Bayesian programme

A new excel programme is now available from the University of Sheffield to convert SF-36 data into the SF-6D utility score estimated using a set of non-parametric Bayesian preference weights. This approach has been shown to perform better in terms of predictive ability (mean absolute error of 0.089 compared to 0.104 on out of sample predictions) and overcomes the bias of the original regression models of under predicting the worst health states (e.g. it predicts a value of 0.203 for the worst SF-6D state compared to 0.301 using the original algorithm). The overall impact on mean health state values was between 0.01 and 0.04 across 4 data sets.

It is recommended that researchers use this algorithm in future work, but use the original algorithm for comparability.

How can the Bayesian programmes for the SF-6D be obtained?

To obtain the computer programs please complete the on-line user licence registration form using the links below or the links on the left-hand side of the page. We will then contact you about your request.

• A licence is available free of charge for all non-commercial applications including work funded by research councils, Government agencies and charities.
• For commercial applications there will be a per study licence (e.g. clinical trial), though an open license for a fixed period is available.

References

Kharroubi SA, Brazier JE, Roberts J, O´Hagan A. Modelling SF-6D health state preference data using a nonparametric Bayesian method. Journal of Health Economics 2007; 26:597-612

Kharroubi S, O'Hagan A, Brazier J. Estimating utilities from individual health preference data: a nonparametric Bayesian method. Applied Statistics 2005; 54:879-895