130 Chapter 5 Construct validity Construct validity was assessed by examining the Spearman correlations between the PROMIS CATs and the SF-36v2 using a multitrait-multimethod (MTMM) analysis. An MTMM analysis can be used to evaluate the convergent and divergent validity when multiple PROs are measured with two or more PROMs. Using the MTMM, we expect to find the highest correlation in the situation when the same PRO is measured (possible multiple times) with the same PROM, followed by when the same PRO is measured with a different PROM, followed by the situation in which both the PRO and PROM differ. In case the same PRO is not measured multiple times with the same PROM, the reliability coefficient can be used as an estimation of the test-retest correlation (e.g., when the same PRO is measured multiple times with the same PROM) 44,45. In this study, we use the MTMM to evaluate the construct validity of the PROMIS CATs by comparing them to certain SF-36v2 domains. While the PROMIS item banks and SF-36v2 domains do not measure the exact same PRO, the PROsetta Stone project and related previous research has shown that the PROMIS CATs can be linked to certain SF-36v2 domains 46,47. We therefore assessed the construct validity of the PROMIS CAT physical function, pain interference, fatigue, anxiety, depression, and ability to participate in social roles and activities by comparing them with the SF-36v2 domains physical functioning, bodily pain, vitality, mental health (for both anxiety and depression) and social functioning, respectively. According to MTMM, construct validity is supported when 1) the correlation between same PRO different PROM (SPDP) (e.g., CAT Physical Function and SF-36v2 physical functioning) is > 0.5, 2) the reliability coefficient is larger than the correlation between SPDP, 3) the correlation between SPDP is larger than the correlation between different PRO same PROM (DPSP) (e.g., CAT Physical Function and CAT Pain Interference), and 4) the correlation between SPDP is larger than the correlation between different PRO different PROM (DPDP) (e.g. CAT Physical Function and SF-36 pain interference) 44,45. If condition 3 or 4 are not met, construct validity can still be supported if a minimum of 75% of the correlations between DPSP or DPDP are smaller than the correlation between SPDP 48.
RkJQdWJsaXNoZXIy MTk4NDMw