Thanks for your comment, Derek. This has been really useful.
Some changes I have made in response:
Changed “death” to “being dead” in my explanation of the DALY scale
Now say that DALYs likely underweight pain, but QALYs may not:
DALYs appear to weight pain very lightly. For example, terminal illness with constant, untreated pain has a disability (DALY) weight of 0.569, which is only 0.029 more than the weight for the same condition with pain medication. QALYs are better at capturing pain: physical pain is the dimension given the highest weight in the EQ-5D, and instrument used to measure quality of life.
Mention that even sufferers may underestimate the badness of depression [with a link to your comment]
A question:
I see from the summary you linked that IHME have used sequelae to identify ailments that are present in multiple health conditions. That seems sensible. I guess the kind of problem I often face is “What will be reduction in someone’s disability weight if they are—protected from getting diabetes / cured of depression / etc. ?”
In the diabetes example, it seems fair to count DALYs averted by not having diabetes and DALYs averted by depression-caused-by-diabetes. Maybe not fair to count, say, obesity, since the increased risk of obesity associated with diabetes is likely to be correlational, not causal. Am I thinking along the right lines?
If we go with the depression example, it seems fair to count both prevented suicide and prevented depression (but not prevented depression-while-dead-by-suicide)
DALYs appear to weight pain very lightly. For example, terminal illness with constant, untreated pain has a disability (DALY) weight of 0.569, which is only 0.029 more than the weight for the same condition with pain medication. QALYs are better at capturing pain: physical pain is the dimension given the highest weight in the EQ-5D, and instrument used to measure quality of life.
You might want to check disability weights for other painful conditions; I don’t remember if they were generally low.
I suspect QALYs still underweight extreme pain, for various reasons, e.g. the arbitrary cap on negative values, and the lack of experience of such states among most respondents (typically the general public in high- and middle-income countries). The distribution of responses typically suggest ‘floor effects’, with some respondents likely to give lower values if it were permitted. The Devlin et al paper I linked to previously gives good evidence of that, but here is a graph from a different paper (UK sample) for illustration (note the cluster at −1).
My point was more that pain gets a high weight relative to other dimensions of the EQ-5D...though not always the highest. As shown in the graph below, the original EQ-5D-3L UK tariff (Dolan, 1997) had pain as second (after mobility) for extreme, and roughly equal first (with self-care) for moderate, based on TTO responses from the general public. (I can give you the Excel version of the graph if you want to modify it.)
The preliminary UK tariff for the newer EQ-5D-5L gave pain the highest weight, followed by depression/anxiety, for the extreme level. Full results below… but note that NICE rejected the value set for methodological reasons so, last I checked, it still recommends mapping the old 1997 3L figures onto the 5L with an algorithm.
even sufferers may underestimate the badness of depression
I think this may be true (given some plausible-to-me philosophical and psychological assumptions), but it’s also more generally that studies done in sufferers likely underestimate the badness. For example, because studies exclude the most severe cases, the badness of severe depression would be underestimated even if the study participants gave fully ‘valid’ responses (and even if an instrument were used that was able to capture the full range of experience).
I see from the summary you linked that IHME have used sequelae to identify ailments that are present in multiple health conditions. That seems sensible. I guess the kind of problem I often face is “What will be reduction in someone’s disability weight if they are—protected from getting diabetes / cured of depression / etc. ?”
In the diabetes example, it seems fair to count DALYs averted by not having diabetes and DALYs averted by depression-caused-by-diabetes. Maybe not fair to count, say, obesity, since the increased risk of obesity associated with diabetes is likely to be correlational, not causal. Am I thinking along the right lines?
If we go with the depression example, it seems fair to count both prevented suicide and prevented depression (but not prevented depression-while-dead-by-suicide)
I don’t remember the details of the DALY/GBD methods, and I don’t know a great deal about diabetes, but I’m pretty sure it can be a cause as well as consequence of obesity. At least insulin therapy can cause weight gain. And obviously you’d want to count only the proportion of diabetics who would have got depressed/gained weight as a result of diabetes.
Not sure I follow the depression example, but yes, you would sum the YLL from suicide (i.e. ‘standard’ or counterfactual life expectancy minus the actual number lived) and YLD (i.e. years lived with depression * disability weight). The formula/steps and examples are here and here.
Thanks for your comment, Derek. This has been really useful.
Some changes I have made in response:
Changed “death” to “being dead” in my explanation of the DALY scale
Now say that DALYs likely underweight pain, but QALYs may not:
Mention that even sufferers may underestimate the badness of depression [with a link to your comment]
A question:
I see from the summary you linked that IHME have used sequelae to identify ailments that are present in multiple health conditions. That seems sensible. I guess the kind of problem I often face is “What will be reduction in someone’s disability weight if they are—protected from getting diabetes / cured of depression / etc. ?”
In the diabetes example, it seems fair to count DALYs averted by not having diabetes and DALYs averted by depression-caused-by-diabetes. Maybe not fair to count, say, obesity, since the increased risk of obesity associated with diabetes is likely to be correlational, not causal. Am I thinking along the right lines?
If we go with the depression example, it seems fair to count both prevented suicide and prevented depression (but not prevented depression-while-dead-by-suicide)
You might want to check disability weights for other painful conditions; I don’t remember if they were generally low.
I suspect QALYs still underweight extreme pain, for various reasons, e.g. the arbitrary cap on negative values, and the lack of experience of such states among most respondents (typically the general public in high- and middle-income countries). The distribution of responses typically suggest ‘floor effects’, with some respondents likely to give lower values if it were permitted. The Devlin et al paper I linked to previously gives good evidence of that, but here is a graph from a different paper (UK sample) for illustration (note the cluster at −1).
My point was more that pain gets a high weight relative to other dimensions of the EQ-5D...though not always the highest. As shown in the graph below, the original EQ-5D-3L UK tariff (Dolan, 1997) had pain as second (after mobility) for extreme, and roughly equal first (with self-care) for moderate, based on TTO responses from the general public. (I can give you the Excel version of the graph if you want to modify it.)
The preliminary UK tariff for the newer EQ-5D-5L gave pain the highest weight, followed by depression/anxiety, for the extreme level. Full results below… but note that NICE rejected the value set for methodological reasons so, last I checked, it still recommends mapping the old 1997 3L figures onto the 5L with an algorithm.
There are many other tariffs from many other countries, for both the 3L and 5L, if you want to compare: https://euroqol.org/information-and-support/resources/value-sets/
I think this may be true (given some plausible-to-me philosophical and psychological assumptions), but it’s also more generally that studies done in sufferers likely underestimate the badness. For example, because studies exclude the most severe cases, the badness of severe depression would be underestimated even if the study participants gave fully ‘valid’ responses (and even if an instrument were used that was able to capture the full range of experience).
I don’t remember the details of the DALY/GBD methods, and I don’t know a great deal about diabetes, but I’m pretty sure it can be a cause as well as consequence of obesity. At least insulin therapy can cause weight gain. And obviously you’d want to count only the proportion of diabetics who would have got depressed/gained weight as a result of diabetes.
Not sure I follow the depression example, but yes, you would sum the YLL from suicide (i.e. ‘standard’ or counterfactual life expectancy minus the actual number lived) and YLD (i.e. years lived with depression * disability weight). The formula/steps and examples are here and here.