Summary: Linear regression was applied to data from 275 persons with osteoporosis-related fracture to estimate EQ-5D-US and SF-6D health state values from the Osteoporosis Assessment Questionnaire. The models explained 56% and 58% of the variance in scores, respectively, and root mean square error values (0.096 and 0.085) indicated adequate prediction for use when actual values are unavailable.
Introduction: This study was conducted to provide models that predict EQ-5D-US and SF-6D societal health state values from the Osteoporosis Assessment Questionnaire (OPAQ).
Methods: OPAQ, EQ-5D, and SF-6D data from individuals at two centers with prior osteoporosis-related fracture were used. Fractures were classified by type as hip/hip-like, spine/spine-like, or wrist/wrist-like. Spearman rank correlations between preference-based system (EQ-5D and SF-6D) dimensions and OPAQ subscales were estimated. Linear regression was used to estimate preference-based system health state values based on OPAQ subscales. We assessed models including age, sex, and fracture type and chose the model with the best performance based on the root mean square error (RMSE) estimate.
Results: Among the 275 participants (198 women), with mean age of 68 years (range 50-94), the distribution of fracture types included 10% hip/5% hip-like, 18% spine/11% spine-like, and 24% wrist/18% wrist-like. The final regression model for EQ-5D-US included three OPAQ attributes (physical function, emotional status, and symptoms), predicted 56% of the variance in EQ-5D-US scores, and had a RMSE of 0.096. The final model for SF-6D, which included all four OPAQ dimensions, predicted 58% of the variance in SF-6D scores and had a RMSE of 0.085.
Conclusions: Two models were developed to estimate EQ-5D-US and SF-6D health state values from OPAQ and demonstrated adequate prediction for use when actual values are not available.