For the purpose of evaluating subjects with neck pain in the Turkish society, a Turkish translation and adaptation of NDI was performed, and the validity and reliability of its use was demonstrated. Our study group for the test which was performed on patients from polyclinic through randomized selection comprised of patients with neck pain. This was a homogeneous population regarding age and gender.
In our study, the duration for completing Turkish version of modified NDI was similar to that in the original study [15
]. Unclear questions were explained by the observer.
As with other studies, there were non-drivers, unemployed and those who did not have leisure time activities [42
]. Ackelman et al. modified this part of the index wherein they added the explanation of "not applicable" [44
]. It was considered appropriate to add the option of "never done" to the test also in our study.
As with most other studies, [18
] test-retest for all NDI domains was found to be high (0.87-1) in our study. In the study by Ackelman et al. the result (0.97) that was found in the test-retest reliability performed with a 2-day interval was high [44
], which was attributed to the additional explanations provided by the investigators. These additions were also provided in our study. The study by Cleland et al., determined an ICC of 0.50, which was the lowest value reported in the literature [46
]. However, this study involved patients with cervical radiculopathy. In the adaptation study conducted in the Netherlands, test-retest ICC was found lower as 0.53 in personal care domain [22
The most important comparison of our results is with the work of Aslan et al. [25
]. They also reported very high test-retest reliability (r = 0.98).
As with our study, in most similar studies, a duration of a day or two was given for test-retest [18
]. In pain studies, a retest interval of one week was reported to be not suitable as treatment administration and would be unethical [18
In NDI internal consistency studies, values were found to be between 0.74-0.93 by investigators. The internal consistency we determined in Turkish version of modified NDI was consistent with those found in the studies by these investigators [18
]. Aslan et al. [25
] did not report on the internal consistency of their Turkish translation of the NDI.
In our study, factor analyses revealed one dimension. Similar factorial structures of the questionnaire were observed in its Greek, Brasilian, Canadian [47
], Spanish versions. But, in French version, two factor were found. The percentage of explanation of the single factor was higher than in the Greek version, but similar to that found in the Brasilian and Canadian versions (84%) [18
]. Aslan et al. [25
] did not report on the factorial structure of their Turkish translation of the NDI.
In validity studies, pain was generally evaluated using VAS. Similarly, the results of our study was found good [21
]. Mousavi et al. found a correlation with VAS as r = 0.71 [21
]. In these studies methodological basis was similar. In addition, results are parallel with those found in many studies that followed different methodologies [23
]. Aslan et al. reported only moderately good (r = 0.51-0.62) correlations with the pain VAS [25
The SF-36 was used for validity also in other studies [18
]. In the Portuguese translation study conducted in Brazil, no correlation was found with physical role, emotional role and pain subtitles of SF-36 [18
]. Similarly, no correlation was found with physical role in our study, but there was a weak correlation with emotional role and pain. This could be associated with the number of participants. Riddle et al. found equivalence at strong correlations with physical and mental parts of SF-36 [48
]. In the translation study for the scale conducted in Iran, only emotional role was found not to be correlated [21
]. This result also supports our study.
The HAD test was used in the translation study conducted in France. Investigators, who found correlation between HADS depression scores and NDI, stated that sense of pain was closely associated with psychology [24
]. In our study, HADS was found to be correlated with both anxiety and depression scores. In studies conducted on patients with neck pain, it was shown that anxiety and stress might either be the cause or the result of neck pain [49
In the comparison with Aslan et al., [25
] no correlations with self-rated questionnaires for any other important health-related variables was reported. They did report a high correlation with the Turkish version of the Neck Pain Disability Index, which would be expected.
With respect to the work of Aslan et al. [25
], in summary, our study has extended that work by including analyses of internal consistency and factorial structure as well as analyses of convergent validity with the pain VAS, HAD and SF-36.
This study has limitations. As this was a study of a modified version of the Turkish NDI, an Item Response Theory (IRT) approach [31
] was not adopted; however, with regard to the reliability study, individual item analyses were undertaken. With regard to validity, only current pain, anxiety/depression and quality of life were assessed for concurrent validity. This was deemed to be an appropriate profile of separate constructs with which to evaluate the modified version of the Turkish NDI. Most of the correlations with these instruments did conform to our moderately high predictions. Additional testing with other constructs such as catastrophization or fear-avoidance beliefs would be interesting.
An additional limitation was the lack of assessment of measurement error and responsiveness. We intend to pursue this is a separate study involving a treatment phase.