Thai version of the foot function index: a cross-cultural adaptation with reliability and validity evaluation

A cross-sectional design was used in this study. Data were collected from May to December 2017.

The FFI questionnaire was translated to Thai version, following a cross-cultural adaptation recommended guideline [32]. The translation steps consisted of: 1) forward translation from the English version to a Thai version by two bilingual Thai and English translators independently; 2) reconciliation of the two forward versions of the FFI by an expert committee using synthesis into the FFI-Th to reduce discrepancies; 3) backward translation of the FFI-Th to English by two independent United States citizens who understood both Thai and English for checking the translated texts; 4) harmonization by developing the pre-final version of the FFI-Th through consensus of a group meeting of health professionals; 5) cognitive interview, by administering the pre-final version of the FFI-Th to 20 general individuals to check the understanding of the question items and 6) proofreading of the final version of the FFI-Th by a linguist. The final version of the FFI-Th was then launched in trials. A summary of the translation procedure is presented in Fig. 1.

Similar to the original version of the FFI, the FFI-Th contains 23 items, including 3 subscales, which are pain (items 1st to 9th), disability (items 10th to 18th) and activity limitation (items 19th to 23rd). Each item is rated on a 10-cm horizontal line visual analogue scale (VAS). The left and right ends of the line are labelled as “no pain” and “worst pain imaginable” for the pain item, “no difficulty” and “so difficult - unable” for the disability item, and “none of the time” and “all of the time” for the activity limitation item, respectively. Participants marked their levels of pain and discomfort at any point on the line that corresponded to their foot symptoms. The participant’s response was measured by ruler on a continuous scale of 10 cm of the VAS line. Therefore, the maximum score for the 23 FFI-Th items was 230, with 90, 90 and 50 being the maximum scores for the pain, disability and activity limitation subscales respectively. A higher score represents a higher severity of pain, disability, and activity limitation [12]. Participants were asked to rate the score in all items of the FFI-Th. When they did not experience the situations of items, they were advised to check the “NA” (not applicable) box instead. The scoring system for the total score and sum of three subscales of FFI was converted to 0–100 scores, by using the formula: sum of score from all items answered by participants, divided by the total score possible, and multiplied by 100 [12].

The pain-VAS, a uni-dimensional measure of pain intensity is used in diverse populations and settings due to its simplicity and adaptability. The pain-VAS is a continuous scale, usually 10 cm anchoring with two different verbal descriptions at each end of the line [35, 36]. Instructions and verbal descriptions may differ, depending on the intended use. As a self-assessment report respondents required no training. Respondents can rate their symptoms by marking on the line [35, 37]. A higher score indicates greater pain intensity. Test-retest reliability of the pain-VAS has been proved to be high [38] and construct validity with a numeric rating scale was good to excellent [39].

Sample size was calculated from the study of Jorgensen et al. in 2015 [24] who tested reliability of the FFI-Denmark version (FFI-DK). Excellent test-retest reliability of the FFI-DK was reported for each subscale [pain (ICC = 0.98 and 95% CI = 0.97–0.99), activity limitation (ICC = 0.95 and 95% CI = 0.91–0.98), and disability (ICC = 0.97 and 95% CI = 0.95–0.98)] and the total score of FFI-DK (ICC = 0.95 and 95% CI = 0.91–0.98). Sample size was calculated from the formula shown below.

Where m_repeat was the number of samples, z²_1-α/2 was set at 1.96. The ξ was a chance error and Wξ was the desired width, which was set at 0.10 and 0.15 respectively. Based on this calculation, the sample size was 48. Thus, the number of participants recruited in the study satisfies this requirement.

Cronbach’s alpha (CA) was used to test the internal consistency of the FFI-Th. The CA values ranged from 0 (no internal consistency) to the full score of 1 (perfect internal consistency). Test-retest reliability of the FFI-Th was evaluated between the first and the second administrations with 1 week apart by the ICC_{3, 1} and 95% confidence interval (95% CI). The ICC values ranged from 0 (no agreement) to the full score of 1 (perfect agreement). The ICC < 0.5 was considered as poor, 0.5 to 075 was moderate, 0.75 to 0.9 was good, and > 0.9 was excellent reliability [40]. Construct validity of the FFI-Th was tested using Principal Component Analysis with the varimax rotation method and considering the eigenvalue was greater than 1. Criterion validity of the pain-VAS and the EQ-5D-5L were investigated using Pearson’s correlation coefficient (r), with the possible range of coefficient was 0 to 1. Interpretations of these coefficients were r < 0.3 (weak correlation), 0.3 to 0.5 (moderate correlation), and > 0.5 (strong correlation) [41]. The positive and negative values indicated the positive and negative directions of correlation.

Before conducting this study, participants were informed about the details of the study and signed informed consent forms approved by the institutional ethics committee (MU-CIRB COA no. 2016/173.3012). Inclusion criteria of the participants were: 1) aged between 20 and 80 years; 2) had pain or tenderness at the plantar surface of the foot during rest and/or during prolonged weight-bearing activities of at least 1 month; 3) able to read and communicate in the Thai language and 4) had no visual problems that could not be corrected by lens or glasses. They were excluded if they had: 1) pain in any other areas of the lower extremity; 2) history of systematic inflammatory disease or neurological disease and 3) received any kind of treatment during participation in the study. There were 51 individuals with plantar foot complaints who passed the selection criteria. Of these, 2 participants did not rate the score for the EQ-5D-5L questionnaire. Therefore, 49 participants were included in this study. Demographic data and clinical symptoms of foot pain are presented in Table 1.

The FFI-Th score was calculated from the score marked on a 10 cm horizontal line for each item. At the end of each question, there was the “NA” box using for the ones to rate if the question was not consistent with their previous experience leading to inability to rate the score. Missing values were excluded and the data were computed using the scoring system following recommendation [12] and method used in previous studies [18, 20, 22]. Following these guidelines, the totals and subscale sum scores were all reported in the range of 0 to 100. Table 2 presents the mean and SD of the FFI-Th scores from the completed analysis and the numbers and percentages of respondents who reported floor score, ceiling score, and NA answer. With the scoring system, an averaged total score of FFI-Th of the participants was 27.60 ± 20.30. Averaged subscales of pain, disability and activity limitation were 33.42 ± 21.60, 31.17 ± 23.57, and 14.34 ± 15.19, respectively. One to three participants reported floor scores for the items in the pain and disability subscales, whereas 7 to 16 participants reported floor scores for the items in the activity limitation subscale. For ceiling score, only one participant reported it in the 1st and 2nd items. For N/A score, there were 26 participants could not rate the score for the 7th and 8th items and 19 for the 22nd and 23rd items.

As presented in Table 3, good to excellent internal consistencies were found for the total score (CA = 0.974), subscales of pain (CA = 0.946), disability (CA = 0.975), and activity limitation (CA = 0.714). Between two trials of testing, good to excellent test-retest reliabilities were found for the total score (ICC_{3, 1} = 0.942, p < 0.001), subscales of pain (ICC_{3, 1} = 0.904, p < 0.001), disability (ICC_{3, 1} = 0.938, p < 0.001), and activity limitation (ICC_{3, 1} = 0.866, p < 0.001).

The correlation matrix showed that the extraction was appropriate. This was observed using the Kaiser-Meyer-Olkin and Bartlett’s sphericity tests (the Kaiser-Meyer-Olkin Measure = 0.856, p < 0.001 and the Chi-square value = 402.978). The computation was performed after missing values were replaced by the mean values. Figure 2 shows the scree plot of all items of the FFI-Th. Four-factor numbers showed eigenvalues greater than 1.

Table 4 shows the total variance explained by factors which were extracted by the Principal Component Analysis using the varimax rotation method. For the extraction sums of squared loadings column, the first factor accounted for 60.393% of the variance, when the second, third and fourth merely accounted for 9.320, 6.623, and 5.220%, respectively. For the rotation sums of squared loadings column, the first factor accounted for 50.527% of the variance, when the second, third, and fourth merely accounted for 11.346, 9.867, and 9.816%, respectively.

Table 5 shows the loading of the four factors after performing exploratory factor analysis. The bold numbers (more than 0.6) show the high level of relationship of the questionnaire items to a single extracted factor. With the first factor, item load ranged from 0.370 (item 23rd) to 0.936 (item 11th). Items load ranged from − 0.244 (item 1st) to 0.827 (item 22nd) for the second factor, − 0.388 (item 8th) to 0.658 (item 20th) for the third factor, and − 0.412 (item 23rd) to 0.550 (item 7th) for the fourth factor.

As presented in Table 6, strong correlations were found between the total score of the FFI-Th with the pain-VAS (r_p = 0.695, p < 0.001), EQ-5D-5L (r_p = − 0.712, p < 0.001), and EQ-VAS (r_p = − 0.508, p < 0.001). The pain subscale of FFI-Th showed strong correlations with the pain-VAS (r_p = 0.755, p < 0.001), EQ-5D-5L (r_p = − 0.626, p < 0.001), and EQ-VAS (r_p = − 0.460, p = 0.001). The disability subscale of FFI-Th showed strong correlations with the pain-VAS (r_p = 0.640, p < 0.001), EQ-5D-5L (r_p = − 0.660, p < 0.001), and EQ-VAS (r_p = − 0.552, p < 0.001). The activity limitation subscale showed strong correlation with the EQ-5D-5L (r_p = − 0.760, p < 0.001) but showed weak and moderate correlations with the EQ-VAS (r_p = − 0.292, p = 0.042) and pain-VAS (r_p = 0.364, p = 0.010).

The participants in the present study all had plantar foot complaints, an average age of 47.22 ± 14.35 years, ranging from 20 to 78 years. Everyone was able to read and answer the questionnaire. Their mean and standard deviation of the total FFI-Th score (27.60 ± 20.30) agreed well with the average total FFI score determined in Rheumatoid Arthritis from a previous result (FFI total score of 28.09 ± 23.26) [12].

For the items of FFI-Th that were tested from this study, firstly, high ratios of NA answers were found in items 7th and 8th (n = 26 for each item), followed by the items 22nd and 23rd (n = 19 for each item). Secondly, one participant reported ceiling scores in the pain subscale. Thirdly, a relatively low internal consistency was found in the activity limitation subscale when compared with the other subscales. In addition, we encountered difficulty in the translation of some items of the original FFI including the distance of a block, which was uncertain in Thailand and unrecognized by Thais. Moreover, 500 m or around 1 bus stop distance was substituted for the distance of roughly four blocks. For this aspect, the differences in culture and livelihood between locations should be researched more.

For the NA answering, this problem had also been reported in earlier studies [42, 43]. Agel et al. found 20% or more of NA answers in four items of the pain and activity limitation subscales [42]. In contrast to the study of Venditto et al., 92% of NA answers were reported in six items of the pain subscale (3rd, 4th, 5th, 6th, 7th and 8th) [21]. According to the irrelevant items, this may make researcher modified the questionnaire as the shortened version [14, 17, 22, 43]. Among the questionnaire items, individuals who had less severity were unable to answer the questions of items 7th and 8th because they had no experiences of using orthotics or devices [14, 42, 43]. However, eliminating these items may have reduced the item numbers and may have affected the psychometric properties. Thus, it remains advisable to use the full 23-item questionnaire to determine all points of consideration of foot function [18].

The validity and reliability of the questionnaire may be jeopardized when high floor or ceiling effects are present. Discrimination between subjects is decreased on an item where there are many answering with the lowest or highest possible scores [44]. In this study, floor and ceiling effects were determined by counting the number of individuals who obtained the lowest (0) or the highest (10) scores.

From our findings, floor scores were observed in almost all items, except for the 1st item (worst foot pain). The highest number of respondents gave floor scores in the activity limitation sub-set, especially in items 22nd and 23rd (n = 16 or 32.65% for both items). It is possible that most of the participants were still carrying out their routine activities and could walk independently without using devices. According to the inclusion criteria, participants who had pain at the plantar surface of the foot during rest or weight-bearing activities were selected. Among those participants, most reported pain at the bottom part similar to the clinical symptoms in plantar fasciitis. Observing floor scores for the disability and activity limitation subscale in this study is consistent with the previous findings in studies that included plantar fasciitis cases [18].

The FFI-Th demonstrated excellent internal consistency for the total score and subscales of pain and disability and provided CA values similar to the previous translation reports [17, 18, 22]. However, a relatively low internal inconsistency in the activity limitation subscale was observed in this study similar to the related studies [14, 18, 42]. This led one study to omit this subscale from the questionnaire [43], but it may reduce the comprehensive nature of overall domains.

For the test-retest reliability, the total score and sum scores of the pain and disability subscales showed excellent reliability, while the activity limitation subscale showed good reliability. This finding was similar to the previous results that showed somewhat lower reliability of the activity limitation than the other subscales [16]. When considering each item, the lowest reliability (ICC_{3, 1} = 0.534) was found in the item 23rd (Limiting activities). This may be the result of uncertainty in rating this item and the majority of participants had no problem of activity limitation.

In the original version, the three subscales included pain (evaluated by the items 1st – 9th), disability (assessed by the items 10th – 18th), and activity limitation (assessed with the items 19th – 23rd) [12]. However, when observing factor extraction performed through the Principal Component Analysis with the varimax rotation method, except for the items 7th and 8th, almost all items of the pain and disability subscales showed values more than 0.6 in the first factor, whereas only two different items load in the second and third factors. Factor analysis of the FFI-Th showed results similar to the previous article [22]. Exploratory factor analysis showed little relationship between the activity limitation subscale and the other two subscales, but a stronger relationship between the pain and disability subscales was found as shown in the first factor. Therefore, diminishing the activity limitation subscale and items 7th and 8th (pain related to orthotic use when walking and standing) for further study in case of similar populations in the study may be recommended.

For criterion validity, many studies focused on the correlation between the FFI and different testing variables. In this study, we investigated the correlation of the FFI-Th with pain-VAS and with the EQ-5D because they indicated severity and health status. The findings demonstrated strong correlations between the total score of FFI-Th and subscales of pain and disability with the pain-VAS, EQ-5D-5L, and EQ-VAS. However, the activity limitation subscale demonstrated a strong correlation with EQ-5D-5L, when low and moderate correlations were found with EQ-VAS and pain-VAS. A strong correlation between the FFI-Th and EQ-5D-5L showed a congruence of foot function and health status determined from five domains. These validity results were similar to those related studies that investigated variable dimensions. Moderate to high correlations were found with pain-VAS [17, 20], function-VAS [17], physical and mental health status [17, 20, 45], activity scale [17] and health-related quality of life [15, 46]. A study of a Spanish version FFI reported a moderate correlation of the questionnaire with the EQ-5D and with the pain-VAS and weak correlation with the 12-item short-form health survey (SF-12) [47]. A German version of the FFI [17] had a strong correlation with pain, a moderate correlation with the physical SF and a weak correlation with the mental SF. Whereas the Chinese version of the FFI [18] had a strong correlation with physical aspects of the 36-item short-form health survey (SF-36), but a weak correlation with the mental SF-36.

Regarding feasibility, a report about the difficulty in the 18-item German version of the FFI found a mean difficulty of only 2.4 on a 10-point VAS with the average time required for completion of 8 min [17]. However, this study did not capture the time of completion or ask about the difficulty of interpreting the items. Nevertheless, all individuals completed the questionnaire without complaint.

The study was limited by a low degree of foot pain severity and clinical symptom characteristics of the participants as well as a small sample number. This may affect the generalizability for the use of this tool in other kinds of population. The psychometric properties including responsiveness or sensitivity to change need to be studied. In addition, future studies should compare the original FFI with the revised shortened version and test with a range of foot problem severities.