OBJECTIVE: DisQ-24 is a 14-item questionnaire with a 6-point Likert-type scale that measures disability in the 24 hours following onset of headache pain. Previous analysis on data from transformed migraine sufferers suggested item iterations may improve psychometric properties. The objective of this study was to evaluate the Disq-24 using classical test theory (CTT) and item response theory (IRT) methods to provide recommendation on further revisions. METHODS: Disq-24 data from 647 subjects who participated in two clinical trials of acute migraine treatment were analyzed. An exploratory factor analysis (EFA) using maximum likelihood estimation and oblique rotation was performed to determine factor structure. Based on results from EFA, CTT and Likert scaling assumptions of equal means, variances and weights, internal consistency and item discriminant validity were examined using the Multi-trait/Multi-item analysis software (MAP-R). Item-characteristic curves, item discrimination ('a') and difficulty parameters('b') based on the Samejima's graded IRT model were estimated using Multilog software. RESULTS: EFA suggested a 2-factor structure but communality estimates for Item 8 (0.41) and Item 10 (0.46) suggested a high proportion of variance in these items is unexplained in the common factors. MAP-R analysis showed most items met CTT assumptions, except items 8 and 10 had lower means (0.87 and 1.17 respectively) and weights (1.7 and 1.8 respectively) as compared to other items within the same subscale. All items exhibited high item-scale correlations (0.61-0.89) and high internal consistency (>0.9), which may suggest item redundancy. IRT analysis showed most items provided high discriminatory power (range 1.8-2.9) in differentiating varying levels of disability and measured disability over a broad range with no floor/ceiling effects, except items 8 (a=1.09) and 10 (a=1.10) had low discrimination and contributed little to the test information function. CONCLUSIONS: CTT and IRT methods provided similar results that two items may be deleted or revised to improve psychometric properties. Further development may consider deleting items with low discriminatory power to reduce redundancy. Additional analysis is needed to confirm improvement of data fit after revisions.