vaccinated? Respondents who had not been vaccinated were not used as research samples. In addition, when screening the questionnaires, we deleted some samples of responses that were obviously unreasonable. Moreover, samples with a large number of missing values were excluded. In total, 768 members participated in the study. The effective Page 11/22 sample included 561 individuals after the data screening. Table 2 provides more detailed descriptive statistics. Table 2 Descriptive statistics of the respondents’ characteristics Demographics Category Frequency Percentage Gender Male 269 48.0 Female Junior college 88 15.7 University 284 50.6 Master’s degree or higher 157 28.0 Province Shandong 78 13.9 Henan 70 12.5 Chongqing 61 10.9 Hunan 60 10.7 Sichuan 57 10.2 Heilongjiang 55 9.8 Guangdong 43 7.6 Jiangsu 34 6.0 Others 103 18.4 Results Reliability and validity test Before hypothesis testing, the reliability and validity of the scale were tested. Using SPSS 17.0 and Lisrel 8.7 software, we conducted an exploratory factor analysis. In a series of analysis results, the KMO value Page 12/22 was 0.89, which suggests that the sample data were suitable for factor analysis [40]. Furthermore, this analysis extracted 12 factors that explained 88.66% of total variance, which indicates that these factors represent most of the information of the 32 items [41]. In addition, the factor loadings of all 32 items were not less than 0.7, which demonstrates that these items can be analyzed for reliability and validity [42]. Table 3 shows the specific factor loading of each item. Page 13/22 Table 3 Statistical results of some indicators Factors Items Loadings AVE CR Cronbach α Note: OE = outcome expectancy; RP = risk perception; PS = perceived self-efficacy; SPC= social positive cues; VH = vaccine hesitancy; CF = confidence; CP = complacency; CV = convenience; VI = vaccination intention; VB = vaccination behavior; SAT = satisfactory; CVN = continuous vaccination; PU = perceived usefulness. Page 14/22 Factors Items Loadings AVE CR Cronbach α Note: OE = outcome expectancy; RP = risk perception; PS = perceived self-efficacy; SPC= social positive cues; VH = vaccine hesitancy; CF = confidence; CP = complacency; CV = convenience; VI = vaccination intention; VB = vaccination behavior; SAT = satisfactory; CVN = continuous vaccination; PU = perceived usefulness. The composite reliability (CR) values and Cronbach α values were used to evaluate the reliability of the scale. Detailed analysis results are shown in Table 3. Based on this table, it can be seen that CR values and Cronbach α values of all factors exceeded 0.7, which indicates that the scale had high reliability [43]. In the validity test, we intended to test the convergent validity and discriminant validity of the scale respectively. The average variance extracted (AVE) was used to analyze the convergent validity. The analysis results were also shown in Table 3 We can see that all AVE values were greater than 0.5, which demonstrates that this scale had satisfactory convergent validity [44]. In addition, the confirmatory factor analysis (CFA) was used to test the discriminant validity of the scale. After the CFA, we discovered that the square roots of AVE values were greater that these coefficients, which indicates that the discriminant validity of the scale meets the requirements for analysis [45]. Hypothesis tests Using Lisrel 8.7 software, structural equation modelling was used to test our hypotheses. The analysis results of this model are shown in Table 4. We learned that all the fit indices results were acceptable, which demonstrates that our hypothesis tests that are used in this model were reasonable and effective [45]. Page 15/22 Table 4 Fitting Indices and Evaluation Criteria of Research Model Fitting indices Absolutely indices Parsimony indices Incremental indices χ 2 /df GFI AGFI RMSEA PNFI PGFI CFI NFI IFI Values of fitting indices 2.76 0.90 0.82 0.07 0.87 0.72 0.98 0.98 0.98 Evaluation criteria values [45] < 3 > 0.9 > 0.8 < 0.08 > 0.5 > 0.5 > 0.9 > 0.9 > 0.9 Figure 2 shows the model’s normalized path coefficients. In the motivation stage, we learned that perceived self-efficacy had a positive significant influence on vaccination intention (β = 0.1, p < 0.05), but this effect did not appear in the relationship between outcome expectancy and vaccination intention (β = 0.04, p > 0.05) and between risk perception and vaccination intention (β = 0.03, p > 0.05). Therefore, H3 was supported, but H1 and H2 were not. In the intention stage, we found that social positive cues (β = 0.13, p < 0.01) and VH (β=-0.67, p < 0.001) exerted a positive and negative significant impact on vaccination intention, respectively, which supports H4 and H5. Moreover, vaccination intention positively affected vaccination behavior significantly (β = 0.75, p < 0.001), so H6 was supported. In the behavior stage, we noted that vaccination behavior had a significantly positive impact on perceived usefulness (β = 0.62, p < 0.001) and satisfaction (β = 0.26, p < 0.001), respectively. Thus, H7 and H8 were supported. In addition, perceived usefulness produced a positive significant influence on satisfaction (β = 0.73, p < 0.001) and continuous vaccination (β = 0.88, p < 0.001), respectively. Thus, H9 and H10 were supported. However, the positive effect of satisfaction on continuous vaccination was not significant (β=-0.06, p > 0.05), so H11 was not supported. Discussion By combining the