PURPOSE: It is not always possible to collect utility-based outcome data, like EQ-5D, needed for conducting economic evaluations in populations of older people. Sometimes, information on other non-utility outcome measures may have been collected. This paper examines the possibility of mapping the EQ-5D from a non-utility-based outcome, the Barthel index. METHODS: Data for 1,189 UK intermediate care patients were used. Ordinary least squares (OLS), censored least absolute deviations (CLAD) estimator and multinomial logistic (ML) models were used. The mean absolute error (MAE) and root-mean-squared error (RMSE) were used to estimate the predictive accuracy of eight regression models. Validation of primary models was carried out on random samples of data collected at admission and discharge. RESULTS: Models where the EQ-5D was entered as a continuous dependent variable and Barthel dimensions used as explanatory variables performed better. CLAD performed best on MAE and OLS on the RMSE, while the ML performed the worst on both measures. The CLAD predicted EQ-5D scores that matched the observed values more closely than the OLS. CONCLUSIONS: It is possible to reasonably predict that the EQ-5D from the Barthel using regression methods and the CLAD model (4) is recommended.
PURPOSE: It is not always possible to collect utility-based outcome data, like EQ-5D, needed for conducting economic evaluations in populations of older people. Sometimes, information on other non-utility outcome measures may have been collected. This paper examines the possibility of mapping the EQ-5D from a non-utility-based outcome, the Barthel index. METHODS: Data for 1,189 UK intermediate care patients were used. Ordinary least squares (OLS), censored least absolute deviations (CLAD) estimator and multinomial logistic (ML) models were used. The mean absolute error (MAE) and root-mean-squared error (RMSE) were used to estimate the predictive accuracy of eight regression models. Validation of primary models was carried out on random samples of data collected at admission and discharge. RESULTS: Models where the EQ-5D was entered as a continuous dependent variable and Barthel dimensions used as explanatory variables performed better. CLAD performed best on MAE and OLS on the RMSE, while the ML performed the worst on both measures. The CLAD predicted EQ-5D scores that matched the observed values more closely than the OLS. CONCLUSIONS: It is possible to reasonably predict that the EQ-5D from the Barthel using regression methods and the CLAD model (4) is recommended.
Authors: Nalin Payakachat; Kent H Summers; Andreas M Pleil; Matthew M Murawski; Joseph Thomas; Kristofer Jennings; James G Anderson Journal: Qual Life Res Date: 2009-06-19 Impact factor: 4.147
Authors: I-Chan Huang; Constantine Frangakis; Mark J Atkinson; Richard J Willke; Walter L Leite; W Bruce Vogel; Albert W Wu Journal: Health Serv Res Date: 2008-02 Impact factor: 3.402
Authors: Stavros Petrou; Oliver Rivero-Arias; Helen Dakin; Louise Longworth; Mark Oppe; Robert Froud; Alastair Gray Journal: Pharmacoeconomics Date: 2015-10 Impact factor: 4.981