Published in:
Open Access 01-12-2013 | Research
Mapping health outcome measures from a stroke registry to EQ-5D weights
Published in: Health and Quality of Life Outcomes | Issue 1/2013
Login to get accessAbstract
Purpose
To map health outcome related variables from a national register, not part of any validated instrument, with EQ-5D weights among stroke patients.
Methods
We used two cross-sectional data sets including patient characteristics, outcome variables and EQ-5D weights from the national Swedish stroke register. Three regression techniques were used on the estimation set (n = 272): ordinary least squares (OLS), Tobit, and censored least absolute deviation (CLAD). The regression coefficients for “dressing“, “toileting“, “mobility”, “mood”, “general health” and “proxy-responders” were applied to the validation set (n = 272), and the performance was analysed with mean absolute error (MAE) and mean square error (MSE).
Results
The number of statistically significant coefficients varied by model, but all models generated consistent coefficients in terms of sign. Mean utility was underestimated in all models (least in OLS) and with lower variation (least in OLS) compared to the observed. The maximum attainable EQ-5D weight ranged from 0.90 (OLS) to 1.00 (Tobit and CLAD). Health states with utility weights <0.5 had greater errors than those with weights ≥0.5 (P < 0.01).
Conclusion
This study indicates that it is possible to map non-validated health outcome measures from a stroke register into preference-based utilities to study the development of stroke care over time, and to compare with other conditions in terms of utility.