This letter highlights several issues in the literature regarding the metric properties of the Compliance Questionnaire on Rheumatology (CQR),1 which demonstrate a strong essential weakness of the studies that use them, together with conclusions extracted from them. The CQR is possibly one of the adherence measures that is being incorporated into most frequently into studies. In a recent bibliographic review,2 5 out of the 9 selected studies used it. Furthermore, recommendation 16 of the Spanish Rheumatology Society,3 suggests developing validation studies for treatment adherence measures with methotrexate (MTX), and shows that the CQR is the only validated method.
In the CQR4–9 validation literature it may be determined that highly frequent types of specific evidence validity refer to patient classification, and the convergent and divergent relationship with other criteria. However, the less used, more ignored method is that of internal structure validity – i.e. the scalability of the CQR which analyses the item-score relationships to define the configuration of the total scoring. Only 2 replication studies have researched their internal structure,6,10 and in them the single dimensionality of the CQE was not corroborated. In other words, the 19 items were not grouped into a single statistical dimension.
This implies 3 things: firstly, that there was no empirical justification to use a single score derived from the simple or weighted sum of the items. Second, some equivalence problems of the CQR compared with other similar measures may be caused by this problem of scalability (see Marras et al.4). Third, the situation seems to describe validity induction,11 which occurs when the validity affirmation is supported by a tool using 1) studies from other cultural contexts (for example, citing the original study1 or another made in another country, with a different type of participant, and/or clinical situation), 2) selecting the type of irrelevant evidence (for example, declaring that Cronbach’s alpha reliability indicates the validity of the scoring) or 3) exchanging specific validity evidence (for example, inferring that the predictive capacity of the CQR is proof of satisfaction with treatment or with the validity in general.
For the latter implication, although it has been recognised that the CQR obtains appropriate levels of prediction in studies with patients who suffer from arthritis and has acceptable levels of reliability, this evidence is not exchangeable with the evidence required to demonstrate that the total score may be interpreted as a unit. In any described situation, the evidence of inducing validity is the omission of analysing the CQR scalability to uphold the interpretation of a single score.
How can the use of a single score be justified if the empirical evidence that must uphold it is missing or questionable? Can the studies which question the single dimensionality of the CQR6,10 be taken into account? The strongest outcomes to date on the internal structure or scalability of the CQR do not guarantee that a single score be used, unless they reduce the number of items10; another option of interpretation is to accept the multidimensionality of the CQR,6 and that the rheumatologist must assess whether they wish to use it in this way or not.
We conclude that, due to the apparent inconsistency in the factorial CQR structure, studies with the CQR must include corroboration of their internal structure, take advantage of current validation recommendations,12 and highlight any international collaborations within them.
Please cite this article as: Merino-Soto C, Angulo-Ramos M. Estudios métricos del Compliance Questionnaire on Rheumatology (CQR): ¿un caso de inducción de la validez? Reumatol Clín. 2022;18:497–498.