Psychometric properties of the Spanish version of the Fear of COVID-19 scale in Paraguayan population.

OBJECTIVE: The aim of this study is to test the psychometric properties of the Spanish validation of the Fear of COVID-19 Scale (FCV-19S) in a Paraguayan population.
METHODS: Participants were recruited through an Internet-based survey. All participants whose scores in the Hospital Anxiety and Depression Scale (HADS) and The Fear Questionnaire (FQ) were greater than zero were included. 1245 subjects responded voluntarily: 1077 subjects, scoring >0, were considered.
RESULTS: To establish construct validity of the FCV-19S, an exploratory factor analysis was performed using the KMO test, which was adequate, and the Bartlett sphericity test, which was significant (p <.0001). The CFI, NFI, GFI, TLI and RMSEA indices were used to evaluate the model and showed good adjustment. Cronbach's α showed valid internal consistency (α = 0.86). This validation was supported by significant correlation (p <.001) with the HADS scale for anxiety and depression and with the FQ scale for specific phobia.
CONCLUSIONS: The Spanish version of the FCV-19S is a 7-item scale with two dimensions, psychological symptoms and physiological symptoms, which demonstrated robust psychometric properties in a Paraguayan population.

Introduction

Since the end of 2019, the world has faced a new threat to public health consisting of a respiratory disease caused by a new coronavirus (Corman et al. 2020), named as COVID-19 by the World Health Organization (Wu et al. 2020).

Moreover, due to its high transmission rates and the interconnectedness of the modern world, the disease has rapidly spread to become a pandemic that is taking a huge toll on public health systems. Mental health is one of the main areas that this pandemic is undoubtedly affecting (Torales et al. 2020). Different governments have taken unprecedented measures in order to safeguard the health of their citizens, including a decrease in social contact and isolation, sometimes even involuntary, of large sectors of the population (Usher et al. 2020). It has been observed that these measures have led to an increase in fear (Ornell et al. 2020), anxiety (Peteet, 2020), depressive symptoms (Stein, 2020), hopelessness (Shaw, 2020) and adjustment disorders. These circumstances have already been described in countries where the pandemic has overwhelmed the health system (Kang et al. 2020; Mantica et al. 2020; Mowbray, 2020; Sindhu & Gupta, 2020), as well as in those areas where disease outbreaks were contained (Ho et al. 2020). The related increase of mental disorders (Mamun & Griffiths, 2020; Troyer et al. 2020) has also added to the pressure on health systems in terms of resources and the need to rapidly design strategies to contain the outbreak of COVID-19.

For these reasons, it is important to have a scale to measure how fear of COVID-19 can affect individuals. The Fear of Covid-19 Scale (FCV-19S) aims to determine these variables (Ahorsu et al. 2020). The objective of this study was to translate the FCV-19S to Spanish and to validate it in the Paraguayan population.

Methods

Participants

Participants were recruited using an Internet survey advertised and published on social networks and on the official social media accounts of the School of Medical Sciences of the National University of Asunción. All participants whose scores in the HADS (Hospital Anxiety and Depression Scale) and The Fear Questionnaire (Marks & Mathews, 1979) were greater than zero were included. The survey was open from March 19 to March 21, 2020 and 1245 subjects responded voluntarily: 1077 subjects, scoring >0, were considered. During this period, the COVID-19 morbidity rate was 0.9 patients per 100 000 inhabitants, while the mortality rate was 0.04 per 100 000 inhabitants.

The Internet-based survey approach was used taking into account that there is supporting evidence that responses to online surveys may provide similar findings to those reported through “in person” samples (Gosling et al. 2004).

Measures

The following data was collected as part of the survey:

Socio-demographics Parameters: information about socio-demographic factors of participants (e.g., age, gender, residence area) was collected.

Hospital Anxiety and Depression Scale: The HADS - Spanish version (Herrero et al. 2003) is a 14-item scale comprising seven items related to anxiety and seven items related to depression. Items are answered on a 4-point response format with a total score ranging from 0 to 21 for each of the two subscales. Cronbach´s α for the scale were 0.835 for the total scale, 0.722 for the anxiety, and 0.721 for depression subscales.

The Fear Questionnaire: Four scores are obtained from the Fear Questionnaire (Marks & Mathews, 1979): level of avoidance caused by specific target phobia identified in writing (question 1, score range 0–8). A total phobia score indicating the extent of avoidance for 15 common phobias (questions 2–16, score range 0–120). This score is made up of 3 separate phobia subscores, each including five items and having a score range of 0–40 (agoraphobia items 5, 6, 8, 12, 15; blood-injury phobia items 2, 4, 10, 13, 16; and social phobia items 3, 7, 9, 11, 14). A rating of associated anxiety and depression obtained from five common non-phobic symptoms found in phobic individuals (questions 18–22, score range 0–40). A global phobia rating reflecting distress and avoidance (final scale on the questionnaire, score range 0–8).

Fear of COVID-19 Scale: The FCV-19S (Ahorsu et al. 2020) is a seven-item scale assessing the fear of COVID-19. The seven items are rated on a 5-point scale from 1 (strongly disagree) to 5 (strongly agree) with scores ranging from 7 to 35.

Translation process and validation

These steps were followed to guarantee the validity of the translation: The English version of the questionnaire (v1) was translated into Spanish by the researchers (v2). Then, the Spanish version (v2) was translated back into English by a bilingual expert (v3). Finally, one of the authors of the original English version compared the back-translated English version (v3) with their version (v1), to determine if they were equivalent in meaning. Subsequently, changes were made according to suggestions of the original authors and the Spanish version was employed in a pilot test with a sample of 15 people, after which, the final Spanish version was obtained (see Annex 1).

Statistical analysis

To establish validity of the construct, the pertinence of conducting an exploratory factor analysis (EFA) was initially analyzed using the Kaiser–Meyer–Olkin (KMO) sample adequacy tests and the Bartlett’s sphericity test. The sample was randomly divided into two subsamples, in which EFA was performed with the SPSS version 23 statistical software and confirmatory factor analysis (CFA) with the JASP 0.11.1 statistical program, respectively. The extraction method was principal axis factoring and varimax rotation with Kaiser normalization.

Chi square test was performed, and the following fit indices were used to assess the model: CFI, NFI, GFI, TLI, and RMSEA. For the validity of model, CFI, NFI, GFI, and TLI values had to be at least 0.90 and RMSEA value below 0.05 (Herrero, 2010). Reliability was measured with Cronbach’s α (a value of ≥0.70 was considered acceptable). For construct validity, correlation tests were used between the scores of the different scales.

Results

1077 subjects were included, 68.71% were women (n = 740), 30.73% men (n = 331) and 0.56% preferred not to mention their sex (n = 6). The mean age was 30.95 ± 10.07 years. 40.76% of the subjects were from the Central department (the area near the capital city), 30.73% from Asunción (the capital city) and 28.51% from the other parts of the country. The mean score for the FCV-19S was 15.84 ± 5.53. Table 1 shows statistics performed on the Spanish version of the FCV-19S.

Table 1.

Statistics on the items of the Spanish version of the FCV-19S

	Corrected item – total correlation	Mean	s.d.	Skewness	Kurtosis
Item 1Item 2Item 3Item 4Item 5Item 6Item 7	0.640.660.510.670.670.610.65	2.82.61.72.52.61.81.9	1.141.220.731.241.220.880.99	0.200.251.260.520.301.151.13	−0.85−1.092.64−0.84−1.021.350.77

Kaiser–Meyer–Olkin (KMO) test was adequate (KMO = 0.85) and sphericity tested significantly (p < 0.0001). These results confirmed that the sample was adequate for a factorial analysis. Secondly, the sample was randomly divided into two subsamples (subsample 1, n = 517; subsample 2, n = 560), in order to perform EFA and CFA.

Factorial analysis

Only the first two factors had raw eigenvalues greater than the parallel random values, thus two factors were retained. Before extraction, these factors explained 71.86% of the total variance. After extraction, the two first factors explained 61.09% of the total variance. All the items have factorial weights associated to only one of the factors (Table 2).

Table 2.

Results from the exploratory factor analysis with varimax rotation on the Spanish version of the FCV-19S

Items	Factor 1	Factor 2	Communality
Item 1	0.847		0.558
Item 2	0.696		0.487
Item 3		0.691	0.415
Item 4	0.681		0.533
Item 5	0.639		0.474
Item 6		0.797	0.541
Item 7		0.710	0.535

Note: Pattern matrix shows values greater than 0.4.

The items 1, 2, 4, and 5 corresponded to factor 1 (emotional fear reactions or “psychological symptoms) and the items 3, 6, and 7 corresponded to factor 2 (symptomatic expressions of fear or “physiological symptoms”). Factor 1 explained 32.94% as well as factor 2 explained 28.15% of variance. Eigenvalue of the first factor was 3.918 and 1.112 for the second.

Since the items were distributed in a non-normal way (see skewness and kurtosis on Table 1), parallel analysis/diagonal weighted least squares method (DWLS) was used. The two-factors model, as found in the EFA performed on subsample 1 (n = 517), was assessed with CFA in subsample 2 (n = 560). The model adjustment was valid, according to all fit indices (S-B χ 2=19.872, df = 13, p = 0.098; RMSEA = 0.031; CFI = 0.996, NFI = 0.989, GFI = 0.996, TLI = 0.994).

Validity, internal consistency, and correlation between factors

Cronbach’s α showed a valid internal consistency (α = 0.86), and could not be improved by removing any item, as found for both factors: psychological symptoms (α = 0.84) and physiological symptoms (α = 0.819). Correlation between FCV-19S total score and the scores of each factor were significant (factor 1: r = 0.948, p < 0.001; and factor 2: r = 0.721, p < 0.001). Mean scores and standard deviations were 10.46 (s.d . = 3.96) and 5.38 (s.d . = 2.24) for factor 1 and 2, respectively.

Concurrent validity was supported by the HADS for anxiety and depression and the FQ for specific phobia as indicated by the significant positive correlation with both scales (p < .001). FCV-19S positively correlated with the HADS (r = 0.330) and FQ (r = 0.262).

Discussion

This research reports on the psychometric properties of the Spanish version of the FCV-19S, testing a sample of 1077 people. This sample included more participants than those involved in the English version of the scale (Ahorsu et al. 2020), as well as versions in Hebrew (Bitan et al. 2020), Italian (Soraci et al. 2020), Arabic (Alyami et al. 2020), Russian (Reznik et al. 2020) and Indian (Doshi et al. 2020). However, our sample was smaller than the validation-sample for the Turkish version (Satici et al. 2020), the Greek version (Tsipropoulou et al. 2020) and the Bangla version (Sakib et al. 2020).

The descriptive statistics of the scale show good levels of corrected item - total correlation ranging from 0.51 to 0.67, the sample was adequate to perform a factor analysis when the assumptions were fulfilled, so it was randomly divided into two sub-samples.

In the first subsample (n = 517) factorial loads and eigenvalues were studied to discover the structure of the scale, two well-differentiated factors were found, with large factorial loads in only one of the factors and with eigenvalues greater than 1 and with good level of explained variance, since this model explains 61.09% of the total variance. The items that were grouped in the first factor are those containing words such as: afraid, uncomfortable, nervous and anxiety, so this factor was named “psychological symptoms”. The items grouped in the second factor describe words like: clammy, sleep and heart racing, so this factor was named “physiological symptoms”. At this point it should be remembered that psychometry studies psychological constructs based on statistical models: this two-factor structure seems to recognize the constructs that are intended to be measured.

This two-factor model was confirmed using subsample 2 (n = 560), according to which all the adjustment indices have been correct, indicating that the scale has two well-defined factors. This has also been found in the validation of the scale in Hebrew (Bitan et al. 2020). It is important to mention that the original authors of the scale stated that more studies are necessary to confirm the structure of the scale (Ahorsu et al. 2020).

Similarly, for the Italian version (Soraci et al. 2020), the authors stated the importance of carrying out more research with a larger sample size to confirm the structure of the scale, since they reported that they found covariance between items 1 and 5, which in our sample remain at factor 1. This difference found does not contradict what has already been published, but, in our study, it was possible to carry out an EFA first and then a CFA, both with quite optimal sample sizes compared to the publications mentioned above.

Although the Turkish version also reports a single factor and has a good sample size, its authors directly carried out a CFA, like others researchers did (Sakib et al. 2020; Satici et al. 2020; Tsipropoulou et al. 2020). This could explain why the Turkish version authors did not find two factors, since they directly performed CFA on the original one-dimensional model (Ahorsu et al. 2020).

The authors of the Hebrew version have found the same structure as this version of the scale; however, they have been criticized for apparently conducting a principal component analysis (Bitan et al. 2020; Pakpour et al. 2020). Our validation was carried out with an exploratory analysis, specifically factorization of main axes. Although it is true that a CFA is more robust than an Exploratory Factor Analysis EFA, the best way to empirically determine the structure is the EFA, since for the CFA it is necessary to assume an already established model. Likewise, it should be noted that the original authors used the item response theory and not an EFA (Ahorsu et al. 2020).

There are three published versions in Spanish. The first version was validated in the general population of Peru (Huarcaya-Victoria et al. 2020), the second one was tested in university students from Spain (FCV-19S mean score = 16.79) (Martínez-Lorca et al. 2020), and the last one among Colombian physicians (Mercado-Lara et al. 2021). The Peruvian validation reported the same two factors that were found in our study, whereas in the Spanish and Colombian validations a one-dimensional model was found. The difference found between our FCV-19S mean score (15.84) and the mean score obtained by Martínez-Lorca et al. (2020) (16.79) (Martínez-Lorca et al. 2020) might be explained by the epidemiological situation of each country at the time of sampling, so it would be recommended to make a global comparison between the data to look for those factors explaining this variability.

Regarding the internal consistency of the full scale and the factors, adequate Cronbach’s α values were obtained, the first being similar to the other adaptations and to the original scale (Ahorsu et al. 2020; Satici et al. 2020; Soraci et al. 2020). Construct validity was performed by comparing the scores of the FCV-19S with the scores obtained in the HADS and the FQ, where both scales correlated positively, but weakly. The explanation for this might be that the scores in specific phobia, anxiety and depression were low in our sample, taking into account that at the time of data collection, the country was managing COVID-19 sufficiently well. This was also noted in the Russian validation of the scale (Reznik et al. 2020).

Limitations of this study included that the sample was obtained in a non-probabilistic way and from the general population, which means that it was not selected on the basis of a specific diagnosis, and scale sensitivity and specificity were not studied. Also, the two-factor model should be studied in other samples with adequate sample sizes to properly determine the scale structure. Another limitation may include that we do not report information that allows us to know whether those who completed the survey were university students or members of the general population. Furthermore, we do not have information of how many people accessed the social network and social media accounts in order to get a sense of what proportion of individuals who saw the survey may have completed it. Finally, the presence of a potential selection bias must be taken into account, which could have influenced the results obtained.

In conclusion, the Spanish version of the FCV-19S, validated in Paraguayan population, is a 7-item scale with two dimensions: psychological symptoms (items 1, 2, 4, and 5) and physiological symptoms (items 3, 6, and 7), and which demonstrated robust psychometric properties in the current study.