OBJECTIVES: This study experimentally evaluated the valuation methods of the EQ-5D-5L bolt-ons, scaling factor model and conventional approach, by using three cognitive dimensions, the cognition bolt-on (hereafter, EQ-5D-5L+C) and additional two previously experimentally developed cognitive dimensions, added to the EQ-5D-5L.

METHODS: Adults (≥18 years) living in Tokyo, Osaka, and Kumamoto were recruited from Japanese general population using snowball sampling. Face-to-face interviews were conducted in June-August 2023. Consenting participants were randomized to Arm 1 (EQ-5D-5L+C), Arm 2 (EQ-5D-5L+Remembering things, previously developed), and Arm 3 (EQ-5D-5L+Thinking clearly, previously developed). Preferences were collected using composite time trade-off (cTTO). Each participant valued a block of 10 health states and the worst possible health state (55555). The cTTO data were modeled for each arm independently, using the scaling factor model, with EQ-5D-5L disutility weights estimated from the existing Japanese value sets. We also fitted the cTTO data into the 24-parameter conventional main-effects model. Both models used a Tobit model and maximum likelihood estimation, and the model performance was assessed using indices of fit, including mean absolute errors (MAE). Lower MAE values indicate higher prediction accuracy.

RESULTS: In total, 864 Japanese participated in this study. After adding each cognitive dimension to the existing EQ-5D-5L, the estimated coefficients for the scaling factor model expanded in all arms. Mean QOL values of the scaling factor model were slightly discrepant from the predicted values, but large discrepancies were not observed in all three arms. The MAE of the scaling factor model and conventional model was 0.30031 and 0.29568, 0.27950 and 0.27297, and 0.27153 and 0.26403 in Arms 1, 2, and 3, respectively.

CONCLUSIONS: The prediction accuracy of the two models did not appear greatly different, and that of the three bolt-ons was comparable. Further experimental studies on valuation using the recently proposed approach and cognition dimensions are warranted.