Presentation (CTI)

OBJECTIVES: Previous studies used the scaling factor model to estimate bolt-on values, fixing the effects of EQ-5D-5L dimensions and estimating only the bolt-on. While this model outperformed conventional models, prior research included only one bolt-on and did not apply a published value set. This study assessed the scaling factor model in valuing psoriasis bolt-on health states with two additional items, using the published Chinese EQ-5D-5L value set.
METHODS: We selected 58 EQ-5D-5L health states with psoriasis attributes using an orthogonal array, divided into four blocks (15 states each). General population samples were recruited via quota sampling from five Chinese cities and interviewed using the cTTO method per EQ-VT protocol. Core EQ-5D-5L coefficients were fixed using both the published (2011) and latest (2023) Chinese value sets. Three models were tested: one conventional additive model and two scaling factor models. Performance was assessed using mean absolute error (MAE) between predicted and observed values.
RESULTS: From January to March 2024, 401 participants completed interviews. The conventional model yielded an MAE of 0.061. The scaling factor model showed improved performance, with MAEs of 0.060 (published value set) and 0.045 (new value set). Coefficients for level 5 of itchiness and confidence were 0.222 and 0.100, respectively. The addition of bolt-on items altered the rank order of core EQ-5D-5L dimensions.
CONCLUSIONS: The scaling factor model consistently outperformed the conventional model, particularly when using the latest EQ-5D-5L value set. Shifts in dimension ranking suggest perceived interactions between bolt-ons and core dimensions, which may challenge assumptions of independence.