Background
Pain-related fear has been shown to be a valid predictor of chronic pain and disability [
1,
2]. Additionally, “kinesiophobia” refers to “an excessive, irrational and debilitating fear of physical movement and activity that results from a feeling of vulnerability in regard to a painful injury or reinjury” [
3] and is conceptualized as a fear of movement/(re)injury by Vlaeyen et al. [
4]. According to cognitive-behavioral models, such as the fear-avoidance models [
4,
5] or the avoidance-endurance model of chronic pain [
6], painful experiences will elicit a fear of movement/(re)injury in certain individuals, which often leads to behavioral avoidance and, in the long run, disuse, depression and increased disability. Other individuals will respond to pain with cognitions of minimization and/or thought suppression, and endurance behavior accompanied with low levels of pain-related fear and avoidance [
7].
The Tampa Scale for Kinesiophobia (TSK) [
3] was developed in order to assess fear of movement/(re)injury. The psychometric properties of the TSK have been tested in different languages [
5,
8‐
12] and for different pain disorders (e.g., CLBP [
5,
13], osteoarthritis [
14], fibromyalgia [
15], and neck pain [
16]). Construct validity has been demonstrated using measures of disability and other fear of pain questionnaires [
5,
17,
18].
There are several versions of the TSK available, with 17, 13, 11 and 4 items [
19], respectively. The 17-item version includes 4 inversely coded items, which are not included in most previous studies of the TSK due to their low factor loadings. Studies with all 17 TSK items arrive at different factor solutions with one (Miller et al., 1991, unpublished), three [
13] or five [
9] factors. Studies with the 13-item TSK usually arrive at a two-factor structure [
20‐
24], although the distribution of the items on the factors varies. The 11-item TSK consistently reveals a two-factor structure, and the items that are used, as well as the item distribution, are invariable across studies [
16,
25]. A 12-item version showed a four-factor structure [
5], and a 4-item version revealed a one factor structure [
15]. Due to the many models that are currently in use, it is important for future research to examine the existing models for their accuracy and usefulness. Previous studies that used a confirmatory factor analysis to compare Vlaeyen’s four-factor model (12 items), Clark’s two-factor solution (13 items) and two one-factor models (13 and 17 items) showed that the two-factor model by Clark et al. [
20] provided the best fit [
14,
25,
13], which was further found to be invariant across different patient groups (e.g., chronic low back pain and fibromyalgia [
23,
24]). In a 2007 study, Roelofs [
16] presented a new two-factor structure that was based on the TSK-11 by Woby et al., 2005 [
25]. This factor structure also proved to be invariant across pain diagnoses and countries [
16,
26].
Because a German version of the TSK has not been available until now, the main objectives of this study were threefold: First, several models from previous studies were examined in a confirmatory factor analysis in order to determine the best fit. Second, the psychometric properties of a German version of the TSK were examined in a sample of patients with low back pain. Finally, construct validity was further investigated by exploring the relation to cognitive-affective and behavioral avoidance, as well as to endurance variables. Criterion-related validity was explored with respect to pain intensity, disability and general distress.
Discussion
The present study evaluated the factor structure, the internal consistency and the validity of a German version of the TSK (TSK-GV) using a sample of patients with low back pain (LBP). The inverse items were eliminated from the TSK-GV after determining that they were detrimental to a good fit. Important issues refer to certain new aspects of validation with respect to behavioral avoidance and pain-related endurance variables, such as Positive Mood despite Pain, Thought Suppression, and Behavioral Endurance.
Between the models that were examined in the present study, the four-factor model by Vlaeyen et al. (1995) [
5] and the two-factor model by Roelofs et al. (2007) [
16] emerged as the models with the best fit in the confirmatory factor analysis. An examination of the reliability of the TSK-GV that was built after each model showed that the two-factor solution by Roelofs et al. (2007) [
16] produced better results. Therefore, the two-factor solution was chosen as the final model for the TSK-GV. This 11-item TSK-GV model also seems to be the economically sound decision because it is the shortest reliable possibility and, therefore, reduces the patients’ burden. It is also supported by previous studies [
16,
31,
59].
In the present study, adequate levels of internal consistency were found for the TSK-GV total score (α = 0.73). The subscales SF and AA showed internal consistency values that were slightly below the desired value of .70, with SS’s alpha = .64 and AA’s alpha = .63. Shorter scales that have less than 10 items are still adequate with an alpha above .60 [
60], and French et al. (2007) [
13] state that they see the reduced subscale reliabilities that they also found in their study with the TSK-13 as a reflection of the small number of items on the scales rather than of problems with the items per se. Nevertheless, the reliability results indicate that it could be adequate to use the TSK-GV’s total score in clinical practice, especially because the total score shows better reliability than the subscales. Unlike French et al. (2007) [
13], the present study did not find a very high subscale intercorrelation (see Table
3), which suggests that the factors in the 11-item version of the TSK measure rather distinct concepts within the main concept of fear of (re)injury. Both subscales showed very high correlations with the total score. These results support the subscales as valid parts of the main concept of fear of (re)injury.
Another emphasis of the present study was on the thorough examination of the validity of the TSK-GV total score and the separate factors. According to prior studies [
4,
13,
20,
24], fear of movement/(re)injury was expected to positively relate to depression, catastrophizing, pain anxiety and fear-avoidance beliefs regarding physical activity and disability. Higher TSK-GV total scores were indeed significantly correlated with higher levels of general pain anxiety, fear-avoidance beliefs and emotional, cognitive and behavioral fear-avoidance responses, such as help-/hopelessness, catastrophizing and avoidance of social and physical activities. Activity-related fear-avoidance beliefs showed a stronger correlation with the total score than the work-related ones, which is consistent with French et al. (2007) [
13]. The TSK-GV total score was negatively associated with measures of endurance responses, such as
Positive Mood despite Pain and
Minimization, while
Endurance Behavior and
Thought Suppression were not significantly related. In general, the correlation pattern was consistent with the construct “kinesiophobia” and with different cognitive-behavioral models of pain [
4‐
6]. Positive correlations between several fear-avoidance variables, general distress and the TSK-GV support the assumptions of a pathway from pain, cognitions of catastrophizing and/or help-/hopelessness via pain-related fear of movement/(re)injury to behavioral avoidance. Negative correlations between the TSK-GV and
Positive Mood despite Pain and between the TSK-GV and cognitions of
Minimization support one of the endurance pathways, which suggests that low levels of pain-related fear and avoidance are accompanied by high eustress-endurance [
6,
7].
Concerning the first TSK-GV subscale
Somatic Focus (SF), we found a correlation pattern that mainly matched the one for the TSK-GV total score. The subscale
Activity Avoidance (AA) differed from this pattern slightly, as it was mostly related to pain anxiety, activity-related fear-avoidance beliefs, help-/hopelessness, avoidance of social and physical activities and negatively to minimization. The correlations between the TSK-GV’s AA subscale and the avoidance subscales of the FABQ and the KPI, as well as the negative correlations between the AA subscale and the KPI endurance scales, support the validity of AA being a measure of avoidance. The low or insignificant correlations with endurance are consistent with previous research: Tkachuck and Harris (2012) [
59] found low but significant negative correlations between the TSK-11’s AA subscale and measures of physical functioning (stair climb and sit-stand). AA was also able to uniquely predict performance in these tasks of physical functioning. Because avoidance measures are scarce, AA may be useful in this regard. In sum, the high correlations with pain anxiety, the moderately positive correlations with avoidance and fear-avoidance beliefs and the moderately negative correlations with certain aspects of pain-related endurance support the construct validity of the TSK-GV.
In accordance with previous findings [
13,
20], positive correlations between the TSK-GV total score and pain-related disability and depression were found, supporting criterion related validity. The SF subscale again showed the same correlation pattern as the TSK-GV total score. For the AA subscale, only a significant correlation with disability that was measured by the ODI was found. Disability that was measured by the PDI and depression did not remain significantly correlated with the TSK-GV after a Bonferroni correction.
Regarding the pain variables, significant correlations between the TSK-GV total score and its subscales and pain duration or pain intensity could not be observed. This finding is in line with previous research [
20], while other studies [
5,
13] found that the TSK scores were positively related to self-report of clinical pain. Vlaeyen (1995) [
5] did not find pain intensity to predict fear of (re)injury and concluded that fear of (re)injury most likely occurs independently of current pain intensity. This suggestion supports the notion of Crombez et al. (1999) [
17], who proposed that the expectation of pain may be more debilitating than the actual pain. This could indicate that the extent of kinesiophobia in patients with low back pain is independent of the duration and average pain intensity of the current pain that they are experiencing.
Limitations
The results of the current study are limited to patients with low back pain. The study was part of a large multicenter study about back pain; therefore, the sample was limited to patients with back pain. Because the data are cross-sectional, conclusions of cause and effect cannot be drawn. The reliability that is stated in this study only refers to internal consistency, as repeated measures for test-retest reliability were not included.
Competing interests
The authors declare that they have no competing interests.
Authors’ contributions
All of the listed authors (ACR, NK, DH, JH and MIH) contributed to the present study by providing substantial contributions to the study design, data collection, data analyses and interpretation of data. Furthermore, all of the included authors were involved in drafting the article and critically amending the first drafts. Before they gave their final approval for the present article to be published, all of the authors discussed the results and commented on the manuscript.