I think I read somewhere that GiveWell doesn't tend to report these figures because the QALY assessment system is so subjective; they instead, for charities focusing on morbidities other than death, report specific results such as "cost per case of blindness averted" or "cost per additional year of school."
Hm interesting. It's plausible this is a way to avoid utility monsterization?
Like hypothetically, let's say an average high-income life is twice as enjoyable than a low-income life. If that were the case, then using cost-per-QALY might give an advantage to saving high-income lives - they were all set to have a fun life as a rich person until they happened to die from some disease. Whereas the low-income lives were going to have a miserable time anyway, so it "doesn't matter as much if they die." I am not endorsing this interpretation - just articulating what using QALYs might imply if we were to use them for comparison purposes.
You can read their GiveWell's public document on DALYs here
A thing that might be worth noting is that according to GiveWell ~40% of AMF's value comes from increasing the long-term income of children, not from health outcomes
Interesting. That is not apparently taken into account in the $4500/life figure they publish which is based exclusively on deaths averted.
"per QALY saved."
Hm, are QALYs saved or more like created?
I don’t think there’s a substantive difference. It’s just counterfactual reasoning. If I give you penicillin that saves your life so you live to age 70 instead of dying at 25, did I just save or create 45 years of extra life for you? I think that is just a matter of word choice.
I'd like to be able to compare various biomedical interventions to GiveWell's estimate of the cost to save a life via AMF ($4500). This is the cost "per death averted," whereas I would like to get the cost "per QALY saved." This would allow for direct comparisons with the biomedical literature, where cost-per-QALY is the standard metric used in cost-effectiveness analysis.
Does GiveWell have cost-per-QALY for their own interventions listed somewhere?
One blunt tool (which I have used) is just to divide the dollars per life saved by by 80 or so (Average-ish QALYs in a healthy life) . So $4500/80 = $56 per QALYs. This is a close enough estimate I think to be useful for comparison to the public health literature, but I'm open to arguments against it.
The public health literature obviously has its own estimates for mosquito nets, which are very much in the ballpark of Givewell
I would go with this, though using more like 60 years
I would also discount a bit for counterfactual mortality.
Agreed that it probably makes sense to be closer to 60 years, or maybe even a bit lower (though if there are major advances in life-extension over the coming decades, then it could be much higher for young children who will have the most chance to benefit).
I'd note that health-related quality of life is likely to be less than 1 per year, perhaps 0.7 or 0.8.
Regarding counterfactual mortality, wouldn't this largely be taken into account of in the overall estimate of life expectancy? Though this overall estimate probably doesn't include things like major catastrophes (devastating pandemics, x-risks)
GiveWell's spreadsheet says that the conventional value would be 37 based on discounting and age-weighting
I am not clear on if that number is meant to be interpreted in terms of years or just as a unitless "moral weight" for comparison purposes. I am hesitant to compare it with QALYs or other metrics without knowing the units.
The spreadsheet in the second link says YLL (years of life lost) which I think can be considered equivalent to QALYs in this case.
It's also consistent with the comment on the 0.40 on the conventional value of increasing income in the first spreadsheet
Yes close enough - in reality it would be YLL + Other DALYs (suffering but not death caused by malaria) which would be equivalent to QALYs, but YLL would be close enough I think!
Thanks for finding that!
Yep that works too @NunoSempere there are different arguments for between about 35 and 80 years, the argument for the higher end e.g. 80 (made interestingly by the WHO) is that we should ascribe everyone their full future potential life expectancy with each year valued at or close to 1.0, no matter where they are in the world. This value every human life the same and recognises that much of what stops people reaching this potential are modifiable factors
The argument for lower-mid end (perhaps more in line with effective altruism thinking) is that we should use the real world situation as of right now to estimate the likely QALYS someone would have experienced in average life in that country. In this case you might get more like 35-40 for somewhere like Uganda like Givewell did
Or you can hedge your bets and go somewhere in between
Either way It makes about a 2x order of magnitude difference which is managable.
Note that we shouldn't be considering total health-adjusted life expectancy (HALE), but rather the HALE of the average person saved - hence you probably want to half that as a typical rule of thumb (not fully accurate, since disease burden is higher for old people, even as the average person saved is probably younger than the midpoint given global age distributions; evens out to some extent, but not clear how much so)