German version of the Cornell Musculoskeletal Discomfort Questionnaire (CMDQ): translation and validation

In order to establish the cultural equivalence of the original version of the CMDQ, we followed previously published guidelines for translation and cross-cultural adaptation of health status measures [25, 26]. The questionnaire was translated from English into German by two professional translators who worked independent of each other and whose native language is German. One of them was familiar with the concept of the questionnaire.

Then the written reports were discussed in a consensus panel between the two translators and a third person. All discrepancies in the written reports were checked against the original questionnaire and against each other regarding content analogousness, and critically judged in terms of everyday language to arrive at the preliminary version.

The translation of this version back into the original language was done by another professional translator whose native language is English and by a German physician who has been living in the USA for 25 years. Afterwards, any discrepancy between the original version, the translation, and the backwards translation were discussed on the basis of a structured interview in committee. Two translators, two occupational physicians, and two psychologists who were familiar with the intent of the measure and the concepts attended the committee. Again, first the semantic equivalence was checked and then all alternative suggestions were compared concerning unambiguity and popularity. Part of the review was also the instruction and the equivalence of steps in the scales. Subsequently, an interim final version was developed.

Key words were then underlined within the scales in order to improve the questionnaire’s clarity. In a final step, the questionnaire’s graphics were revised (see Figs. 2 and 3). In an earlier validation study [23], errors occurred during questionnaire completion due to mistakes in the horizontal dimension within the questionnaire. The lines representing single anatomical regions were thus marked off by gray shades in the new version.

The German version of the CMDQ (D-CMDQ) was pretested on 44 subjects because a sample size between 30 and 50 participants enhances the probability of detecting even rare problems, e.g. unclear questions, unfamiliar words or ambiguous syntax [27]. The inclusion criteria covered 18- to 67- year-old native German speakers in either a regular job or in an academic or professional education. Non-native German speakers were excluded from the whole study. 59.1 % of the participants were females. The pretest’s main purpose was to validate the comprehensibility of the adapted version, thus the pretest included subjects with different educational backgrounds and different occupations, for example kitchen porters, cleaners, office employees, musicians, students, and research assistants. The proportions of the subjects with high, intermediate, and low educational status were about 47.7, 31.8, and 20.4 %, respectively. The participants were contacted in their respective professional environments, were informed both orally and in writing about the purpose of the research study and data protection, and were asked to indicate occurring difficulties with regard to the instructions, the questionnaire’s structure, or the comprehensibility of certain terms. After completion the participants were asked again whether, they had experienced difficulties in understanding the instruction, the three questions or a single word. In total, only one of the 46 contacted persons declined to participate in the pretest: a 56-year-old male engineer declined due to protection of privacy. The participation of a 46-year-old female cleaner was interrupted due to functional illiteracy.

Inconsistency was encountered when subjects claimed they did not feel pain or complaints in certain anatomical regions within the frequency scale (“never”) but indicated severity and work interference for the same anatomical regions.

Taken as a whole, none of the pretest’s subjects claimed that they had difficulties with regard to clarity and comprehensibility when filling out the questionnaire. In some cases, items were omitted when subjects felt no complaints and thus left out certain anatomical regions in the questionnaire. Omitted items were the most frequent cause of missing data. With 20 items and 44 participants there were 880 possible items. Missing data occured in only 2 % of all cases. Inconsistency was found in many more cases (15 %).

The most frequent error was found when subjects answered the questions within the work interference scale “If you experienced ache, pain, discomfort, does this interfere with your work ability?” with “not at all” although they had claimed earlier that they did not experience pain and discomfort. This, however, did not lead to a substantial contradiction. Knowing that the regional score, as suggested by Hedge, is calculated as the product (frequency x intensity x work interference) and “never” equates to the value “0”, this error had no consequence, because in this case the result would remain “0” [28]. For this reason, the expert committee decided against further modifications in this particular case.

Besides missing data and inconsistent findings, the study participants provided two further indications in the interviews. The first indication concerned the fact that pain in the elbow could not be directly assigned by the subjects. Furthermore, it was apparent that a subdivision of the hip/buttocks into right-side and left-side would have been useful. Both indications were been added to the final version of the D-CMDQ after they had been discussed within the expert committee and after consultation with the developer of the CMDQ. Finally, the illustration of the female body was slightly modified in order to visualize the neck area more precisely. Later the male version of the standing worker was adapted. By doing so, the cross-cultural adaption process was completed.

Numeric Rating Scales (NRS) have been reported to yield reliable, valid, and sensitive measurements of pain intensity [29]. In this study an 11-point NRS (0–10) was used as a criterion for current overall pain intensity to test the construct validity of the D-CMDQ. For that the correspondence between the two methods was proved. The NRS assesses pain intensity from 0 (no pain) to 10 (worst imaginable pain), and study participants were asked to check the correct number describing the worst pain perceived during the previous week for each body area. The participants were then asked to complete the D-CMDQ. The responses regarding the occurrence of pain were then compared between both questionnaires in the same subjects. It was also examined whether all participants who had reported pain in the NRS did so in the D-CMDQ frequency scale as well. Those participants who had declared “no pain” in the NRS were expected to check “never”in the D-CMDQ frequency scale. Further NRS scores were expected to correlate positively with D-CMDQ severity scores.

The agreement between responses given in NRS and the D-CMDQ frequency scale was analyzed by Kappa coefficients (κ). There are different recommendations for the interpretation of the Kappa statistic [30‐32]. Following Landis & Koch [30] Kappa values between 0.61 and 0.80 were found to be substantial. Values between 0.00 and 0.20 were believed to be slight, values between 0.21 and 0.40 were believed to be fair, and values between 0.41 and 0.60 were regarded as moderate.

The Spearman rank correlation statistic was used to determine the correlation between NRS scores and D-CMDQ severity scale scores. Spearman’s correlation coefficients allow the analysis of the strength of association between variables of ordinal measurement levels in a single value between -1 (negative association) and +1 (positive association). The values can be interpreted as follows: very low association between 0.00 and 0.20; low association between 0.20 and 0.50; moderate association between 0.50 and 0.70; high association between 0.70 and 0.90; very high association between 0.90 and 1.00 [33]. These values are considered to be recommendations. The interpretation of a value always depends on the scientific question.

Test-retest reliability for self-administrated tests is measured by presenting a questionnaire twice to a person separated by a given time interval in order to assess stability over time. In this study, a time interval of 7 days was used as a reasonable compromise between memory bias and clinical change [34, 35]. For sample size determination the general recommendation of 2 to 20 subjects per item of the instrument scale was used [36, 37].

First test-retest reliability was calculated by Spearman rank correlation coefficients using mean sum scores of each scale (frequency, severity scale, and work interference scale) [38]. In general, high correlations can be expected for time-stable attributes only. Additionally, Kappa coefficients were calculated to analyze the test-retest-reliability for the responses given on the frequency, the severity and the work interference scale for each body area separately. In musculoskeletal research, ratings for clinical diagnosis or classification often lie on an ordinal scale. For this data the kappa statistic is an appropriate measure of reliability, provides valuable information and is commonly used even in the clinical setting [39]. The Kappa value is influenced by the prevalence of the outcome [40] and depends on the number of categories. The CMDQ frequency scale includes 5 categories; the CMDQ severity and work interference scale include 3 categories, respectively. The prevalence effect is related to the probabilities of “yes” and “no” and can lead to low Kappa values in the calculation of Kappa statistics despite of high agreement. Thus in this study an heuristic approach was used to overcome this problem by calculating a maximum Kappa (κ_max) for each measurement to compute the arithmetical ratio (κ/κ_max) subsequently [41]. Kappa maximum is the highest obtainable agreement for a specific data set to relativize the Kappa correlation coefficient. Additionally, the proportion of observed agreement (P_O) and the proportions of positive (p_pos) and negative agreement (p_neg) were calculated to obtain more information about response consistency (Table 3). p_pos is the number of positive responses agreed on for both measuring points divided by all positive responses for both tests, and likewise for p_neg [42].

Finally, the internal consistency of each scale was tested by calculating Cronbach’s alpha statistic. For empirical investigations an alpha-lower limit of 0.70 is considered as satisfactory [43].

All procedures performed in the study involving human participants where in accordance with the Helsinki declaration or comparable ethical standards. The design and protocol of the study was approved by the Ethics Committee of the University of Rostock. Participants were informed about the study purpose, methods, and confidentiality of data. Written informed consent was obtained from all participants included in this study.