Examining the Similarities and Differences of OMERACT Core Sets Using the ICF: First Step Towards an Improved Domain Specification and Development of an Item Pool to Measure Functioning and Health

METHODS

Four OMERACT core sets, psoriatic arthritis (PsA), rheumatoid arthritis (RA), ankylosing spondylitis (AS), and longitudinal observational studies (LOS) in rheumatology^4,5,6,7, were conveniently selected and compared. First, the selected domains for each were compared directly. Next, concepts contained in the instruments that were considered or selected to measure the OMERACT domains were linked to the ICF categories, which served as the external reference for comparison^7,8,9,10. Linking refers to the procedure that follows established rules and aims to match a domain or concept to the most precise ICF category that represents that concept or domain^11,12. Comparability of OMERACT core sets is discussed in view of the difficulties encountered in comparing the domains and concepts.

RESULTS

Table 1 presents the comparison of the “core domains” of the OMERACT core sets for PsA, RA, AS, and LOS in rheumatology^4,5,6,7. First, core sets differed in the way in which proposed domains were organized. The RA core set selected one set of domains that represent the minimal to be included in every clinical study, with one additional domain (radiology) that should be included in studies lasting one year or more⁵. The PsA core set distinguished a minimal number of domains to be included in every clinical study, a number of optional domains that can be of further interest, and a number of domains that need further study before including them⁴. The core set for LOS in rheumatology first selected subdomains that were then grouped into broader domains, some of which are optional⁷. Finally, the AS core set distinguished 3 settings — symptom-modifying antirheumatic drugs (or SMARD) or physiotherapy, disease-modifying antirheumatic drugs, and clinical record-keeping — with increasing number of (obligatory) core domains⁶.

Second, a difference in the number of domains or subdomains was seen. For the core set of PsA, 6 core domains were selected, 8 optional domains, and 5 domains for the research agenda; for RA, 7 (8 for studies of more than 1 year); for LOS, 11 subdomains within 5 broader domains (2 of which are optional); and for AS a total of 9 domains for the 3 settings together.

The names of the domains were surprisingly difficult to compare and could not easily be classified under the proposed “five D” structure of domains. Specifically, the domains in the core set for LOS were difficult to compare with other core sets: the starting point of the core set for LOS considered not only the “five D’s”² but also included the “impairment→disability→handicap model” or the International Classification of Impairments, Disabilities and Handicaps model of the World Health Organization (ICIDH)¹³, which preceded the ICF and aimed to make a distinction between process versus damage. This resulted in names and subdomains that differed substantially from other core sets. Further, the domain “health-related quality of life” (HRQOL) posed some challenges. In the LOS, HRQOL was synonymous with health state, and together with Disease process and Damage, part of the domain “health.” HRQOL (or health status) domains included symptoms, physical function, and psychosocial function. In the PsA core set, HRQOL was selected as a domain in addition to the domains pain, physical function, global assessments, and participation. When taking into account the specifications of the subdomains that were proposed for the (sub)-domains of the LOS core set, then pain, physical function, patient global assessment of disease activity, and joint assessment were domains common to all 4 core sets. Damage (including radiographic damage) and acute-phase reactants were presented in all core sets except for the PsA core sets, where it is part of the research agenda. Fatigue was a separate domain in the AS core set, was mentioned among the domain “symptoms” of the LOS core set, was not included in the RA core set, and was on the research agenda for the PsA Core Set. The core set for LOS was the only one to include psychosocial function as a specific domain (within HRQOL). The core set for psoriatic arthritis was the only one to include the domain “skin.” When considering the “five D” model, it is clear the OMERACT core sets emphasize Discomfort, Disability, and Damage. The LOS was the only one to include Drug toxicity and Death. Costs were included only as an optional domain in the core set for LOS.

When trying to compare the specific concepts of “what to measure” contained in the domains of the different core sets, it was assumed that the content of domains is specified by the concepts included in the items of the selected instruments. However, only for the core set of AS were measures formally selected by consensus¹⁰. Notwithstanding, the PsA core set and the core set for RA, LOS, respectively, considered or proposed measures for domains^7,8,9. For the purpose of the present study, these “considered or proposed” measures were linked to the ICF categories. A separate table that represents the full result of this linking exercise is available from the authors on request.

Linking to the ICF revealed that the majority of concepts addressed in the OMERACT domains were pertinent to pain functions, movement/mobility, self-care, and participation. In addition, categories referring to energy and drive and emotional functions were addressed. To illustrate the advantage of linking, Table 2 presents the comparisons of the constructs included in the physical component of the Medical Outcome Study Short-Form 36 (SF-36)¹⁴ (considered in the domain “physical function” of the OMERACT core set for PsA), the Health Assessment Questionnaire (HAQ)² (considered for the domain “physical function” in the RA and LOS core set, and the Bath AS Functional Index (BASFI) (selected for the domain “function” in the OMERACT core sets for AS). Consistent with the development for RA, the HAQ included more activities of the hands and upper limbs, and the BASFI included ICF categories referring to “maintaining and changing body positions,” which is more specific to patients with back problems. The SF-36 also contained categories on mental function and participation and thus exceeds the “traditional” domain of “physical function.”

DISCUSSION

The comparison of the OMERACT domains and more specific concepts represented within the domains across 4 selected OMERACT core sets proved useful. We learned that “what to measure” is not consistently defined across core sets and lacks specifications. There are 2 findings that we believe are relevant for further discussion within OMERACT.

First, within OMERACT core sets, the domains are grouped and named following different approaches. Although OMERACT suggested the “concept of five dimensions” as the starting point, this is not consistently reflected in the OMERACT core sets. Further, in the core set for LOS, subdomains are grouped in core domains comprising “health,” “process,” “damage,” “toxicity,” and “death.” Specifically, the domains “process” and “damage” represent a different construct, since it is a human interpretation that goes beyond the experience at the level of the patient. On the other hand, in the core set for AS, domains were grouped according to the setting of application [SMARD studies, disease-controlling antirheumatic therapy (DC-ART) studies, or clinical record-keeping]. Finally, in the core set for PsA and LOS, a number of optional domains are selected in addition to the “core domains.” As a consequence, in the core set for PsA, acute-phase reactant and radiographic damage are not in the core, while they are core domains for RA (radiographs in studies longer than 1 year), for AS (in DC-ART studies and clinical record-keeping) and for LOS (radiographs are a subdomain in the domain “damage”). Specifically challenging was the domain “quality of life” (QOL). While QOL is part of the core set for PsA and LOS, the domain covers different content. In the core set for LOS, HRQOL is the umbrella for “health” and includes symptoms as well as physical and psychological functioning. In the core set for PsA, HRQOL is a domain separate from pain and physical function.

Second, to be able to compare the specific concepts that are represented in the domains using the ICF as external reference, it was aimed to use the constructs in the items of selected questionnaires that need to represent the OMERACT domains. Remarkably, only the core set for AS agreed upon core instruments. Therefore, it was decided to also include in the comparison the instruments that were considered or recommended in the development of the OMERACT core sets. Comparison of the concepts in the items using the ICF again proved to be useful, especially for the domain “physical function.” The ICF allowed us to gain insight into which aspects of functioning are “typical and relevant” across all diseases versus those for specific diseases.

Historically, the choice of OMERACT to define “what to measure” in broad domains was a logical choice. First, a detailed framework that represented all possible aspects of health was not available. Second, the OMERACT process wanted, for domain specification, to make use of existing instruments that had proven usefulness and had been extensively validated. The need addressed by OMERACT was primarily to agree on selected instruments, in order to improve harmonization (uniformity) in the world of outcome measurement.

The question therefore remains whether there is a need to revisit the question “what to measure.” In the new structure of OMERACT, the selection and further specification of domains are given careful attention¹⁵. With improved insight into the “concept of health” and the emergence of new approaches to development of instruments, it appears that reexamining the domains and domain specification makes sense. Despite their proven usefulness for progress in rheumatology, the existing measures also have some drawbacks. Item response theory has already been applied to the HAQ (included in the core set for RA and considered for PsA) and BASFI (included in the core set for AS) and showed that although instruments are reasonably unidimensional, neither of them (HAQ or BASFI) are true interval measures and include items that show differential functioning and non-ordered responses^16,17,18,19. Moreover, they lack items that help to distinguish between patients with lower levels of disability (i.e., floor effect). The ICF offers a classification by the WHO that is necessary to describe functioning and health. The ICF framework can be used as the external standard and therefore the starting point of defining “what to measure” for either one of the rheumatological conditions or within a particular setting. Moreover, the ICF could facilitate the management of a pool of items that are clinimetrically valid and could be used in the development of new instruments. For AS, effort is going towards the development of a pool of items covering the categories of the ICF core set for AS^20,21. In the area of the development of item pools, the Patient Reported Outcomes Measurement System (PROMIS) is sophisticated and useful^22,23,24, and it would be a beneficial initiative for OMERACT to look at the active interface between the ICF and PROMIS¹⁵.

Other fertile ground within the OMERACT community is to consider the application of the ICF by way of the so-called ICF core sets. Different from OMERACT core sets, ICF core sets refer to a short list of ICF categories that are relevant to a health condition or health-related event. These are selections of ICF categories that are necessary to classify functioning of an individual using extensive and multiple methodologies. They encompass the perspective of patients, existing research, and health professionals. Such core sets are available for several rheumatologic or musculoskeletal conditions — RA²⁵, AS²¹, osteoporosis²⁶, chronic low back pain²⁷, chronic widespread pain²⁸, and osteoarthritis²⁹. These core sets could be the starting point to specify OMERACT domains.

In addition to the comprehensive core sets, there are current initiatives to develop a statistically derived or combined statistical experts-derived Brief ICF Core Set based on the comprehensive core sets. Although the ICF (Core Sets) can be useful as external reference, we need to recognize that the ICF has limitations, such as the lack of universally agreed operationalization of its categories, and the lack of classification of personal factors. However, these limitations do not outweigh the advantage of having a common conceptual language and classification to measure functioning and disability domains across diseases. The ICF initiative within OMERACT is a starting point to propose for each “domain” the best (set of) concepts (ICF categories) to specify the domain.

In conclusion, we learned, by doing a comparison of some OMERACT core sets, that domain conceptualization and definition vary across diseases and domain specification is insufficient, if not lacking. A common framework to assess outcome and specification of domains could further improve outcome measurement as already addressed in the OMERACT report on domain selection¹⁵. The ICF proved useful in specifying the domains referring to physical function in the OMERACT core sets, but can also provide specification for domains that are not part of OMERACT core domains but are relevant to outcomes assessment.