An Introduction to Confirmatory Composite Analysis for Patient-Reported Outcomes with Illustrative Analysis

Objectives
To introduce confirmatory composite analysis (CCA) using the refined Henseler–Ogasawara (H-O) specification and illustrate its application to patient‑reported outcomes (PROs) by analysing the Tinnitus Functional Index (TFI).

Methods
CCA was applied to data from 200 adults with tinnitus. Four CCA models were estimated with a robust maximum‑likelihood estimator: (1) eight correlated TFI domains, (2) those domains plus Wellbeing modelled as latent variable, (3) those domains with Executive Functioning modelled as latent variable, and (4) those domains with both Wellbeing and Executive Functioning modelled as latent variables simultaneously. Model identification followed the refined H-O specification; global fit, (standardised) composite loadings, (standardised) weights, and criterion validity were subsequently examined.

Results
All models demonstrated good to acceptable fit (Model 1 RMSEA = 0.077, CFI = 0.959, SRMR = 0.046). All 25 standardised composite loading and weight estimates were stable across models. Some TFI domains (e.g., Intrusive) displayed heterogeneity of weight estimates, suggesting that items within this domain contribute unevenly to the resultant emergent variable. Correlations between the TFI domains modelled as emergent variables and Wellbeing ranged from r = 0.621 (Emotional) to 0.317 (Auditory); correlations between the TFI domains and Executive Functioning ranged from r = -0.355 (Emotional) to −0.214 (Sleep, Auditory).

Conclusion
CCA provides a rigorous SEM‑based approach for evaluating PROs that behave as emergent variables and enables the evaluation of structural validity for composite models, as well the ability to derive weights for their scores.