| To Register SMDM Homepage | |
|
|
|
||||
Purpose: The purpose of this study was to examine the convergent and divergent validity of three utility elicitation methods used commonly in medical decision making- the standard gamble, time tradeoff and visual analog scale.
Method: A multitrait multimethod (MTMM) matrix was used for this validation exercise. Three health states- blindness, AIDS and stroke- were used for the MTMM exercise. In addition, the relationship between specific individual differences (construct-irrelevant) variables and utility responses was tested using a structural model.
101 subjects provided utilities using the iMPACT3 protocol. Data were also collected about individual differences such as numeracy, time preference and risk preference using several attitudinal instruments.
Results: The MTMM analysis showed the methods achieving convergent validity. In other words, the different methods converged on a given health state. However, the methods failed to achieve divergent validity, i.e., given different health states, methods were unable to discriminate between the health states.
This research was the first to study construct-irrelevant variables- such as numeracy, experience with illness and risk preference (including several types of risk)- cited in the literature as likely to affect responses to utility elicitation methods. A linear structural model was built. The model achieved a good fit and showed that physical risk was positively associated with Standard Gamble utilities. The model also showed that professional risk was negatively associated with Standard Gamble and Time Tradeoff utilities.
Conclusion: This is the first empirical study to examine the convergent and divergent validity of utility elicitation methods and it finds that the methods fail to achieve divergent validity. The adequacy of convergence and the lack of divergence achieved by the methods points to two important conclusions. Firstly, that the debate as to the usefulness of these methods has been unsettled perhaps, because of this duality in the nature of these methods. Secondly, that method variance is a predominant characteristic of utility elicitation further strengthening the notion that utilities might be constructed rather than elicited by these methods.
This is also the first study to examine the relationship between construct-irrelevant variables and utility responses and it finds that specific sub-types of risk are related with utility responses on the standard gamble and the time tradeoff instruments. This research highlights the need to study these sub-types of risk more closely.
See more of Oral Concurrent Session B - Clinical Effectiveness and Quality of Life
See more of The 26th Annual Meeting of the Society for Medical Decision Making (October 17-20, 2004)