Microsoft Word - Edited-GJPHM-2021- ADULT IMMUNIZATION.docx 496 GLOBAL JOURNAL OF PUBLIC HEALTH MEDICINE 2021, VOL 3, ISSUE 2 gggggglo Original Research DEVELOPMENT AND CONSTRUCT VALIDATION OF QUESTIONNAIRE ASSESSING THE ADULT IMMUNISATION PERCEPTIONS AMONG MALAYSIAN POPULATION Siti Nor Mat, Shamsul Azhar Shah, Syafiq Taib, Norzaher Ismail & Mohd Rohaizat Hassan Department of Community Health, Faculty of Medicine, Universiti Kebangsaan Malaysia, 56000 Cheras, Kuala Lumpur, Malaysia *Corresponding author: sitinor.mat@gmail.com ABSTRACT Introduction: Immunisations are one of the most effective public health interventions, reducing or eliminating the burden of many infectious diseases. This study aims to establish the construct validity of a newly developed adult immunisation perceptions questionnaire among Malaysians. Methods: The Adult Immunisation Perceptions-Questionnaire (AIP-Q) was created following literature reviews on The Health Belief Model. Primarily, 64 questions were pooled, followed by face validity by experts, pre- tested via 20 healthcare personnel and later translate into the Malay language. A total of 305 respondents were selected for the construct validation process. Varimax rotation method used in the analysis for exploratory factor analysis (EFA) and confirmatory factor analysis (CFA) done using AMOS software. Results: Ten constructs were produced as predicted in EFA: health believes, experience, knowledge, attitude, perceived severity, perceived susceptibility, perceived benefits, perceived barriers, and 2 cues for action. Thirteen items with low factor loading and unrelated to the recovered domains were removed from being included in CFA. In path analysis, the scale fitted χ2/df=1.943 (n=305) =p<0.001, CFI=0.908, IFI=0.909, TLI=0.901 and RMSEA=0.056. Strong factor loading was found across the final items, ranged from 0.53 to 0.94 with a good reliability test (Cronbach Alpha, AVE and CR values) for all constructs. Conclusion: The final AIP-Q consists of 10 domains with 45 items that give a promising psychometric property. This questionnaire can measure the perceptions of adult immunisation among the Malaysian population and can be utilized for the nationwide study. Keywords: Adult immunisation perception questionnaire, factor analysis, construct validation 497 GLOBAL JOURNAL OF PUBLIC HEALTH MEDICINE 2021, VOL 3, ISSUE 2 gggggglo INTRODUCTION Immunisations are one of the greatest public health achievements of the twentieth century (Alexa et al.,2016). Adult immunisation is not commonly practiced in most countries as well as in Malaysia. Even though there are significant morbidity and mortality due to Vaccine-Preventable Diseases (VPD) within this age group, the awareness of the benefits of immunisation for adults is still lacking (Malaysia Adult Immunisation Guideline, 2014). Hence, less attention has been given to adult immunisation, even in developed countries with strong public health infrastructures (Levine et al.,2011). Data from the Ministry of Health Malaysia, 2011 showed that VPD are still commonly encountered in Malaysia, as the incidence rates for measles, hepatitis B and pertussis were 5.42, 4.32, 0.86 per 100,000 populations respectively (Malaysia Adult Immunisation Guideline, 2014). Whereas in Singapore, the proportion of influenza-associated deaths was 11.3 times higher in persons aged 65 years and above (Singapore Report, 2014). Meanwhile, in the United States, the VPD caused death in approximately 50,000 adults every year: 36,000 from influenza, over 6,000 from invasive pneumococcal disease and 5,000 from hepatitis B (US Report, 2009). However, fewer than 500 children die from VPD in the United States each year. Various factors have been recognized to influence the perception towards adult immunisation and vaccine preference comprising socio-demographic, socio-economic, health-related factors, social impacts, disease/vaccine-related factors, common attitudes towards health and vaccines, custom, awareness and knowledge, real-world barriers and promoters, and humanity (Wheelock et al.,2013). Although strategies have been identified and few programmes have been initiated to encounter the low vaccine coverage in adults (Poland et al.,2010), little progress has been achieved. This study aims to determine its reliability and validity based on structural equation modeling (SEM) within a CFA which provide reliable instrument and valid measurement for components of health belief. In addition this questionnaire should provide a useful tool to measure the perceptions of adult immunisation among Malaysian population with age group 18 and above. METHODS Respondents This is a cross-sectional study, recruiting 305 respondents aged ≥18 years in April 2018. The study involved two health clinics in Gombak District, Selangor among people who are attending the health clinics regardless of patient or caregiver. All the respondents who fulfilled the criteria of (1) age 18 years and above (2) understand English or Malay language, and (3) willing to participate in the survey, were invited to participate in this study. The researcher cross-checks the answer thoroughly, ensuring no missing responses of the items and respondents allowed to verify the difficult or confusing questions. The respondents were told if they felt that they did not have enough information or knowledge about any of the items to select the best answer closest to their perception. This is an important element of the instructions so that the respondents would feel comfortable about answering the questions. Instruments The initial 64 items of the Adult Immunisation Perception Scale (AIPS) were generated to fill the gap in perception and its associated factors about adult immunisation among Malaysians. The Health Belief Model was used in developing items in the questionnaire where it includes the health belief model domains: susceptibility, severity, barriers, benefits, and cues to action as well as its associated factors including health belief, past medical experience, knowledge, and attitude towards adult immunisations. There are 4 steps to generate the items which are: (1) identify appropriate constructs, (2) form item pool, (3) define a format for measurement, and (4) all item pool should be revised by experts, and later pre-tested. (De Vellis, 2003) Description for each step is summarized as follow: 1) Identifying Appropriate Constructs The questions were generated after wide exploration from quantitative and qualitative literature on knowledge, attitude, and perception about adult immunisations. Several instruments for measuring knowledge, attitude and perception regarding immunizations have been developed in various countries 498 GLOBAL JOURNAL OF PUBLIC HEALTH MEDICINE 2021, VOL 3, ISSUE 2 gggggglo at different times [Rashwan et al.,2011; Antoinette et al.,2012; Halperin et al.,2015; Donadiki et al.,2014 & Halperin et al.,2015]. The framework was based on Health Belief Model and any questionnaire related to perceptions with or without The Health Belief Model were used. Development of instrument items to measure the constructs in this study is based on the Lazarsfeld Scheme which involves four stages namely imaginary concept, specification concept, index selection and index construction. 2) Generating Item Pool Originally, the item pool consisted of 64 items, in which five items to measure health belief, six items to measure past experience, 15 items measuring knowledge, six items measuring attitude and 32 items to measure perception which include four items to measure susceptibility, six items to measure severity, four items measuring benefits, four items measuring barriers, 11 items measuring cues for action. 3) Define Format for Measurement Items were scored on a tenth-point agreement level; using Likert scale; 1 = Strongly Disagree to 10 = Strongly Agree. According to Zainudin et al.,2015, to determine how much agree or disagree of the respondent towards particular questions, the long scale such as 10 points is way better than the short scale. The respondents have also been told to choose the best answer closest to their perception about the topic even though they do not have enough knowledge or information about it. This instruction was important to make the respondents feel comfortable while answering the questions. 4) Experts review Three experts were chosen among Public Health Consultant from Ministry of Health Malaysia and local universities to review the scale and content validity. According to Rosnah et al.,2013, the experts are better from healthcare professionals since they have enough information and updated with the objective and conceptual basis measure. They were evaluating the clarity of the items, decide whether they are relevant to the domain, and justify whether the items should remain in the pool or not. The assessment of the questionnaire was continued pretested by selected 20 healthcare workers at a health clinic. They were asked to constructively comment on each item by evaluating its objective, suggest items for deletion, alteration, or recommend new items. All their perceptive thinking and probing results were recorded. Finally, the item pool was translated from English to Malay using the simplified back-translation method (Brislin, 1976). Analysis of Data The EFA used to assess items for its psychometrically, replicating Naing’s suggestion (Naing, 2010). The EFA was commenced using principal component analysis (PCA) and promax rotation with eigenvalues more than 1; factor loading <0.40 was removed. It was then followed by an examination of rating scale quality and later by CFA. Analysis was performed using IBM SPSS Statistics version 22. The value of the score scale groups was observed to create; (i) the group regularities exhibited a steady distribution (i.e., uniform, normal, bimodal, or slightly skewed), (ii) at least ≥10 responses per group item, (iii) the adjacent threshold distance between 1.4 and 5.0 logits, (iv) the regular measures increased monotonically through the score scale, (v) a different probability curve graph in each response group, and (vi) the outfit was <2 to measure the suitability of the ten-point Likert scale response group. The analysis replicated Rasch model analysis using Winstep (Bond & Fox, 2015). Maximum likelihood estimate used in assessing model fitness to the covariance matrix of the CONFIRM data set, namely comparative fit index (CFI) > 0.9, goodness of fit index (GFI)> 0.9, normed fit index (NFI)> 0.9, root mean square error of approximation (RMSEA) range 0.05 to 0.1, and chi-square difference (chisq/df) <5.0 (Zainudin, 2012). The items that persistently stable after EFA and CFA analysis were reserved (Hair et al.,2010). 499 GLOBAL JOURNAL OF PUBLIC HEALTH MEDICINE 2021, VOL 3, ISSUE 2 gggggglo RESULTS Respondents Total of 305 respondents involved in validity and reliability study. Since this study was conducted using one-to-one interview, the response rate was 100%. Respondents age was between 18 to 71 years old, with mean (±s.d.) of 30.67 (+10.73) years. Female respondents were 190 (62.3%) compared to 115 (37.7%) males. The majority of the respondents were Malay, 257 (87.0%). In terms of work, 186 (61.1%) of them were working while 109(35.7 %) were unemployed and the rest were retired. 248 (81.3%) stayed in city area (Table 1). Table 1: Details of socio-demographic characteristics of respondent participated in the validation study of Adult Immunisation Perception-Questionnaires (AIP-Q) CHARACTERISTICS FREQUENCY (n) PERCENTAGE (%) AGE (YEARS) Mean (+SD): 30.67 (+10.73) GENDER Male Female 115 190 37.7 62.3 RACE Malay Chinese Indian Bumiputera Sarawak Bumiputera Sabah 257 26 17 1 4 84.3 8.5 5.6 0.3 1.3 MARITAL STATUS Single Married Divorced 118 186 1 38.7 61.0 0.3 EDUCATION LEVEL Illiterate Primary Secondary Tertiary 1 4 62 238 0.3 1.3 20.3 78.0 WORKING STATUS Yes No Pensioner 186 109 10 61.0 35.7 3.3 WORKING SECTOR Government Private Self-employment Unemployed Student 55 104 39 55 52 18.0 34.1 12.8 18.0 17.0 MONTHLY INCOME RM7000 107 112 70 16 35.1 36.7 23.0 5.2 PLACE OF LIVING Urban Rural 248 57 81.3 18.7 MEDICAL ILLNESS Yes No 47 258 15.4 84.6 500 GLOBAL JOURNAL OF PUBLIC HEALTH MEDICINE 2021, VOL 3, ISSUE 2 gggggglo Exploratory Factor Analysis A preliminary PCA was carried out to explore the data set. The result was divided into 2 subgroups: 1) Adult immunisation perception scale and 2) Adult immunisation domain scale. The Kaiser-Meyer-Olkin (KMO) of Domain scale and Perception scale was 0.859 and 0.887 respectively with Bartlett’s test of sphericity was significant at p <0.001. It showed that this data is suitable and useful to proceed with factor analysis. Initially,8 components with eigenvalues above 1.0 were generated in Domain scale whereas 7 components generated in Perception scale. A total of 15 constructs were not exhibited an unacceptable reliability alpha of 0.7, as well as the items representing five other constructs also failed to be theoretically meaningful. The deleted item in Domain scale and Perception scale are shown in Tables 2 and 3. Table 2: The initial factors extraction by the EFA using PCA extraction method with promax rotation and reliability analysis of each construct. ITEM MEAN (±SD) PERCEIVED SUSCEP- TIBIITY PERCEIVED SEVERITY PERCEIVED BENEFITS PERCEIVED BARRIERS CUES 1 CUES 2 ITC IIC CRON- BACH ALPHA F1i 6.01 (2.363) 0.967 0.796 0.58- 0.85 0.865 F1ii 5.97 (2.255) 0.942 0.824 0.62- 0.85 F1iii 6.18 (2.228) 0.683 0.623 0.58- 0.62 F1iv 6.12 (2.303) Item deleted F2i 5.55 (2.393) 0.777 0.769 0.59- 0.89 0.934 F2ii 5.68 (2.290) 0.804 0.836 0.62- 0.89 F2iii 5.49 (2.416) 0.921 0.816 0.63- 0.80 F2iv 5.64 (2.402) 0.938 0.841 0.65- 0.80 F2v 5.92 (2.411) 0.904 0.833 0.64- 0.80 F2vi 5.53 (2.576) 0.823 0.727 0.59- 0.69 F3i 7.24 (2.300) 0.891 0.686 0.57- 0.67 0.905 F3ii 7.41 (1.972) 0.824 0.833 0.66- 0.78 F3iii 7.60 (1.873) 0.850 0.837 0.67- 0.79 F3iv 7.40 (1.936) 0.805 0.777 0.57- 0.77 F4i 7.83 (2.205) 0.812 0.657 0.44- 0.64 0.829 F4ii 7.41 (2.420) 0.798 0.636 0.48- 0.64 F4iii 5.89 (2.658) Item deleted F4iv 5.30 (2.642) Item deleted F4v 7.50 (2.397) 0.789 0.632 0.44- 0.67 501 GLOBAL JOURNAL OF PUBLIC HEALTH MEDICINE 2021, VOL 3, ISSUE 2 gggggglo F4vi 7.90 (2.232) 0.839 0.693 0.48- 0.67 F5i 7.89 (2.320) Item deleted 0.885 F5ii 6.88 (2.054) 0.554 0.540 0.39- 0.50 F5iii 6.71 (1.887) 0.837 0.715 0.46- 0.74 F5iv 6.37 (1.880) 0.758 0.636 0.43- 0.74 F5v 7.52 (1.947) 0.730 0.746 0.43- 0.79 F5vi 7.24 (2.143) 0.886 0.817 0.45- 0.86 F5vii 7.26 (2.133) 0.869 0.742 0.39- 0.86 F6i 7.09 (2.451) 0.690 0.531 0.39- 0.48 0.824 F6ii 7.31 (2.171) 0.758 0.621 0.39- 0.55 F6iii 6.56 (2.286) 0.836 0.627 0.43- 0.52 F6iv 7.44 (2.075) 0.742 0.639 0.41- 0.58 F6v 7.70 (2.008) 0.658 0.659 0.46- 0.58 Table 3: The initial factors extraction by the EFA using PCA extraction method with promax rotation and reliability analysis of each construct. ITEM MEAN (SD) HEALTH BELIEVE PAST EXPERIENCE ATTITUTE GENERAL KNOWLEDGE ITC IIC CRONBACH ALPHA B2i 4.51 (1.892) 0.633 0.297 0.16-0.29 0.571 B2ii 6.66 (2.566) 0.711 0.502 0.22-0.48 B2iii 5.02 (2.420) 0.629 0.238 0.16-0.22 B2iv 5.21 (2.317) Item deleted B2v 6.46 (2.772) 0.645 0.412 0.16-0.48 C1i 8.57 (1.787) 0.859 0.738 0.49-0.78 0.874 C1ii 8.44 (1.951) 0.888 0.765 0.58-0.78 C1iii 7.40 (2.127) Item deleted C2i 8.70 (1.750) 0.863 0.754 0.61-0.68 C2ii 8.38 (2.027) 0.783 0.655 0.49-0.68 C2iii 7.48 (2.084) Item deleted E1 6.32 (2.832) Item deleted 0.706 E2 6.94 (2.066) 0.734 0.557 0.32-0.58 502 GLOBAL JOURNAL OF PUBLIC HEALTH MEDICINE 2021, VOL 3, ISSUE 2 gggggglo E3 7.42 (1.906) 0.557 0.593 0.32-0.58 E4 6.46 (2.534) 0.865 0.389 0.24-0.37 E5 7.28 (2.212) Item deleted E6 7.38 (2.125) 0.500 0.407 0.24-0.43 D1 7.85 (1.800) 0.901 0.779 0.45-0.78 0.941 D2 7.84 (1.808) 0.885 0.768 0.40-0.74 D3 7.96 (1.727) 0.844 0.765 0.48-0.75 D4 7.86 (1.882) 0.892 0.802 0.51-0.75 D5 8.14 (1.806) 0.755 0.774 0.48-0.71 D6 7.28 (2.194) Item deleted D7 6.81 (2.187) Item deleted D8 6.04 (2.901) Item deleted D9 8.22 (1.796) 0.622 0.717 0.47-0.65 D10 3.82 (2.536) Item deleted D11 8.07 (1.889) 0.712 0.671 0.40-0.66 D12 7.85 (1.862) 0.852 0.786 0.51-0.88 D13 7.87 (1.780) 0.884 0.829 0.49-0.88 D14 7.61 (1.866) 0.684 0.697 0.44-0.70 D15 7.94 (1.772) 0.650 0.598 0.44-0.55 Note: • Initial total items were 64, 13 items were deleted during EFA. Only 51 items left for CFA. • The item with bold text was removed from being included in the CFA To further ensure that the EFA process to be accurate, the scale of reliability analysis of those 10 constructs were performed to the remaining 51 items which had factor loading >0.4. The reliability test showed that the ITC and IIC values of each item towards their respective constructs were acceptable except for items B2i and B2iii. However, this item remained since it has good factor loading (>0.6) with an acceptable Cronbach Alpha value of 0.57. Scale reliability was 0.89 with good subscale reliability ranging from 0.57 to 0.94 (Table 2 and 3). Confirmatory Factor Analysis The CFA was performed to test the exploratory factor structure using AMOS software. CFA was conducted on the 51 items identified in the exploratory data set and was found to provide an acceptable, but relatively poor fit based on the goodness of fit statistics: χ2/df=2.674 (n=305) = p<0.001, CFI=0.810, IFI=0.811, TLI=0.800 and RMSEA=0.074 (Figure 1). 503 GLOBAL JOURNAL OF PUBLIC HEALTH MEDICINE 2021, VOL 3, ISSUE 2 gggggglo Figure 1: The path analysis of the CFA showing standardized estimates of the correlations between the five second order constructs (ellipse), six first order constructs (ellipse) and the fifty-one items (rectangle), and their respective residuals (circle). The numbers on the double headed arrows indicate the calculated correlation values by the path analysis. Note: Certain fitness indexes do not achieve the required level. Upon examination, items B2i, B2iii, E4, F5ii, F5iv had low factor loading and item deleted. According to Zainudin, 2012, a discriminant validity which other criteria for measurement model, the items must be free from redundancy. Any identified items with a high value of modification index (MI) were either deleted or constrained as “free parameter estimate”. The measurement model was run again after MI assessment was carried out since fitness index still not acceptable even after removing items with low factor loading. Finally, the result showed a significant improvement, and fit the 45-item model over the 51-item model, better chi-squared value, and goodness of fit statistics. The shown model was estimated against the recent data sets and produced χ2/df=1.943 (n=305) =p<0.001, CFI=0.908, IFI=0.909, TLI=0.901 and RMSEA=0.056 (Figure 2). Strong factor loadings were found across the 45 items, ranging from 0.53 to 0.94 (Table 4). 504 GLOBAL JOURNAL OF PUBLIC HEALTH MEDICINE 2021, VOL 3, ISSUE 2 gggggglo Figure 2: The final measurement model of Adult Immunisation Perception Questionnaires (AIP- Q) Another crucial requirement of discriminant validity is the value of correlation between exogenous constructs must not >0.85 (Zainudin, 2012). Four out of five factors were managed to get out correlation values below 0.85 but only one factor just having slightly above the limit value which is between Knowledge and Perception (0.86). Overall, the exogenous constructs were not redundant or had a multicollinearity issue. Figure 2 explains the five domains; assumed to cause variation and covariation in the measurement of perception. There is a double-headed arrow between the five domains, indicates that the five domains are correlated. 505 GLOBAL JOURNAL OF PUBLIC HEALTH MEDICINE 2021, VOL 3, ISSUE 2 gggggglo Table 4: The CFA report for every construct in the measurement model with factor loadings of each item towards respective domain- Average Variance Extracted and Composite Reliability. CONSTRUCT ITEM FACTOR LOADING CRONBACH ALPHA COMPOSITE RELIABILITY (CR) AVERAGE VARIANCE EXTRACTED (AVE) Health Belief B2i B2ii B2iii B2v Item deleted 0.75 Item deleted 0.64 0.645 0.65 0.49 Past Experience C1i C1ii C2i C2ii 0.91 0.85 0.72 0.70 0.871 0.88 0.64 Attitude E2 E3 E4 E6 0.65 0.86 Item deleted 0.53 0.699 0.73 0.48 Knowledge D1 D2 D3 D4 D5 D9 D11 D13 D14 D15 0.81 0.82 0.80 0.83 0.81 0.76 0.67 0.80 0.70 0.63 0.933 0.93 0.59 PERCEPTION P. Susceptibility P. Severity P. Benefit P. Barrier Cues to Action 1 Cues to Action 2 0.46 0.31 0.88 0.26 0.71 0.60 0.67 0.33 Perceived susceptibility F1i F1ii F1iii 0.89 0.94 0.66 0.865 0.88 0.70 Perceived Severity F2i F2ii F2iii F2iv F2v F2vi Item deleted 0.80 0.88 0.90 0.89 0.76 0.925 0.93 0.72 Perceived Benefit F3i F3ii F3iii F3iv 0.72 0.90 0.89 0.86 0.900 0.91 0.72 Perceived Barrier F4i F4ii F4v F4vi 0.62 0.57 0.77 0.88 0.827 0.81 0.52 Cues to Action 1 F5ii F5iii F5iv F5v F5vi F5vii Item deleted 0.71 Item deleted 0.89 0.87 0.73 0.894 0.88 0.65 Cues to Action 2 F6i F6ii F6iii F6iv F6v 0.58 0.69 0.67 0.74 0.78 0.819 0.82 0.48 506 GLOBAL JOURNAL OF PUBLIC HEALTH MEDICINE 2021, VOL 3, ISSUE 2 gggggglo Cronbach’s alpha for the AIP-Q was beyond 0.6 in all domains, representing adequate level of internal consistency. Table 4 showed that internal consistency for all factors is supported by composite reliability (CR) and the average variance extracted (AVE). In order to achieve the CR, the value should be >0.6. As for AVE, it indicates the average percentage of variation explained by the measuring item for a latent construct. The value of AVE must be >0.5 for every individual construct. DISCUSSION Immunisation can prevent infectious diseases, decrease morbidity as well as mortality of some diseases. At present, there are 26 infectious diseases that can be protected by vaccines (Andre, 2003). Nevertheless, adult immunisations often remain under-utilized when most of them consider those vaccines are recommended but not required (US Report, 2000). Several factors influenced the immunisation perception in adult population. By recognizing the factors and groups that give unsatisfactory feedback in adult immunisation, appropriate communications and delivery of messaging can be tailor specifically to those groups. Numerous tools for assessing knowledge, attitude and perception about immunisations have been established in many countries at different times (Rashwan et al.,2011; Antoinette et al.,2012). By using these tools and/or comparing them with few research, it can provide interesting results but it must be done with careful supervision. It is due to many factors that can influence knowledge, attitudes, and their perception regarding adult immunisation where it should be taken into consideration, such as their different socioeconomic background, cultural and religious beliefs of different populations, epidemiological differences, and policy support (Ammar et al.,2014). The developed questionnaire (in the English language) was translated into Bahasa Melayu, the common spoken language of Malaysians. The skilful bilingual translator then certified that the translation was precise and correct by review on both versions (English and Malay). The development and validation of the original items made the questionnaire relatively simple and practical to use among the Malaysian population. Ten points Likert scale were used in this questionnaire compared to normal 5 or 7 points. According to Zainudin et al., 2015, to determine how much agree or disagree of the respondent towards particular questions, the long scale such as 10 points is way better than the short scale. The scale of measurement is supposedly reflecting the actual intention of the respondent towards the question submitted. Since this study is using SEM for validation, both measurement and structural models can be assessed successfully with 10 points of Likert scale in determining the construct validity. The statistical indices from SEM were used for a final version questionnaire which consists of 45 items in order to get a suitable degree of reliability and validity (Hafizah et al.,2002). The Malay version of AIP-Q from this study also indicate that it has acceptable psychometric properties to measure the knowledge, attitude, and perception among adult population in Malaysia regarding immunisation. The overall observations were adequate; ensuring stability during the EFA, where its able to produce usable results from 64 items to 51 items. According to William et al.,2012, it is considered necessary for the item to have a rotated factor loading of at least 0.4 (meaning >+0.4 or <−0.4) to make it significant. Therefore, if the item was shown not to have validity even after some additional criteria had been measured, the items were eliminated if the factor loading constantly less than 0.4 (Costello, 2009). Since there was no previous model to test against this model, absolute fit indices were used in this study (Schermelleh et al.,2003). The CFA is a method to identify certain variables that belongs to certain factors and it is likely to identify further which factor items are loaded to. Meanwhile, factor analysis allows the factor scores to be explained by common factors which only contain variance (Williams, 2012). Thus, the effect of measurement error can be removed by calculating the factor scores and putting them into path analysis, Fortunately, this is the principle of SEM which offers several advantages in performing CFA (Schermelleh et al.,2003). The reliability testing was higher than acceptable values indicate that an acceptable internal consistency was observed in the full model of AIP-Q. The reliability assessment of every construct based on Cronbach's alpha, AVE and CR shows acceptable internal consistency tested among adult population. 507 GLOBAL JOURNAL OF PUBLIC HEALTH MEDICINE 2021, VOL 3, ISSUE 2 gggggglo The reasons of good internal consistency in most of the factors could be due to respondent had a good understanding of the item (Hair et al.,2010; Tabachnik & Fidell, 2001), hence less random error. Several strengths were recognized throughout this validation. Preliminary evidence shows its initial reliability and validity towards developing a new questionnaire among Malaysian concerning perceptions towards adult immunisations. Even though there are few existing questionnaires already developed based on the five-health belief domains, but this questionnaire is a new tool covering the whole spectrum of adult immunisation and design specifically for a Malaysian population. It is essential to find out their perceptions in general rather than specific vaccination since most of Malaysians had minimal exposure to adult immunisation. Therefore, this new questionnaire fills a gap of knowing their knowledge, attitude, and perceptions towards adult immunisation. There are few limitations in this validation study which is lack of test-retest reliability assessment of the instrument and it remains to be further tested in future studies. Besides, no concurrent validity evaluation to show its correlation to other measures, such as the health behaviour measure. They might be a connection of recorded vaccine uptake with the immunisation perception which need to be prioritized in future studies. Meanwhile, perceptions may change over the time and represent a snapshot of the study period. Hence, search for respondents who are very similar can be done, but for one specific domain or variable could be difficult to find. CONCLUSION The self-administered Adult Immunisation Perception-Questionnaire (AIP-Q) is a valid and reliable instrument for providing psychometric properties. It is a valid measurement for components of health belief, experience, knowledge, attitude, perceived severity, perceived susceptibility, perceived benefits, perceived barriers, and cues for action. This questionnaire should provide a useful tool to measure the perceptions of adult immunisation among Malaysian population with age group 18 and above. Acknowledgement The research was funded by the Dana Fundamental PPUKM (FF-2017-342) and ethically approved by the National Medical Research Centre of Malaysia (NMRR-18-1617-40854 (IIR)). A special thanks to those direct and indirectly involved in the research. List of Abbreviation Adult Immunisation Perceptions-Questionnaire (AIP-Q) Exploratory factor analysis (EFA) Confirmatory factor analysis (CFA) Principal Component Analysis (PCA) Goodness of Fit Indexes (GFI) Modification indices (MI) Comparative Fit Indexes (CFI) Inter Total Correlation (ITC) Composite reliability (CR) Inter Item Correlation (IIC) Root mean square error of approximation (RMSEA) Average variance extracted (AVE) Vaccine-Preventable Diseases (VPD) Structural Equation Modelling (SEM) Conflicts of Interest The author declares no conflicts of interest. 508 GLOBAL JOURNAL OF PUBLIC HEALTH MEDICINE 2021, VOL 3, ISSUE 2 gggggglo REFERENCES • Adult Immunisation Guideline (Malaysia) - 2nd Edition 2014 • Alexa M. Sevin, Cristina Romeo, Brittany Gagne, Nicole V. Brown, & Jennifer L. Rodis. (2016). Factors influencing adults’ immunisation practices: A pilot survey study of a diverse, urban community in central Ohio. BMC Public Health, 16:424. https://doi10.1186/s12889- 016-3107-9 • Ammar Ihsan Awadh, Mohamed Azmi Hassali, Omer Qutaiba Al-lela, Siti Halimah Bux, Ramadan M Elkalmi, & Hazrina Hadi. (2014). Immunisation knowledge and practice among Malaysian parents: A questionnaire development and pilot-testing. BMC Public Health, 14, 1107. • André, F.E. (2003). Vaccinology: Past achievements, present roadblocks, and future promises. Vaccine, 21(7-8), 593-595. • Antoinette, B. Coe, Sharon B.S., Gatewood, Leticia, R. Moczygemba, Jean-Venable “Kelly” R. Goode, & John, O. Beckner. (2012). The use of the health belief model to assess predictors of intent to receive the novel (2009) H1N1 influenza vaccine. Inov Pharm, 3(2), 1–11. • Bond, T. & Fox, C.M. (2015). Applying the Rasch Model: Fundamental measurement in the human sciences. Routledge, New York. • Brislin, R.W. (1976). Comparative research methodology: Cross- cultural studies. Int J Psychol, 11(3), 215–29. • Costello, A.B. (2009). Getting the most from your analysis. Pan, 12(2), 131–146. • De Vellis, R.F. (2003). Scale Development, 2nd edition. Sage, London. • Donadiki, E.M., Jime´nez-Garcı´a, R., Herna´ndez-Barrera, V., Sourtzi, P., Carrasco- Garrido, P., Lo´pez de Andre´s, Jimenez-Trujillo, & Velonakis, E.G. (2014). Health Belief Model applied to non-compliance with HPV vaccine among female university students. Public Health 128:268-273. • Hafizah, P., Zaleha, M.I., & Shamsul, A.S. (2002). The validation of questionnaire on risk perception of developing five most common non-communicable diseases in Malaysia. J Nurs Health Sci, 1(2), 29–35. • Hair, J.F., Black, W.C., Babin, B.J., & Anderson, R.E. (2010). Multivariate data analysis: A global perspective. Pearson Education Inc., New Jersey. • Halperin, B.A., MacDougall, D., MacKinnon-Cameron, D., McNeil, S.A., Langley, J. M. & Halperin, S.A. (2015). Universal tetanus, diphtheria, acellular pertussis (Tdap) vaccination of adults: What the Canadian public knows and wants to know. Human Vaccines & Immunotherapeutics, 11(9), 2167–2179. • Levine, O.S., Bloom, D.E., Cherian, T., de Quadros, C., Sow, S., Wecker, J., Duclos, P. & Greenwood, B. (2011). The future of immunisation policy, implementation, and financing. Lancet, 378, 439–448. • Naing, L. (2010). Reliability analysis. First UKMMC Intermediate to Advance Biostatistics in Medical Sciences Workshop. Anjuran Jabatan Kesihatan Masyarakat, Faculty Perubatan, Universiti Kebangsaan Malaysia, Cheras, Kuala Lumpur. • National state and local area vaccination coverage among children aged 19–35 months - United States. (2009). MMWR Morb Mortal Wkly Rep, 59(36), 1171–7. • Poland, G.A., Belmin, J., & Langley, J. (2010). A global prescription for adult immunisation: time is catching up with us. Vaccine, 28, 7137–9. • Rashwan, H., Lubis, S.H. & Ni, K.A. (2011). Knowledge of cervical cancer and acceptance of HPV vaccination among secondary school students in Sarawak, Malaysia. Asian Pacific Journal of Cancer Prevention, 12(7), 1837–1841. • Rosnah, I., Hassim, I.N., & Shafizah, A.S. (2013). A systematic translation and cultural adaptation process for Three-Factor Eating Questionnaire (TFEQ-R21). Med J Malaysia, 68(5), 425. • Schermelleh Engel, K., Moosbrugger, H., & Müller, H. (2003). Evaluating the fit of structural equation models: tests of significance and descriptive goodness-of-fit measures. Methods Psychol Res Online, 8(2), 23–74. • Singapore Burden of Disease Study Report 2010-Published in 1 June 2014. • Tabachnik, B.G., & Fidell, L.S. (2001). Using multivariate statistics. Allyn and Bacon, London. • U.S. Department of Health and Human Services. (2000). Healthy People 2010: 509 GLOBAL JOURNAL OF PUBLIC HEALTH MEDICINE 2021, VOL 3, ISSUE 2 gggggglo Understanding and Improving Health. (2nd ed.). Washington, DC: U.S. Government Printing Office. • Wheelock, A., Thomson, A., & Sevdalis, N. (2013). Social and psychological factors underlying adult vaccination behavior: lessons from seasonal influenza vaccination in the US and the UK. Expert Rev Vaccines, 12, 893–901. • Williams, B., Brown, T., & Onsman, A. (2012). Exploratory factor analysis: A five-step guide for novices. Australas J Paramed, 8(3), 1. • Zainudin Awang, Asyraf Afthanorhan, & Mustafa Mamat. (2015). The Likert scale analysis using parametric based Structural Equation Modeling (SEM). Computational Methods in Social Sciences, 4(1), 13-21. • Zainudin, A. (2012). A handbook on Structural Equation Modeling (SEM) using Amos. Bangi: MPWS Publication Sdn Bhd.