As alluded to in yesterday’s journal round-up, on reading a recent article by Versteegh and Brouwer, I have had some thoughts about the way we think about the the debate between the use of either patient or public preferences for health state valuation.
When it comes to valuing health states, NICE (and some of their counterparts) advise the use of preferences from the general public. An alternative argument is that we might use patient preferences, because the public probably do not have an accurate understanding of what it’s like to live in a particular health state. In their new paper, Versteegh and Brouwer outline the key arguments in favour of using public preferences but highlight the limited nature of these arguments. One thing they discuss is the notion that public preferences are ex ante while patient preferences are ex post. It’s analogous to preferences vs satisfaction, or decision utility vs experienced utility. The authors outline some limitations to this interpretation. In this blog post I’d like to build on this discussion. My main focus is on defining what we actually mean when we talk about ‘patient preferences’.
Before and after what?
Ex ante means ‘before the event’, and ex post after it. But when we are valuing health states there is no event before or after which utility can be estimated. We are trying to value a state, not preferences regarding an event. We may contrive an event – such as the onset of a particular health state – but that is theoretically quite a different thing to value. Indeed, this contrivance of an event taking place may be a problem.
We should probably do away with these terms and just speak in English, but let’s be realistic. At the very least, we need to be clear about what ex ante and ex post mean in this context; the ‘event’ in question is experience of the given health state.
But then, health state valuation isn’t about just one health state – it’s only possible to value health states in relation to one another and in particular ‘full health’ and a state equivalent to being dead. Furthermore, there is little doubt that a person’s valuation of past or future health states relates to their current health state. Chances are that any individual completing a health state valuation will be valuing some states from an ex ante position and some from an ex post position, both of which are influenced by their current health status.
Ultimately, whether preferences being elicited are ex ante or ex post has nothing to do with whom is being asked, and everything to do with what they are being asked about.
But that isn’t the crux of the matter anyway. What we really want to do here is differentiate between ‘patient preferences’ and ‘public preferences’. ‘The public’ is easy to define. It’s everyone. We usually try to get a representative sample because we cannot ask everyone to do a TTO exercise. But we need to be clearer about how we define patients. Patients are not ex ante – that we can agree on. Or can we? What if we ask an individual about an inevitable future health state associated with disease progression, of which they have a good understanding? What’s worse, patients might also not be ex post, depending on our definition of these terms.
It seems far more intuitive and accurate to describe patients as ex tempore: essentially meaning ‘at the time’. Patients’ health state preferences are neither retrospective nor prospective, but explicitly in relation to their current health state. Crucially, it is that current health state that we are trying to value.
So, a person valuing their own health state is doing so ex tempore, and that’s usually what we mean by ‘patient preferences’. But I hope it’s clear by this point that an individual patient’s preferences need not necessarily be ex tempore either.
People who have never experienced a given health state are necessarily stating their preferences ex ante, but they could still be a patient or not. Meanwhile, somebody who does have experience of a health state could be valuing it from any of the alternative temporal positions. They may, for example, be valuing a future in a health state that they have previously experienced. Versteegh and Brouwer provide a nice taxonomy of the arguments for the use of public preferences. I’d like to provide my own taxonomy here, of the different types of preferences we might elicit. I see it as follows:
|Experience of health state||No experience of health state|
|ex ante||ex tempore||ex post||ex ante||ex tempore||ex post|
There are 4 types of responder (A, B, C and D), determined by whether they are a patient and whether they have previously experienced the health state currently being valued. Similarly, there are 3 different types of health state valuation, depending on whether the state being valued is a past, present or future state. For any given person valuing any given health state, the elicited preferences will be one of the labelled boxes. Ask that same person to value a different health state, or ask a different person to value the same health state, and the elicited preference may well differ.
There may of course be other ways in which individuals differ, such as the extent to which they have adapted to their current health state. But while that’s an important consideration in determining from whom we ought to elicit preferences, I don’t think it’s a key question in identifying patient preferences as opposed to public preferences.
Patient vs patients
One implication of this is that we have (at least) two types of patient preferences. Patient preferences could be A+B. That is, we value a particular health state in all patients, regardless of their current health state. That might be done in a sample representative of the current population of people considered to be a patient, however that might be defined. It strikes me that this is the true definition of patienthood as might be used in other contexts.
The kind of patient we talk about when we discuss ‘patient preferences’ is, I think, just those people falling into ‘A2’; patients valuing their own current health state.
Versteegh and Brouwer seem to suggest that any valuation of current health – that is, ex tempore – represents patient preferences. In practice this will likely be ‘A2’ through the identification of participants, but it’s important to consider the existence of ‘C2’. Just because a person is experiencing the health state of interest does not necessarily make them a patient in any practical sense of the word.
For what it’s worth, I think that public preferences are the least bad option for now. But Versteegh and Brouwer’s suggestion that we should report both is a good one, which could lead to more research that may very well change my mind. I think it will also force this issue of clearer definition of ‘patient preferences’.
Photo credit: Tori Cat (CC BY-NC-ND 2.0)
[…] researchers think we should be using values from people actually in those health states. I’ve written before about some of the problems with this debate. In this study, the authors try to bring some clarity […]
[…] ought to heed its advice. However, I still think it’s limited. As I described in a recent blog post and discussed at a recent HESG meeting, I think the way we talk about ex ante and ex post in this […]
[…] a paper by John Brazier and co, which fortuitously related to a paper that I previously discussed here on the blog. I was badly behaved, going well over time, but there were a lot of issues to grapple with around […]
[…] argument is also sometimes skirted around with regards to the valuing of health states. Should it be the preferences of healthy people, or the experienced utility of sick people, that determine health […]
[…] public, patient and experienced-based values. Matthijs Versteegh, whose paper featured in a recent blog post, set out arguments in favour of using patient values in addition to general public values, and […]
*autocorrect, this = issues
Sorry am late to the party! I tweeted to Werner just after publication applauding the paper and the investigation of non-public valuation. The “too few people in bad states” issue is in fact CRUCIAL, particularly since discrete choice based tasks are increasingly used. I should know in time for the 2017 EuroQoL plenary whether experience is associated with different preferences for EQ-5D-5L but I know it certainly does – quite a lot – for ICECAP-O. The only way you can do this properly is to use quota, not representative, sampling.
Of course the fact that people with experience of/in an impaired state have radically different preferences does not mean the general population tariff would change radically because this much smaller group’s preferences are averaged out, but the fact is this has never ever been properly empirically tested in a discrete choice framework (til now).
Anyway I won’t go on and on since I’m all in favour of this type of work – I discussed similar issues myself (but more from the perpsective of the logistics of using a discrete choice based framework – I steered clear of the normative issues you guys discuss as I got repeatedly slammed down LOL) in my papers from 2010 onwards.
Looking forward to discussing these and other this in Berlin next week!
Thank you for taking the time to read and reflect on our article! As Werner Brouwer mentioned, our main aim was more discussion. I’m beginning to think that a blog might have been a more successful strategy to meet our goal, rather than painstakingly meeting the word limit of a journal (and as you know SS&M is generous with word limits!). But then again, we needed our reasoning to be out there in the academic community. And with Twitter and all, we might even draw some attention to it.
After reading the conclusion of your blog I thought to myself: is solely using public preferences indeed the least bad option for now? As we mentioned in our paper, the argumentation by the Gold panel seems to point to healthy individuals (who have not encountered any serious health problems), rather than a community sample, as does the insurance principle. Arguably, this between distinction between general public and healthy individuals is merely a theoretical one, and it is empirically shown to be irrelevant in terms of effect on value sets: our forthcoming paper in HE shows that a general public sample for a tariff study has so few individuals with health problems (who do have different valuations), that it does not affect the mean valuations of the sample (but does alter the individual valuations, which suggests that had there been more patients, health state values would have been different). In my opinion: the theoretical foundation for general public preferences is less strong than that for healthy individuals, but empirics save the day as the distinction could be irrelevant from a consequentialist point of view.
So are public preferences the least bad option, or is it in principle better to ALSO have all EQ-5D health states valued by individuals who are currently in the state under valuation and use both in economic evaluations? As you nicely argue, the patient preference elicitation process would not be without issues and assumptions, but, as you also point out, neither is the general public elicitation process. Perhaps the relevant question is whether we should or should not include patient preferences in economic evaluations, rather than having the ‘how and what’ of it dominate the discussion -despite those issues deserving much attention. Perhaps practical (and theoretical?) issues with demarcating and eliciting patient preferences turn out to be a good reason not to include them in economic evaluations, but then, those arguments should be put forward and discussed. Until then, I would say that the least bad option is to include both patient and general public preferences in economic evaluations. Indeed, both perspectives probably ignore the reference point brought about by previously experienced health states, which would be a very interesting research topic.
Thank you once again for devoting your time and blog of yesterday to a topic that seems to have our mutual interest.
Best wishes from The Netherlands,
Thanks, Matthijs. I think we broadly agree. Certainly, the basis for both positions is limited. I suppose I am more worried about adaptation than about the ‘inaccuracy’ of public preferences, hence my (weak) support for public preferences. Doing as you suggest seems like a smart way to progress. You are welcome to contribute to the blog if there is more that you would like to say. We may enforce word limits, but there is always room for more blog posts!
Thank you Chris. The September EuroQoL plenary of 2016 will devote some presentation time to the topic, including our SSM article and a very interesting paper by John Brazier. I’ll await the feedback and get back to you after that. And word limits are a blessing in disguise I know 😉