Skip to content

Multiattribute utility theory

Resurrecting Multiattribute Utility Function: Developing a Value Set for Health Utility for Glaucoma. Kevin Kennedy, PhD, Simon Pickard, PhD, Jean-Eric Tarride, PhD, Feng Xie, PhD. VALUE HEALTH. 2023; 26(8):1249–1257

The publication ”Resurrecting Multiattribute Utility Function: Developing a Value Set for Health Utility for Glaucoma” by Kennedy, et al. caught my attention as this provides a clear explanation of the process for valuing health states using multiattribute utility theory (MAUT). MAUT is one of the valuation methods used in the development of a function that predicts utility values for health states described by preference based measures of health. I thought this publication was useful, for a generalist, as this is the first explanation that I have found that outlines the process in a tangible and step wise manner.

The background to this lies in the economic evaluation of health technologies using cost utility analysis (CUA). Results of CUA are based on QALYs, which are calculated using estimates of health state utility. Health state utility values are a 0 to 1 weighting that represents the health related quality of life experienced for a given period of time.

Health state utility values are commonly obtained via the use of generic preference based measures (GPBM), such as EQ-5D-5L or condition specific preference based measures (CSPBM), like the Health Utility for Glaucoma (HUG-5) measure, the focus for the publication. There are other methods that can be used to obtain health state utility values including: mapping, vignettes and direct valuation.

The design of GPBMs/CSPBMs and their descriptive systems mean that there are many more health states that would require valuation than can feasibly be elicited, for example, the EQ-5D-5L can generate over 3000 different health states. To practically obtain utility values for the health states described by these measures, developers will ordinarily value a subset of health states described by the measure and then develop a function that predict values for all the possible health states. 

Values can be elicited via cardinal valuation data using techniques such as standard gamble or time trade off. Other techniques such as discrete choice experiments, that generate ordinal data can also be used; this data is then used to estimate cardinal values. When developing a function based on cardinal valuation data, developers have the choice of using statistical modelling or MAUT. 

In this publication, the authors clearly outline, in figure 1, the design, selection of health states, sequencing of the valuation tasks and provide clear examples; the steps were as follows:

  • Step 1 – Reflective framing exercises
  • Step 2 – Choosing the preferred states – death or the pits state
  • Step 3 – Valuing single attribute health descriptors via a visual analog scale e.g. 21111, 31111, 41111, 51111
  • Step 4 – Standard gamble exercises:
    • A warm up exercise for a single health state
    • Valuation of corner states – where one attribute is changed in each dimension to reflect the most severe problems. E.g. 11151
    • Valuation of random marker states – e.g. mild or advanced disease
    • Valuation of respondents least preferred health states (death or pits) by examining the relationship between preferred health state and the chance of death, perfect health and the pits state.

The relevance of this from an evidence perspective is that, MAUT uses core, choice-based valuation techniques and in this case, standard gamble, which is the classical approach, grounded in economic theory, although potentially more complex for respondents compared with time trade off. MAUT is also grounded in economic theory, although the tangible benefit of this is unclear with conflicting evidence regarding the comparative performance of MAUT versus statistical modelling. From a practical perspective, MAUT is said to require direct valuation of fewer health states which is potentially an advantage for descriptive systems with a large number of possible health states.

In summary, MAUT is one of the valuation methods used in the development of a function that predicts utility values for health states described by preference based measures of health. MAUT and the valuation technique used here are grounded in economic theory and MAUT potentially offers advantages over statistical modelling for larger descriptive systems. This publication, I think as a generalist, is really useful as it provides a clear and tangible example of how valuation using MAUT is applied in practice.

Leave a Reply

Your email address will not be published. Required fields are marked *