We compared the item content of ten scales and item banks assessing RF and SF from five PRO measures commonly used in cancer patients, the EORTC CAT Core, the EORTC QLQ-C30, the FACT-G, the SF-36 and the PROMIS measure assessing the Ability to Participate in Social Roles and Activities. Relying on the ICF framework, we found that a large majority of items were linked to ICF categories within the component ‘d - Activities and participation’, mostly covering ‘d6 Domestic life’, ‘d7 Interpersonal interactions and relationships’, ‘d8 Major life areas’ and ‘d9 Community, social and civic life’.
Comparing the item content coded in the ICF categories, we found that the EORTC CAT Core RF item bank consisted mostly of the content in the first-level category ‘d6 – Domestic Life” and ‘d9 – Community, social and civic life”. The FACT-G Functional Well-being scale covered mostly ‘b1 Mental functions”, ‘d6 Domestic Life’ and ‘d8 Major life areas’. For the SF-36 Role-Emotional scale ‘d2 General Tasks and Demands’ was the most common first-level category, followed by ‘d8 Major Life Areas’. The SF-36 Role-Physical scale consisted as wellof content from ‘d2 General Tasks and Demands’ and ‘d8 Major Life Areas’. The PROMIS Ability to Participate in Social Roles and Activities covered predominantly ‘d7 Interpersonal interactions and relationships’, ‘d9 – Community, social and civic life”, and ‘d6 Domestic Life’.
For the Social Functioning measures, we had similar heterogeneous observations. Whereas the FACT-G Social/Family Well-being scale consisted mainly of content in ‘d7 Interpersonal interactions and relationships’, ‘e3 Support and Relationships”, and ‘e4 Attitudes”. The EORTC CAT Core SF item bank focused mainly on ‘d7 Interpersonal interactions and relationships’ and ‘b1 Mental Functions’, and the SF-36 SF scale covered, aspects within ‘d7 Interpersonal interactions and relationships’ and ‘d9 Community, social and civic life’. For the EORTC measures, our analysis provides further evidence for the conceptual distinctness of RF and SF. The EORTC CAT Core SF item bank was found to cover primarily ‘d7 Interpersonal interactions and relationships’, while the RF item bank covered mostly ‘d6 Domestic life’, ‘d9 – Community, social and civic life’, and ‘d8 Major life areas’. In more detail, the most frequent second-level categories for SF were ‘d670 Family relationships’ and ‘d750 Informal social relationships’, while for RF this was ‘d920 Recreation and leisure’ and ‘d640 Doing housework’. There was (some) content overlap between the QLQ-C30 RF and SF scales, but between the corresponding EORTC CAT Core item banks, this overlap was less pronounced Moreover, quantitative analyses of scale structure have confirmed the distinctness of the two concepts in the QLQ-C30 [32, 33].
The PROMIS Ability to Participate in Social Roles and Activities item bank was found to cover ‘d7 Interpersonal interactions and relationships”, ‘d6 Domestic Life" and ‘d9 Community, social and civic life’ to a similar extent, with the most frequent second-level categories being ‘d920 Recreation and leisure’, ‘d640 Doing housework’, ‘d750 Informal social relationships’, ‘d760 Family relationships’, and ‘d859 Work and employment, other specified and unspecified’. The combination of these categories in a single item bank shows the broader perspective on social health taken by PROMIS[16] which distinguishes social function and social relationships, but not RF and SF as the EORTC measures do. In the PROMIS model, the content of the Ability to Participate in Social Roles and Activities item bank is considered to cover SF specifically.
For the FACT-G our results corroborate previous findings [34, 35] that highlighted the conceptual differences between the FACT-G and EORTC approach to measuring social health. The Functional Well-being scale was found to include quite heterogenous content ranging from ‘b Body Functions" (sleep and emotional function) to ‘d Activities and participation’. The FACT-G Social/Family Well-being scale was the only scale included in this analysis that also covered environmental factors (besides one coding for the EORTC CAT RF scale), i.e. ‘e3 Support and Relationships’ and ‘e4 Attitudes’ (of people other than the patient). This finding supports the previous recommendation by Luckett et al. [36] to use the QLQ-C30 if the research focus is on social activities and to rely on the FACT-G if relationships and support are of interest.
In contrast to the broader PROMIS approach to measuring social health, the SF-36 not only distinguishes social and role functioning, but also further separates a role-emotional and role-physical domain. Compared to the EORTC and the PROMIS measures, the SF-36 role domains were found to focus more on ‘d2 General tasks and demands’ and ‘d8 Major Life areas’, whereas the SF-36 social function scale, with its focus on ‘d7 Interpersonal interactions and relationships’ and ‘d9 Community, social, and civic life’, was more congruent with the PROMIS item bank and the EORTC SF domain.
While the ICF framework has been used frequently in the literature for the analysis of item content of PRO measures, our study presents the first such analysis for the EORTC CAT Core item banks for RF and SF and provides a content comparison against other common PRO measures used in cancer patients. For the PROMIS item banks, Tucker et al. [37] provided a comprehensive analysis mapping the PROMIS measures on ICF categories, and their results largely align with ours. The comparability with other content analyses of PRO measures in cancer patients is limited, because results were usually not reported per subscale, but only for the overall questionnaires [38, 39].
Our study clearly confirms that the differences in content assessed by PRO measures reflect the heterogeneity of the underlying concepts of social health. This heterogeneity has also been discussed by Hahn et al. [16] who related it to the fact that social health has been less a focus of outcome measurement than the physical and emotional domains. These authors also emphasised that not all social factors are necessarily outcomes. Depending on the type of intervention being evaluated, a concept such as social support may provide contextual information or may be considered as effect modifier rather than serving as an informative outcome parameter. However, SF assessed with the EORTC QLQ-C30 has been shown to differ frequently between treatment arms in cancer trials [12], suggesting that aspects of social health can be important outcome parameters in clinical trials.
Given the diversity of social health models and measures we argue that a detailed analysis of scale (and item bank) content is especially important to better understand what is being measured and how different PRO measures may be different or similar. It is important to bear in mind though that our comparison relies on the social health model of the ICF framework, and its hierarchical structure determines how content is mapped. Applying other models may lead to conclusions regarding (dis)similarities of PROs that are not fully consistent with our analysis. A further limitation of our study is that the selection of PRO measures included in our study was based on their common use in cancer research but did not follow a more systematic approach for identifying a larger number of such measures. In particular, PROMIS offers a number of additional measures for other social health aspects that were not included in this analysis (e.g. PROMIS item banks for Emotional Support, Satisfaction with Social Roles and Activities, or Social Isolation [28]). Furthermore, we would like to highlight that our content analysis does not provide information about the content validity of the measures in the cancer patient population, which would require a different methodological approach [40].
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 [41] or implementing content balancing in computer-adaptive assessments [42]. Furthermore, our results will inform the development of possible linking procedures for score conversion of PRO measures for social and role function.