Mapping utility scores from the Barthel index.

Area of Science:

• Health Economics
• Outcome Measurement
• Geriatric Care

Background:

• Economic evaluations in older populations often require utility-based outcome data (e.g., EQ-5D).
• Collecting such data can be challenging, necessitating alternative approaches when only non-utility measures are available.
• The Barthel index is a commonly used non-utility outcome measure in healthcare settings.

Purpose of the Study:

• To investigate the feasibility of mapping the EQ-5D from the Barthel index.
• To assess the predictive accuracy of different regression models for this mapping.

Main Methods:

• Utilized data from 1,189 UK intermediate care patients.
• Employed Ordinary Least Squares (OLS), censored least absolute deviations (CLAD), and multinomial logistic (ML) models.
• Evaluated predictive accuracy using Mean Absolute Error (MAE) and Root-Mean-Squared Error (RMSE), with validation on admission and discharge data.

Main Results:

• Models using EQ-5D as a continuous dependent variable and Barthel dimensions as predictors performed best.
• The CLAD model demonstrated superior performance based on MAE, while OLS was better for RMSE.
• CLAD predictions more closely matched observed EQ-5D values compared to OLS.

Conclusions:

• Predicting EQ-5D scores from the Barthel index is feasible using regression techniques.
• The CLAD model is recommended for its accuracy in mapping EQ-5D from the Barthel index.
• This approach can aid economic evaluations when direct EQ-5D data is unavailable.