PSYCHOMETRIC PROPERTIES OF NECK DISABILITY INDEX – A SYSTEMATIC REVIEW AND META- ANALYSIS

Abstract

Purpose

To evaluate the data on the psychometric properties of the Neck Disability Index (NDI).

Materials and methods

Medline, Embase, PsychINFO, Web of Science, and Scopus were searched in April 2023.The records were screened independently and the risk of systematic bias was assessed. The random effects meta- analysis was conducted when possible.

Results

The search resulted in 492 records. Of them, 79 records were included into analysis. The were 34 different translations and the sample sizes varied from 17 to 2,070 respondents. Of the samples, 76% were predominated by women. The average age varied from 26 to 65 years. Of the studies 61% have been conducted among patients with neck pain lasted over three months and without red flags. Overall, 70 studies were considered to be of low risk of systematic bias. Internal consistency has been assessed by 34 studies, and alpha was over 0.81 in 33 studies. The intraclass correlation coefficient (ICC) have been reported by 38 studies with pooled mean of 0.91 (95% CI 0.90 to 0.93). The test-retest correlation coefficients varied from 0.56 to 0.98. The convergent validity of the NDI has been tested against different scales. Of them, the most frequently used scales were pain numeric rating scale (NRS) or pain visual analogue scale (VAS). Their convergent validity varied from 0.38 to 0.89. Exploratory factor analysis has been employed by 28 studies. Of them, 13 have found the NDI to be unidimensional, 14 – two-dimensional and one study has reported on a three-dimensional structure. The use of confirmatory factor analysis was rare and the results were inconsistent. Every third study has calculated a minimal detectable change (MDC). The minimal detectable change varied from 3% to 27% and minimal clinically important difference (MCID) between 5% and 33%. The meta-analysis was conducted when numerical data for mean estimates and their variances were available. This way, pooled test-retest intraclass correlation and Area Under the Curve were estimated. Clinical relevancy has mostly been evaluated by reporting Area Under the Curve (AUC). The mean pooled (12 studies) estimate was 0.74 (95% CI of 0.68 to 0.80). Every third study has reported floor and/or ceiling effects and 25 out of 26 studies have not seen neither floor nor ceiling effect. or Rasch analyses have been rare. Item Response Theory has been employed by two studies: sex-related differential item functioning has been present in one study and absent in another. The difficulty and discrimination of the NDI items were good. Rasch analysis has been used by one study stating that the properties of the item ‘headaches’ differed significantly from the other items.

Conclusions

The NDI demonstrated good internal consistency and test-retest reliability without floor or ceiling effect. In most of the situations, the NDI could be considered a unidimensional scale. The NDI correlated especially well with such scales as Neck Pain and Disability scale (NPAD) and Patient- Reported Outcomes Measurement Information System (PROMIS). The minimal clinically important difference (MCID) and minimal detectable change (MDC) were around 15% (7.5/50 points). The NDI was able to discriminate people who have improved from those who have not. The NDI score may slightly depend on the sex.