DEVELOPMENT AND PSYCHOMETRIC TESTING OF KNOWLEDGE, ATTITUDE, AND PRACTICE ON COVID-19 OUTBREAK QUESTIONNAIRE (KAPCovQ) FOR GENERAL COMMUNITY

Objective: To develop a valid and reliable questionnaire about knowledge, attitudes, and practices towards Covid-19 (KAPCovQ) in the general community based on psychometric properties. 
Methods: This study consisted of item development, scale development, and scale evaluation. Item development was based on literature review and content validity by experts. Scale development was conducted by pre-testing ten respondents. Scale evaluation was assessed using 375 respondents. Scale evaluation included construct validity with exploratory factor analysis (EFA) followed by confirmatory factor analysis (CFA) and reliability test with Cronbach’s α, composite reliability, and test-retest reliability. 
Results: The final KAPCovQ consisted of 3 domains with 31 items. Twelve items of the knowledge domain met the acceptable range for item analysis. Three factors of attitude domain and one factor of practice domain showed that 59.13% and 57.97% of the total variance respectively were identified in EFA. The result of the CFA for both attitude and practices domain indicated acceptable fit indices for the proposed model. The CFA model fit indices of attitude domain were χ2/df 2.05, p-value 0.01, GFI 0.92, RMSEA 0.07, TLI 0.90, CFI 0.92, and PNFI 0.64 and practices domain were χ2/df 1.18, p-value 0.28, GFI 0.98, RMSEA 0.03, TLI 0.98, CFI 0.99, and PNFI 0.54. Knowledge and all factors in the attitude and practice domain had an acceptable range in internal consistency reliability and test-retest reliability. 
Conclusion: The finding of this study demonstrated that KAPCovQ is valid and reliable for measuring the KAP on Covid-19 in the general community.


INTRODUCTION
Covid-19 is a potentially fatal disease and a major global public health problem. The development of the virus is very fast, positive cases and the number of deaths continues to increase [1]. Currently, there is no specific treatment for Covid-19, therefore, nonpharmaceutical interventions, especially prevention, should be maximized to slow the transmission of the virus spreading. Socialization and intensive health promotion efforts need to be carried out intensively and thoroughly that will have an impact on changes in cognitive, affective, and psychomotor communities in preventing Covid- 19 [2]. Knowledge of the mode of transmission, symptoms, and prevention of infectious diseases plays an important role in efforts to suppress the spread of the virus. Knowing the causes and sources of transmission of disease can increase people's awareness of the spread of infectious diseases and take preventive measures to slow transmission. Appropriate knowledge plays an important role in shaping beliefs and attitudes toward specific behaviors [3,4].
Increasing the number of Covid-19 cases must be an important concern of what the increasing cases cause and how the transmission of Covid-19 occurs, including the knowledge, attitudes, and practice of the community towards Covid-19. The public community plays an important role in minimizing the spreading of Covid-19. Public knowledge, attitude, and practice (KAP) assessments of Covid-19 are very important to evaluate and design more appropriate interventions to improve public KAP and to decrease the number of Covid-19 spreads. To conduct an assessment of public KAP, a valid and reliable questionnaire is needed. However, the availability of valid and reliable questionnaires to measure KAP for the general community towards Covid-19 is still limited. Previous studies have not explained the use of valid and reliable questionnaires or there was no information about the development, validity, and reliability of the questionnaire process [5][6][7][8][9]. It is necessary to develop a valid and reliable questionnaire that can measure the general community KAP towards Covid-19 (KAPCovQ) to ensure the accuracy of the data collected and ensure the validity of the measurement results of the KAP general community towards Covid-19. The objective of the study was to develop and validate a KAP general community towards Covid-19. This questionnaire could serve as an instrument to assess KAP towards Covid-19 in the public setting and as a baseline to tailor the appropriate intervention. This questionnaire could be used to evaluate the results of the intervention as well.

MATERIALS AND METHODS
This cross-sectional study was conducted in three districts in Central Java consist of Pemalang, Pati, and Temanggung district, Indonesia from August to October 2020. The inclusion criteria were a member of the general community who was equal or more than 17 y old, having ability in writing and reading, and willing to be respondents by signing the inform concern and completing the questionnaire. The exclusion criteria were having a health-related education background or working in health care facilities. The development and validation of the questionnaire took place in several steps. This study consisted of the item and scale development, and scale evaluation.

Item and scale development
This step consisted of item generation, content validity, and pilot study. Identification of domain and item generation were based on the deductive methods: extensive literature and previous studies about KAP towards Covid-19. Content validity examined by panel experts consists of a physician, a community pharmacist, and an academician by reviewing the pool items. Panel experts evaluated for representativeness, content relevancy, clarity, simplicity, and technical quality of each item. The assessment of experts was quantified using a formula with a content validity ratio (CVR) [10,  . Content validity ratio range from-1.0 (total disagreement) to 1.0 (total agreement). The number of panel experts in this study was three, the items pertain if the CVR ≥0.99 [10,12]. Before the survey is administered, a pilot study was conducted on ten target populations to ensure that all items are meaningful and can be understood by the target population.

Scale evaluation
This step included item analysis, internal consistency with Cronbach's α, and test-retest reliability for the knowledge domain. Construct validity with exploratory factor analysis (EFA) and confirmatory factor analysis (CFA), internal consistency with Cronbach's α, composite reliability, and test-retest reliability were run for the attitude and practice domain.

Item analysis
Item analysis is one of the bases used to maintain or remove the items based on the item difficulty index (IDI) and discrimination index (DI). The item difficulty index determines the proportion of correct answers given per item by respondents. IDI is calculated by comparing the number of respondents who answered correctly with the total number of respondents. IDI score<20% or>90% was recommended to be dropped because of too easy or too difficult [11,12]. The higher the DI score, the better the item can determine the difference performance between two groups of respondents (higher and lower performance). Items that have a DI score<0.2 should be dropped or rewritten [12,13].

Construct validity
The construct validity test measures the extent to which a research instrument can measure the construct (the framework of a concept) which is analyzed using EFA and CFA [14]. The minimum sample size for EFA and CFA is 100 and 200, respectively, or a 1: 5-10 ratio per item [12,[15][16][17][18]. In this study, 175 respondents were analyzed for EFA and 200 respondents for CFA. EFA was conducted for the attitude and practice section using principal component analysis and Varimax rotation. Knowledge items were not meant for factor analysis because of known facts and were not abstract ideas [19]. The appropriateness for conducting factor analysis was based on the Kaiser-Meyer-Olkin (KMO) with range 0.5-1.0 and Bartlett's test of sphericity score should be<0.05. Retaining the number of the factors was considered by the eigenvalue>1, loading factor>0.400, scree plot, and reliability of factor [18]. The assessment of the goodness of fit the proposed model was based on the result of χ2/df ≤2 or ≤5, p -value>0.05, goodness of fit index (GFI)>0.90, root mean square error of approximation (RMSEA)<0.08, Tucker-Lewis index (TLI)>0.90, comparative fit index (CFI)>0.90, and parsimonious normal fit index (PNFI)>0.60 [20].

Reliability test
Reliability is a measure of the consistency of a research instrument if it is used with different methods, conditions (place and time) [21].
The reliability test used in this study was internal consistency with Cronbach's and composite reliability. The intraclass correlation coefficient (ICC) was used to analyze test-retest reliability. The testretest reliability was evaluated by correlating between the first and second questionnaire scores from the same respondents in a twoweek interval [22]. The minimum sample for test-retest is more than one over ten of the total respondents. Forty-two respondents participated in fulfilling the questionnaire twice. To reduce bias, forty-four respondents were not notified when they fill out the questionnaire twice. Respondents were taken randomly.

Ethical approval
Health Research Ethics Committee of Faculty Medicine of Universitas Muhammadiyah Surakarta approved the protocol of study with reference number 3011/B.1/KEPK-FKUMS/VIII/2020. All respondents were asked to obtain and sign the informed concern form before they conducted the study.

Data collection and statistical analysis
The sampling method used was non-probability sampling with an incidental sampling technique. Data were analyzed using Statistical

RESULTS
The initial draft consisted of demographic information (7 closedended and multiple-choice questions) and three domains namely knowledge (30 items), attitude (18 items), and practice domain (7 items). The response option of knowledge was yes/no/don't know with a score of 1 for the correct answer. Five Likert scales were used for attitude from strongly disagree to strongly agree with the score from 1 to 5 and practice domain from never/seldom/sometimes/often/always with the score 1 to 5. Several items were reverse scoring.
Three items of knowledge and two items of attitude were removed because they were not valid and CVR<0.99 based on the result of content validity [10,12]. In the pilot study, ten community members completed the questionnaire within 10-15 min and provided feedback. Some items were modified and reworded to make the respondent easier to understand without changing the meaning. Ambiguity or fuzzy expression of items was revised and changed, for example quarantine was revised to be "isolation" because "quarantine" is not familiar for some respondents.
Three hundred and seventy-five respondents participated in this study for analysis of item analysis, construct validity, and reliability tests. The response rate of this study was 98.68%. Male respondents were more than women and a quarter of the respondents were 40-49 y old, almost 35% of their background education was in elementary school, and 70% worked. The demographic characteristic of respondents is displayed in table 1.    test-retest reliability) to produce a valid and reliable questionnaire based on the psychometric properties [11,12,20]. The use of questionnaires in previous studies has no information or inadequate information on the validation and reliability test process [8,9]. Validation in Abdelhafiz, 2020, and Tamornpark, 2020 study consisted of content validity by three experts and pretesting on 10 to 20 participants for assessing the internal consistency with Cronbach's alfa and test-retest reliability [5,6]. Rahman's study explained the Cronbach's alfa was 0.64 and no information on the process of validation and test of the questionnaire [7].
All of the items met content validity with CVR>0.99 after removing three items of knowledge and two items of attitude, suggesting that all items could well reflect, be relevant, and representative to assess the KAP in the public community [11]. Based on the item analysis result, 12 items have an IDI score of 20% to 90%. As many as 23.08%, 30.77%, and 46.15% of all items were excellent (DI ≥0.4), good (DI: 0.30-0.39), and acceptable (DI: 0.2-0.29) DI score, respectively. It means that those items are qualified; those items are not too difficult or not too easy and can measure the difference performance between two groups (lower and upper performance) [11][12][13]. The construct validity test used exploratory factor analysis (EFA). In this study, EFA was feasible because of KMO>0.5 and Sphericity<0.05. EFA was used to determine the optimal number of factors and domains that fit a set of items [11]. The eigenvalue>1, loading factor>0.40, and scree plot are considered for retaining the number of the factors and items [18]. A three-factor solution has resulted from EFA with a cumulative contribution of 59.13% for the attitude domain and one factor with a cumulative contribution of 57.97% for the practices domain. In general, the acceptability of factor and meaningfulness of the relationship between the items and factors based on the cumulative contribution of more than 40%, and each item on the corresponding factor owns enough loading (>0.4) [23].
To validate or confirm whether the previous hypothetical structure from the EFA result fits the items and to estimate the independent cluster model, the researcher conducted CFA [11]. The assessment of the goodness of fit the proposed model based on the result of χ2/df ≤2 or ≤5, p -value>0.05, GFI>0.90, RMSEA<0.08, TLI>0.90, CFI>0.90, and PNFI>0.60 [20]. The result of the three-model of attitude domain and one-model of practices domain after respecification based on the modification indices suggestion was an acceptable fit model. Meaning that all of the factors and items of attitude and practices domain met the goodness of fit indices.
Cronbach's α reflects the internal consistency of the questionnaire. Cronbach α of 0.65-0.70; 0.70-0.80; and 0.80-0.90 considered as minimum acceptable value, good, and best respectively [24,25]. In this study, Cronbach's α of knowledge was 0.775, the attitude domain ranged, 0.659 to 0.852, and the practice domain was 0.741, revealing that KAPCovQ had an acceptable and best internal consistency reliability. The composite reliability of the attitude domain ranged from 0.627 to 0.802 and the composite reliability of the practice domain was 0.698 meaning that all factors had acceptable or good composite reliability. Composite reliability 0.60-0.70 is the minimum acceptable value and>0.70 is good composite reliability [20]. The test correlation coefficient of this study ranged from 0.793 to 0.909, 0.880, and 0.883 for the attitude, practice, and knowledge domain respectively, revealing that the test-retest reliability of the questionnaire was up to the requirements of psychological measurement and showing good stability for testing in the different time. Intraclass correlation coefficient 0.60-0.74 categorized good and>0.75 categorized excellent reliability [26].
The strength of this study was the process of developing and validating the questionnaire has been conducted comprehensively. On the other hand, because Indonesia consists of many islands with different cultures, this result may not be generalized to all populations in Indonesia. Modifying the items adapted to the local culture are suggested if this questionnaire is used in a different place or country.

CONCLUSION
The final KAPCovQ consisted of 3 domains with 31 items. The result of validity and reliability analysis demonstrated that the KAPCovQ had strong psychometric properties. The KAPCovQ supplements the limitation of the currently available KAP towards the Covid-19 questionnaire and this questionnaire can be used to identify the KAP general community especially in the Central Java population towards Covid-19. The data may be used as a baseline of the general community KAP and assess the result of intervention on improving the KAP in the general community especially in preventing and controlling the increasing spreading of the Covid-19 infection.