Background

Chronic pain is a serious socio-economic and medico-biological problem of modern society [1-3]. According to DALYs parameter (Disability Adjusted Life Years), musculoskeletal pain is at the third place in the world and lower back and neck pain prevail in the overall structure of pain [3]. In Russia, lower back and neck pain is among the top ten leading causes of poor quality of life [4].

Multidisciplinary program is one of the preferable approaches with evidenced effectiveness in the treatment of chronic pain. This approach implies coordinated work of a team of specialists (physicians, psychologists, social workers, nurses, etc.) united by the same purpose (to ensure chronic pain relief) [2, 5]. At the same time, the latest systematic review has shown that clinical studies of multidisciplinary chronic pain management programs should be continued and their quality is required to be improved [6].

An important approach towards improving the quality of clinical trials is clinical introduction of tools measuring how a patient assesses own health, quality of life, effect of disease, pain intensity, effectiveness of treatment, etc. [7]. In the English-language literature, these tools are called patient-reported outcomes (PRO) [8]. Thus, this assessment must come from the patient himself, without interpretation by a doctor or anyone else [8, 9]. PROs are assessed using scales and questionnaires. Combination of several scales and questionnaires is recommended in clinical studies of chronic pain: one-dimensional pain scales (visual analogue or numerical rating scale); multidimensional questionnaires of a general type assessing pain intensity and various aspects of life (Brief Pain Inventory); Likert scale (Global perceived effect scale); disease-specific questionnaires [8-12].

Neck Disability Index (NDI) is one of disease-specific questionnaires recommended for research of chronic neck pain. Its psychometric properties are well understood. It is known that these properties depend on inclusion criteria [13]. For example, patients with non-specific neck pain, neck pain accompanied by radiculopathy, cervicogenic headache differ qualitatively, despite the common symptom (neck pain). These qualitative differences affect the quantitative indicators of NDI, for example, minimal detectible changes (MDC).

Initially, the Russian-language version of the questionnaire (NDI-RU) was adapted in a heterogeneous group of patients with neck pain [14]. Therefore, psychometric properties of NDI-RU in certain homogeneous groups of patients, for example, in people with cervicogenic headache seem relevant. To date, NDI can be considered the only disease-specific tool for cervicogenic headache. Indeed, the questionnaires developed for primary headaches do not fit this function. Moreover, Migraine-Specific Quality of Life Questionnaire (MSQ v2.1) is the only reliable tool to assess the impact of headache on the quality of life [15].

In population studies, prevalence of cervicogenic headache varies from 0.17% to 4.1% [16]. In a specialized pain clinic, patients with cervicogenic headache made up 2.6% [17]. Cervicogenic headache was diagnosed in 20% of patients with chronic headaches and in 53% of cases after whiplash injury [18, 19]. Effective treatment of cervicogenic headache requires a multidisciplinary approach. Therefore, reliability of the questionnaire specific for cervicogenic headache seems to be relevant for various specialists.

The purpose of the study was to assess reliability of the Russian-language version of the Neck Disability Index (NDI-RU) in patients with cervicogenic headache.

Material and methods

The sample consisted of patients with non-specific neck pain and associated diseases, as well as those with cervical radiculopathy. A subgroup of patients with cervicogenic headache (CHA) was distinguished from this sample. We used the International Classification of Headaches [20] and Cervicogenic Headache International Study Group criteria [21]. Patients with specific neck pain were excluded.

Neck Disability Index

The questionnaire consists of 10 items (pain intensity, personal care, lifting, reading, headache, concentration, work, driving, sleeping, recreation). Two items reflect CHA-related symptoms (neck pain and headache), eight other items — functions or activities that may be impaired with CHA (concentration, sleeping, work). Each item contains 6 statements arranged in an order of progressive pain impairment (no pain; mild; moderate; severe; extremely severe; the most severe one can imagine) and dysfunction (I can read as much as I want; ...; I cannot read at all). Similar to verbal rating scale, NDI is a 5-score scale. Each statement is assigned a score from 0 to 5, where 0 — no symptoms (or dysfunction), 5 scores — maximum intensity of a symptom (or dysfunction). The minimum score is 0, maximum — 50. Neck Disability Index is assessed as a percentage if the patient skips one item (most often — driving) [14].

Psychometric properties

Reliability — Internal Consistency

Internal consistency is one of the quality criteria of the questionnaire characterizing its reliability as a measuring instrument. Internal consistency of the questionnaire shows consistency of its sections with each other and the questionnaire as a whole. This value determines how much each item of the questionnaire measures a feature that is measured by the entire questionnaire. Cronbach’s α was calculated for assessment of NDI-RU internal consistency (α> 0.6 — doubtful value; α> 0.7 — sufficient; α> 0.8 — good; α> 0.9 — very good). The optimal point-to-point correlation does not exceed 0.5 (optimal = 0.35; satisfactory range 0.2—0.5)

Test-retest reliability

Another criterion of the questionnaire quality is test-retest reliability. This value characterizes resistance to measurement errors. To assess test-retest reliability, we performed repeated measurements every 2-4 days and calculated intraclass correlation coefficient (ICC). This value characterized consistency between two measurements. The test was considered reliable in case of high correlation of the first and the second measurements.

Competitive validity

To assess the competitive validity, we analyzed the Pearson correlation coefficient r between NDI-RU, neck pain and headache intensity measured by 11-point numerical rating scale.

Minimal detectable changes

Minimal detectable changes (MDC) for 95% confidence interval were calculated using the equation MDC = SEM × √2 × 1.96, where SEM — standard error of measurement or standard error of mean; 1.96 — coefficient for 95% confidence interval (since 95% of sample means are within the 1.96 standard deviation (SD) of the population mean). SEM is an index of reliability indicating variability of values obtained from measurement to measurement [22].

SEM was calculated in two fashions. The 1^st method was an equation SEM = Sd / √2, where Sd — standard deviation of NDI-RU score difference obtained in test-retest method. The 2^nd method was an equation SEM = s √1–r, where s — mean standard deviation of the total NDI-RU score in the first measurement (test) and repeated measurement (retest), r — Pearson correlation coefficient between the total NDI-RU scores obtained by test-retest method.

Constructive validity. Variate analysis.

Variate analysis was carried out via selection of the main components through Varimax rotation. The numerical values were presented by Pearson correlation coefficient r between the corresponding items and the identified factors. The questionnaire factors in the CHA group were compared with the structure of the questionnaire factors in the entire sample of patients with neck pain and associated diseases [23].

Data processing. The calculations were performed using the Statistica 8.0 software package (StatSoft®, Inc.).

Results

The study was conducted at the Manual Therapy Center and included 670 patients (490 women and 180 men) aged 18 — 60 years (mean 39.4 ± 10.7 years). All patients had non-specific neck pain and associated diseases, as well as cervical radiculopathy. Distribution of NDI values were normal: Kolmogorov-Smirnov test d = 0.07 (p <0.05); Shapiro-Wilk test W = 0.98 (p = 0.000).

The CHA group made up 20.3% of the entire sample: 136 patients (103 women 33 men) aged 18 — 60 years (mean 42.1 ± 11.1). The group was representative to the entire sample regarding both age and gender. Distribution of NDI values were also normal: Kolmogorov-Smirnov test d = 0.13 (p <0.05); Shapiro-Wilk test W = 0.96 (p = 0.001). There was moderate impairment of vital activity in patients with CHA (17.9 ± 6.8 scores).

Internal consistency

Data on internal consistency of the questionnaire for CHA are presented in Table 1.

Table 1. Spearman’s correlation coefficient and Cronbach’s alpha for each item of the questionnaire (n=136)

Items of the NDI-RU	Spearman correlation coefficient	Cronbach's α
1. Neck pain intensity	0.54	086
2. Self-care	0.69	0.86
3. Lifting	0.63	0.85
4. Reading	0.62	0.85
5. Headache	0.45	0.86
6. Concentration	0.54	0.86
7. Working	0.77	0.84
8. Driving	0.69	0.88
9. Sleeping	0.45	0.87
10. Recreation	0.66	0.86

As one can see in the table, good (> 0.8) values of Cronbach's α were obtained for all items of the questionnaire that indicates good consistency of each item with the questionnaire as a whole.

Overall internal consistency of the questionnaire: between-class correlation coefficient r_S = 0.44; Cronbach's α = 0.87. Split reliability: correlation between even and odd items: r_S = 0.44; Cronbach's α = 0.87; correlation between the first and second parts of the questionnaire: r_S = 0.44; Cronbach's α = 0.87. These data indicate a good internal consistency of the questionnaire used in patients with CHA.

Test-retest reliability

Assessment of test-retest reliability is shown in Table 2.

Table 2. ICC values for each item and the whole scale (n=136)

Items of the NDI-RU	ICC	95% CI
1. Neck pain intensity	0.72	0.62-0.80
2. Self-care	0.76	0.67-0.83
3. Lifting	0.65	0.54-0.74
4. Reading	0.70	0.59-0.78
5. Headache	0.86	0.80-0.90
6. Concentration	0.79	0.72-0.85
7. Working	0.77	0.68-0.84
8. Driving	0.73	0.59-0.82
9. Sleeping	0.87	0.82-0.91
10. Recreation	0.84	0.77-0.88
Total score	0.93	0.90-0.95

Good (> 0.7) and excellent (> 0.9) correlation between repeated measurements was obtained that indicates a good test-retest reliability of the questionnaire used in patients with CHA.

Competitive validity

There was moderate correlation between neck pain intensity (according to numerical rating scale) and disability (according to NDI-RU): r_S = 0.58 (t = 8.27; p = 0.00). Correlation between headache intensity (according to numerical rating scale) and disability (according to NDI-RU) was also moderate: r_S = 0.49 (t = 6.46; p = 0.00). Correlation between neck pain intensity and headache intensity measured by numerical rating scale was characterized by r_S value of 0.70 (t = 11.47; p = 0.00). Thus, we found a moderate significant correlation between NDI-RU score in CHA patients and neck pain intensity (as well as headache) measured by numerical rating scale. This indicates a good competitive validity of the questionnaire.

Variate analysis

Variate analysis revealed a 2-factor structure of the questionnaire with eigenvalues over 1 (Table 3).

Table 3. Varimax-Rotated Factor Matrix of the Neck Disability Index—Russian Language Version in patients with cervicogenic headache (n=136)

Items of the NDI-RU	Factor 1Sensory and Cognitive Functions, Daily and Social Activity	Factor 2Sensory, Cognitive, and Mental functions
3. Lifting	0.872	0.014
7. Working	0.766	0.359
2. Self-care	0.753	0.270
10. Recreation	0.735	0.258
4. Reading	0.706	0.170
8. Driving	0.625	0.450
1. Neck pain intensity	0.614	0.237
6. Concentration	0.219	0.793
5. Headache	0.163	0.745
9. Sleeping	0.211	0.665

The first factor (all items of the questionnaire except 5, 6, 9) was comprised by all types of daily activity (lifting, working, self-care, driving), social activity (recreation) and cognitive function (reading). Eigenvalue for the 1^st factor was 4.63; this factor can explain 49.05% of the total variance. For the 1^st factor, Cronbach's α was 0.88; Spearman's r — 0.54. We named this factor “Daily activity and cognitive functions” since loadings were maximal for these functions (> 0.70).

Sensory (headache), cognitive (concentration) and mental (sleeping) functions comprised the second factor. We included the items with loadings over 0.35 into the second factor to enhance its internal consistency (working and driving items). This method maximizes reliability of the questionnaire (10). We named the second factor as sensory, mental and cognitive functions, since loadings were maximum for these functions. Eigenvalue of the 2^nd factor was 1.15; this factor can explain 60.58% of the total variance. For the 2^nd factor, Cronbach's α was 0.77 (acceptable value), r_S — 0.42.

The results of variate analysis can depend on the sample size. Therefore, we compared the results in the CHA group (n = 136) with the data of the entire sample (n = 670). In case of samples size over 500, the results of analysis can be interpreted as very good even at low loadings [24]. Variate analysis of the entire sample is shown in Table 4.

Table 4. Varimax-Rotated Factor Matrix of the Neck Disability Index—Russian Language Version in the whole sample (n=670)

Items of the NDI-RU	Factor 1Sensory and Cognitive Functions, Daily and Social Activity	Factor 2Sensory, Cognitive, and Mental functions
3. Lifting	0.808	0.014
7. Working	0.789	0.071
2. Self-care	0.725	0.372
10. Recreation	0.655	0.066
4. Reading	0.645	0.434
8. Driving	0.617	0.514
1. Neck pain intensity	0.612	0.376
6. Concentration	-0.044	0.741
5. Headache	0.172	0.749
9. Sleeping	0.278	0.579

The first factor was named daily and social activity, cognitive and sensory functions. Eigenvalue for the 1^st factor was 4.45; this value can explain 44.53% of the variance. Cronbach α coefficient for the 1^st factor was 0.86; interclass correlation coefficient — 0.49. The second factor was named headache, cognitive and mental function. Eigenvalue for the 2^nd factor was 1.24; it can explain 56.94% of the variance. For the 2^nd factor, Cronbach's α was 0.81; interclass correlation coefficient — 0.40.

Comparison of Tables 3 and 4 revealed that an increased sample size did not fundamentally affect the structure of the questionnaire, although factor loadings were increased for the items “working” and “driving”. Thus, we can conclude that the questionnaire has the same factor structure among patients with CHA as in the entire sample of patients with neck pain.

Minimal detectable changes

According to the first method of SEM calculation (SEM = Sd/√2), this value was 1.84 (2.6/1.41), according to the second method (SEM = s√1–r) — 1.76 (7.2 × √1–0.94). In the first case, MDC value (1.96 × √2 × SEM) was 5.09 (1.96 × √2 × 1.84); in the second case — 4.86 (1.96 × √2 × 1.76). If we will round off the hundredths up to the whole number, MDC will be equal to 5 scores. Thus, if vital activity of a patient with CHA is improved by more than 5 points (for example, after treatment) or impaired by more than 5 points (for example, after painting the ceilings), these changes can be considered true.

Discussion

The purpose of our study was to assess the psychometric properties of the Neck Disability Index-Russian-language version containing the items necessary for patients with CHA. We found adequate psychometric properties (internal consistency, test-retest reliability, competitive and structural validity) of the Russian-language version of the questionnaire. Comparison with our previous results showed that internal consistency and test-retest reliability of the questionnaire are higher in patients with CHA [14]. This can be explained by more homogeneous sample of patients with CHA compared to people with neck pain and associated diseases.

As in our previous study, variate analysis revealed a two-factor structure of the questionnaire with similar factors and small differences in loadings for each item [14]. The first factor reflects impairment of sensory functions, daily and social activity, and the second factor — impairment of cognitive, sensory and mental functions in patients with CHA.

If the 2^nd factor contains only 3 items with high loadings (in our case, headache, concentration, sleeping), it is allowed to include the items with a loading over 0.32 into the second factor. It is necessary to achieve the overall number of items up to 5. Therefore, we included 2 items (driving, working) with a loading over 0.35 into the 2^nd factor and increased internal consistency of the 2^nd factor. This is also justified from a clinical point of view, since headache and impaired concentration can be accompanied by impaired working, deterioration of driving quality and sleep disturbances. Thus, these items can be combined into the same factor.

Only one clinical trial analyzed the psychometric properties of NDI (sensitivity, specificity, sensitivity to changes, minimal detectable changes) in patients with CHA [25]. In our study, MDC value was the same as in the study by Young et al. (5.3 scores). Although calculation methods were different, the time interval between the first and the second measurements was approximately the same (in our case 2-4 days; study by Young et al. — 7 days). Thus, disability changes over 5 points should be considered true in patients with CHA.

Comparison of other results is difficult due to different designs of our studies. Indeed, efficacy of CHA management should be analyzed for such comparison [26]. This study would make it possible to evaluate such properties of the questionnaire as sensitivity, specificity and sensitivity to changes. Moreover, the questionnaire might be regarded as a specific tool for CHA.

Conclusion

The Neck Disability Index — Russian-language version is a reliable tool for assessment of disability of patients with cervicogenic headache.

The authors declare no conflicts of interest.