A total of 911 HP students completed the self-administered questionnaires from 1,000 registered students, giving a response rate of 91.1%. As shown in table 1, the mean age was 20.78 years old and the majority were female (75.4%), pharmaceutical students (62.7%), and Vietnamese students (94.6%). About 50% of students self-rated their health status from good (29.4%) to very good (20.9%). Nearly 60% of students have received a seasonal influenza shot in the past 6 months. In terms of knowledge about the COVID-19 vaccine being used in Vietnam (AstraZeneca), less than 50% of HP students knew the time interval between 2 doses of vaccine (42.8%) and the dangerous adverse effects (36.5%). In summary, 72.4% of students achieved the level of good knowledge (average scores ≥ 3 points out of 5 points).
As seen in Table 2, more than 50% of HP students agreed on beliefs about the susceptibility to COVID-19 infection and vaccine side effects, as well as the benefits of vaccination. More than 40% of HP students agree with their beliefs about the severity of COVID-19. Agreements on beliefs about barriers ranged from 20% to 40%. In cues to action, about 60% of HP students agreed on confirmation of the effectiveness and safety of the COVID-19 vaccine from the government and the media. The proportion of HP students who would accept, hesitate and refuse COVID-19 vaccination was 58.0% (95% CI: 54.7% - 61.3%), 40.4% (95% CI: 37.2% - 43.7%), and 1.5% (95% CI: 0.8% - 2.6%), respectively.
In table 3, the multinomial logistic regression model predicted HP students into one of three categories of COVID-19 vaccine acceptability. By default, SPSS uses the highest-numbered category as the reference category. Therefore, vaccine acceptance with the highest number (525 cases) was used as a reference group [24]. For 2 multinomial logit models, the first for “Vaccine hesitancy” relative to “Vaccine acceptance” included 8 statistically significant predictors (p < 0.05). The second for “Vaccine refusal” relative to “Vaccine acceptance” consisted of 5 statistically significant predictors (p < 0.05). The regression coefficients of these predictors were statistically different from zero (p < 0.05). In this study, no predictive value was found by the multinomial logistic regression models between COVID-19 vaccine acceptability and some demographic and personal factors such as age, gender, health status, and knowledge about COVID-19 vaccines (p > 0.05).
In the first model for “Vaccine hesitancy” versus “Vaccine acceptance”, 2 regression coefficients (B) or log-odds of "Mass media appreciating effectiveness and safety of vaccines", and "HP students get serious complications of COVID-19 if not vaccinated" were negative such as -0.657, and - 0.294. If an HP student increased the agreement of these statements by one point, the multinomial log-odds of choosing "Hesitancy" over "Acceptance" would be expected to decrease by 0.657, and 0.294 units, respectively. For recent seasonal influenza shots (B = - 0.484), HP students who recently received a flu shot were less likely to hesitate about COVID-19 vaccination than to accept it. These predicting variables had odds ratio (OR) less than one (OR<1) to be suitable for a negative regression coefficient. For “Mass media appreciating effectiveness and safety of vaccines” with OR = 0.518, HP students were 0.518 times less likely to choose hesitancy of vaccination than acceptance. Similarly, for the belief of "HP students get serious complications of COVID-19 if not vaccinated", it would be 0.745 times less likely to select hesitancy of vaccination than acceptance. Then, HP students having a recent flu shot were 0.616 times less likely to choose “Hesitancy of COVID-19 vaccination” than students not getting a flu shot. Three remaining variables in the first model, which had positive regression coefficients, were "Manufacturers not disclosing adverse effects of vaccines", "Vaccines having little efficacy & serious adverse effects", and "Adverse effect causing death". If the agreement scales were increased by one point, the multinomial log-odds of selecting "Hesitancy" versus "Acceptance" would be expected to increase by 0.330, 0.671, and 0.504 units, respectively. Interpreting with the odds ratio (OR), if the agreement scales of these perceptions were increased, the probability of hesitancy of vaccination would be 1,390, 1,957, and 1,656 times higher than acceptance of vaccination, respectively. For demographic variables like nationality, and majors, Vietnamese and pharmaceutical students were 5.22 times and 2.21 times more likely to choose "Hesitancy" over "Acceptance" compared to Laos and physiotherapy students, respectively. Students of nursing and medical laboratory technology were not predictive of “Vaccine hesitancy”.
In the second model for “Refusal” versus “Vaccine acceptance”, a cue to actions such as “Manufacturers not disclosing adverse effect of a vaccine” and a perceived barrier like “Adverse effect causing death” with positive log-odds (B) were 5.29 times and 10.25 times more likely to choose “Refusal” than “Acceptance”, respectively. Other predictors with negative log-odds such as "Mass media appreciating effectiveness and safety of vaccines", "Unvaccinated HP students feasibly infected by coronavirus during hospital internship ", and "HP students get serious complications of COVID-19 if not vaccinated" were 0.33, 0.34, and 0.43 times less likely to select "Refusal" than "Acceptance", respectively. In general, the rate of “acceptance of the vaccine” among HP students was determined with the highest correct rate (78.3%) compared with the correct rate of hesitation (65%) and refusal (30.8%).
In summary, the predictors in the "Hesitancy versus Acceptance" and "Refusal versus Acceptance" models of which positive regression coefficient (B) were statistically different from 0 (p < 0.05) included "Manufacturers do not disclose adverse effects of vaccines", "Vaccines have little efficacy & serious adverse effects", "Adverse effects causing death", "Vietnamese nationality", and “Pharmacy major”. In contrast, predictors with a negative regression coefficient (B) included “Mass media appreciate vaccine efficacy and safety”, “Unvaccinated HP student is likely to contract COVID-19 during hospital internship”, “HP students will develop serious complications of COVID-19 if not get vaccinated”, and “Currently seasonal flu shot”. Nationalities (Vietnamese versus Laos, OR= 5.221, 95% CI: 2.177-12.520, p=0.0005), and majors (Pharmacy versus physiotherapy, OR= 2.215, 95% CI: 1.145-4.285, p= 0.018) were the strong predictors of "vaccine hesitancy". Strong HBM predictors of vaccine refusal were "Manufacturers not disclosing adverse effects of vaccines" (OR= 5.299, 95% CI: 1.687-16.641, p= 0.004), and "Adverse effect causing death" (OR= 10.255, 95% CI = 3.528-29.814, p= 0.0005).