This study employed a multicenter cross-sectional design to assess influenza VH among older adults in Shanghai, China. Figure 1 displayed a theoretical model diagram based on WHO definition of vaccine hesitancy, referring to delay in acceptance or refusal of safe vaccines despite availability of vaccination services (30). Following the WHO definition of VH, participants were asked to indicate their willingness to receive the influenza vaccine by responding to the question: “Are you willing to receive the influenza vaccine?” The response options were: “Refuse all,” “Refuse but unsure,” “Refuse some,” “Delay,” “Accept some,” “Accept but unsure” and “Accept all,” a 7-point scale recommended by the SAGE working group on vaccine hesitancy (23). “Refuse all” is an extreme form of vaccine hesitancy, which is also classified as vaccine hesitant in many studies (31–33). In order to conduct more in-depth research, this study further subdivided vaccine hesitant people into vaccine acceptance with doubts and vaccine refusal and refusal with doubts. (see Figure 1) “Refuse all,” “Refuse but unsure,” “Refuse some” or “Delay” were categorized as vaccine refusal and refusal with doubts, indicating reluctance or refusal toward vaccination despite some level of concern. Similarly, participants who selected “Accept some” or “Accept but unsure” were considered as vaccine acceptance with doubts, meaning they were willing to receive the vaccine but still harbored uncertainties. Those who chose “Accept all,” corresponding to a direct acceptance, were categorized as non-hesitators, signifying complete vaccine acceptance without concerns. Using the same classification approach, we also assessed participants’ hesitancy toward COVID-19 vaccination.

A structured questionnaire assessed VH, 3Cs model components (confidence, complacency, convenience) (see Table 1), VL via the Health Literacy about Vaccination in Adults (HLVa-IT) scale (see Table 2). The Geriatric Depression Scale (GDS-15), a validated tool for assessing depression in older adults (34, 35) (Cronbach’s alpha ≈ 0.80), was used to measure depressive symptoms, with scores ≥5 indicating potential depressive symptoms, but not a definitive diagnosis. Previous research defines complacency as encompassing low perceived risk of contracting influenza, limited knowledge of influenza and its vaccine, and prejudices against vaccination (36). Our questionnaire retained core elements of complacency, including perceived severity (Items 1–3) and susceptibility (Items 4–5) to influenza and co-infection. To better capture beliefs prevalent among older adults, we added four items (Items 6–9): reliance on influenza-specific medications, belief in personal immunity, perceived lack of necessity for older adult vaccination, and trust in non-pharmaceutical interventions (NPIs) like masks based on previous research (37). Additionally, the survey incorporated other potential determinants of influenza VH, including engagement with community health workers (e.g., family doctor sign-up rates, frequency of vaccine-related health communication), history of chronic illnesses and falls, allergy history, exposure to negative vaccine-related information and previous experiences with vaccination. A supplementary questionnaire was developed to assess general vaccine knowledge among older adults, with a specific focus on common misconceptions about influenza vaccines, adapted from prevalent online misinformation sources. Additionally, VL was assessed using the HLVa-IT tool, which evaluates functional, interactive, and critical VL (see Table 2). Scores were calculated based on the mean values of responses, ranging from 1 to 4. Higher scores on functional VL indicate higher levels of vaccine literacy, whereas higher scores on interactive and critical VL are associated with lower levels of vaccine literacy.

Lastly, a pilot survey was conducted with 30 older adults to evaluate the clarity and readability of the questionnaire, with minor modifications made based on feedback. The final survey required about 15 min for completion.

According to the Law of the People’s Republic of China on the Protection of the Rights and Interests of the Elderly (38), citizens aged 60 and above are defined as older adult in China. A stratified random sampling approach was used to ensure a representative sample of older adults (≥60). The study was conducted in the Xuhui district, the largest urban district in Shanghai, China from January to June 2024. Participants were randomly recruited through a community-based management system, an official registry that includes residents of all ages and is commonly used for sampling in population-based surveys. We set the following inclusion Criteria: (1) Community-dwelling adults aged ≥60 years in Xuhui District; (2) Permanent residency status (≥6 months residence in Shanghai during the preceding 12 months, irrespective of household registration); (3) Preserved cognitive capacity with ability to provide informed consent; (4) Verbal competence for clear self-expression; (5) Voluntary participation. Exclusion Criteria: (1) Diagnosed cognitive impairment or severe psychiatric disorders; (2) Decompensated chronic conditions (cardiac failure, respiratory failure, etc.); (3) Functional dependency; (4) Restricted mobility. Prior to survey participation, all respondents provided written informed consent. This document explicitly assured that: (1) All responses would remain strictly confidential through anonymized data processing; (2) Participation would entail no negative consequences for individuals; and (3) The study involved no human experimentation. Collected data served exclusively to assess vaccine literacy levels among older adult residents in Shanghai and inform evidence-based health policy development.

A total of 1,313 older adult individuals aged ≥60 years were initially recruited. Before data collection, 26 trained enumerators from 13 community centers received standardized training to address inquiries during the administration of questionnaires for the older adult. Surveys were conducted via face-to-face interviews, during which enumerators provided clarification on survey questions and directly entered responses into an electronic questionnaire platform.¹

The questionnaire was designed to take a minimum of 15 min to complete. Of the 1,313 recruited participants, four declined to participate, and nine participants failed to finish the questionnaires. A total of 1,300 fully completed questionnaires were included in the final analysis. The required sample size was estimated using the standard formula for cross-sectional studies with an α error of 0.05 and a permissible margin of error (δ) of 0.05:

N represents the sample size, μ is the standard normal deviation of α, μ_α = 1.96 when α error of 0.05, and p represents the estimated prevalence of influenza VH. Based on a prior study, the influenza VH rate in mainland China was 37.18%, which informed the sample size calculation (39). Design effect (deff) equals 2 in this study.

We performed factor analysis to verify the reliability and validity of each questionnaire scale. Confirmatory Factor Analysis (CFA) was employed to validate the predefined factor structure. Items demonstrated standardized factor loadings >0.50, exceeding the minimum threshold, were considered as acceptable item retention. Composite reliability (CR) > 0.7 and squared multiple correlations (SMC) > 0.4 were considered as high aggregation validity (40). Model fit indices can further confirm structural validity (e.g., CFI > 0.90, RMSEA < 0.08). Additionally, Cronbach’s alpha coefficients exceeding 0.70 were considered as robust internal consistency reliability (41).

Descriptive statistics were used to summarize participant characteristics and VH status. The chi-square test was applied to examine the associations between influenza VH and key determinants, including socio-demographics, the 3Cs model components, VL levels, and other potential influencing variables. Multinomial logistic regression was performed to identify independent predictors of VH, with “No hesitancy” serving as the reference group, adjusting for confounders including sociodemographic variables, self-report health status and self-report vaccination experiences. After that, we employed structural equation modeling (SEM) to examine the underlying factors contributing to vaccine hesitancy and quantify their interrelationships. SEM assumed the effect of 3C factors as three latent variables on influenza VH combined with other factors screened by multinomial logistic regression. After fineness optimization and further modifications, the final integrated model for influenza VH was established.

We also used Stepwise Regression and calculated the Variance Inflation Factor (VIF) for all predictor variables to identify and eliminate variables that did not contribute much to the model, thus reducing the effect of multicollinearity. Factors with VIF values greater than 10 imply multicollinearity and will be excluded from the model. Adjusted Odds ratios (aORs) and 95% confidence intervals (CIs) were reported for each included determinant. A two-sided p-value <0.05 was considered statistically significant. Data analyses were conducted using R software (version 4.4.2).

Based on stepwise regression results and the values of VIFs, we excluded several variables due to multicollinearity and model fit concerns: COVID-19 vaccine hesitancy, Confidence items 4–5, and Convenience items 7–8. For the retained 3Cs model items, standardized factor loadings ranged from 0.642 to 0.887, greater than the 0.50 threshold (see Table 3). Model fit indices (CFI = 0.921, TLI = 0.901, RMSEA = 0.097, SRMR = 0.075) were marginally acceptable. Cronbach’s α of confidence, convenience and complacency were 0.889, 0.841, and 0.754, respectively, all exceeding the 0.70 reliability standard. For the HLVa-IT scale adapted for older adult populations, standardized factor loadings spanned 0.719–0.917 (all > 0.50) (see Table 3). Model fit indices (CFI = 0.947, TLI = 0.935, RMSEA = 0.097, SRMR = 0.042) were acceptable fit. Both scales demonstrated robust structural validity and internal consistency (Table 4).

One thousand three hundred older adult participants completed the survey, with 53.54% male and 46.46% female (see Table 5). The average age of the respondents was 70.2, with the highest proportion of participants in the 66–75 age group (57.7%). Education levels varied, with 55.3% of participants having completed high school or technical school, while 38.6% had not received secondary education. Most respondents were married (88.38%) and living with family (91.92%). In terms of financial status, 72.69% of respondents reported a monthly income between 3,500 and 6,500 RMB, and 9.08% earned below 3,500 RMB, suggesting moderate economic diversity.

Health-related characteristics also varied within the sample. 73.54% of the older adult had undergone physical examinations in the past year, and 61.4% reported having chronic diseases. 21.1% of the older adult population scored five or above on the GDS-15, suggesting a potential predisposition to depression. Most respondents self-rated their health status as normal (56.3%) and relatively good (35.5%), while 8.2% perceived their health as poor. 64.8% had experienced influenza-like symptoms in the past year, and VH was significantly higher in this group (p < 0.05). A considerable proportion of participants reported needle phobia (60.8%) and concerns about vaccine side effects (78.1%), both of which were strongly associated with hesitancy (p < 0.05). Vaccination history showed high coverage of COVID-19 vaccines (82.69%), exceeding the WHO target of ≥70% coverage for total populations by mid-2022 (42). While only 17.77% had received an influenza vaccine, and 11.3% had received a 23-valent pneumococcal polysaccharide vaccine (PPSV23). Although most participants reported no direct side effects from vaccination, 15% indicated hearing negative information about vaccines from neighbors, which contributed to greater hesitancy (p < 0.05). Regarding engagement with healthcare providers, 73% of the older adult had registered with a family doctor, yet over 50% reported infrequent health communication with their physicians. Additionally, over half of the respondents did not actively enquire about influenza vaccination with their family doctors, indicating potential gaps in vaccine promotion (p < 0.05).

Only 21.62% of respondents demonstrated high vaccine knowledge, with higher proportions in the non-hesitant group (n = 74) compared to refusers (n = 35). Conversely, 39.3% had low vaccine knowledge, with the highest proportion among hesitancy refusers (n = 110). Figure 3 displays the distribution of VL scores across different hesitancy status. Participants with no hesitancy exhibited the highest scores across all three domains of VL. Functional literacy remained relatively stable, but interactive and critical VL levels were significantly lower in refusers, indicating difficulty in analyzing and interpreting vaccine-related information. Regression analysis (see Table 6) confirmed that lower vaccine knowledge significantly increased hesitancy. Individuals with moderate vaccine knowledge were 2.09 times more likely to hesitate (95% CI: 1.08–4.06), while those with low knowledge had 1.85 times greater odds (95% CI: 1.73–4.41). Additionally, the older adult who had higher critical vaccination literacy were more likely to refuse influenza vaccines (aOR = 0.36, 95% CI: 0.29–0.45).

Older adults who had needle phobia (aOR = 3.93; 95% CI: 1.51, 10.25) were more likely to be vaccine refusal. Also, the effect of potential depression symptoms was more pronounced in the vaccine refusal older adult population (aOR = 2.72; 95% CI: 1.14, 6.53).

Table 5 presents the goodness-of-fit indices for the integrated model of influenza vaccination hesitancy. All indices fell within acceptable ranges, indicating adequate model fit (43). Each pathway was statistically significant using standardized estimated coefficient (β) based on the best final integrated model (see Figures 4, 5).

Among vaccine acceptors with hesitancy, confidence (β = −0.026), convenience (β = 0.284) and complacency (β = 0.192) all significantly and directly affected VH. Vaccination history and vaccine knowledge also contribute positively to vaccine hesitancy, with standardized coefficients of 0.429 and 0.149, respectively. There were significant covariances between confidence and complacency (β = 0.247), vaccine knowledge and confidence (β = 0.658), influenza vaccination history and convenience (β = 0.570). Among vaccine refusers, confidence (β = 0.314), convenience (β = 0.5) and complacency (β = 0.234) all significantly and directly affected vaccine refusal. Depression, side-effect experience and needle phobia also contribute positively to vaccine hesitancy, with standardized coefficients of 0.11, 0.297 and 0.038, respectively. Critical vaccine literacy negatively affected vaccine refusal (β = −0.114) and confidence (β = −0.350). There were significant covariances between confidence and complacency (β = 0.846), side-effect experience and complacency (β = 0.293), side-effect experience and depression (β = 0.294), convenience and depression (β = 0.293), side-effect experience and needle phobia (β = 0.362).

Figure 6 illustrates the distribution of VH status for influenza and COVID-19 vaccines. Among all participants, the proportion of non-hesitators (those who fully accepted the vaccine) was lower for influenza (14.8%) compared to COVID-19 (19.92%), suggesting greater reluctance toward influenza vaccination. The largest subgroup among both vaccines was refusers with doubts, accounting for 45.2% for influenza and 37.0% for COVID-19. Absolute refusers were also slightly more prevalent for influenza (16.2%) than for COVID-19 (14.68%), reinforcing the stronger hesitancy toward influenza vaccines.

The 3Cs framework illuminated the central role of confidence, complacency, and convenience in driving influenza VH. A lack of trust in vaccine effectiveness and safety emerged as a key barrier, reflecting broader concerns about healthcare systems and information credibility. Complacency was evident in the underestimation of influenza’s risks, particularly its potential to affect family members, which was further compounded by reliance on alternative preventive measures. Convenience issues, such as limited family support and access to vaccination services, also hindered uptake.

Notably, critical VL emerged as a dual-force factor negatively influencing vaccine refusal and confidence, underscoring the necessity for customized literacy interventions that target specific drivers of vaccine hesitancy. Additionally, the study identified a significant influence from COVID-19 VH spillover, likely stemming from the high acceptance of COVID-19 vaccines compared to influenza vaccines. This spillover appears to shape perceptions of disease severity and necessity, aligning with the complacency dimension, and underscores the interconnected nature of vaccine attitudes in the post-pandemic era (44).

Confidence in vaccine efficacy emerged as the strongest predictor of hesitancy, reinforcing previous research that trust in vaccines and healthcare providers is a crucial determinant of vaccine acceptance (45, 46). Participants who distrusted the effectiveness of influenza vaccines were over 25 times more likely to refuse vaccination, a pattern not observed for COVID-19 vaccines. Unlike the widespread trust fostered by COVID-19 vaccination campaigns, skepticism toward influenza vaccines persists, possibly due to lingering misinformation or historical perceptions of vaccine efficacy (47–49).

Additionally, concerns about vaccine safety played a role, though its impact was less pronounced than efficacy-related concerns. This aligns with prior studies indicating that, among older adults, the perceived effectiveness of vaccination influences decision-making more than safety concerns (50). Addressing these concerns requires targeted communication strategies that reinforce trust while acknowledging safety considerations, potentially leveraging community health workers to bridge the confidence gap.

Complacency emerged as a significant contributor to VH, driven by a tendency to downplay the severity of influenza and its complications, especially within family contexts. This perception was exacerbated by the availability of influenza-specific medications and the belief that NPIs, such as wearing masks or social distancing, provided sufficient protection. Previous studies have highlighted similar patterns, where low perceived risk reduces the motivation for vaccination (51). The post-COVID-19 context further complicates this, as the heightened focus on COVID-19 may have diminished attention to influenza risks, including the dangers of co-infection (5). Given the ongoing circulation of both viruses, public health messaging should emphasize the added risks of co-infection and highlight how influenza vaccination complements COVID-19 vaccines in reducing disease burden. Moreover, older adults who accept vaccines with doubts were more likely to mistakenly believe vaccines become unnecessary with age (e.g., “I’ve lived long enough”; aOR = 2.55, 95% CI:1.01–6.40). Public health efforts should therefore emphasize the cumulative risks of influenza and the complementary role of vaccination, particularly for older adults who may misjudge the benefits due to age-related assumptions.

Despite VL’s theoretical importance, its role in directly influencing influenza VH was limited, prompting a deeper exploration of its interactions with the 3Cs dimensions. Functional VL, which involves basic comprehension of vaccine information, did not independently reduce hesitancy, possibly because basic knowledge alone does not address deeper attitudinal barriers like distrust (28). However, functional VL likely interacts with confidence, as improved understanding of vaccine benefits could enhance trust in efficacy, though this effect may have been overshadowed by the pandemic context. Interactive VL, which involves seeking and discussing vaccine information, showed a nuanced effect, appearing more relevant among acceptors with doubts, potentially by facilitating engagement with CHWs and reducing uncertainty (52). Unexpectedly, higher critical VL, which reflects the ability to critically evaluate vaccine information, was associated with greater hesitancy, possibly due to increased skepticism or susceptibility to misinformation among those who critically appraise health messages (53, 54). This suggests a complex interaction with complacency, as critical evaluators may be more likely to endorse alternative protective measures like NPIs. These VL-3Cs interactions indicate that VL’s impact is not uniform but varies depending on the attitudinal and contextual factors captured by the 3Cs model, highlighting the need for tailored literacy interventions that address specific hesitancy drivers.

Structural barriers significantly influenced VH, with inadequate family support and limited access to vaccination services posing substantial obstacles. The role of social networks in decision-making aligns with previous research, highlighting the need for family involvement in health interventions (55). Interestingly, the lack of engagement with community health workers (CHWs) appeared to have a complex effect, potentially indicating that refusers sought information from alternative sources (e.g., social media platforms or interpersonal networks with peers/relatives). This is particularly concerning, given that over 50% of participants had infrequent or no discussions about influenza vaccination with their family doctors. This gap underscores the compounding role of provider engagement: prior research has demonstrated that consistent, clinician-led counseling remains a pivotal driver of vaccine confidence (56). To disrupt this cycle, healthcare systems must mandate provider-initiated vaccine dialogues during clinical encounters, particularly for high-risk groups with fragmented access to credible health information. Embedding such discussions into standard care protocols could transform passive patient-provider interactions into trust-building opportunities, ensuring that vaccine hesitancy is preemptively addressed rather than passively observed.

The negative spillover from COVID-19 VH significantly shaped influenza VH, as the higher trust in COVID-19 vaccines did not translate into confidence in influenza vaccines (57). This spillover likely amplified complacency toward influenza, as older adults prioritized COVID-19 risks, underestimating influenza’s threat. The distinct hesitancy patterns-influenza vaccines facing more outright refusal compared to COVID-19 vaccines attracting more hesitant acceptors-suggest that lessons from COVID-19 vaccine campaigns, such as government transparency and media outreach, could be adapted to improve influenza vaccine uptake.

This study has several strengths. It is one of the first studies in China to examine influenza VH using the 3Cs model alongside VL factors. The large, community-based sample enhances the generalizability of the findings. Additionally, the inclusion of COVID-19 VH as a comparison provides important insights into the broader vaccine confidence landscape. However, some limitations should be acknowledged. This study was conducted exclusively in Shanghai, a metropolitan area with higher socioeconomic status and advanced healthcare infrastructure. The findings may not fully represent populations in rural regions where distinct challenges exist, including differential access to healthcare services and variations in vaccine literacy levels. Caution should therefore be exercised when generalizing these results to non-urban settings. Future multi-regional studies are needed to provide a more comprehensive understanding of the older adult’s hesitancy toward influenza vaccines across China’s diverse populations. Additionally, the cross-sectional nature of the study prevents causal inferences about VH determinants. Future longitudinal studies are needed to assess changes in vaccine attitudes over time, particularly in the evolving post-pandemic landscape.