [41].

The findings emphasize the need for geographically and demographically tailored public health campaigns to mitigate vaccine hesitancy and vaccine refusal. In rural areas, strategies should focus on improving access to healthcare services, enhancing health education, and engaging trusted local healthcare providers or other stakeholders to build trust in vaccines and increase awareness in vaccine preventable diseases risk perception. Furthermore, addressing specific hesitancy drivers—such as fear of adverse effects, distrust in vaccines, and perceived lack of necessity—will be crucial. The distinct patterns identified through LCA suggest that interventions should be customized for different subgroups to maximize their effectiveness.

While this study provides important insights, several limitations should be considered. First, the reliance on self-reported data may introduce bias, and the sample size may limit the generalizability of findings. Second, the cross-sectional design captures attitudes at a single point in time, which may not fully reflect dynamic changes in hesitancy over time. Future research should explore longitudinal trends in vaccine hesitancy and evaluate the impact of tailored interventions on improving vaccine acceptance.

Conclusions

Vaccine hesitancy remains a multifaceted challenge, with rural populations being disproportionately impacted. Implementing targeted and context-specific interventions that address key factors such as distorted risk perception, fear of adverse effects and distrust in vaccines is essential for enhancing vaccination coverage. These results highlight the necessity of tailored public health strategies to ensure equitable vaccine access and acceptance.