The impact of risk perception and institutional trust on COVID-19 vaccine hesitancy in China

ABSTRACT

Vaccination has become the primary means for citizens to prevent severe morbidity and mortality during the pandemic. However, vaccine hesitancy poses a major threat to global public health security. Based on the data from Chinese General Social Survey in 2021, this study aims to explore the socio-political aspects of COVID-19 vaccine hesitancy, focusing on the relationship between COVID-19 risk perceptions, institutional trust and vaccine hesitancy. Among the samples, 39.8% of the respondents exhibited COVID-19 vaccine hesitancy, 48.9% had a high awareness of the risk of COVID-19, and 74.6% presented a high level of trust in institutions. The results showed that higher risk perception and institutional trust are negatively correlated with vaccine hesitancy (p < .001). Institutional trust had no statistically significant moderating effect on the association between risk perception and vaccine hesitancy, but the role of institutional trust in influencing vaccine hesitancy is more significant at a lower level of perceptions of COVID-19 risk. Furthermore, regional variations in the factors contributing to vaccine hesitancy were demonstrated in China. These findings have important implications for developing strategies to address vaccine hesitancy.

KEYWORDS:

• COVID-19
• vaccine hesitancy
• institutional trust
• risk perception
• region
• China

Introduction

Vaccination is widely recognized as a critical measure for preventing pandemic infectious diseases. However, vaccine hesitancy (VH), defined by the WHO as the “delay in acceptance or refusal of vaccines despite availability of vaccine services,”^1,2 poses one of the most significant threats to global public health security.³ Previous research has demonstrated that VH is a complex and dynamic issue that is influenced by a range of individual-level factors, as well as contextual, social, cultural and political factors.^4,5 Recently, the COVID-19 pandemic has brought even greater attention to this issue across the world.^6–10 Similar to the United States, the United Kingdom and some other Western countries, China has witnessed a substantial body of research on COVID-19 VH.^11–13 However, the majority of existing research is based on specific populations (such as people with underlying diseases,¹⁴ healthcare worker,¹⁵ college students,¹⁶ parents of minors¹⁷ and individuals from one or several provinces,¹⁸ with very few nationally representative samples. Furthermore, few studies have focused on the socio-political aspects of VH, particularly the relationship between risk perception (RP), institutional trust (IT) and VH, which represents significant issues in VH research. There are basically no studies exploring regional differences of VH in China. Therefore, this study aimed to fill these gaps.

Method

Hypotheses

As mentioned above, available evidence indicates that COVID-19 VH may be explained by various factors, including not only medical factors but also socio-political ones. Thus, our main focus is on the relationship between RP, IT and VH.

RP refers to an individual’s estimation of their future likelihood of contracting a particular disease, considering behavioral and temporal factors.^19,20 People’s catching and even fear of COVID-19 can have a serious adverse impact on their physical and mental health.²¹ Thus, people who perceive high risks of catching the COVID-19 may be more likely to have a positive attitude toward vaccination, as vaccines can be effective in reducing fear and preventing severe or even fatal disease.²² So we propose our first hypothesis, H1a: RP is negatively associated with the degree of VH.

Trust has been given a multitude of definitions and dimensions within the socio-medical literature.²³ IT means that citizens have trust in institutions such as government, judicial system, medical institutions and scientific institutions to behave in a predictable, fair and transparent way, and to act in a manner that is in the interests of citizens.²⁴ It functions to reduce uncertainty and complexity when dealing with novel crisis such as COVID-19, and increase people’s compliance with government campaigns.^25,26 Therefore, people who trust their institutions may have a more positive attitude toward getting the COVID-19 vaccine.²⁷ Conversely, people who distrust institutions may refuse vaccination because they question the safety and efficacy of vaccines.^28,29 Thus, we derive the following hypothesis, H1b: IT is negatively associated with the degree of VH.

It is noteworthy that the link between risk perception and VH may not necessarily be straightforward. Some extant studies have shown that despite individuals having some perceptions of risk, they seldom take appropriate preparations. The relationship may be influenced by other intervening factors, an important one of which is trust. When facing an extremely uncertain and complex crisis like the COVID-19, individuals may not be able to handle with all the threats, which make them employ trust in authorities, experts and other external systems to manage personal risk.²⁵ Consequently, we hypothesize that H2: IT positively moderated the relationship between RP and VH.

Additionally, VH represents “a complex behavioral phenomenon specific to vaccines, contexts, time and place.”³⁰ In this sense, we expect that geographical locations and regional factors may exhibit an impact on COVID-19 VH. Some existing studies suggest that COVID-19 acceptance rates and the causal factors of VH differ both between and within countries.^6,26 Hence, we propose that H3: RP and IT have different impacts on VH in different regions (such as the east, central and west in China).

Data and samples

The data for this study were collected from the 2021 Chinese General Social Survey (CGSS2021), which is a national, comprehensive and ongoing academic survey project in China. The survey is mainly conducted by Renmin University of China, with focus on adults and families across mainland China. The CGSS has a strong scientific foundation and has been widely used in scientific research, teaching and government decision-making. Due to the outbreak of COVID-19, a special topic on vaccination and VH was incorporated into the survey conducted in 2021. The samples were located in various provinces and cities, including Beijing, Hebei, Liaoning, Jiangsu, Zhejiang, Fujian, Shandong and Guangxi in the east; Shanxi, Inner Mongolia, Anhui, Jiangxi, Henan, Hubei and Hunan in the central region; and Chongqing, Shaanxi, Gansu and Ningxia in western China. These provinces and cities are representative of their respective regions. A total of 8148 respondents were interviewed in CGSS2021. After applying the inclusion criteria – being an adult (age ≥18 years) and living, working, and/or receiving healthcare in China, 2717 samples were finally eligible for analysis. Missing values for the education level variable were imputed using the mean difference imputation method. Other missing values were generally eliminated. Among the samples, 45.1% were males and 54.9% were females. The age of the respondents ranged from 18 to 99 years, with a mean age of 52.0 ± 17.6.

Measurement

COVID-19 VH was assessed using the question “Which of the following statements best describes your situation?” Respondents were provided with four response options: (a) Initially reluctant, but eventually accept vaccination; (b) Initially hesitant, but ultimately decided to get vaccinated; (c) Unwilling to provide an answer; and (d) Initially intended to get vaccinated. To identify VH, responses were dichotomized, with options (a), (b), and (c) coded as positive (1), and option (d) coded as negative (0).

RP of COVID-19 infection was measured by asking respondents, “How worried are you about getting COVID-19?” Respondents could choose from the options “Very worried,” “Some worry,” “Not too worried,” and “Not worried at all.” The degree of RP was graded using a four-point Likert scale,³¹ with these options coded from 4 to 1 respectively. “Very worried’ indicated the highest level of risk perception, while “Not worried at all” indicated lowest level of risk perception. The scale of 1–4 represented an increasing level of risk perception. Further, the data was analyzed using the method of entropy weight calculation of composite scores. Scores ranging from 1 to 2 were categorized as low risk perception (1), while scores ranging from 3–4 were categorized as high risk perception (2).

IT was measured across three dimensions: the perception of social trust, confidence in the health care system, and confidence in the government. For ease of measurement and explanation, each dimension was evaluated using a five-point Likert scale.³² Social trust was assessed by the question, “In general, would you agree that most people in society can be trusted?” Respondents were allowed to specify their levels of agreement on a five-point Likert scale, namely “Strongly disagree”(1), “Disagree”(2), “Neither agree nor disagree” or “Don’t know” or “Refusing to answer”(3), “Agree”(4) and “Strongly agree”(5). Higher scores indicated greater trust. Trust in the healthcare system was measured by asking, “How have the measures the healthcare system in our country has taken to address the COVID-19 pandemic changed your confidence?” Trust in the government was measured by asking, “How have the measures the government has taken to deal with the COVID-19 pandemic changed your confidence?” For both questions, respondents could choose from the following options: “Increased significantly,” “Increased slightly,” “Remain basically unchanged,” “Unable to choose,” “Refused to answer,” “Decreased slightly,” and “decreased significantly,” which were coded as 5, 4, 3, 3, 3, 2, and 1, respectively. Higher scores indicated higher levels of trust. Following Cheng, et al.³³ the weight was calculated using the entropy method, which enabled the computation of the composite index for IT.

VH may exhibit regional differences.³⁴ To test whether the influencing factors of VH are different across China, provinces are divided into the eastern, the central, and the western region.³⁵

The study also incorporated several control variables including sex, age-group, ethnicity, income, education and residency registration. Sex was categorized as nominal, with men assigned as 1 and female as 0. Respondents were categorized into two groups according to their age: age-group was given a value of 1 if a respondent was 70 years old or above and a value of 0 otherwise.³⁶ Ethnicity was classified as Han, Mongolian, Manchu, Hui, Tibetan, Zhuang, Uygur, and others. The Han ethnicity was assigned a value of 1, while other ethnicities were assigned a value of 0. Income is divided into five ranking categories (in yuan): less than 4000, 4000 to less than 10,000, 10000 to less than 20,000, 20000 to less than 50,000, and 50,000 or more. Education was categorized as primary school and below, junior high school, high school, and college degree or above. Residency registration is classified into agricultural and non-agricultural registrations, with the former recorded as 1, while the latter recorded as 2.

Statistical analysis

SPSS20.0 was used for data processing and analysis. In a first step, hierarchical regression analysis was employed to examine the association between RP, IT, and VH. In a second step, the moderating effect of IT on the correlation between RP and VH was analyzed by incorporating an interaction term. In a third step, the whole samples were divided into two sub-groups (high RP and low RP), and group regression analysis was carried out to explore the effects of IT on VH. Finally, the entire samples were divided into three groups based on the eastern, central and western regions of China, and comparisons were made to reveal the regional differences of the factors associated with VH. What we conducted are two-sided statistical tests. A two-sided p-value of less than .1 indicates a statistically significant difference.

Results

Descriptive statistics

Table 1 presents descriptive statistics for the sample. 60.2% of the respondents reported that they were likely or extremely likely to receive the COVID-19 vaccine if it were accessible to them; however, 39.8% reported that they were unlikely or very unlikely to do so. 48.9% of the respondents had a high perception of risk, while 51.1% had a low perception of risk. 74.6% of the respondents expressed a high level of trust in institutions, whereas 25.4% remained distrustful (Figure 1).

Figure 1. Proportions of vaccine hesitancy, risk perception and institutional trust. VH: vaccine hesitancy; RP: risk perception; IT: institutional trust. Abscissa: Categories;Ordinate: proportion.

Table 1. Descriptive statistics.

Variables	Mean	SD	Range
Vaccine hesitancy (VH)	0.4	0.5	0–1
Risk perception (RP)	2.5	1.1	1–4
Institutional trust (IT)	4.0	0.7	1–5
sex	0.5	0.5	0–1
age-group	0.2	0.4	0–1
ethnicity	0.9	0.3	0–1
education	2.3	1.1	1–4
income	3.2	1.6	1–5
residency registration	1.4	0.5	1–2

Vaccine hesitancy (VH) is a binary variable; risk perception (RP) has an increasing scale from 1 to 4; institutional trust (IT) is a composite indicator, with higher scores indicating higher levels of trust; sex is a binary variable, with 1 representing male while 0 representing female; age-group is a binary variable, with 1 denoting elder people of 70 years old or above, and 0 otherwise; ethnicity, a binary variable, is given a value of 1 if an individual belongs to the Han ethnicity, and 0 otherwise; education is an increasing categorical variable, ranging from 1 to 4, representing primary school and below, junior high school, high school, and college degree or above, respectively; residency registration is a binary variable, with 1 representing an individual who has an agricultural registrations, and 0 otherwise.

Hierarchical regression analysis

The results of hierarchical regression analysis were presented in Table 2, revealing the associations between RP, IT, other individual-level factors and VH. The results of Model 1 showed that age-group (p = .000), ethnicity (p = .014), and residency registration (p = .000) exhibited a significantly positive correlation with VH, while education (p < .1) had a negative correlation with VH. Furthermore, the coefficients of sex and income were not found to be statistically significant.

Table 2. Hierarchical regression analysis (dependent variable: VH).

	Stratification 1			Stratification 2			Stratification 3
	b	t	p	b	t	p	b	t	p
RP	/	/	/	−0.023***	−2.642	.008	−0.023***	−2.605	.009
IT	/	/	/	/	/	/	−0.060***	−4.255	.000
Constant	0.235***	5.472	.000	0.313***	6.008	.000	0.539***	7.258	.000
sex	0.003	0.155	.877	0.000	−0.022	.982	0.003	0.161	.872
age-group	0.267***	10.559	.000	0.258***	10.122	.000	0.268***	10.484	.000
ethnicity	0.079**	2.281	.023	0.069**	1.985	.047	0.073**	2.102	.036
education	−0.017*	−1.707	.088	−0.019*	−1.908	.057	−0.017*	−1.735	.083
income	−0.008	−1.253	.210	−0.008	−1.284	.199	−0.007	−1.171	.242
residency registration	0.074***	3.451	.001	0.072***	3.365	.001	0.072***	3.373	.001
N	2717			2717			2717
R^2	0.058			0.06			0.067
F	F (6,2710) = 27.747, p = .000			F (7,2709) = 24.833, p = .000			F (8,2708)  = 24.129, p = .000
△F	F (6,2710) = 27.747, p = .000			F (1,2709) = 6.983, p = .008			F (1,2708) = 18.108, p = .000

*p < .10, **p < .05, ***p < .01. RP，Risk Perception; IT, Institutional Trust. VH, Vaccine Hesitancy.

Based on Model 1, Model 2 added RP, which resulted in significant changes in F (p = .008), indicating that the inclusion of RP had explanatory significance for the model. The coefficient of RP was −0.023, which was statistically negatively significant (p = .000), indicating that RP had a significantly negative effect on VH, after controlling for such factors as age-group, sex, ethnicity, income, education, and residency registration.

Furthermore, based on Model 2, Model 3 included IT and demonstrated significant changes in F (p = .000), suggesting that the incorporation of IT has explanatory value for the model. Specifically, the coefficient of IT was −0.060, which was statistically negatively significant (p = .002), indicating that IT had a significantly negative effect on VH, after controlling for RP as well as other individual-level variables.

Moderation analysis

In order to further explore the moderation effects of IT on the relationship between RP and VH, we incorporated into Model 3 an interaction term of RP and IT. Thus, we constructed a new regression model as follows:

$\mathrm{V}\mathrm{H}=\hspace{0.17em}{\beta }_0+{\beta }_1\cdot \mathrm{R}\mathrm{P}+{\beta }_2\cdot \mathrm{I}\mathrm{T}+{\beta }_3\cdot \mathrm{R}\mathrm{P}\times \mathrm{I}\mathrm{T}+\beta \cdot \mathrm{C}\mathrm{V}+\epsilon .$

In this model, CV represents other control variables, while the coefficient β₃ is the primary focus of our attention. Specifically, β₃ indicates the strength and direction of the moderating effects of IT on the relationship between RP and VH. The regression results were given in Table 3, which indicated that after controlling for other variables, IT had no statistically significant moderating effect on the association between RP and VH (${\beta }_3=0.016,$ p = .218).

Table 3. Moderating analysis.

	b	t	p
RP	−0.023***	−2.62	.009
IT	−0.060***	−4.279	.000
RP*IT	0.016	1.233	.218
constant	0.542***	7.298	.000
sex	0.003	0.134	.894
age-group	0.268***	10.495	.000
ethnicity	0.072**	2.075	.038
education	−0.017*	−1.718	.086
income	−0.008	−1.202	.230
residency registration	0.072***	3.369	.001
N	2717
R^2	0.067
F	F (9,2707) = 21.622, p = .000
△F	F (1,2707) = 1.521, p = .218

*p < .10, **p < .05, ***p < .01. RP，Risk Perception; IT, Institutional Trust. VH, Vaccine Hesitancy.

Grouped regression analysis

To gain a more nuanced understanding of how IT and RP jointly affect VH, we further divided the whole samples into two separate groups: high RP and low RP. The results of grouped regression analysis were reported in Table 4, which shows that when people’s perceived risk of COVID-19 infection was high, IT was statistically significantly associated with VH (β = −0.047, p = .018). Likewise, when RP is low, IT was also substantially related to VH (β = −0.074, p = .000); however, the coefficient of IT has considerably increased, and the statistical significance level has also notably increased, indicating that the role of IT in influencing VH is more evident at a lower level of perceptions of COVID-19 risk.

Table 4. Group regression analysis.

	High RP			Low RP
	b	t	p	b	t	p
IT	−0.047**	−2.37	.018	−0.074***	−3.678	.000
constant	0.367***	3.992	.000	0.587***	5.532	.000
sex	−0.029	−0.029	.275	0.035	1.347	.178
age-group	0.267***	6.935	.000	0.269***	7.871	.000
ethnicity	0.066	1.564	.118	0.085	1.389	.165
education	−0.011	−0.748	.454	−0.023*	−1.659	.097
income	−0.006	−0.636	.525	−0.008	−0.886	.376
residency registration	0.104***	3.447	.001	0.042	1.412	.158
N	1329			1388
R^2	0.057			0.072
F	F (7,1321) = 11.415, p = .000			F (7,1380) = 15.225, p = .000

*p < .10, **p < .05, *** p < .01. RP，Risk Perception; IT, Institutional Trust. VH, Vaccine Hesitancy.

Regional differences analysis

To explore the regional differences of associated factors of VH, we analyzed the data based on the eastern, central and western regions in China, the results of which were given in Table 5. Regarding RP, the coefficients in the eastern (p = .309) and central (p = .193) regions were negative but not statistically significant, while the coefficient in the western region was statistically significantly negative (β = −0.033, p = .045). Regarding IT, the coefficients were statistically significantly negative across all the three regions; however, the values of the coefficients increased from eastern (β = −0.089) to central (β = −0.055) and to western (β = −0.044) regions, whereas the significance level sequentially decreased from eastern (p = .000) to central (p = .018) and to western (p = .081) regions.

Table 5. Hierarchical regression analysis by region.

	East			Central			West
	b	t	p	b	t	p	b	t	p
RP	−0.015	−1.018	.309	−0.019	−1.303	.193	−0.033**	−2.008	.045
IT	−0.089***	−3.826	.000	−0.055**	−2.372	.018	−0.044*	−1.748	.081
constant	0.918***	6.510	.000	0.507***	3.962	.000	0.559***	4.247	.000
sex	−0.008	−0.253	.800	0.028	0.883	.378	0.007	0.194	.846
age-group	0.240***	5.824	.000	0.312***	7.228	.000	0.242***	5.082	.000
ethnicity	−0.037	−0.454	.650	0.087	1.150	.250	0.036	0.802	.423
education	−0.007	−0.474	.635	−0.031*	−1.908	.057	−0.031	−1.627	.104
income	−0.02*	−1.83	.068	−0.019*	−1.871	.062	0.007	0.610	.542
residency registration	0.04	1.180	.238	0.063*	1.742	.082	−0.017	−0.383	.702
N	1098			913			706
R^2	0.056			0.095			0.067
F	F (8,1089) = 8.052, p = .000			F (8,904) = 11.835, p = .000			F (8,697) = 6.285, p = .000

*p < .10, **p < .05, ***p < .01. RP，Risk Perception; IT, Institutional Trust. VH, Vaccine Hesitancy.

Discussions

Vaccination is not solely a medical issue, but also a socio-political one. COVID-19 VH has been widely documented in the literature.^37–39 Despite extensive literature on the reasons for COVID-19 VH, relatively limited attention has been focused on its socio-political aspects, particularly the role of IT in the context of VH in China. In this study, we fill this gap by analyzing how the interaction of COVID-19 RP and trust in institutions influenced VH, based on a nationally representative survey dataset (the CGSS2021) in China.

First, we found that both higher RP and higher IT are negatively correlated with COVID-19 VH. The results generally align with previous research findings.^40–42 In general, individuals with higher RP should exhibit high protective behaviors and/or low risk behaviors.²⁰ Nusair et al. found that persons at higher risk for infection were expected to be more willing to take preventive measures and to be more likely to be vaccinated than their peers.⁴³ Truong et al. found that RP and trust in the institutional and health system were closely linked to the extent of vaccine immunization.⁴¹ Du et al. found that public trust in the vaccine system and the government can enhance vaccine confidence and reduce vaccine hesitancy.⁴⁴ Overall, the findings indicate that to promote willingness to receive vaccination, effective measures should be taken to increase the public’s risk awareness regarding infectious diseases and to establish and maintain public trust.

Second, we found that IT had no statistically significant moderating effect on the association between RP and VH, but that the role of IT in influencing VH is more evident and significant at a lower level of perceptions of COVID-19 risk. This is a more interesting and nuanced aspect that has been understudied in previous research. When people’s RP is very high, they may pay more attention to and worry about the risk of COVID-19 itself, rather than the relevant factors of vaccine and vaccine administration. In this case, even if IT increases, it may not significantly change people’s attitudes toward vaccination because people’s concerns and worries about the risk of COVID-19 exceed the relevant factors of vaccine and vaccine administration. However, in low-risk situations, people may pay more attention to and consider the relevant factors of vaccine administration (such as safety, effectiveness, etc.); at this time, the increase of IT may more easily influence people’s views and attitudes toward vaccine administration. Additionally, people with low RP do not initiate vaccination but may cooperate to get vaccinated based on trust in the government, healthcare institutions and society.⁴⁵ This implies that continuous attention and adaptation to changes in public RP and trust levels are necessary to ensure sustained vaccination efforts.

Third, the study has demonstrated regional variations in the factors contributing to VH, confirming that VH is a complex phenomenon with significant geographical and contextual associations.^6,26,30 More specifically, no significant correlations were observed between RP and VH, yet notable negative correlations were evident between IT and VH in the eastern and central regions. This may be attributed to the fact that, although the epidemic situation in the eastern and central regions is potentially more severe where the population density is higher compared to the western region, medical resources and healthcare quality are even advanced, leading to a relatively less severe concern regarding the risk of COVID-19. Additionally, individuals in the eastern and central regions tend to have a greater emphasis on the safety and efficacy of vaccines, coupled with a higher level of trust in the institutions.⁴⁶ On the contrary, in the western regions, there is a significant negative correlation between RP and IT with VH. This may be because, despite the relatively mild epidemic situation in the western regions, medical resources and healthcare quality are even worse, leading to a higher focus on the risk of COVID-19. In addition, IT levels in the western regions may be relatively lower compared to the eastern and central regions, resulting in a less significant impact of IT on VH.⁴⁷ Thus, further research is needed to explore these influencing factors.

Limitations and implications

This study has a limitation in that the data collected were cross-sectional, which prevents us from determining the causal relationship between associated factors and COVID-19 VH over different time periods. The limitation was partially mitigated by the extensive and diverse sample, which closely mirrored the distribution of VH across the country. Future studies should aim to investigate the influencing factors with more rigorous designs, such as using time-series data or experimental methods. Another limitation is that it did not incorporate such factors as personal health status, interpersonal network, healthcare resources, and social media exposure⁴⁸ into the analysis. Follow-up analysis incorporating these factors will further enrich our understanding of the issue. Additionally, although all the respondents are adults, some of them are parents or caregivers of children, thus our findings remain relevant to children’s VH.

Furthermore, it is worthy to note that 39.8% of our samples were hesitant about getting vaccinated in 2021, but it was reported that the vaccination rate in China reached 97.6% by 2023.⁴⁹ To some extent, this phenomenon is similar to the situation in India, where the government’s openness and transparency led to a gradual decline in VH.⁵⁰ Thus, future research is expected to explore how China managed to achieve such a high vaccination rate, examine the role of RP and IT play in this process, and compare the differences in working mechanisms between China and India, two of the largest developing countries in the world.

With these limitations and future directions in mind, our results provide valuable implications for medical policymakers to recognize that vaccination is not merely a medical issue, but also a socio-political matter. To enhance vaccine coverage, corresponding measures aimed at bolstering the public’s trust in institutions should be implemented, tailored to the specific circumstances of different regions. In this regard, our research also carries practical implications for addressing other public health concerns.

Author contributions

All authors contributed to the study conception and design. Conceptualization: [Guiwu Chen]. Methodology: [Guiwu Chen] [Yiyun Yao]. Assessment: [YaoWen Zhang]. Formal analysis and investigation: [Yiyun Yao]. Writing-original draft preparation: [Fei Zhao]. Writing-review and editing: [Fei Zhao], [Guiwu Chen]. Funding acquisition: [Guiwu Chen]. Supervision: [Fei Zhao].

Disclosure statement

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

Additional information

Funding

This research was supported by the Fundamental Research Funds for the Central Universities [No. 23IJKY02].