Related Experiment Videos
Estimating the association between SF-12 responses and EQ-5D utility values by response mapping
Alastair M Gray1, Oliver Rivero-Arias, Philip M Clarke
1Health Economics Research Centre, Department of Public Health, University of Oxford, Old Road Campus, Headington, Oxford, OX3 7LF, UK. alastair.gray@dphpc.ox.ac.uk
Summary
This study explores a new method for mapping health status measures to utility values, offering more descriptive health effect data than traditional regression. While the mean squared error is higher than ordinary least squares, it provides valuable insights for health economics.
Area of Science:
- Health Economics
- Biostatistics
- Psychometrics
Background:
- Health economists require reliable methods to convert health status measures into utility values.
- Current methods predominantly use ordinary least squares (OLS) regression, prompting an investigation into alternative approaches.
- This study focuses on mapping between the SF-12 and EQ-5D generic health status instruments.
Purpose of the Study:
- To evaluate an alternative approach for mapping generic health status measures to health state utilities.
- To compare the proposed method with existing ordinary least squares (OLS) regression techniques.
- To assess the predictive accuracy and descriptive capabilities of the new mapping method.
Main Methods:
- Multinomial logit regression was employed to predict EQ-5D responses using SF-12 data.
- Monte Carlo simulations generated predicted EQ-5D responses, with utility scores (tariffs) subsequently applied.
- The novel approach was validated against direct OLS mapping using large datasets from the US Medical Expenditure Panel Survey and the Health Survey for England.
Main Results:
- The developed method yielded an in-sample and out-of-sample Mean Squared Error (MSE) of 0.03, compared to 0.02 for OLS.
- Mean Absolute Error (MAE) for the new method was comparable to OLS.
- The approach successfully predicted group mean utility scores and distinguished between groups with and without existing illnesses.
Conclusions:
- The proposed mapping method, while exhibiting a higher MSE than OLS, offers richer descriptive data regarding health effect domains.
- Further research involving out-of-sample prediction is recommended to rigorously validate the method's utility.
- The findings suggest a promising alternative for health economists seeking nuanced utility value estimations.