BackgroundThe purpose of this secondary analysis of clinical databases of people with neck pain was to use a mixed unique conceptual and statistical approach to develop a brief version of the Neck Disability Index (NDI).MethodsAn a priori framework of neck-related function based on the International Classification of Functioning, Disability and Health was used to identify items from the original 10-item NDI that do not conceptually fit. Remaining items were subject to Rasch analysis to identify items that did not statistically fit with axioms of quantitative measurement. Finally, approaches drawn from classical test theory were used to compare stability, responsiveness and concurrent validity of the original NDI, the new brief NDI and the linearly-transformed brief NDI.ResultsConceptual analysis identified 3 items that did not fit with the construct of self-reported ability to perform activity: pain intensity, headache, and sleeping. These items were removed, and responses to the remaining 7 items drawn from an assembled database of 316 physiotherapy patients with neck pain were subject to Rasch analysis. Two items were removed due to either considerable differential item functioning (reading) or statistical redundancy (lifting). The remaining items were considered the NDI-5. Test-retest reliability, responsiveness, sensitivity to change, and concurrent validity were all comparable across the original NDI, NDI-5 and linearly-transformed NDI-5. Sensitivity to change over a 1-month period of physiotherapy was the notable exception, where the linearly-transformed NDI-5 showed superiority over the other two forms.ConclusionsA shortened version of the NDI, the NDI-5, has been constructed that is conceptually and statistically sound. Implications for research and clinical practice are discussed. Comparison with the NDI-8 is provided that suggests overall similar function across the forms, although the latter may be more sensitive to change.