Content comparison of the EORTC CAT Core, SF‐36, FACT‐G, and PROMIS role and social functioning measures based on the International Classification of Functioning, Disability and Health

In line with the World Health Organizations' health definition, patient‐reported outcome (PRO) measures frequently cover aspects of social health. Our study aimed to evaluate the role functioning (RF) and social functioning (SF) contents assessed by PRO measures commonly used in cancer patients.


| BACKGROUND
The World Health Organization (WHO) defines health not merely as the absence of disease or infirmity but as a state of physical, mental and social well-being. 1 This broad perspective is also reflected in the WHO definition of health-related quality of life (HRQOL) that refers to the "individual's perception of their position in life […] incorporating in a complex way individuals' physical health, psychological state, level of independence, social relationships, personal beliefs and their relationships to salient features". 2 The social aspect of health including the social roles of individuals is well recognised in both definitions and has also been highlighted in early publications on the use of HRQOL measures for the evaluation of health and treatment outcomes in cancer patients. [3][4][5] For social health a number of conceptual models are available that differ with regard to the concepts included and distinguished. 6,7 Two domains that are commonly assessed in cancer patients are role functioning (RF) and social functioning (SF). 8 RF can be viewed as individual capacity to cope with environmental requirements, as interaction in social systems, or as capacity to cope with activities that are specific to age and social responsibility. 6 SF entails an individual's ability to adequately interact within a social network such as family, friends or working colleagues. 9 Further definitions describe the ability to develop and maintain social relationships as the main aspect of SF. 10 The US Food and Drug Administration (FDA) acknowledges the importance of patients' ability to work and carry out daily activities and has consequently included RF in a recommended core set of patient-reported outcomes (PROs). 11 A recent systematic review showed that RF is sensitive to differences between treatment arms in cancer clinical trials, reflecting the impact of treatment burden and disease control on this domain. 12 The European Medicines Agency (EMA) acknowledges that the assessment of SF might provide important contextual information for primary endpoints in clinical trials, furthermore stating that SF is considered to be of importance to patients. 13 Both the EMA and FDA highlight the importance of using valid and reliable measures when assessing RF and SF. 11,13 For the measurement of RF and SF in cancer patients several multidimensional PRO measures are available, such as the widely used EORTC QLQ-C30, 14 the FACT-G, 15 and the SF-36. 10 In addition, the PROMIS initiative has developed measures of social health including an item bank for the assessment of the ability to participate in social roles and activities. 16 The fairly new EORTC CAT Core 17,18 provides item banks that measure the same concepts as the RF and SF domains of the QLQ-C30. Whilst the availability of different measures may be an advantage when selecting the most appropriate instrument for a specific application, the differences regarding the underlying concepts and frameworks make comparisons of results from different measures challenging.
To facilitate comparison of results from studies that use different PRO measures but assess similar concepts, linking rules and common metrics have been developed. These rely on sophisticated statistical methods 19 to make scores comparable and allow data pooling and meta-analysis of trial data. However, such quantitative analysis need to be complemented by equally important qualitative evaluations allowing for a better understanding of conceptual differences and similarities across PRO measures as reflected by the content of their questions. 20 The International Classification of Functioning, Disability and Health (ICF) 21 provides a structured framework, which has not only been used to develop conceptual models for RF and SF 6,22 but also to compare contents of PRO measures. The standard methodology for such comparisons follows the linking rules by Cieza et al. [23][24][25] that allow categorisation of item content in the ICF framework to investigate overlaps and discrepancies of different PRO measures.
The aim of our ongoing project 26,27 is to evaluate the possibilities to link scores from commonly used PRO measures in cancer research, with a focus on the EORTC CAT Core. 17 This work comprises a qualitative assessment of the content of the various measures to investigate conceptual (dis)similarities, followed by quantitative analyses on the actual linking of scores from these measures. Since such information is key for understanding conceptual differences between PRO measures, the objective of this study was to compare the item content of the RF and SF domains of the following frequently used PRO measures using the ICF framework 28   The PROMIS initiative has developed several PRO measures to assess social function and social relationships, 30 which are intended to assess patients with chronic diseases and are not specific to cancer patients. 31 For this analysis we selected the PROMIS item bank Ability to Participate in Social Roles and Activities 2.0, which was considered most likely to have the largest conceptual overlap with RF and SF. 32 This item bank comprises 35 items assessing limitations related to work, family, friends, and recreational activities. The questions have no specific recall period and are answered on a 5point rating scale with response categories from "Never" to "Always".  Table 1.
To link the item content of the PRO measures under investigation to the respective ICF categories we relied on the methodological approach proposed by Cieza et al. [23][24][25] First, all meaningful concepts of an item are identified. Each of these meaningful concepts are then linked to the corresponding ICF category, that is, one or more specific ICF codes are assigned to the item. Note that an item may be linked to more than one ICF category if more than one meaningful concept is covered by the item content. Cieza et al. [23][24][25] acknowledged that some meaningful concepts cannot be linked to a specific ICF category. They suggest the following coding in these instances: 'not covered' (nc) for item content beyond the ICF coverage (e.g. quality of life) and for content such as 'health condition' (specific diagnosis) 'personal factors' such as age, gender, or race, which are part of the contextual factors but lack specification; and 'not definable' (nd) if an item is too broad for a specific category or lacks adequate information (e.g. general health) For this study, the item content of each HRQOL measure was linked to the respective ICF categories independently by two reviewers, following the described linking procedure. Disagreements between the codings were discussed by the reviewers, and a third reviewer helped to find consensus, if needed. Interrater agreement is provided as total agreement (%) on the second level.
Following this linking process, we descriptively analysed the number of items linked to ICF categories and the number of items not being covered by the ICF for each HRQOL measure under consideration. Furthermore, to facilitate the content comparison across PRO measures, we determined the percentage of meaningful concepts coded in each first-level category separately for each PRO measure. Codings for the item stems (e.g. "Has your physical condition or medical treatment interfered with"…) and the codings "not covered" and "not definable" were not considered for calculating the percentages for comparisons of measures. Please note, that a single item could contain meaningful concepts from more than one ICF category, thus the total number of codings was higher than the total number of items.

| RESULTS
Across the five PRO measures, with altogether 10 scales or item banks under investigation, 85 items were assigned to three compo-

| Clinical implications
The

| CONCLUSION
In conclusion, the results from our analysis highlight conceptual differences between PRO measures for RF and SF and provide insight into the content covered by each measure. Such information on the concepts covered by the PRO measures may help to select the most appropriate measure for a specific application, and may also support item selection when creating static short-forms from item banks 43 or implementing content balancing in computer-adaptive assessments. 44 Furthermore, our results will inform the development of possible linking procedures for score conversion of PRO measures for social and role function.