Recently, I’ve witnessed an ever-increasing number of condition-specific utility measures being developed. Much work is being carried out by The University of Sheffield on measures such as the AQL-5D for asthma, and the EORTC-8D for cancer. I struggle to see their use.
Why do they exist?
It’s fair to say that generic preference-based measures of health aren’t always appropriate. For this reason condition-specific measures are used, which might be more sensitive or responsive to different health states within this condition. The problem arises when we want to compare interventions using these measures, which are not comparable with each other. Many have therefore elicited preference weights from the health states defined by these measures to allow the generation of a utility value…
Apples and oranges?
But surely these values are not comparable with QALYs generated using measures such as the EQ-5D or SF-6D, so what’s the point? To be comparable, condition-specific preference-based utilities must represent the same thing as generic utilities, but how can this be? It seems to me that they represent asthma-, cancer- or dementia-related quality of life, not health generally. For this reason I do not see their purpose. They may be of interest at a condition-specific level, but in this case it would surely not be meaningful to elicit preferences from a general population.
Condition-specific measures often need to be compared alongside generic measures, this is certain. However, I fail to see the benefit of eliciting preferences for condition-specific measures over the alternative method of mapping values from a condition-specific measure to a generic measure. It seems to me that this movement risks a regression back to a time when interventions for two different health concerns were not comparable in terms of cost-effectiveness.
Am I missing the point of such measures? Have you used them? What are the advantages and disadvantages of these measures?
I agree with you on the comparative component of your argument Chris, that it is important to use generic measures to compare apples with apples, so to speak. My key problem is when our preferred measure, such as the EQ-5D, does not measure a key component of the disease state in question, most notably it has no fatigue dimension. You are right, that one thing you can do then is map, but as peceditor has highlighted when you map you don’t capture all of the variance and this is even worse if the key thing you are trying to map i.e. fatigue is not even measured by the EQ-5D. To me it comes down to the key word – preferences! The great thing about the EQ-5D and SF-6D is that they are preference based health related quality of life measures. If it is important to patients that certain aspects of their illness are relieved, such as fatigue, then that should be included in the valuation. To not include them would mean we are no longer taking into account patient preferences when we are making resource allocation decisions. I’m going to cop out here though and say I have no idea what the best way we should do this is, be it preference elicitation through TTO or redesign of generic measures, except that for me TTO is the far easier option given that the EQ-5D is unlikely to be overridden any time soon partially due to the amount of work it would take (although I think that work needs to be done!).
Thanks Rachael, I totally agree. I think it is good that people are trying to find new ways to deal with the problems of current QALY measures, I just think condition-specific QALYs are a step in the wrong direction. They appear to simply paint over the cracks, so to speak.
I agree with you and the previous commenter, that mapping is not ideal. This is exactly my point: that condition-specific QALY measures seem no more useful than mapped results! I may be wrong though… I have never used them after all!
I agree with the sentiment you express Chris. However, mapping also introduces another layer of uncertainty and can hamper comparability. I personally think its time for a major rethink on the generic measures currently used
So would you argue that we should be using generic measures, just not the ones we currently do?
If so, do you have any ideas you’d care to share?