BIFS is composed of 20 items divided into two factors according to the original proposal; the first is ‘fear of injections’ (Items 1–12) and the second “fear of blood” (Items 13–20).2 Responses use a 5-point Likert scale (1=Strongly agree; 2=Agree; 3=Neither agree nor disagree; 4=Disagree; 5=Strongly disagree).2

Three bilingual translators, whose native language is Brazilian Portuguese, performed the translation of the BIFS, independently. The two researchers responsible for the study, who are also bilingual, compared the three translated versions. Thus, the Portuguese version of BIFS was defined.

The Portuguese version of the instrument was sent to a translator fluent in Brazilian Portuguese whose native language is English to perform the back translation. This translator was not aware of the content of the original scale and was not informed that it was a back-translation. The back translation was important to confirm that the translation of the instrument did not change the meaning of the items according to the original proposal.

After these steps, the Portuguese version of the BIFS (BIFS-P) was obtained and had its psychometric properties evaluated in this study.

To evaluate the objectivity and relevance of the BIFS-P, the content of its items was assessed independently by a panel of eight judges, health professionals, with experience in hematology transfusion medicine and/or in professional services in primary health care. Judges were informed about the objectives of the study and were asked to classify each item of the instrument as adequate or not adequate for use in the target population. Suggestions of modifications and simplification of items were requested in these cases. The content validity index (CVI) was calculated for each item considering that IVC≤0.70 is indicative of the need of a new translation (i.e., 30% or more of the judges classified the item as not adequate).11 According to the suggestions of the judges, some items were modified aiming to simplify and improve the understanding prior to the application of the instrument in the target population.

In a second phase of the content analysis, a panel of 20 judges with experience in hematology and transfusion medicine was invited to classify, also independently, each item of the instrument according to its essentiality (essential, useful but not essential or not necessary) to evaluate ‘fear of blood or injections’. The content validity ratio was calculated according to the proposal of Lawshe.12 The significance of the decision was in accordance with the proposal of Wilson et al.,13 adopting a level of 5%.

Before starting data collection, the three interviewers responsible for this step participated in a training session on how to apply the instrument and all the strategies regarding the approach, explanation of the research objectives and instructions to patients were discussed.

The pre-test of the BIFS-P in the target population was conducted in 30 primary health care users of Ribeirão Preto in a single healthcare clinic located in the central region, which caters for different medical specialties. The incomprehension index (proportion) of each item was calculated and, if necessary, the item was rewritten.

The sample was composed by adult individuals (≥18 years), residents in Ribeirão Preto, southeastern Brazil, who are primary health care users in the municipality. Ribeirão Preto is a medium sized city of São Paulo State that has approximately 670,000 inhabitants. The municipality is divided into five health districts, North, West, Central, South and East. At the time of data collection, there were 41 primary government healthcare clinics in the municipality, which were classified according to the district in which they are located and to the Paulista Index of Social Vulnerability (PSVI) prevalent in their area. The PSVI is developed by the SEADE (State System of Data Analysis) from census data, with the geographic areas being classified into six distinct groups of social vulnerability.14 Healthcare clinics were grouped into 12 strata according to the district where they are located and the predominant IPVS. The clinics where the PSVI was equal to 1 or 2 were grouped in the same stratum and the clinics with PSVI equal to or greater than 4 were grouped in another stratum.

It should be noted that this study is part of a wider project that aims to evaluate the association of some variables with blood donation. Thus, the sample size was obtained considering a stratified sampling plan,15 a 95% confidence level and an absolute precision of 3% for the estimative of the proportion of blood donors. This proportion, for maximization of variance, was considered equal to 50% in each stratum. Thus, a sample size of 1054 interviews was estimated, with the number of interviews to be conducted in each stratum being proportional to their estimated population size. The healthcare clinics where the interviews were conducted were selected randomly within each stratum. In addition to the BIFS-P, a questionnaire about sociodemographic characteristics and general health of the participants was used to characterize the sample.

Factorial validity of the BIFS-P was assessed using exploratory factor analysis (EFA) and confirmatory factor analysis (CFA). EFA was performed using polychoric correlation methods implemented using the SAS software. The CFA was also performed using the polychoric correlations matrix implemented in the software Mplus, version 6.0. The estimation method used in CFA was the weighted least squares mean and variance adjusted (WLSMV). The ratio of Chi-square to its degrees of freedom (χ2/df), comparative fit index (CFI), Tucker–Lewis index (TLI) and root mean square error of approximation (RMSEA) were considered as goodness of fit indices. The fit of the models was considered adequate when χ2/df was ≤2.0, CFI and TLI was ≥0.90 and RMSEA was <0.10.16 Items with factor weights (λ) lower than 0.40 were considered underweight items for the corresponding factor. Modification indices, obtained from the Lagrange multipliers, were used to verify the existence of correlations between errors.16 Comparisons between the tested models were performed considering their respective fit in general, through the goodness of fit indices, with emphasis on the RMSEA index.17 RMSEA estimates show how well the parameters of the model reproduce the population covariance and can be considered a parsimonious correction index. Thus, lower values of RMSEA are indicative of better models.17,18

Factorial invariance of the models was assessed using multigroup analysis considering the Chi-square difference (Δχ2). This analysis is used to evaluate the stability of the models in different samples. For this, the total sample was randomly divided into two equal parts the ‘test sample’ and the ‘validation sample’. In this analysis the equivalence of factorial weights [metric invariance (λ)], the factorial weights and intercepts [scalar invariance (Int)], and the factorial weights, intercepts and variances/covariances [strict invariance (Cov)] were evaluated.16,19

This study was approved by the Ethics Committee on Human Research of the Hospital das Clínicas in Ribeirão Preto (CAAE: 38148814.2.0000.5440), with data collection being approved by the Department of Health of Ribeirão Preto. In addition, this study was conducted in accordance with the Helsinki Declaration as revised in 2008. Only adult individuals (≥18 years) who agreed and signed informed consent forms participated. The authors of the original scale authorized the translation and validation of the instrument.

The first step of the content validity process of the BIFS-P was performed by a panel of four male and four female judges with a mean age of 48.8 (SD 13.3) years who have been healthcare professionals for 20.6 years on average (SD 15.1). Of these, 37.5% work in hematology/transfusion medicine services and 62.5% are engaged in primary healthcare activities. According to the CVI values (CVI≥0.75 for all items), there was no need for new translations of any item of the BIFS-P. However, some important suggestions were made and considered to simplify and adapt the instrument for the target population.

The second step of the content validity process was performed by 20 judges, who did not participate in the first step of this evaluation; 80% were female, the mean age was 36.1 (SD 9.6) years and the judges have worked as healthcare professionals for 10.6 (SD 8.2) years on average. Of these, 10% are engaged in activities of attending the population in primary healthcare services, 10% are nursing staff in transfusion medicine or blood donation services, and 80% are engaged in other types of activities in services of hematology/transfusion medicine or blood donation. The content validity ratio (CVR) was estimated and is presented in Table 1.

Only Items 1, 4, 8, 11, 15, and 18 were considered essentials by the judges.

In the pre-test of the scale in the target population, 30 patients were interviewed (female: 73.3%; mean age: 36.97±15.1 years) in a single healthcare clinic located in the central region of the municipality which attends different specialties. The incomprehension index was null for all items, indicating that there was no need for rewrites.

In the data collection step which aims to evaluate the psychometric properties of the BIFS-P, 1055 users of primary healthcare services in the municipality were interviewed in 12 healthcare clinics (female: 79.7%; mean age: 40.6±15.2 years). Regarding the perception of health, 69.7% classified their health as ‘good’, 26.5% as ‘regular’, and 3.8% as ‘bad’. Arterial hypertension and diabetes mellitus were the most prevalent chronic diseases in the sample at 22.0% and 8.0%, respectively.

Figure 1 presents the polychoric correlations matrix for the 20 items of the BIFS-P. Higher correlation coefficients (>0.90) were detected between Items 11 and 12, 13 and 14, and 16 and 17. Items 2 and 19 were not strongly correlated with any other item of the BIFS-P, with correlation coefficients between 0.39 and 0.62.

Figure 1.

Polychoric correlations matrix for the items of the Portuguese version of the Blood/Injection Fear Scale. Darker colors were attributed to higher correlation coefficients.

(0,32MB).

Table 2 shows the results of the EFA. According to the EFA, the items can be grouped in up to five factors, with Factors 4 and 5 being responsible for providing a more specific feature of the BIFS-P. These factors are related to fear and/or reactions on seeing their own blood (Factor 4) and to fear and/or reactions to seeing the blood of others (Factor 5).

Thus, three factorial structures of the BIFS-P were tested in the CFA: (1) original factorial structure with two first-order factors (fear of blood/fear of injections); (2) factorial structure with three main first-order factors resulting from the EFA; (3) factorial structure with 5 first-order factors resulting from EFA. These structures are called Model 1, Model 2 and Model 3, respectively.

Table 3 presents the results of the CFA for the factorial structures tested for the BIFS-P. It should be mentioned that it was necessary to exclude Item 19 due to its low factorial weight in all analyses and lower correlation coefficients with all the other items of the scale in order to have adequate fits of all models (Figure 1).

Despite the appropriate factor weights of all items in each factor in the three models tested (λ>0.60), the original structure with two factors (Model 1) does not present an adequate fit to the data considering the goodness of fit indices. Even after the exclusion of Item 19 and the model re-specification (seven correlations between errors considered), Model 1 presented results of RMSEA higher than adequate and high modification indices, which can be considered a poor adjustment (Table 3). On the other hand, Models 2 and 3 presented adequate fit to the data, with goodness of fit indices within acceptable limits (after re-specification of the models and the exclusion of Item 19). Despite high levels of χ2/df, this index is overestimated in larger samples, and should be evaluated together with the other indices to make decisions on the fit of the models. In Models 2 and 3, observing the RMSEA values and their confidence intervals, it is clear that no model has a better fit as both models are properly adjusted to the data and both structures can be considered in the assessment of fear of blood and injections in this sample.

Distribution of the items in each factor considered in the confirmatory factor analysis of Models 2 and 3 and the factorial weights (λ) are presented in Figure 2. The names of the factors were adopted in accordance with the theoretical content of the items when grouped.

Figure 2.

Distribution of items of the Portuguese version of Blood/Injection Fear Scale in each factor and the local fit (λ) of Models 2 and 3 of the data obtained in the confirmatory factor analysis. FIG: fear of injection in general; FFI: fainting due to fear of injection; FBG: fear of blood in general; FRI: fear of receiving injection; FORI: fear of seeing others receiving injection; FFI: fainting due to fear of injection; FBO: fear of seeing blood of others; FOB: fear of own blood.

(0,4MB).

All items presented adequate factorial weights for the corresponding factors (λ≥0.66), which confirms an adequate distribution of items in the exploratory factor analysis.

Model 2 (structure with three factors) presented equivalence of factorial weights (metric invariance – λ: χ2=18.7; p-value=0.29), of factorial weights and intercepts (scalar invariance – Int: χ2=85.2; p-value=0.10), and of factorial weights, intercepts and variances/covariances (strict invariance – Cov: χ2=67.7; p-value=0.09) considered very stable in different samples. Model 3 (structure with five factors) presented equivalence of factorial weights in different samples (metric invariance – λ: χ2=20.6; p-value=0.11) with poor stability.