Validation and investigation of cross cultural equivalence of the Fremantle back awareness questionnaire - German version (FreBAQ-G)

The study was designed as a multicentre, cross-sectional study. Data was collected as part of a study evaluating lumbar movement control in persons with NSCLBP. All participants provided written informed consent and all procedures conformed to the Declaration of Helsinki. To investigate cross-cultural validity the data of this study were pooled with those collected by Wand et al. [20] .

Participants were recruited in seven physiotherapy practices in Germany between April and September 2019.

Participants had to meet the following inclusion criteria: age ≥ 18 years; sufficient German language ability to complete the questionnaire; currently experiencing NSCLBP with or without leg pain (leg pain above the knee and main pain had to be localized below the costal margin and above the inferior gluteal folds) and duration of symptoms ≥3 months. The pain level, calculated as the mean of the actual pain intensity and the average pain intensity during the last 3 months, measured on an 11-point numeric rating scale (NRS), needed to be above 0. Participants were excluded if they had signs and symptoms indicating specific spinal pathology [24].

Data collection and data analysis was conducted by different persons to minimize potential risk of bias.

Participants provided basic demographic information and completed a self-developed questionnaire to collect information about LBP characteristics. Pain related disability during daily activities, leisure time and work, as well as pain intensity, were assessed using 11-point numerical rating scales (NRS 0 = no pain / disability - 10 = worst imaginable pain / disability). For overall pain related disability we calculated the mean of the impairment scores during daily activities, leisure time and work. Pain related fear was estimated using the German version of the Fear Avoidance Beliefs Questionnaire (FABQ) [25]. Finally, the participants completed the FreBAQ-G [22]. The FreBAQ-G consists of nine items measuring back specific self-perception on a five point rating scale with a range 0–36 (higher values indicating greater levels of impairment).

For questionnaires with ordinal scaled items, polytomous item response models are recommended [26]. The COSMIN (Consensus-based Standards for the selection of health status measurement instruments) checklist advocates a minimum sample size of 200 participants for IRT based Rasch analyses [23]. However, for polytomous IRT models the sample size should be at least 250, but 500 for accurate parameter estimates is recommended [27]. To assess the psychometric properties of the German Version we aimed to recruit a sample greater than 250. To evaluate cross cultural validity we pooled our German data set with the English-language data set collected by Wand et al. [20]. The sample size of the English data set consists of 251 participants with NSCLBP. So the overall sample size to investigate cross cultural validity meets the recommendation of 500 participants.

Two hundred seventy-two patients with NSCLBP were included of which 271 completed all questionnaires. Table 1 gives a summary of the demographic data of this sample and of the sample used by Wand et al. [20] for the validation of the original English version of the FreBAQ.

At the time of measurement, the average pain intensity within the last 3 months was 3.8 (SD 2) and the actual pain intensity was 3.4 (SD 2.1) measured on a 0–10 NRS. The average pain level was 3.6 (SD 1.7) and the average pain related disability was 3 (SD 2.1) on a 0–10 NRS.

The average total FreBAQ-G score was 8.0 (SD 6, range 0–27), with a median of 7.0 (interquartile range 3–12). Figure 1 shows the distribution of the FreBAQ-G sum-scores.

Two hundred sixty-four subjects had no missing values, 6 subjects had 1 missing value each, 1 subject had 5 missing values and 1 subject had 9 missing values (see also Table 2).

The frequencies, median responses and missing values to each FreBAQ-G item are given in Table 2. Response option 4 (always or most of the time feels like that) was not chosen for items 7 and 8. For items 1 to 8, the distribution of the responses to each category is left skewed towards the option 0 (never feels like that), whereas for item 9 the responses are more equally distributed. Loevinger Hj coefficients are between 0.34 and 0.48 (homogeneity of the scale: 0 represents no correlation and 1 represents a perfect Guttmann scale), with the lowest value for item 9 and the highest value for item 4. Cronbach’s alpha for the scale is 0.84 (test score reliability coefficient: 0 represents no interitem correlation 1 represents perfect interitem correlation).

After fitting the GR IRT model, test characteristics were examined. The test characteristic curve (Fig. 2) shows the expected overall observed sum-score of the FreBAQ-G at each given θ-value of the underlying trait (back specific self-perception), which is plotted on the x-axis. It can be seen that the test characteristic curve is an increasing, nonlinear function of the underlying trait. This was anticipated because subjects with a higher perceptual impairment (right side of the x-axis) will be expected to score higher on the FreBAQ-G sum score. The test characteristic curve shows that 95% of the people with NSCLBP will score between 1 and 21 on the classical FreBAQ-G sum-score. In people with NSCLBP, who have an average self-perception, the sum-score will be 7, whereas a person with a self-perception 1 SD worse than average will reach a value of 14, and if the self-perception is 1 SD better above average, the score will drop to 2. Keeping in mind that higher values are an expression of more impairment, the figure also shows that the questionnaire can better differentiate people with a higher degree of impaired back specific self-perception, whereas people with a lower degree of perceptual impairment, the questionnaire offers only 7 points for discrimination (between zero points and seven points).

The test information function curve (Fig. 3) confirms that the FreBAQ-G is most informative for persons with a self-perception of the back worse than average (Theta between 0 and + 4).

The graph also shows that for persons with a self-perception of the back worse than the average, the SE (standard error of measurement) is lower than for those with a better than average self-perception of the back. The overall reliability of the FreBAQ-G in this study is 0.84. The reliability for self-perception of the back better than average is 0.797 and 0.884 for worse than the average.

Table 3 gives an overview of the item characteristics for the FreBAQ-G. All discrimination values of the nine items are above the desired value of 1. The highest discrimination parameter showed item 4, the lowest item 9. Because item discrimination and item information are related concepts it can be expected that item 4 offers the highest information whereas item 9 offers the lowest information about individual back specific self-perception. Table 3 also gives the item difficulty parameters for each item and category. The interpretation of each of the four item difficulty parameters per item is that a person equal to it has a probability of 50% for responding in the pertinent category or higher. For example, looking at the estimate of 0.87 for item 1 (“My back feels as though it is not part of the rest of my body”) category ≥2 means that a person with a back specific self-perception (Theta value) of that value (that is 0.87) has a probability of 50% to answer with category 0 or 1 versus response category 2 or higher (being 2, 3 or 4). Similarly, someone with a Theta value of − 0.07 has the same probability to answer 0 as to answer 1 or higher.

Figure 4 shows the category characteristic curve of item 4 (“When performing everyday tasks, I don’t know how much my back is moving”). It can be seen that only the first and last response categories are monotonically decreasing and increasing. Furthermore, the categories can be considered as ordinal measures, so if one moves from left (not impaired) to the right (highly impaired) on the x-axis the probability of the response to the question of item 4 changes from category 0 (never feels like that) to category 4 (always, or most of the time feels like that). The crossing points can be described as transitions points. For instance, a Theta value of − 0.17 indicates a person whose back self-perception is 0.17 SD better than the average. In addition, for a person with a Theta value > − 0.17 the probability of choosing category 1 (rarely feels like that) becomes for the first time higher than choosing the reference category 0 (never feels like that). It is important to notice that each category has an interval where the probability for this response category is highest and that the intervals between the response categories are reflecting the ordinal structure of the questionnaire.

Examination of all category characteristic curves showed that for item 6 category 2 and for item 7 category 1 did not have a clear interval in the latent trait (see Fig. 5). Furthermore, none of the participants in the current German sample had sufficiently disrupted self-perception to choose category 4 (always, or most of the time feels like that) for items 7 and 8. The response categories of all items are shifted to the right side of the x-axis (e.g. see Fig. 6), except the categories of item 9 who are more equally distributed (see also Table 3).

The calibrated mean item difficulties were plotted with the German items on the y-axis and the English items on the x-axis (Fig. 6). All difficulty parameters, except for item 7, lie within the 95% CI borders indicating equivalence of the item difficulty parameters of the English and German-language versions.

Item discrimination parameters (a) of both populations were plotted in a similar way with the line of identity and 95% CI (Fig. 7). Again, all items lies within the 95% CI borders, indicating equivalence of the English and German versions. Values higher than 1 are desirable and indicate better discrimination. Items 4, 5 and 6 have the highest discrimination parameter in both populations.

After accounting for the uniform differential item function (DIF) of Item 7 the graded response model (GRM) estimation of both groups showed a mean Theta of 0 with a variance of 1 for the English-language population (anchor) compared to a mean of − 0.1006 with a variance of 0.8348 for the German-language population. The test characteristic curve (Fig. 8) shows that individuals in the German population with a back specific self-perception below-average (Theta between 0 and + 4) will score one point lower on the FreBAQ than individuals with the same level of body perception in the English-language population. For individuals with above-average body perception (Theta 0 to − 4) there is no obvious difference in the expected sum-score.

Self-perception of the back showed significant, moderate correlations with pain intensity, disability and fear avoidance beliefs (Table 4).

The average total FreBAQ-G score in our sample was slightly lower with an average of 8.0 (SD 6.0) compared to Wand et al. [20] with 9.8 (SD 6.6). This difference can be explained by the slightly lower pain intensity and FABQ-PA level of our sample, which may also be a reason for the observed floor effect of the sum-scores (compare to Fig. 1). For items 7 and 8, none of the participants in the German-language population scored category 4 (always, or most of the time feel like that). In line with Wand et al. [20], item 9 had the lowest mean difficulty parameter and item 8 the highest.

The calculated reliability index of our study was r = 0.84 (range 0.79–0.88) and Cronbach’s Alpha was 0.84. These values are comparable to those found by Wand et al. [20] (CA = 0.80, r = 0.74). Loevinger Hj coefficients, referring to the questionnaires ability to differentiate persons into score groups [29], showed moderate scalability for all items except for items 8 and 9. Here the values were slightly lower (below 0.4).

In line with Wand et al.’s study, we found that the scale of the FreBAQ-G reflects an unidimensional construct. Compared to Wand et al. [20], the PCA in our study showed more robust results. The eigenvalue of the first factor was greater than 4 and explained 53% of the variance, compared to Wand et al. [20] with an eigenvalue greater than 2. Based on X² statistics the assumption of local independence could not be rejected. The use of the IRT-graded response model (GRM) was supported by the model-fit and item-fit of the data.

The FreBAQ-G is suitable to differentiate between people with back specific self-perception between Theta − 1 and + 2 (− 4 stands for not impaired and + 4 stands for highly impaired) (see also Fig. 3). For people with a better back specific self-perception perception (Theta values below − 1), other items have to be developed. This view is supported when looking at the FreBAQ-G sum-score that ranges from 0 to 45. People with an average impairment of self-perception of the back had a median score of 8. This means that for patients with worse than-average self-perception the score ranges from 8 to 45 whereas for patients better than-average, the available score range is only from 0 to 8.

Furthermore, the classical sum-score of the FreBAQ-G should be used with care. The sum-score can only have integer values, and because of the non-linear relationship between the sum-score and the theta score, the assumption of equal distances between scores is violated. Therefore the sum-score is not interval scaled and even its ordinal scale can be questioned.

We used the graded response model (GRM) to compare the performance of both language versions. The English-language version showed very good model fit for the graded response model (GRM). All items showed comparable item discrimination parameters. In both populations items 4, 5 and 6 showed the highest discrimination. Therefore, in both language versions, the responses to these three items are offering the highest amount of information about the back specific self-perception of the respondent. Only in the English-language version Items 8 and 1 had discrimination parameters lower than 1. We found no non-uniform differential item function (DIF). However we found uniform differential item function (DIF) in item 7. This does not automatically indicate that translation of these items was not accurate. Differential item function (DIF) can also occur due to chance, different sample sizes, group differences in age, sex or disease characteristics, different administration modes and real differences between cultural and language settings [31]. As our total sample size was large and evenly distributed between the two language groups, differential item function (DIF) is unlikely to be due to chance or sample size differences. Differences in age, sex, or disease characteristics between the German and the English-language group are more likely explanations. There are differences between the two language groups in terms of age, BMI, pain intensity and the FABQ. The German language group was significantly younger, had a lower BMI, lower pain intensity and a lower mean score in the FABQ. Figure 6 shows that for eight items the calibrated mean item difficulty parameter lies above the line of equality in the German-language population. We think that this is a sign of a systematic difference between these two study populations. Therefore we believe that the differential item function (DIF) observed in item 7 as well as the other observed differences in the performance of the FreBAQ between these two populations were due to the described sample differences.

Some limitations of our study need to be discussed. First of all, the sample size is at the lower margin of sample sizes recommended for IRT studies. However, our results are sufficiently precise to assume that a larger sample size would not alter their magnitude or direction. Also, levels of pain intensity were relatively low in comparison to Wand et al’s study [20], therefore comparability between results is compromised.

Our results indicate that the FreBAQ-G gives the most valid results for persons with NSCLBP and physical impairment above average. The item discrimination values of item 4, 5 and 6 show that these provide the most information in regards to impaired back specific self-perception.

For clinical interpretation of FreBAQ-G sum-scores, a common metric is very helpful, especially when the results of different measurement instruments have to be compared [36]. However, the theta-scale has negative and positive values and it might be difficult to communicate their meaning. To aid the interpretation of the individual trait level estimates we recommend a T-transformation. The T-transformation is defined as follows:

T-scores have a range from 0 to 100 with a mean of 50 and a SD of 10. Higher values indicate more pronounced perceptual impairment of the back. T-scores are interval-scaled, easy to interpret and T-scores from different questionnaires can easily be compared. E.g. if a person with NSCLBP has a T-score of 65 we know that their body perception is 1.5 SDs worse than average, or if another person has a T-score of 40 the body perception is 1 SD better than average. Assessment of impaired self-perception could add information to the complex clinical picture of NSCLBP and thereby help clinicians to select targeted treatment options.

Future longitudinal studies should investigate sensitivity to change of the FreBAQ. It would add confidence to the validity of the questionnaire, if improvement in outcomes such as pain, function or health-related quality of life correspond to changes in self-perception of the back. Furthermore, the FreBAQ could be used as an outcome measure in controlled trials investigating interventions that target impaired self-perception of the lower back, such as tactile discrimination training or visual feedback training, results may help to explore how treatments work or why treatments do not work.