Cross-cultural adaptation and validation of Spanish version of The Foot and Ankle Ability Measures (FAAM-Sp)

The cross-cultural adaptation process was undertaken using the guidelines derived from the methodology recommended by the International Society for Pharmacoeconomics and Outcomes Research (ISPOR) for the translation and validation of patients reported outcome measures [14]. Cross-cultural adaptation involved eight stages: (1) forward translation, (2) reconciliation; (3) back translation, (4) back translation review, (5) harmonisation, (6) pilot testing/cognitive debriefing, (7) pilot testing review/review of cognitive debriefing results and (8) proof reading. Fig. 1 summarises the cross-cultural adaptation process.

Two forward translations in Spanish were undertaken from the original English version of the FAAM. The translations were undertaken by two independent health professionals who were native residents of Spain and fluent in both Spanish and English. Both translators based their translations on the guidelines provided by ISIS outcomes for the translation and linguistic validation process [15].

The two forward translations were reconciled into one version (draft 1) by two original translators, a third independent translator and feed-back from the principal researcher in case of doubts [GGN].

The reconciled Spanish language version (draft 1) was back translated into English independently by two professional English native translators resident in Spain. The translators had no prior knowledge of the FAAM and were not given the original wording of the English version of the FAAM.

The project lead [GGN] and a native Spanish speaker resident in UK and fluent in both languages reviewed the back translation for any discrepancies in meaning or terminology used. Any problematic item was discussed until the discrepancies were resolved. This process resulted in a refined second draft of the Spanish translation (draft 2).

To produce the final Spanish language translation, a harmonisation meeting was undertaken involving three Spanish translators, the project lead [GGN] and the developer of the original USA version of the FAAM [RRM]. During this meeting, any discrepancies or issues that were highlighted from the back translation were discussed, the translated version of the FAAM was evaluated and a final version agreed.

Once the translation process was completed, the translation was formatted to match precisely the original American language version. The translated FAAM version was initially assessed for comprehensibility in five patient participants, who were Spanish residents and native speakers, met the inclusion criteria described above and had a low educational background without being illiterate. At this stage, each participant was asked by the in-country investigator [PCG] to carry out the following tasks:

To improve the performance of the translated questionnaire, the pilot testing results were reviewed by the in-country investigators [PCG]. At this stage, any item that caused comprehensibility difficulties for more of the 40% of the participants was reviewed, and any modifications suggested by the respondent’s comments were incorporate to the final translated version.

The project lead [GGN] and a translator [ABOA], who was not involved in the translation process previously, independently proofread the final formatted translation, and any suggested changes were discussed with the project lead [GGN]. Furthermore, the Flesh Reading Ease test and the Flesh Kincaid Grade Level were calculated for readability [16].

Following this process, a final draft of the FAAM translated and culturally adapted into ‘Spanish for Spain’ was finalized and entered into the cross-cultural validation phase (final draft).

Before completing the questionnaire, the following data for each participant were recorded: age, sex, professional status, education level, affected extremity and diagnosis. Podiatrists working in the seven different private clinics of podiatry collected the questionnaires. They addressed any possible concern of the participants. Items were numbered from 1 to 29. Items 1 to 21 were from Activities of Daily Living (ADL) subscale and items 22 to 29 were from the Sports subscale. Each question is rated on a scale of 0 to 4, with 0 being “unable to do” and 4 being “no difficulty”. None answer responses are not counted. Thus, each subscale has a maximum potential score (84 ADL and 32 Sport subscales). The score for each subscale is divided by the maximum potential score of the subscale and multiplied by 100 to get a percentage, with a higher score meaning a higher level of function on each of the two subscales.

To test a two-factor model, assuming the null hypothesis of a mean square error of approximation (RMSEA) from 0.04 to 0.085, with an alpha value of 0.05, a statistical power of 0.8 and a maximum of 26 degrees of freedom, as suggested by MacCallum et al. [17], a sample of 196 participants was required which was over-estimated by 10% to cover possible losses. The calculations were performed with the Statistica 12 software [18].

Descriptive statistics were carried out with means, standard deviations, and absolute and relative frequencies. Analysis of normality of distributions was evaluated by the Kolmogorov-Smirnov test, symmetry analysis and kurtosis. Internal consistency was calculated using Cronbach’s alpha. A Cronbach’s alpha between 0.70 and 0.95 was considered “good” [19]. Moreover, item-total and inter-item correlations were assessed. Pearson correlations were calculated to assess the convergent validity between FAAM and EuroQol-5D Spanish version [20]. Content validity was assessed by the distribution of the scores and occurrence of floor-ceiling effect. The floor-ceiling effect was estimated through the endorsement frequency, with a maximum accepted value of 85%. To assess factor structure, a confirmatory factor analysis was performed, the evaluated model was fit with the following parameters: the penalizing function (chi square/ df), which is indicative of good fit with values less than 3; Root Mean Square Error of Approximation (RMSEA) and confidence intervals (CI 90%), taking the value 0.05 as cut-off of good fit; Normed Fit Index (NFI), the Comparative Fit Index (CFI), and Goodness of Fit Index (GFI) with a minimum value of good fit of 0.90. Multinormality was evaluated with the Mardia’s coefficient (multivariate curtosis), which could not be over “p” (p + 2), where “p” are the number of observed variables [21].

Reliability was performed using the Intraclass Correlation Coefficient Type 2.1 (ICC 2.1) [22] test-retest methodology in the full sample recorded at baseline and 1 week (7 days) following a period of no treatment.

All the analyses were performed with SPSS 23 [23] and AMOS 21 [24].

The FAAM was translated into Spanish and culturally adapted to provide the new FAAM-Sp (Fig. 1). In the pilot testing phase, results showed no discrepancies in meaning or terminology used in the translated version of the FAAM. Hence, no modification of this version was performed. Participants did not request assistance in interpretation of the questionnaire or any of its items. The time needed to fill out the questionnaire was 3.83 min (SD 0.99). The result for Flesh Reading Ease test was 56.4 and 7.8 for the Flesh Kincaid Grade Level.

Items 5, 9, 12, 15, 16, 17 and 21 showed backup frequencies between 85.4 and 89.6%.

A total number of 204 participants were included in this study. However, 10 participants were excluded from data analysis because they did not fully complete the FAAM-Sp. Hence, 194 participants (130 women and 64 men), with a mean age of 38.45 (SD: 16.04) were assessed. The characteristics of the sample are presented in Table 1.

For exploratory factor analysis, the correlation matrix for the FAAM-Sp ADL was determined as suitable from the Kaiser–Meyer–Olkin values (0.879) and Barlett’s test of sphericity (p < 0.001). A three factor solution was obtained, after a principal axis factor extraction and oblique rotation, with item loading (Table 2). A confirmatory factor analysis (CFA) was then carried out to test this three-factor structure. The 21 items solution did not obtain a good fit (x ²/df 4.28, p < 0.001, GFI 0.76, NFI 0.85, CFI 0.88, RMSEA 0.13 IC 90%: 0.12 to 0.14).

An additional model of 15 items was tested, after deleting items which have had a bad adjustment in the model, evaluated through their standardized residual covariances (items 2, 7, 13, 14, 18, and 20). This model obtained the best fit: relative chi-square (x ²/df) of 2.46, GFI 0.90, CFI 0.95, NFI 0.93, and RMSEA 0.08 (90% CI: 0.04 to 0.09), and fulfilled the multinormality criteria (Fig. 2).

This instrument obtained a Cronbach’s alpha of 0.95, with an inter-item mean correlation of 0.73 (Range: 0.57 to 0.88). The mean scoring was 11.59 (SD: 5.98). Item-total correlation was above 0.70 in all the items.

For exploratory factor analysis, the correlation matrix for the FAAMSp Sport was determined as suitable from the Kaiser–Meyer–Olkin values (0.905) and Barlett’s test of sphericity (p < 0.001). A one factor solution was obtained, which explained 76.70% of total variance. CFA corroborated this model with an excellent goodness of fit: relative chi-square (x ²/df) of 0.80, GFI 0.99, CFI 1.00, NFI 0.99, and RMSEA 0.00 (90% CI: 0.00 to 0.75), and fulfilled the multinormality criteria (Fig. 3).

No floor-ceiling effect was observed. Only items 27, 28 and 29 obtained high endorsement frequencies in the lowest value, but under 85%.

This version of the 15 items FAAM-Sp ADL showed a good degree of internal consistency for each factor (Cronbach’s alpha 0.90, 0.92 and 0.90 respectively, and 0.95 overall).

The inter-item correlation matrix is detailed in Table 3. Mean Inter-item correlation was 0.60 (range 0.44 to 0.80), 0.77 (0.72 to 0.88), and 0.71 (0.66 to 0.74), for factor 1, 2, and 3, respectively. Item-total correlation was 0.65 (Table 4). Mean values and standard deviations of each factor were: 10.04 (4.60) for factor 1, 4.81 (2.03) for factor 2 and 4.85 (2.06) for factor 3.

Floor effect was observed in items 5, 9, 12, 15, 16, 17, and 21, with endorsement frequencies between 85.4 and 89.6%.

Statistical analyses revealed correlation coefficients of − 0.596 between the ADL subscale and the EuroQol5D (p < 0.001) and − 0.472 between the Sports subscale and the EuroQol5D (p < 0.001).