Vaccinating with Valor: A Risk Preventive Model to Explain Factors in Parents’ Choice to Vaccinate Their Children for COVID-19

ABSTRACT

Childhood immunization can effectively control and prevent infectious diseases; however, not all parents choose to vaccinate their children against vaccines, including COVID-19. This study aimed to determine potential factors influencing people’s willingness to vaccinate their children. An online survey was conducted with 509 adult parents/guardians of children to test our hypotheses. Based on the TPB model with the RISP model as the antecedent, results indicated that people’s systematic risk information processing, trust in science, and concerns about the disease positively influenced their cognitive structure, further impacting their attitude toward vaccinating their children. The results also verified that attitude toward vaccination and perceived behavioral control are both significant predictors of parents/guardians’ intention to vaccinate their children. The results contribute to health risk communicators creating effective strategies to better communicate with adults and increase intentions to vaccinate their children.

GRAPHICAL ABSTRACT

Introduction

Childhood immunization has proven to be the most effective public health strategy to control and prevent disease, which became a key intervention to reduce childhood morbidity and mortality during the COVID-19 outbreak (World Health Organization, 2020). Some studies indicated vaccination against COVID-19 can benefit minors; although mild, local adverse reactions may occur, serious adverse events caused by vaccination are rare (Fowlkes et al., 2022; Hause et al., 2021; Hause, Baggs, et al., 2022; Hause, Marquez, et al., 2022). Although the pros of vaccination outweigh the cons, by the end of 2022, more than half of children from six months to 17 years old in the United States have not received their first dose of the COVID-19 vaccine (American Academy of Pediatrics, 2022). The decision to vaccinate is often not the child’s decision, it is the decision of the parents/guardians.

Vaccination is a common practice for many, yet some individuals may hesitate to receive vaccines, a phenomenon observed in different diseases (MacDonald, 2015; Marti et al., 2017); this decision, particularly concerning their children, can be even more challenging, with the topic of vaccine hesitancy heightened attention during the COVID-19 pandemic. Vaccine hesitancy refers to the choice to delay acceptance or refusal of getting vaccines despite the availability of vaccination services (MacDonald, 2015; World Health Organization, 2014). People’s hesitation or refusal of vaccines is thought to be the main reason for low immunization coverage and the high risk of outbreaks of vaccine-preventable diseases (Dubé et al., 2013). When the number of people vaccinated is not high enough to reach a certain threshold, it cannot achieve herd immunity and completely eradicate the infectious disease (Fine et al., 2011). More than 70% of countries reported that vaccine hesitancy was a problem (Marti et al., 2017); therefore, increasing public willingness to receive the COVID-19 vaccine for adults and their children has become an urgent concern worldwide.

Risk communication is an effective way to allow the public acquire a correct understanding of science and reduce vaccine hesitancy (MacDonald, 2015). Many statements against COVID-19 vaccines indicated the distrust of science and suspicion of risks caused by the vaccine (Küçükali et al., 2022; Pullan & Dey, 2021). Although these risks may be untrue, they mirror the viewpoints of people with vaccine hesitations and are similar to research findings about other vaccine hesitancy. Distrust of vaccine safety, lack of knowledge and awareness of infectious diseases and prevention, and other culturally contextual factors (such as religion, gender, or socioeconomic status) were cited as the most common predictors of vaccine hesitancy (Benin et al., 2006; Bond et al., 1998; Dubé et al., 2013; Marti et al., 2017; McKee & Bohannon, 2016; Salmon et al., 2009). While these factors to elude to predictors of vaccine hesitancy, it is necessary to explore additional factors that contribute to vaccine hesitancy (MacDonald et al., 2018) to help inform communication strategy.

In addition to cultural context factors (i.e., religion, gender, or socioeconomic status), trust in science and risk perceptions have been found to have relationships with vaccination status (Baker et al., 2023; McKee & Bohannon, 2016). Individuals with higher trust in science are more likely to have decreased vaccine hesitancy and are more encouraged to get the vaccine (Baker et al., 2023). In contrast, those who have stronger concerns about the risks tend to be less likely to vaccinate (Troiano & Nardi, 2021).

Moreover, although previous studies have examined the public’s intent to vaccinate themselves (Li & Zheng, 2022; Nah et al., 2023; Yang et al., 2022; Zhou et al., 2021), no evidence has shown the relationships between them and people’s decisions to vaccinate their children during COVID-19. Given the absence of prior research focusing on parents’ intentions to vaccinate their children during the pandemic, this study addressed a significant gap. Considering the distinct dynamics involved in deciding to vaccinate oneself versus others, as highlighted by Xiao and Wong (2020), it becomes imperative to explore what drives parental decision-making in the context of child vaccinations.

This investigation is not only essential but also valuable, as it aims to uncover key factors that influence parents’/guardians’ intentions to vaccinate their children against COVID-19. Our research integrated key variables identified from numerous studies on COVID-19 vaccination to construct a comprehensive risk prevention model. This model elucidated the primary factors driving parental decisions to vaccinate their children against COVID-19. The insights gained from this study are expected to significantly enhance the development of effective health communication strategies specifically tailored for parents, thereby contributing to more informed and potentially increased vaccination rates among children.

Theoretical framework

The Theory of Planned Behavior (TPB) model has been extensively utilized to elucidate the decision-making process in behaviors where individuals exercise self-control and to predict their intention to engage in specific behaviors (Ajzen, 1991, 2012). Within the TPB framework, potential influencing factors extend beyond the foundational elements of attitude toward the behavior, subjective norms, and perceived behavioral control. The model also encompasses additional components, such as cognitive structure and normative belief structure (Figure 1). These later additions enrich the model’s capacity to capture a more comprehensive array of determinants influencing behavioral intentions.

Figure 1. TPB model (Ajzen, 1991, 2012).

This shows the factors in the Theory of Planned Behavior model.

In environments characterized by risk, the TPB model can be used to predict individuals’ intentions for risk-preventive behavior. Taking the Risk Information Seeking and Processing (RISP) model as the antecedent can further delve into the dynamics of how individuals navigate and manage known risks within such environments (Griffin et al., 1999).

The TPB model with the RISP model as the antecedent is well-suited in this study for explaining why people choose to vaccinate their children against COVID-19 due to its comprehensive approach to assessing how individuals process and respond to health risks. This model integrates various psychological and social factors that are crucial in shaping parental decision-making and can effectively explore and predict parental behaviors in the context of a global health crisis like the COVID-19 pandemic.

The original TPB model posits that intention to perform a behavior – and subsequently the performance of that behavior – can be predicted by one’s attitude toward the behavior, their subjective norms regarding the behavior, and their perceived behavioral control (Ajzen, 1991, 2012). According to the TPB, attitude toward the behavior refers to the individual’s experiential and instrumental beliefs regarding that behavior. Subjective norms refer to the societal pressures one feels regarding the behavior. Finally, perceived behavioral control encompasses one’s confidence in their capacity and autonomy to perform the behavior successfully. Empirical research has found a positive association between the TPB variables and intentions to perform preventive health acts. For example, Arkorful et al. (2023) and Yang and Kim (2023) found that attitude, perceived behavioral control, and subjective norms were all linked to intentions to perform health-related behaviors.

The TPB model has demonstrated its efficacy in elucidating individuals’ vaccination intentions (Dillard, 2011; Li & Li, 2020; Xiao & Wong, 2020; Yang, 2015). Attitudes toward vaccines, subjective norms, and perceived behavioral control collectively shape the intention to vaccinate, underscoring the model’s robust applicability and critical relevance in understanding and predicting vaccination behaviors. Therefore, we expected that parents’ and guardians’ attitudes, subjective norms, and perceived behavioral control would all be positively associated with their intentions to have their children vaccinated for COVID-19.

The extended TPB model includes cognitive and normative belief structures. In the context of risk communication, cognitive structure refers to the schema in which individuals’ beliefs are organized (Griffin et al., 1999). Previous studies have indicated that when the RISP model is positioned as a precursor in the TPB framework, one’s cognitive structure can be influenced by various factors. First, trust in science and concerns about risks are pivotal in shaping cognitive structure, which in turn indirectly impacts attitudes and intentions toward engaging in preventive behaviors (Griffin et al., 1999; Yang et al., 2010). Trust in science refers to the confidence and belief individuals place in the credibility, integrity, and reliability of scientific information, institutions, and processes (Hendriks et al., 2016). Risk perceptions, or concerns, refer to the level of worry individuals have about potential hazards or uncertainties associated with a particular situation, action, or decision (Wolff et al., 2019).

To be more specific, vaccine research has revealed that distrust in medical systems can significantly influence individuals’ perceptions, often leading to vaccination hesitancy (Nah et al., 2023). Conversely, heightened awareness of disease risks tends to foster positive attitudes toward vaccinations and increase the likelihood of vaccine uptake (Chew et al., 2021; Li & Li, 2020; Xiao & Wong, 2020). In light of these insights, our study incorporated trust in science and COVID-19 concerns into the model to extend the TPB framework; we expected trust in science and scientists and COVID-19 concern to be positively associated with cognitive structure regarding COVID-19 and, indirectly, attitudes toward childhood COVID-19 vaccinations (Griffin et al., 1999).

Additionally, the systematic processing of risk information also contributes to one’s cognitive belief structure and attitudes toward an act (Griffin et al., 1999). Systematic information processing involves consciously and thoroughly analyzing information (Chen & Chaiken, 1999). When people process risk information systematically rather than in a heuristic way, they often make considerable cognitive efforts to actively digest and evaluate information (Kahlor et al., 2003); this trait further impacts people’s risk-related cognition and prevention intentions (Griffin et al., 1999; Li & Zheng, 2022). Thus, we expect the systematic processing of risk messages to also contribute to one’s cognitive structure and, subsequently, their attitudes and intention to have their child vaccinated for COVID-19.

Normative belief structure refers to the set of beliefs and perceptions individuals hold about social norms and pressures (Griffin et al., 1999). One’s normative belief structure positively impacts one’s subjective norms regarding risk-preventive behavior (Griffin et al., 1999). Based on this theoretical postulate, we expect one’s normative belief structure to be positively associated with one’s normative beliefs and, subsequently, one’s intention to vaccinate one’s children for COVID-19.

While the TPB model has been successfully applied to predict vaccine-related intentions in lots of empirical studies, its application to the specific context of parental decisions to vaccinate children against COVID-19 remained unexplored. This gap presented a critical opportunity for our research to extend the model’s application to a highly relevant and timely health issue. By doing so, this study aimed to provide a more comprehensive understanding of the factors influencing vaccination decisions, potentially improving health risk communication strategies (MacDonald, 2015). In addition, our study enhanced the novelty and practicality of the model in the health communication domain by incorporating factors such as trust in science, COVID-19 concerns, and systematic processing of risk information. This innovative approach offered valuable insights into parental decision-making regarding child vaccination, finding the niche in the current health communication literature.

Research purpose and hypotheses

This study aimed to determine what potential factors influence people’s willingness to vaccinate their children. This study tested and extended the existing TPB model by applying parts of the RISP model with trust in science/scientists and concern for the pandemic (Griffin et al., 1999). Based on the literature review, this study proposed eight research hypotheses and built a predictive model in deciding to let the children of citizens in the United States get the COVID-19 vaccine (Figure 2):

Figure 2. The hypothesized model.

The model shows the measurement and structural model of this study. Each latent variable was explained by two to five observed variables, and the hypotheses between the variables were shown.

H1:

Systematic risk information processing positively impacts cognitive structure.

H2:

The extent of trust in science/scientists positively impacts cognitive structure.

H3:

The extent of pandemic concern positively impacts cognitive structure.

H4:

Normative belief structure positively impacts subjective norm.

H5:

Cognitive structure positively impacts attitude toward acts.

H6:

Subjective norm positively impacts behavioral intention.

H7:

Attitude toward acts positively impacts behavioral intention.

H8:

Perceived behavioral control positively impacts behavioral intention.

Materials and methods

Data collection

To better understand how to effectively increase Americans’ intentions of vaccinating their children during the pandemic, online survey research was utilized to address the purpose. This study obtained a non-probability opt-in sample of United States adults who are parents/guardians of children that currently live together on a full or part-time basis by Qualtrics, an online platform. This study used an online survey to gather data from a non-probability sample of adults in the United States. Non-probability sampling, though not fully representative, was suitable for making population estimates in this context (Creswell & Creswell, 2018). It is a valuable tool for studies with limited resources or time constraints, and research has shown it can yield comparable results to probability sampling under certain conditions (Baker et al., 2013). This method also easily reaches members of the population of interest (Lamm & Lamm, 2019).

Respondents for this study were selected using Qualtrics, which incorporates advanced measures such as internet protocol (IP) address checks and digital fingerprinting technology to prevent duplicate responses and ensure data validity (Qualtrics, 2019). In recognition of the valuable time and input of participants, compensation was provided, encouraging their complete engagement with all survey questions. To maintain the integrity of our analysis, we excluded participants who did not complete the entire survey.

This research is part of a broader COVID-19 public opinion project, wherein data were collected from April 20 to June 7 2022, from 1,774 adults. A crucial aspect of the questionnaire involved identifying respondents who were parents or guardians of children currently residing with them. Those affirming this status were directed to a specialized survey section designed to gather responses pertinent to our research inquiries. To ensure the accuracy of their role as a parent or guardian, we employed a multi-step approach. Initially, the survey included a specific question where respondents were asked to confirm if they were parents or guardians of children currently living with them. This question was designed to be straightforward, requiring a simple “yes” or “no” response. To enhance the reliability of their responses, we integrated follow-up questions that delved into more details about their parenting or guardianship. These questions were aimed at gathering contextual information about the number of children in their care, the age range of these children, and the nature of their guardianship. Inconsistencies or contradictions in responses served as a flag for potential inaccuracies, leading to a review or exclusion of the respondent’s data. Ultimately, we retained data from 509 respondents who confirmed they were parents/guardians of children living with them either full-time or part-time. With the help of Qualtrics, the collected samples match the United States Census characteristics for gender, age, region, race, ethnicity, and socioeconomic status to ensure the survey sample is representative.

The majority of respondents indicated that they live with only one child (n = 262, 51.5%), followed by those living with two children (n = 165, 32.4%), with the number decreasing as the number of children increases. Additionally, the largest group of respondents reported that their children fall within the 5–11 age range (n = 221, 43.4%), followed by those aged 12–17 (n = 188, 36.9%).

Measurement

Participants’ responses were assessed on a 5-point Likert scale unless stated otherwise, ranging from 1 = strongly disagree to 5 = strongly agree. The questionnaire design was divided into the following parts. The specific measures are shown in Table 1, along with the results of Cronbach’s alpha, mean, and standard deviation of variables.

Table 1. Dimensions and items of the survey (N = 509).

Model Dimension	Item		α	M	SD
systematic risk information processing	sys_1	I thought about what actions I myself might take based on what I read.	.90	3.63	.93
	sys_2	I found myself making connections between the message and what I’ve read or heard about elsewhere.
	sys_3	I tried to think about the importance of the information in my daily life.
	sys_4	I thought about how the message related to other things I know.
	sys_5	I tried to relate the ideas in the message to my own personal experiences.
trust in science/scientists	trust_1	I trust scientists to create knowledge that is unbiased and accurate.	.93	3.51	1.07
	trust_2	I trust scientists to create knowledge that is useful.
	trust_3	I trust scientists to advise government officials on policy.
	trust_4	I trust scientists to inform the public on important issues.
concern about COVID-19 (6-point scale)	conc_1	How concerned are you currently about being personally at risk of contracting COVID-19?	.95	3.69	1.60
	conc_2	How concerned are you currently about your loved ones being at risk of contracting COVID-19?
	conc_3	How concerned are you currently about coming into contact with a person infected by COVID-19?
	conc_4	How concerned are you currently about dying from COVID-19?
	conc_5	How concerned are you currently about loved ones dying from COVID-19?
cognitive structure	CS_1	Giving the COVID-19 vaccine to my children is my responsibility and obligation toward other people.	.95	3.13	1.30
	CS_2	Giving the COVID-19 vaccine to my children is my responsibility to society.
	CS_3	Giving the COVID-19 vaccine to my children would feel like making a personal contribution to something better.
	CS_4	Giving the COVID-19 vaccine to my children would feel like the morally right thing.
	CS_5	Giving the COVID-19 vaccine to my children would make me feel like a better person.
normative belief structure	NBS_1	My family members think I should give the COVID-19 Vaccine to my children.	.91	3.07	1.23
	NBS_2	My friends think I should give the COVID-19 Vaccine to my children.
	NBS_3	My colleagues think I should give the COVID-19 Vaccine to my children.
subjective norm	SN_1	When it comes to giving the COVID-19 Vaccine to my children, I want to do what my family members think I should do.	.91	2.48	1.22
	SN_2	When it comes to giving the COVID-19 Vaccine to my children, I want to do what my friends think I should do.
	SN_3	When it comes to giving the COVID-19 Vaccine to my children, I want to do what my colleagues think I should do.
attitude toward acts	ATT_1	I feel positive when I think about giving my children the COVID-19 vaccine.	.95	3.12	1.46
	ATT_2	The COVID-19 vaccine is good for my child’s health.
perceived behavioral control	PBC	For me, giving my child the COVID-19 vaccine is an easy choice.		3.01	1.54
behavioral intention (3-point scale)	INT	Do you think your children should be vaccinated for COVID-19?		2.08	.90

First, three variables outside the TPB model, systematic processing of risk information, trust in science, and concerns about the pandemic, were included (Baker et al., 2023; Griffin et al., 1999). The scale proposed by Kahlor et al. (2003) was used to measure respondents’ propensity to process risk information, which included five questions (α = .90). The scale was used to examine the systematic processing strategies of people applying health risk information. The scale about people’s trust in science/scientists contains four questions (α = .93) from McCright et al. (2013). Respondents’ concern about the pandemic was also measured with a 6-point scale (1 = Not at all concerned to 6 = Extremely concerned) to know how much respondents care about the impact of COVID-19 on themselves and their loved ones (α = .95).

The second part of the survey was based on the variables from the original and extended TPB model (Ajzen, 1991, 2012). Six variables were included, with one to five items in each. This study also controlled for demographic characteristics, and the demographic characteristics of the sample are shown in Table 2.

Table 2. The characteristics of the sample (control variables).

Demographic		n	%
gender	male	212	41.7
	female	296	58.2
	other	1	.2
residency	a farm in a rural area	35	6.9
	rural area, not a farm	187	36.7
	urban or suburban area outside of the city limits	177	34.8
	subdivision in a town or city	74	14.5
	downtown area in a city or town	36	7.1
annual family income	$ 24999 or less	108	21.2
	$ 25000 to $ 49999	154	30.3
	$ 50000 to $ 74999	104	20.4
	$75,000 to $ 149,999	106	20.8
	$ 150,000 to $ 249,999	29	5.7
	$ 250,000 or more	8	1.6
highest educational level	less than 12th grade	18	3.5
	high school graduate (includes GED)	147	28.9
	some college, no degree	104	20.4
	2-year college degree (Associate, Technical, etc.)	79	15.5
	4-year college degree (Bachelor’s, etc.)	97	19.1
	graduate or professional degree	64	12.6
political belief	very liberal	40	7.9
	liberal	72	14.1
	moderate	260	51.1
	conservative	81	15.9
	very conservative	56	11.0
political affiliation	Republican	156	30.6
	Democrat	155	30.5
	Independent	144	28.3
	Non-affiliated	54	10.6

Analysis

The established extended TPB model from Ajzen (1999, 2012) has been confirmed by many studies that it can be used to analyze the relationship between intent variables and the intention to avoid risks by getting vaccines (Dillard, 2011; Xiao & Wong, 2020; Yang, 2015). Therefore, this study tested the data through this model and added several additional elements to explore the potential factors based on the RISP model (Griffin et al., 1999). Since the hypothesized model of this study was expected to confirm the complex relationship among the eight variables, structural equation modeling (SEM) with Mplus software was used to examine the hypothesized model to consider potential relationships simultaneously. SEM combines regression analysis and factor analysis to study both direct and indirect relationships among latent (unobservable) and observed variables. Its main goal is to empirically test theories by creating models that represent theoretical predictions in a structured way (Hayduk et al., 2007). In this study, employing SEM allowed for a thorough analysis in line with the proposed theoretical models, enabling a detailed exploration of the connections between different constructs.

In our analysis, we adhered to rigorous statistical procedures, including the maximum likelihood method with 5,000 bootstrapped samples and 95% bias-adjusted confidence intervals, under the assumption of independent data distribution. Following Anderson and Gerbing’s (1988) two-step strategy, we first conducted confirmatory factor analysis (CFA) to validate the measurement model, followed by SEM to estimate the model’s fit. Model fit was assessed using various criteria: exact fit was evaluated with chi-square and degrees of freedom; for comparative fit index (CFI) and Tucker-Lewis index (TLI), values close to .90 were considered indicative of a reasonable fit; and for root mean square error of approximation (RMSEA), values up to .08 was deemed acceptable (Hu & Bentler, 1999; Kline, 2011). Additionally, reliability and validity were assessed through R-square, composite reliability (CR), and average variance extracted (AVE).

Results

The CFA indicated that the measurement model had a good fit: χ² = 791.67, df = 303, p  < .001, χ²/df = 2.61; CFI = .96; TLI = .96; RMSEA = .06 (90% CI = [.052, .061]). All factor loadings are larger than .80. Additionally, the standardized coefficients are all greater than .73, z-value over 1.96, and p-value less than .001 indicated the parameters are all significant. Both item reliability and convergence validity greater than .5 and composition reliability over .7 also suggested items are ideal for dimensions and have good explanatory power. It should be noted that we initially included three variables for “perceived behavioral control.” However, factor analysis showed that keeping any two of the three items did not fit well. To preserve the validity and clarity of the structural equation modeling (SEM), we chose to retain only the most representative variable, ensuring the model’s accuracy and adherence to the theoretical framework. Table 3 shows the results of CFA.

Table 3. Results of CFA.

Dim.	item		parameters of significant test				item reliability	composite reliability	convergence reliability
		Est.	Unstd. S.E.	Std. S.E.	z-value	p-value	R-square	CR	AVE
ATT	ATT_1	.95	.00	.01	132.59	**	.90	.95	.91
	ATT_2	.96	.02	.01	142.79	**	.92
NBS	NBS_1	.88	.00	.01	71.03	**	.78	.91	.78
	NBS_2	.93	.04	.01	94.96	**	.86
	NBS_3	.83	.04	.02	50.72	**	.68
SN	SN_1	.74	.00	.02	35.50	**	.55	.92	.79
	SN_2	.97	.05	.01	126.05	**	.94
	SN_3	.94	.06	.01	105.90	**	.87
CS	CS_1	.75	.00	.02	37.31	**	.57	.95	.78
	CS_2	.88	.06	.01	77.64	**	.77
	CS_3	.93	.06	.01	127.88	**	.87
	CS_4	.93	.06	.01	132.71	**	.87
	CS_5	.91	.06	.01	103.47	**	.83
sys	sys_1	.77	.00	.02	36.40	**	.59	.90	.65
	sys _2	.76	.06	.02	35.00	**	.58
	sys _3	.81	.06	.02	44.88	**	.66
	sys _4	.85	.06	.02	54.65	**	.73
	sys _5	.85	.06	.02	54.81	**	.73
concern	conc_1	.88	.00	.01	73.46	**	.77	.95	.78
	conc _2	.89	.04	.01	79.39	**	.79
	conc _3	.93	.03	.01	112.30	**	.86
	conc _4	.87	.04	.01	69.01	**	.75
	conc _5	.85	.04	.01	62.60	**	.73
trust	trust _1	.87	.00	.01	66.89	**	.76	.93	.76
	trust _2	.85	.04	.01	59.14	**	.73
	trust _3	.87	.04	.01	67.71	**	.76
	trust _4	.89	.04	.01	76.13	**	.80

*p < .05, **p < .001.

The structural model results suggested that some of the model fit indexes were acceptable: χ² = 2259.75, df = 470, p < .001, χ²/df = 4.81; CFI = .88; TLI = .86; RMSEA = .086 (90% CI = [.083, .090]). While our model demonstrated slightly lower CFI and TLI values than the conventional threshold of .90, it is important to consider these results within the broader context of model evaluation. Hu and Bentler (1999) argued for a more nuanced interpretation of these indices, considering the complexity of the model, sample size, and the nature of the data. Moreover, other values of the model are acceptable within the acceptable range, reinforcing its overall validity. The model achieves significant predictive power and theoretical coherence, as evidenced by the satisfactory levels of R-square, CR, and AVE. Therefore, despite the slight deviation in CFI and TLI, the model presents a credible and meaningful representation of the underlying theoretical constructs.

This study first analyzed the impact of three external variables: systematic risk information processing, trust in science/scientists, and concern about the COVID-19, on the cognitive structure, and then determined the relationships between variables in the risk preventive model. After controlling demographic variables, our H1 states that people’s systematic risk information processing positively influences their cognitive structure, and the results supported this view (β = .13, p = .01). H2, the extent of trust in science/scientists positively impacts the cognitive structure, was also supported (β = .36, p < .001). Similarly, H3, which says that people with more concerns about the harm caused by COVID-19 positively impacts the cognitive structure, received support as well (β = .39, p < .001).

Based on the extended TPB model, our H4 states positive relationship exists between normative belief structure and subjective norm, and the results supported this hypothesis (β = .52, p < .001). The results also showed the cognitive structure positively impacts attitude toward acts; therefore, H5 is supported (β = .85, p < .001). Lastly, this study determined the relationships between attitude toward acts and perceived behavioral control to people’s intention to have their children vaccinated, and the results showed both of them positively impact behavioral intention (β = .68, p < .001; β = .20, p = .002). H7 and H8 are both valid.

However, H6 is not supported. The results showed that the relationship between subjective norm and behavioral intention of getting their kids COVID-19 vaccine is insignificant (β = .04, p =.30). The final SEM results and research model are shown in Table 4 and Figure 3.

Figure 3. Relationships between variables.

The model shows the hypotheses between the variables. Systematic risk information processing, trust in science, and concern about the pandemic positively impact their cognitive structure, which impacts their attitude toward the COVID-19 vaccine. Also, both their attitude and perceived behavioral control can positively impact their intention to vaccine their children.

Table 4. Results of SEM (direct impact).

Hypothesis	DV	IV	EST.	Unstd. S.E.	Std. S.E.	z-value	p-value
H1	CS	sys	.13	.05	.05	2.77	*
H2		trust	.36	.05	.06	6.49	**
H3		concern	.39	.03	.05	7.66	**
H4	SN	NBS	.52	.04	.04	12.37	**
H5	ATT	CS	.85	.06	.02	34.62	**
H6	intention	SN	.04	.03	.04	1.03	.30
H7		ATT	.68	.04	.05	12.98	**
H8		PBC	.20	.03	.07	3.05	*

* p < .05, **p < .001.

Apart from direct impacts, this study also addressed how various variables indirectly impact Americans’ decision to vaccinate their children against COVID-19. The results revealed that systematic risk information processing (β = .07, p = .01), trust in science (β = .16, p < .001), and concern about the pandemic (β = .11, p < .001) all have indirect positive effects on behavioral intention through cognitive structure and attitude in this model. Another key mediation of our study was examining cognitive structure (β = .45, p < .001) and its indirect positive effect on behavioral intention through their attitude toward the COVID-19 vaccine. On the contrary, normative belief structure does not significantly influence their behavioral intention through subjective norms (β = .02, p = .31).

Discussion and Conclusions

The results of this study added to the understanding of people’s perceptions related to risks and the impact of these variables on intentions of giving the COVID-19 vaccine to their children. This study first confirmed positive relationships between Americans’ systematic risk information processing, trust in science/scientists, and concern about COVID-19 to the cognitive structure. Results indicated that individuals who systematically process risk information are more likely to form informed and structured perceptions about the COVID-19 vaccine, which can further impact their attitude and intention toward vaccination. This aligns with previous work showing people’s differences in information processing (Griffin et al., 1999; Li & Zheng, 2022).

People’s trust and concern are also important factors (Benin et al., 2006; Bond et al., 1998; Marti et al., 2017; Salmon et al., 2009) that impact their perception of the vaccine information about the disease they receive. When individuals exhibit greater trust in science or heightened concern about the pandemic, it enhances their cognitive structure toward vaccination and can impact their intention by influencing their attitude (Baker et al., 2023; Chew et al., 2021; Li & Li, 2020; Nah et al., 2023; Xiao & Wong, 2020). Moreover, the impact of people’s trust (β = .36) and concerns (β = .39) on the cognitive structure is even stronger than systematic risk information processing (β = .13). Therefore, parent’s trust in scientific knowledge and their awareness of the impact of COVID-19 both play crucial roles in shaping their beliefs about the advantages and significance of vaccinating their children.

In addition, this study found some differences in demographic variables; for instance, individuals with higher education levels were found to shape their cognitive structure (β = .04, p = .002) and trust in science (β = .05, p = .03), which all indirectly impact their decision to let their children receive vaccines. Individuals with higher annual incomes are more inclined to trust science (β = .05, p = .02), demonstrating greater confidence in the wisdom of vaccines and perceiving the benefits of vaccinating their children as outweighing the risks (Marti et al., 2017; McKee & Bohannon, 2016). Moreover, we observed that parents living in urban areas and those with liberal political views are more likely to be concerned about health issues (for both variables, β = .04, p < .001). They also tend to engage in systematic risk information processing when dealing with pandemic-related information (for both variables, β = .05, p < .001); both may indirectly influence their decision to vaccinate their children. The results emphasized the need to consider demographic factors in vaccine communication strategies (Dubé et al., 2013; Marti et al., 2017). Tailoring messages to specific groups based on their trust levels, concerns, and information processing habits could lead to more effective public health interventions.

An important portion of our research focused on understanding the differences between factors influencing vaccination intentions for their children versus themselves. Our model aligned with previous studies in that, within the TPB model framework, attitude (β = .68) is often the key component influencing individuals’ intentions to vaccinate their children (Baker et al., 2023; Dillard, 2011; Li & Li, 2020; Xiao & Wong, 2020). Furthermore, our findings highlighted that perceived behavioral control (β = .20) can also predict people’s behavioral intentions (Xiao & Wong, 2020; Yang, 2015). This indicated that when forming strategies to increase vaccination rates, especially among parents for their children, enhancing both the positive attitudes toward vaccines and the perception of control over the vaccination process is necessary.

However, part of the model had differences from previous studies. The study confirmed a significant relationship between normative belief structure and cognitive structure in the model (β = .52); however, no significant relationship was shown between subjective norm and intention. This study inferred that subjective norms could not significantly positively impact people’s behavioral intention of getting vaccines to their children because the attitude toward acts has a strong relationship to the intention, making it the most important predictor in this data.

We also inferred this may be because the definition of “important others” in the study is people close to respondents, such as their family members or friends, instead of some experts or authority, such as the Centers for Disease Control and Prevention (CDC) and health agencies in communities. This may indicate that the opinions of these significant others are often not a primary consideration when people make major health-promoting decisions, such as vaccinating children. The influence of immediate social circles might be limited compared to the influence of professional health advice. Parents may view vaccination as a medical decision where expert opinion holds more weight than that of non-experts, even if they are close family or friends. Besides, with the rise of the internet and social media, parents can access lots of information sources beyond their social circles. This means they may form opinions about vaccination independently of friends and family. Future research could examine whether the information provided by health professionals can increase people’s intentions to vaccinate their children to complement the parts that are not covered.

These results highlighted the novelty of our model, demonstrating the limitations of the current model. Our study proposed potential modifications and expansions to existing models better to capture the complexities of decision-making in contemporary contexts. Our risk preventive model implied that communication strategies to persuade parents to vaccinate their children should focus more on influencing their attitudes and perceived behavioral control. The influence of friends and family on parents’ decisions to vaccinate their children may not be as strong as previously thought, possibly due to trust in healthcare professionals and increased access to different information channels. Besides, our model indicated that trust in science, concern about the pandemic, and systematic risk information processing are also potential factors that can change parents’ attitudes and behavioral intentions toward vaccinating their children. This insight is crucial for developing effective public health communication and interventions tailored to the nuanced needs of parents in the current health landscape.

This study has some limitations. First, we only considered people’s risk information processing but did not analyze the influence of information-seeking behaviors on their risk perceptions (Griffin et al., 1999). This provides a gap for future research to fill. In addition, it is ideal to have three to five observed variables under each latent variable, and a latent variable must cover at least two observed variables (Kline, 2011). However, based on our results of CFA, only one item in the perceived behavioral control ultimately remained, which may further impact the final model established in this study. Similarly, behavioral intention consists of only one observation variable, which can be improved. For future improvements, designing multi-item constructs, conducting pilot tests, and consulting experts are all potential solutions to deal with this limitation. Lastly, future studies can consider employing diverse sampling methods, including random probability sampling, to validate and expand upon our findings.

Despite the limitations, this study still contributes to understanding why parents choose to vaccinate their children. This study is the first to create the risk preventive model by combining trust in science, concern about COVID-19, and information processing with the TPB model and analysis with SEM. The results highlighted latent variables that positively impact Americans’ risk perspectives and intentions, which can help health risk communicators create an effective way to provide health risk information about improving people’s intentions to vaccinate their children. In addition to practical implications, the results of this study not only provide partial support and adjustments for the risk preventive model but also provide recommendations for future studies. Future research can apply the model tested in this study to other health or risk topics, such as nutrition or other disease-preventive behaviors. Future research should also extend the findings of this study to focus more on people living in rural areas or with lower educational levels and determine how trust in science/scientists and concerns about illness impact health decisions made by underrepresented groups.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Additional information

Funding

This research [IRB#202000655] was funded by the University of Florida Institute of Food and Agricultural Sciences (UF/IFAS) Center for Public Issues Education in Agriculture and Natural Resources (PIE Center) and the UF/IFAS Office of the Dean for Research.