key: cord-0029613-pd6k5i19 authors: Wang, Qiang; Xiu, Shixin; Yang, Liuqing; Han, Ying; Cui, Tingting; Shi, Naiyang; Liu, Minqi; Yi, Youqin; Liu, Chang; Wang, Xuwen; Zhou, Weijie; Jin, Hui; Lin, Leesa title: Validation of the World Health Organization’s parental vaccine hesitancy scale in China using child vaccination data date: 2022-02-14 journal: nan DOI: 10.1080/21645515.2021.2021060 sha: ac347589afbb3b07fbf9961205582e000a036a3e doc_id: 29613 cord_uid: pd6k5i19 Evidence for the validity and reliability of the World Health Organization’s 10-item vaccine hesitancy scale (VHS) in different settings is not sufficient, especially for criteria validity. This study aimed to assess the validity and reliability of the VHS using child vaccination data in China. A cross-sectional survey was performed with parents of 19–48-month-old children at six vaccination clinics in Wuxi City between September and October 2020. The VHS was revised to category A (expanded program on immunization, EPI) VHS and category B (Non-EPI) VHS. Factor analysis was used to confirm the latent domain and to assess the model structure. The average variance extracted (AVE) was calculated to assess convergent validity, and Cronbach’s α and composite reliability (CR) were used to determine internal consistency. The association between VHS scores and children’s vaccination status was examined to assess criteria validity using logistic regression. The survey response rate was 75.3% (n = 802). Two factors were identified, explaining 64.60% and 63.34% of the common variance in categories A and B VHS, respectively. The Cronbach’s α of > 0.7 and CR of >0.7 in the scale indicated the VHS has acceptable internal consistency. The AVE values indicated that convergent validity was not ideal for the VHS. There were no statistically significant associations between VHS scores and vaccination status, indicating that the criterion validity was not ideal. The VHS needs improvement before becoming a standard survey tool. As early as the 18th century, people hesitated to vaccinate against smallpox. 1 One of the most remarkable pieces of evidence of this is an anti-vaccine picture published in the magazine Punch during this time period. 2 In the picture, people who received smallpox vaccination developed features of cows. In the mid-1970s, antivaccine movements reemerged because it was reported that 36 children experienced serious adverse effects after receiving the diphtheria, tetanus, and pertussis vaccine (DTP). 2 In the 21st century, the Internet has become an unprecedented, rapid, and effective tool to propagate anti-vaccine messages. 3 According to the World Health Organization (WHO), vaccine hesitancy was considered one of the ten issues threatening global health in 2019. 4 Because vaccine hesitancy has the potential to contribute to a decrease in vaccine uptake, it could cause the emergence and reemergence of vaccine-preventable diseases. For example, more than 100,000 pertussis cases were reported in the UK after antivaccine movements against DTP emerged in the mid-1970s. 2 The 2014-15 Disneyland measles outbreak, which reported the highest number of measles cases in the USA in two decades, attracted significant attention, and vaccine hesitancy was considered as the cause of this outbreak. 5 In 2012, WHO defined the term "vaccine hesitancy" as follows: the refusal or delay in acceptance of vaccination despite availability of vaccination services. 6 There is a continuum of vaccine hesitancy, between the two extremes of completely accepting and completely refusing all vaccines. It is influenced by complex factors, including confidence (including lack of confidence in vaccine and healthcare workers), complacency (such as not paying attention to the need for vaccines), and convenience (such as high price). 6 Because of this complexity, it is crucial to develop a standardized survey tool to measure, evaluate, and compare vaccine hesitancy in different settings and over time. Work of this nature has been developed since 2011. Various survey tools, including the parent attitudes about childhood vaccines (PACV) scale with three domains, 7 the vaccine confidence scale with three domains, 8, 9 the vaccination attitudes examination scale with three domains, 10 the knowledge of vaccination scale with one domain, 11 and caregiver vaccination attitudes scale with three domains, 12 have been developed. Two systematic reviews evaluating tools measuring parental vaccine hesitancy found that there existed heterogeneity in the survey tools and many tools were not validated. 13, 14 In 2015, the 10item vaccine hesitancy scale (VHS) was developed by the Strategic Advisory Group of Experts through various methods, including systematic review and expert consultation. 15 This tool has been validated in several countries. [16] [17] [18] [19] [20] However, these findings are inconsistent. Four studies reported two domains being identified in the VHS, [16] [17] [18] [19] and one study reported three domains being identified. 20 Most studies suggested that some items need to be deleted to improve the scale's construct validity, but the items to be deleted in the different studies varied. In addition to internal consistency and construct validity, criterion validity is also an important index. Only one study used vaccination records to explore the association between the scale and children's vaccination status to evaluate the criteria validity of this scale. 20 VHS scores were related to vaccination status and not related to on-time vaccination. 20 The criteria validity of the VHS needs to be assessed in more studies. Meanwhile, this scale needs to be comprehensively validated in China. In China, immunization programs should be considered in the measurement. There are two types of vaccination programs in China: the Expanded Program on Immunization (EPI) and non-EPI vaccines. EPI vaccines, also known as category A vaccines, are free and mandatory, while non-EPI vaccines, also known as category B vaccines, are optional and billed. Different types of vaccines might result in different hesitancy levels and different vaccination statuses. Our study aims to describe and examine the validity and reliability of the VHS using child vaccination data in China, including the scale's structure, internal consistency, convergent validity, and criterion validity. This cross-sectional study was performed in Wuxi City, China, between September and October of 2020. Participants were recruited from local vaccination clinics. The study was approved by the Ethics Committee of the Wuxi Center for Disease Control and Prevention (2020No10). The sampling method for vaccination clinics has been shown in a previous study. 21 Areas (counties) in Wuxi were divided into three groups according to gross domestic product in 2019 and one area was selected randomly in each group; then two clinics in each area were sampled using the simple randomization technique. All parents who brought their children to clinics received a paper-questionnaire and an informed consent form during the survey period. They were informed of the purpose of the investigation and anonymity of the data. The parents who agreed to participate in the survey signed the informed consent form and completed the questionnaire. Children aged 19-48 months were included in the survey because the previous study suggested that this age was a suitable time to collect parental views on vaccines, as the children would have received most of their vaccination doses by this age range. 7 We excluded children who did not have a unique vaccination identification (ID) number or a child who was not born in Wuxi city. Parents were encouraged to respond to all questions and received a gift worth RMB 5 when they completed the questionnaire. The VHS was translated into Chinese. The scale was independently translated into Chinese by two researchers. These two editions were then translated back into English by two language professionals. An expert committee consisting of experienced researchers in vaccination work, language professionals, and sociologists reviewed the two editions and developed the first Chinese edition. The pilot study was performed in Wuxi City (approximately 50 respondents) to assess the first edition. The main purpose of the pilot study was to determine whether parents could understand the scale and the time needed to complete the scale. It was found that some parents were puzzled at items (L5 and L10) and about one minute was needed to complete this scale. The completing time was considered to be acceptable. Without changing the original meaning of the VHS, we adjusted the Chinese expression of items L5 and L10 and final edition was developed. To distinguish category A and B vaccines, we added "category A" and "category B" before "vaccine" in the scale. For example, the description of the L2 item was "Childhood vaccines were effective. We revised it to "Category A childhood vaccines are effective" and "Category B childhood vaccines are effective." A five-point Likert scale was used to answer the VHS (strongly disagree, 1 point; disagree, 2 points; not sure, 3 points; agree, 4 points; or strongly agree, 5 points). In addition to vaccine hesitancy, parents and children's sociodemographics were also collected. Vaccination records were derived from an electronic information system using the ID number on November 18, 2020. There were eight types of category A vaccines in EPI: Bacillus Calmette-Guérin vaccine (BCG), hepatitis B vaccine (HepB), polio vaccine (PV), DTP, measles-containing vaccine (MCV), Japanese encephalitis vaccine (JEV), hepatitis A vaccine (HepA), and meningococcal polysaccharide vaccine type A (MPSV-A)/meningococcal polysaccharide vaccine type A and C (MPSV-AC). 22 There are more than 20 types of category B vaccines in non-EPI, including enterovirus 71 vaccine (EV71), seasonal influenza vaccine (SIV), pneumococcal polysaccharide conjugate vaccine (PCV), Haemophilus influenzae type b vaccine (Hib), rotavirus vaccine, human papillomavirus vaccine (HPV), some combined vaccines (e.g., DTP and Hib combined vaccine; DTP-Hib), and some alternatives to category A vaccines. We calculated the delay status for the children's category A vaccines. The delay calculation method was derived from previous studies. 7, 23 The delay was determined as the actual vaccination date minus the recommended vaccination date. Delays in BCG, HepB1, HepB2, HepB3, PV1, PV2, PV3, DTP1, DTP2, DTP3, MCV1, and JEV1 were included in the analyses. HepA was administered at 18 months, which was close to the minimum age of the included children, making it impossible to calculate this vaccine's delay. The meningococcal polysaccharide-containing vaccines were also not included in the analyses because the recommended schedules of MPSV-A and MPSV-AC were different and there were corresponding alternative category B vaccines, making it difficult to calculate the delay. The vaccination schedules are provided in Table 1 . Vaccination was regarded as the outcome indicator of category B vaccines, and we included EV71, SIV, and PCV in the analyses. 20 Descriptive statistics were used to analyze the sociodemographics and VHS responses. The Kaiser-Meyer-Olkin value and Bartlett's test of sphericity were used to examine correlations between the 10 items (whether approximate to perform factor analysis). The sample was randomly split into two halves, one for exploratory factor analysis (EFA) and another for confirmatory factor analysis (CFA). The sociodemographics between two groups were not statistically significant. In EFA, the factors with eigenvalues >1.0, were extracted with varimax rotation. According to the structure identified in EFA, CFA was used to confirm whether each latent domain (factor) was properly measured. The Chi square/DF index (χ 2 /DF), root mean square error of approximation, goodness-of-fit index, adjusted goodness-of-fit index, comparative fit index, Tucker-Lewis index, and standardized root mean square residual were used to evaluate the model fit. The items with low factor loadings would be deleted to determine whether the model fit could be improved. The average variance extracted (AVE) was calculated using the factor loadings to assess convergent validity in each domain. The square root of AVE and correlations between latent factors to assess the discriminant validity. Cronbach's α and composite reliability (CR) was calculated to determine the internal consistency reliability for the items in each factor. The sum of scores to 10 items was calculated for category A and category B vaccine scales. The answers to L5, L9, and L10 were flipped because these three items were worded negatively and other items were worded positively. 20 The maximum and minimum scores were 50 and 10 points, respectively. Higher scores represented lower vaccine hesitancy. The delaying status for category A vaccines was divided into six groups based on a previous study: 0 days, 1-7 days, 8-31 days, 1-2 months, 3-6 months, and >6 months. 21 The administration status for category B vaccines was divided into four groups: 0 type (received none), 1 type (received one of three vaccines), 2 types (received two of three vaccines), and 3 types (received all three vaccines). By adjusting the children's and parents' socio-demographics, we used logistic regression to examine the association between category A VHS scores and category A vaccine delay, and the association between category B VHS scores and category B vaccine administration status to assess the criteria validity. The association between each item in the VHS and vaccination status was also examined. To avoid too few answers for some options in some items, "strongly disagree," "disagree" and "not sure" were combined into "Not sure and disagree," "strongly agree" and "agree" were combined into "Agree." Statistical significance was defined as a two-sided P value of <0.05. All analyses were performed using SPSS 23.0 (IBM Corp, New York, NY, USA) and AMOS 23.0 (IBM Corp, New York, NY, USA) software. Overall, 802 parents completed the questionnaire, with a response rate of 75.3% (Table 2 ). Approximately 76.7% of respondent parents were mothers, and 68.5% of respondent parents had an educational level of college or higher. More than 90.0% of respondents' parents were not involved in the healthcare occupation. The category A and B vaccine hesitancy scores were 40.81 ± 4.00 and 38.32 ± 5.08, respectively. Less than 1% of parents showed disagreement with L1, L2, L3, L4, L6, and L7 items in category A VHS (Figure 1 ). More than 20% of parents showed agreement with the L5 and L9 items in category B VHS. There were 29.93% of children with no delay in category A vaccines ( Figure 2 Bartlett's test of sphericity (P < .001) showed that correlations between items were sufficient to conduct a factor analysis. Two factors were identified with eigenvalues >1, which explained 64.60% and 63.34% of the common variance of the 10 items in category A and B VHS, respectively (Table S1 ). In category A and B VHS, there are seven items (L1-L4, L6-L8) loading on factor 1, which were related to vaccine confidence and three items (L5, L9, L10) loading on factor 2, which were related to vaccine risk and complacency. Most of the factor loadings of each item in the two scales were >0.7, except for item L8 in category B VHS (0.653). In categories A and B, the model fit was acceptable according to a series index ( Table 3 ). The model fit was not improved when items L2 and(or) L5 were deleted in category A VHS. The model fit was improved when items L8 and(or) L10 were deleted in category B VHS. In category A VHS, the AVE of 0.451 for Factor 1, and the AVE of 0.475 for Factor 2. In category B VHS, the AVE was 0.640 for Factor 1, and the AVE was 0.531 for Factor 2 (Table S2 ). The suggested values for AVE and CR were >0.5 and >0.7, respectively. The convergent validity of the VHS is not ideal. The correlation between two latent factors was significant and was lower than the square root of AVE, which indicated the discriminant validity was good in category A and B VHS. In category A VHS, Factor 1 consisted of seven items with a Cronbach's α of 0.900, and factor 2 consisted of three items with Cronbach's α of 0.718. In category B VHS, the Cronbach's α of the seven items in Factor 1 was 0.894, and that of Factor 2 was 0.727. In category A VHS, the CR was 0.851 for Factor 1, and the CR was 0.727 for Factor 2. In category B VHS, the CR was 0.925 for Factor 1, and the CR was 0.770 for Factor 2. The low correlation between two latent factors being lower than square roof of AVE indicated that VHS has a good discriminant validity. (Table S3 ). The Cronbach's α of >0.7 and CR of >0.7 indicated the VHS has an acceptable internal consistency. We did not find a statistically significant association between category A VHS scores and category A vaccine delay status after adjusting for socio-demographics (Figure 3) . No statistically significant association was found between category B VHS scores and category B vaccine administration status. Category B VHS was not related to EV71 or PCV administration status. However, category B VHS was related to influenza vaccination status. The parents were less likely to show hesitancy in the category B vaccine when their child was vaccinated against influenza (adjusted odds ratio [aOR]: 1.06, 95% CI: 1.02, 1.09, P = .003). Each item in category A VHS was also not found to be associated with a category A vaccine delay (Table S4 ). There was also no statistically significant being found in the association between each item in category B VHS and category B vaccine administration status (Table S5) . Similarly, each item in category B VHS was not related to EV71 or PCV status (Table S6) . Item L3 was statistically associated with influenza vaccination status (aOR: 1.72, 95% CI: 1.01, 2.95, P = .047). Our findings demonstrated that the VHS scale has two domains, and the model fit and internal consistency are acceptable in Chinese settings. The low correlation between two latent factors (