Development and Initial Validation of a Scale for Assessing Affecting Factors on Preventive Behaviors of COVID-19 (AFPB-CO): Using the Protection Motivation Theory

Maryam Khazaee-Pool (  khazaie_m@yahoo.com ) Mazandaran University of Medical Sciences https://orcid.org/0000-0002-2587-3460 Seyed Abolhassan Naghibi (  anaghibi1345@yahoo.com ) Mazandaran University of Medical Sciences Tahereh Pashaei Kurdistan University of Medical Sciences Leila Jahangiry Tabriz University of Medical Sciences Mahbobeh Daneshnia Mazandaran University of Medical Sciences Koen Ponnet Universiteit Gent

recently become Pandemic, so far, the researchers have not found any study and scale based on recent PMT in published documents. Thus, the current study was aimed to development and psychometric assessment of a multi-component questionnaire based on PMT better to understand factors affecting on preventive behaviors about COVID-19. Such scales could support to identify the viewpoints of health experts and policymakers and, in turn, help in developing extended interventional plans for participating of citizens about the rate of the preventive behaviors toward COVID-19.

Study design
The present work was applied in two separate stages. At the start, in the rst stage, item generation and development of the scale were performed, based on the interview with panelists and respondents, as well as a review of the literature in this background relating to the protection motivation theory (PMT) for making an item pool and content domain. The principal dependent variable in the current analysis was the affecting factors on preventive behaviors toward COVID-19. Besides, the independent variables include ve factors, organized into logical framework counting (a) Perceived severity, (b) Perceived vulnerability, (c) Perceived response cost, (d) Perceived response e ciency, and (e) Perceived selfe cacy. Then, the content and face validity assessment was done. In the next stage, the psychometric properties of the established scale were measured through a cross-sectional study on a sample of people. To complete this nal stage, we done the exploratory factor analysis (EFA), con rmatory factor analysis (CFA), and analyzed item-scale correlation. At the end of the following part, the reliability of the AFPB-CO scale was measured, by internal consistency (through Cronbach's' alpha coe cient) and stability of the AFPB-CO scale, via test-retest.

Phase 1: Item generation and scale development
Research design This stage of the study intended to develop a scale to measure the affecting factors of Iranian citizens in the preventive behaviors about COVID-19. We derived items employing the following three procedures: an exploration of the suitable number content domain, items' production, and the instrument design (11).
The content domain discusses content related to the variables being assessed (12). The content domain generation emerged from a review of the related literature, as well as on interviews with the target group and professionals. The data collected from the interviews with the target group were applied to generate the instrument items (13,14). The literature review reveals another method used to identify the content domain (13). In the other phase, item generation was carried out by the comment collected from the content domain. The following step of the study began the process of instrument structure, in which items were modi ed and suitably prepared. All accepted items were so arranged in an effective procedure (15).
The item pool included 76-item at this point. It was tried to make the content of the items more comfortable and removed extra items through discussion. The chief researcher (MK) and other team members read items and deleted additional items. To the end, the rst draft of the instrument was developed and contained with 51-item. Then, content and face validity were assessed. Validity Content validity.
Qualitative and quantitative approaches were applied to assess content validity. In the qualitative phase, a panel of experts, including health promotion expert and epidemiologist, measured the content validity of the AFPB-CO scale. They evaluated phrasing, grammar, wording, item allocation, and scaling of the AFPB-CO scale. In the quantitative phase, the content validity index (CVI) and the content validity ratio (CVR) of the AFPB-CO scale were evaluated. The CVI was assessed by requesting the experts to rate each item according to its simplicity, relevancy, and clarity (15) on a scale ranging from 1 = not relevant, simple, and clear to 4 = very relevant, simple, and clear. The CVI was measured as the proportion of items on the questionnaire that achieved a rating of three or four (16,17). The CVR evaluated the essentiality of each item in the questionnaire. For measuring the CVR, the specialists rated each item as 1 = essential, 2 = useful but not essential, or 3 = not essential. Afterward, according to the Lawshe table, items with a CVR score of .99 or greater were determined to be acceptable and were maintained (15). In the quantitative measurement of the AFPB-CO scale, items with a CVR and a CVI more than .99 and .80, respectively, were preserved. 21-items were removed, resulting in a 39-item pool. The expert panel revised the AFPB-CO scale about wording, item allocation, and grammar as well.
Face validity: Both qualitative and quantitative approaches were implemented for measuring the face validity. A group of laypeople (n=10) was questioned to assess each item of the AFPB-CO scale and to indicate whether they experienced di culty or confusion in answering the item. Afterward, to determine the percentage of laypeople who recognized items as important or quite important on a ve-point Likert instrument, the impact score (importance × frequency) was measured. Items were determined to be suitable if they had an impact score of 1.5 or higher (which equals a mean rate of 50% and mean the importance of 3 out on 5) (18).
To sum up, three items had an impact score lower than 1.5. The range of impact score was from 1.6 to 5.
The rst form of the AFPB-CO scale, including 36 items, was con rmed for the next stage of psychometric assessment. In other words, the group of laypeople showed that they experienced no problems reading and understanding the 36-item of the AFPB-CO scale.
A pre-nal version of the AFPB-CO scale: following the consideration of the above approaches, the prenal version of the AFPB-CO scale, including 36-item, was produced for the subsequent stages (construct validity and reliability of the AFPB-CO scale). Porsline is an online questionnaire creation software. It takes only a few drags and drops to create the questionnaire and form.
After receiving the answer, we will receive instant statistical analysis and graphs.
In the professional use of Porsline, we can direct them to relevant questions at any time, depending on the answer the respondents choose. Additionally, we can prevent one person from recording more than one response on a device, such as a mobile or laptop. The respondent will then be prompted with the message, "You have already answered this questionnaire; it is not possible to re-register the answer." We can also allow for a response but identify responses recorded more than once on a device as spam in the results table.
In the next step, the designed questionnaire link (https://survey.porsline.ir/s/56BXXtB) was shared in cyberspace and social networks (WhatsApp, Telegram, Instagram, LinkedIn, Facebook, and Twitter). One of the items considered at the beginning of the questionnaire was the province and city of residence. All the people who lived in one of the cities of Mazandaran province were included in this study. The sample size was assessed based on Gable and Wolf's reference, which suggests that a sample of ve to 10 people per item is needed to support a hypothetically xed factor structure (20). Thus, the required sample size, i.e., 360 (36 items × ten persons) laypeople for each part of the analysis (EFA and CFA), were engaged using these approaches (overall 720 persons for construct validity).
ii) Statistical analysis Several statistical methods were performed to measure the psychometric properties of the AFPB-CO scale. These were organized in the following instruction.
1. Construct validity: The construct validity was applied using exploratory factor analysis (EFA), con rmatory factor analysis (CFA), and item-scale correlation for the 36-items that remained.
Exploratory factor analysis (EFA) An EFA was accomplished to nd the main factors of the AFPB-CO scale. A sample of 360 lay people completed the online questionnaire of AFPB-CO. A principal component analysis (PCA) with varimax rotation was applied to extract the main factors. Bartlett's Test of Sphericity and Kaiser-Meyer-Olkin (KMO) were used to measure the suitability of the sample for the factor analysis. Additionally, the eigenvalues and scree plot were measured to specify how many dimensions to keep. One principle applicable for describing the number of dimensions to retain is the Kaiser's criterion, which is based on a rule of thumb. According to Kaiser's criterion, the criteria performed to extract the main dimensions had an eigenvalue of 1 or above, which categorized them as satisfactorily enough. Eigenvalues above one and scree plot were accompanied for nding the number of dimensions. Factor loadings ≥ .40 were considered acceptable (21). There are no quick guidelines for de ning the cut-off point, and the particular methods implemented were in uenced mainly by our purposes for the current study. Most texts recommend a cut-off point of 0.4. However, there are no de nite approaches suggested to make such a purpose, and much depends on the scale being used. By contrast, Comrey and Lee (2013) recommended the performance of more stringent cut-off points, counting 0.32 (poor), 0.4-0.54 (fair), 0.55-0.62 (good), 0.63-0.70 (very good), or 0.71 and more (excellent) (22). In the current study, scores of.40 and more were considered as satisfactory for factor loadings.

Con rmatory factor analysis (CFA)
A CFA was performed to measure the coherence between the model and the data. A separate sample of 360 lay people completed the online questionnaire of AFPB-CO. The relative chi-square, comparative t index (CFI), the goodness of t index (GFI), root mean square error of approximation (RMSEA), standardized root mean square residual (SRMR), normed t index (NFI), and non-normed t index (NNFI) were measured to assess the model t (20,23). The CFI, GFI, NNFI, and NFI values range between 0 and 1, but values equal to or greater than .80 are generally considered satisfactory model ts. An RMSEA value between .08 and .10 indicates a mediocre t, and a value lower than .08 indicates a good t. Values smaller than .05 indicate a good t for SRMR, but values between .05 and .08 and between .08 and .10 represent a close t or a satisfactory t, respectively (24).

Item-scale correlation
As a nal point, item-scale correlations of AFPB-CO were evaluated to measure the degree to which each item was correlated to its dimension via the Spearman correlation coe cient. We estimated that, for each dimension of the AFPB-CO, the item scores of the dimension (e.g., perceived threat) would correlate more with the total score of the own dimension (e.g., perceived threat) instead of the total score of another dimension (e.g., motivational protection). Correlation values Zero -.20 are re ected poor; .21 -.40, fair; .41-.60, good;.61 -.80, very good; and beyond .81, excellent (25).

Reliability
The internal consistency was evaluated, applying Cronbach's alpha coe cient for measuring the reliability of the AFPB-CO scale. The alpha values equal to or greater than .70 were considered acceptable (26). Additionally, the intraclass correlation coe cient (ICC) was assessed for evaluating the stability of the AFPB-CO scale. The AFPB-CO scale was read ministered by 25 citizens from Mazandaran province a week after the rst completion of the AFPB-CO scale by the shared link of the AFPB-CO scale (https://survey.porsline.ir/s/56BXXtB) in cyberspace and social networks (WhatsApp, Telegram, and Instagram). ICC values equal to or higher than .40 are considered satisfactory (r's from .81 to 1.0 are excellent, from 0.61 to .80 are very good, from .41 to .60 are good, from .21 to .40 are fair, and from .0 to .20 are poor) (26). The analyses were performed using the statistical program SPSS for Windows version 24.0 and Amos 24.0.

Scoring of AFPB-CO scale
The nal version of the AFPB-CO scale is revealed in Appendix 1. The items were rated on a ve-point Likert-type scale, anchored at the extremes with one =completely agree to ve=completely disagree for seven dimensions including perceived threat, responses costs, responses e cacy, rewards, e cacy, and motivation protection and with 1 = never to 5 = always (for the dimension of behavioral responses).

Results
The study sample In this study, 720 lay people aged 18 years or older completed the AFPB-CO scale (360 participants for EFA and 360 participants for CFA reported that they obtained information from national media, and 19.44% of participants (n = 140) began using information from family and relatives. To buy essential items, 63.1% (454 participants); go to work, 17.8% (128 participants); go to picnic and tourisms, 13.6% (98 participants); and other cases, 5.5% (40 participants) were the most common reason for going out of the house during the days of the Corona outbreak in Mazandaran Province by participants.
a. Exploratory factor analysis The measured KMO was .750, and Bartlett's test of sphericity was signi cant (χ2 = 5276.468, p < .001), presenting the appropriateness of the sample for EFA. An eight-factor solution, with a 36-item scale, arisen, based on eigenvalues more than 1 and a loading level equal to 0.5 or more. Factor loadings of each item and the eight dimensions are offered in Table 2. All items were loaded on their dimensions. The eight dimensions of the AFPB-CO scale jointly accounted for 58.64 % of the observed variance. Additionally, the scree plot directed an eight-factor solution (Fig. 1). In total, an EFA was applied on the 36item of the AFPB-CO scale (cut-off point: .50).

b. Con rmatory factor analysis
We conducted a CFA on the 36-item AFPB-CO scale to assess the tness of the model obtained from the EFA. Based on Figure 2, the best model t was obtained. Covariance matrixes were performed, and t indices were measured. All t indices were found to be appropriate. The relative chi-square (χ2/df) was equal to 2.692 (p < .001). The RMSEA of the model was .069 (90% CI = .06 -.07), and the SRMR was .06. All comparative indices in the structural model, containing GFI, AGFI, CFI, and NFI, were more than .90 (.89, .91, .83, and .80, respectively). Even though the model t was acceptable and good, modi cation indices for the regression weights were tested to nd covariance among the eight factors. There was no achievement of any signi cant enhancement on t indices; thus, no variations were performed, and the model was accepted in its existing form. Figure 2 demonstrates the model. Table 2 shows the item-scale correlation for the AFPB-CO scale. As can be observed, all coe cients are higher than .30, and most of them are higher than 0.50. Perceived threat and e cacy had the highest and the lowest item-scale correlation, respectively.

Reliability
The Cronbach's alpha was tested separately for the AFPB-CO scale and each dimension of the AFPB-CO scale in order to assess the internal consistency of the AFPB-CO scale. The Cronbach's alpha coe cient for the AFPB-CO scale ranged from .62 to .84 for its dimensions, which is more than the acceptable threshold. Additionally, the test-retest analysis was applied to assess the stability of the AFPB-CO scale.
The results showed a satisfactory threshold. Intra-class correlation (ICC) ranged from .73 to .87 for the dimensions of the AFPB-CO scale, giving support for the scale's stability. The results are available in Table 3.

Discussion
In the current study, we de ned the development and psychometric properties of a new scale, called the AFPB-CO, for measuring factors that affect persons' participation in prevention behaviors toward COVID-19. This is the rst project to extend a measure for assessing the factors related to participate in preventive behaviors of COVID-19 in the world, especially in the Iranian population. The content of the AFPB-CO scale items was rst developed based on the interviews with the experts/laypeople and the review of the literature to ensure that this instrument covered all theoretical concepts linked to the protection motivation theory (PMT). One of the approaches to scale design, especially in the area of preventive behaviors, is the use of motivational theories like the protection motivation theory (PMT). Overall, the PMT makes available a valued theoretical structure for dealing with the di culties of preventive behaviors. PMT can be performed to describe behavior associated with threat and coping related to health dangers and behavioral intention. Concepts such as perceived severity, perceived vulnerability, perceived response cost, perceived response e ciency, and perceived self-e cacy are the key constructs of the PMT. The assessment of theoretical constructions is one of the most di cult and necessary components in the study of theory-based health education and promotion (10). By knowing cognitive-behavioral principles, elements in uencing the encouragement to do determined behaviors are recognized and impacted in the intervention researches. In the current project, the PMT components were measured by the complex technique. These components are one of the most common reasons to do the prevention of disease people might consider themselves at risk of possible comorbidities. The PMT may be applied to advise effective communications to reform existing postoperative recommendations (10). Thus, according to the reports mentioned above, PMT can be a suitable theory to design interventional programs.
Overall, the ndings indicated satisfactory psychometric properties for the AFPB-CO scale. The CVI and the CVR indicated that the content validity of the AFPB-CO scale was sensible. The results of the analysis due to the KMO index show adequate sample size and satisfactory factor analysis. Furthermore, the results of the EFA and CFA showed a good structure for the AFPB-CO scale. After EFA, an eight-domain scale including perceived threat, behavioral responses, perceived response costs, fears, perceived responses e cacy, rewards, perceived self-e cacy, and protection motivation extracted accounted for 58.64 of the variance and the maximum stated variations were linked to the perceived threat as a rst domain. Threat references to the extent to which persons perceive they are vulnerable to the health danger and their opinion of the sternness of the health danger (10). This shows that the AFPB-CO scale is useful for introducing various aspects of the health concerns in uenced by participation in preventive behaviors.
The CFA was also used to specify whether there was coherence between the information and the theoretical structure. The CFA revealed good t indices for the existing model, and the convergent validity of the eight subscales of the AFPB-CO scale was acceptable. As such, the nal AFPB-CO scale contained 36 items, with seven items indicating outcome evaluations, six items representing a perceived threat, ve items representing behavioral responses, six items representing responses costs, four items expressing fears, four items describing responses e cacy, three items representing rewards, four items representing e cacy, and three items representing protection motivation of preventive behaviors toward COVID-19.
Additionally, the internal consistency of the AFPB-CO scale, as evaluated by the Cronbach's alpha, displayed suitable reliability for two dimensions. Moreover, after testing 30 laypeople over a week period, our results clearly showed that the AFPB-CO scale has appropriate stability in the short term; however, it has yet to be perceived whether the AFPB-CO scale is stable for a long time.
Overall, the results con rmed that this newly developed AFPB-CO scale is a suitable and valid scale that can be used for measuring participating in preventive behavior toward COVID-19 among lay people who speak Farsi. Developing theory-based scales can be used as a signi cant necessity for the calculation of any educational program. Therefore, we consider the results from the present study to be valuable for clients who are part of a COVID-19 preventive plan. On the other hand, it may be principally useful to health care experts to identify and design processes that are bene cial and targeted to particular conditions. The inclusion of eight domains in the AFPB-CO scale further lets specialists identify and decide how components in which a person can be improved.

Limitations
Although the current study has several strengths, it has some limitations. First, the present study was done among a sample of people from Mazandaran province (north of Iran) to express affecting factors on their participation in performing preventive behaviors about COVID-19. Due to this, we cannot be de nite that the conclusions perform to laypeople in another geographic background. Consequently, further researches may be needed to support the applicability of the theory-based scale of AFPB-CO as a fully con rmed applied and useful measure in the Iranian context. Second, the Cronbach's alpha coe cients of some factors were not satisfactory. Future studies are needed to overcome these problems. The third one is related to the truthfulness of responses and the self-reported nature of the scale. The generalizability and sample size made up another limitation of the present study. The current sample was limited to a group of 720 (both EFA and CFA) laypeople, and it is uncertain whether we would attain the same results if a large sample of both people were employed. Finally, the present study included only persons who were referred to the online link of scale and who completed the questionnaire online and virtual. There may be many populations who do not have a smartphone or Internet access to complete a questionnaire. Furthermore, it can be interesting for future studies with a larger sample to test whether the psychometric properties of the instrument hold with an alternative measure of reliability and validity (e.g., test-retest validity).
In summary, one of the objectives of the Century is increasing participation of people in preventive behaviors toward COVID-19. To do so, we developed the AFPB-CO scale, which was revealed to have acceptable psychometric properties.

Conclusion
Generally, the AFPB-CO scale indicated good construct validity, and the majority of domains showed high internal consistency reliability; thus, the results of the current study offer that a theory-based AFPB-CO scale is a valid and reliable scale for assessing affecting factors on participation of people in preventive behaviors about COVID-19. Furthermore, the Iranian version of the AFPB-CO scale may be helpful for health care workers in nding and planning health strategies that are useful and targeted to particular statuses. Additionally, more studies are proposed to identify the strengths and weaknesses of the AFPB-CO scale when it is performed in other contexts.

Consent for publication
Not applicable.

Availability of data and materials
The datasets produced and analyzed throughout the current study are not publicly available in order to keep the participants' privacy but are available from the corresponding author on reasonable request.

Competing interests
The authors declare that they have no competing interests.

Funding
No external funding sources were provided for this manuscript.   Q8. I get anxious if I come in contact with different surfaces (e.g., elevator door, home door -shopping center -gas station -workplace).

Tables
Q9. I am scared of even going out of the house for essential work.
Q10. If I have Corona, I have to spend long days in the hospital with anxiety.
Perceived Responses costs Q11. Continuous washing of hands at least 20 seconds outside the home is time-consuming and di cult.
Q12. If I do not shake hands or I have a distance with them when communicating with others, they may be angry with me.
Q13. When I wear gloves, my hands sweat.
Q14. It is hard to breathe with a mask.
Q15. Washing regularly my hands with soap makes my skin more sensitive.
Q16. If I stay at home and be a quarantine, I get depressed and bored.

Rewards
Q17. In these days of Corona outbreaks, traveling or leaving home makes me happy.
Q18. When I shake hands with others or kiss them, I feel more intimate.
Q19. Eating cold food and bread is more enjoyable than cooked food and hot bread.
Perceived Responses e cacy Q20. Washing my hands with soap and water for at least 20 seconds is an excellent way to prevent me from Corona disease. Protection motivation Q24. I am going to disinfect all the nuts and the stuff I buy outside the house.
Q25. I want to keep myself and others healthy by being 2 meters away.
Q26. As long as there is a risk of Corona transfer, I plan to keep in touch with family and friends online.
Q27. I'm going to warm the food and bread before eating well.
Perceived Self-E cacy Q28. I can easily wash my hands before contacting my face and after contact with surfaces outdoors.
Q29. If I have had a fever, shortness of breath, or cough, I can go to the hospital immediately or talk to a doctor.
Q30. Without any problems, I can use clean masks and gloves when I get out of the house and into crowded places.
Q31. I can stay home for as long as I have a chance of getting a corona, even if I'm upst and frustrated.
Behavioral Responses Q32. When exposed to different surfaces, I wash or disinfect my hands with soap and water for at least 20 seconds.
Q33. To cut the Corona transmission chain, I put the dirty gloves and masks after use in the trash bin.
Q34. I regularly disinfect different levels of contact (home -car -work -pedestrian card, mobile phone, etc.).
Q35. When touching contaminated surfaces, I take care not to touch my eyes, mouth, and nose with my hands.
Q36. I change clothes outside and put them outside before I get inside the home. Figure 1 Scree plot for determining factors of the designed instrument.