Evaluation of the Pain Impact Index for Community‐Dwelling Older Adults Through the Application of Rasch Modelling. Issue 5 (21st January 2021)

• Record Type: Journal Article
• Authors: Gilmartin‐Thomas, Julia F.‐M.; Forbes, Andrew; Liew, Danny; McNeil, John J.; Cicuttini, Flavia M.; Owen, Alice J.; Ernst, Michael E.; Nelson, Mark R.; Lockery, Jessica; Ward, Stephanie A.; Busija, Ljoudmila
• Is Part Of: Pain practice. Volume 21:Issue 5(2021)
• Page Start: 501
• Page End: 512
• Publication Date: 2021-01-21
• Subjects: chronic pain -- geriatric -- index -- psychometrics -- Rasch analysis; Pain -- Treatment -- Periodicals; 616.0472
• DOI: 10.1111/papr.12980
• Languages: English
• ISSNs: 1530-7085

Abstract

Abstract: Objective: Evaluate the Pain Impact Index, a simple, brief, easy‐to‐use, and novel tool to assess the impact of chronic pain in community‐dwelling older adults. Methods: A Rasch modelling analysis was undertaken in Stata using a partial credit model suited to the Likert‐type items that comprised the Index. The Index was evaluated for ordering of category thresholds, unidimensionality, overall fit to the Rasch model, measurement bias (Differential Item Functioning, DIF), targeting, and construct validity. Results: The four‐item Pain Impact Index was self‐completed by 6454 community‐dwelling Australians who were aged at least 70 years and experienced pain on most days. Two items showed evidence of threshold disordering, and this was resolved by collapsing response categories (from 5 to 3) for all items. The rescored Index conformed to the unidimensionality assumption and had satisfactory fit with the Rasch model (analyses conducted on a reduced sample size to mitigate the potential for overpowering: n = 377, P > 0.0125, power > 77%). When considering uniform DIF, the most frequent sources of measurement bias were age, knee pain, and upper back pain. When considering nonuniform DIF, the most frequent source of measurement bias was knee pain. The Index had good ability to differentiate between respondents with different levels of pain impact and had highest measurement precision for respondents located around the average level of pain impact in the study sample. Both convergent and discriminant validity of the Index were supported. Conclusion: The Pain Impact Index showed evidence of unidimensionality, was able to successfully differentiate between levels of pain impact, and had good evidence of construct validity.