A new SF-6Dv2 value set based on a Hybrid model using SG, cTTO, and DCE data
Thomas G. Poder,Hosein Ameri
DOI: https://doi.org/10.1016/j.socscimed.2024.117632
IF: 5.379
2024-12-15
Social Science & Medicine
Abstract:Objective To develop a value set for the Short-Form 6-Dimension version 2 (SF-6Dv2) by incorporating societal preferences obtained from three distinct approaches: Standard Gamble (SG), composite Time Trade-Off (cTTO), and Discrete Choice Experiment (DCE). Methods Data were gathered from the general population of Quebec, Canada, using the standardized valuation protocol developed by EuroQol for the cTTO and DCE tasks, as well as the valuation protocol developed by Sheffield University for the SG. The SG and cTTO data were analyzed using OLS, GLS, GLS Tobit, and heteroskedastic Tobit models. Conditional logit model was used for the DCE, while hybrid, hybrid Tobit, and heteroskedastic hybrid were applied to analyze the combined data from SG, cTTO, and DCE. The performance of models was assessed using mean absolute error (MAE), the logical consistency of the parameters, and significance levels. Results Over 56,000 observations collected from the SG, cTTO, and DCE were analyzed. The utility values generated by DCE were generally lower than those provided by cTTO and SG. Among the models tested, the heteroskedastic hybrid model demonstrated the best fit in terms of logical consistency and statistically significant coefficients. This model generated a value set ranging from -0.216 for the worst health state (555655) to 1 for full health (111111), with 0.52% of the values being negative. Among dimensions, the largest decrements were consistently found in the pain dimension, highlighting its significant impact on overall health state valuations. Conclusion A heteroskedastic hybrid model using data from SG, cTTO, and DCE was identified as the most effective approach for generating the SF-6Dv2 value set and is expected to provide key input for healthcare decision-making.
public, environmental & occupational health,social sciences, biomedical