Article
What PROMs measure in Dupuytren’s disease: unrecognised issues unmasked by combining PROMs with the Aachen item banking protocol
Search Medline for
Authors
Published: | February 6, 2020 |
---|
Outline
Text
Objectives/Interrogation: A key assumption of a patient-reported outcome measure (PROM) that delivers a summary score is that the items it comprises all measure the same underlying entity, or construct, i.e. that the PROM is unidimensional.
Two of the most popular PROMs for the study of Dupuytren's disease (DASH and URAM) have been examined previously. The URAM has been found to be unidimensional, whereas the DASH has not.
One issue with such studies is that the analysis has to extract the underlying entities and can only do so if enough of the PROM items relate to each underlying factor.
This study hypothesised that the URAM is not truly unidimensional, but that its multidimensionality would only be unmasked when extra information from DASH answers was added to a relevant model. Such analyses form part of "item banking" where items from multiple PROMs are combined.
Methods: This study involved the secondary analysis of cross-sectional data from 1 and 5-year follow ups of aponeurotomy, fasciectomy and dermofasciectomy procedures performed at 5 UK hospitals. Steps from the Aachen Protocol for item banking were employed, using common person equating, as each patient had completed both the DASH and URAM questionnaires.
Factor analyses were performed using principal axis factoring and promax rotation, and with the number of factors extracted in each analysis based on standard heuristics (scree plots, parallel analysis and Eigenvalue>1).
Separate analyses were run for URAM items and DASH items, and then all items from both were analysed together in a single model.
Results and Conclusions: 433 patients completed DASH and 252 patients completed URAM as well. As in previously published results, the URAM appeared unidimensional when studied in isolation, and the DASH did not - the "symptom" items loaded separately from the "task" items.
However, in the combined model the URAM was no longer unidimensional with questions "can you hold a bottle?" and "can you pick up small objects with your thumb and index finger?" loading alongside DASH items, and not with the rest of the URAM items.
This suggests that these two URAM questions might measure a different construct from the others, but because only two questions are affected, this may not be detected when the URAM items are studied alone.
This study suggests that the URAM may be less valid than previously thought. Consideration should be paid to the design of analyses of PROM validity to ensure that issues affecting small numbers of items are still identified.