Building off of Jason's comment: Another way to express this is that comparing directly to the $5,500 Givewell bar is only fair for risk-neutral donors (I think?). Most potential donors are not really risk neutral, and would rather spend $5,001 to definitely save one life than $5,000 to have a 10% chance of saving 10 lives. Risk neutrality is a totally defensible position, but so is non-neutrality. It's good to have the option of paying a "premium" for a higher confidence (but lower risk-neutral EV).
Leaving math mode...I love this post. It made me emotional and also made me think, and it feels like a really central example of what EA should be about. I'm very impressed by your resolve here in following through with this plan, and I'm really glad to have people like you in this community.
Thanks so much for the encouragment, really do appreciate it.
Great point I hadn't thought about risk neutrality vs non-neutrality here and that there might be a pool of people even within EA who would rather pay a "premium" for higher confidence. Outside EA my experience has been that perhaps even the majority of people would probably prefer to pay for higher confidence.
I agree with your last sentence, and I think in some versions of this it's the vast majority of people. A lot of charity advertising seems to encourage a false sense of confidence, e.g. "Feed this child for $1," or "adopt this manatee". I think this makes use of a near-universal human bias which probably has a name but which I am not recalling at the moment. For a less deceptive version of this, note how much effort AMF and GiveDirectly seem to have put in into tracking the concrete impact of your specific donation.
Orthogonally, I think most people are willing to pay more for a more legible/direct theory of impact.
"I give $2800, this kid has lifesaving heart surgery" is certainly more legible and direct than a GiveWell-type charity. In the former case, the donor doesn't have to trust GiveWell's methodologies, data gathering abilities, and freedom from bias. I've invested a significant amount of time and thought into getting to my current high level of confidence in GiveWell's analyses, more time than most people are prepared to spend thinking about their charitable donations.
And I think most people -- including myself -- have a prior that projections and analyses of all sorts tend to be overinflated in comparison to reality. How many building projects come in on time and under budget? How many IT projects? The less complex the theory of impact, the less the lightly-researching donor will end up implicitly discounting the organization's claims on account of their background skepticism of projections and analyses.
I think that’s such a good point about risk neutrality - and when looking outside of EA spaces this is important to keep in mind, I think risk neutrality is so accepted in EA spaces it’s sometimes forgotten that outside it is not necessarily so.
Also, just a really moving post. Trying to do the best you can in an emotional yet rational way. I really appreciate this.
I suspect the limiting factor on scalability here may be having people in-country who have the technical skills, access to information, and mindset to pull something like this off. I'm sure that saying no was excruciating. Potentially, one could scale (at some probable tradeoff for cost-effectiveness) after a while with algorithms. For instance, you might determine that providing transport and ancillary supports for moderate-to-severe malnutrition cases up to $150 is cost effective if the treatment was unlikely to happen otherwise.
Conditioned on who the likely donors are, I think using the GiveWell bar is too demanding. As you mentioned, this project hits the fuzzies hard. It's quite legible to non-EA donors: identifiable beneficiary with a story,[1] very understandable theory of change. The amounts at play are small enough that a donor who gives $50-$100 would feel like a significant part of potentially changing that child's life. This feels like an easy thing to share with people who are not EAs or even sympathetic to EA principles.
If the counterfactual is that the donor would have given the money to a big NGO, to almost any charity in their own country, or would have spent the money on non-charitable purposes, then the cost-effectiveness for this program would not need to be particularly high. So I'd be interested in hearing a little bit about cases number 4 through 6, which may give an idea of how much the cost-effectiveness might drop upon broader scope.
In many cases, you'd need to fund the treatment out of reserve funds and fundraise retroactively, but I don't think that would be a major problem if transparently disclosed. Doesn't seem to hurt the disaster-relief charities' fundraising right after a natural disaster.
Yes lack of that combination of skills, information and mindset would definitely be a barrier. There's plenty of people who could be trained in cost-effectiveness calculations though, and exactly like you say with the right algorithms and checks and balances. Great example of malnutrition - although in that case there might well be a good argument you are better off funding more RUTF or better malnutrition centers.
Love the insight about the GiveWell bar maybe being too demanding, food for thought indeed.
Onto the cases 4 through 6... Before getting to that there were some cases we dismissed almost out of hand, because the counterfactual of whether they would get the care without us seemed so tiny it was hard to even estimate.
A couple of the cases we thought about and refused were cancer treatment - which could be cost-effective in kids but these were adults. Cancer treatment in Uganda is very poor and expensive. The obvious factor which made situations far less cost-effective even if they were cured were just age, life saving treatment in kids.
Then there was other was treatment for a thyroid condition that might actually have been cost-effective but another local NGO ended up helping so we didn't need to.
Then there were a couple non life-saving treatments that could maybe have been cost effective. One was eye surgery in a kid (not cataracts), and one a surgery to correct a broken bone that had left someone unable to work in the garden.
I hate to say it but it wouldn't describe it as excruciating for us refusing to help, for a couple of reasons. First I treat many patients here that passively decide not to help (not people asking, just sick patients who can't afford the best treatment option). I'm used to seeing people not get the care they need and I suppose I'm someonewhat numb to the implications. Not proud of that but hard to avoid.
Second we obviously get loads of requests for medical and non-medical things we turn down all the time, so we're used to saying no a lot as well. In this case it was actually easier in a way, we were being sent cases of sick people that we never actually told we might helpe, so we didn't actually have to refuse a personal cry for help.
Great example of malnutrition - although in that case there might well be a good argument you are better off funding more RUTF or better malnutrition centers.
I think there's a real difference in funging risk between your intervention and funding ready-to-use therapeutic food (RUTF) or malnutrition centers.
If we donate to RUTFs or malnutrition centers, some of that may be offset by big NGOs and governments giving a little less. That's pretty undesirable, given the relatively low value of what many of them would do with the extra money.
If we donate to the nine-month old's family, there is the possibility that the $100 for transport would have been raised by the family and the surrounding community. But unlike in the NGO/government scenario, your added value in the scenario in which the family/community raised the money is hardly ~zero. There's now counterfactually $100 more money in the pockets of (presumably) poor individuals, which is a good thing and is probably roughly equivalent to the impact of GiveDirectly distributing $100. Moreover, in that hypothetical, the infant's family has probably exhausted its financial resources and gone into debt -- factors that may reduce the risk of successful aftercare and a good ultimate outcome. So even the scenario where your donation was 100% subject to funging is not such a terrible outcome.
Cancer treatment in Uganda is very poor and expensive. The obvious factor which made situations far less cost-effective even if they were cured were just age, life saving treatment in kids.
My lay understanding is that successful treatment of pediatric cancers often allows for a ~normal lifespan and life, while even initially successful treatment of adult cancers often carries a significant 5-10 year mortality rate[1] and involves significant continuing morbidity (e.g., residual pain, lymphedema). So initial successful treatment of a pediatric cancer seems to be a more good result than initial successful treatment of an adult cancer, even if one disregards years of life saved.
In this case it was actually easier in a way, we were being sent cases of sick people that we never actually told we might helpe, so we didn't actually have to refuse a personal cry for help.
I think this touches on another probable advantage to your model. I've read and heard that Westerners have to be careful doing relief work, lest their actions disrupt the processes the community has developed for mutual aid and support by instead leading everyone to look to the NGO to fix the problem with their money. This problem seems much less likely to develop under your system, as you had no application process to attract candidates, a large catchment area for selecting beneficiaries, and a very low rate of selection.
I'm assuming that most of the cancers identified were not early-stage cancers due to limited levels of medical care and specialized diagnostic equipment.
I have for a time wanted something like a list of "personal" interventions likely to be ~GiveWell cost effective. This means when I or people in my social circle encounter someone struggling to pay for e.g. fixing a blocked aortic valve, I can feel somewhat confident about donating money directly to the afflicted person.
I imagine many of us EAs know several people who live or travel in poor countries where they every now and then encounter people in desperate situations. I myself am somewhat frequently contacted by friends about paying for someones surgery or school fees.
Basically I understand our current EA global health initiatives to require a high density of people to help. But this requirement might be bypassed if instead we have networks of people we trust that can identify people in situations we know are likely to be cost-effective to support financially. I agree with other commenters that this is not scalable, but it might make EAs have better standing in their respective social circles and could have other positive effects. This is almost like a personal re-granting program.
I love your post - I think it goes a long way towards understanding if such a list might be useful. And I, like the other commenters, feel admiration and sadness after reading your piece.
Thanks Ulrik, this is a really interesting question about a list of interventions and you might be right there is scope for more work here. I would imagine @CE (charity entrepreneurship) have thought about this before, and might have a list of specific medical interventions that could be cost-effective.
I might start by splitting these into "life saving" interventions, where most of the benefit comes from one potential life "saved" - situations like malnutrition, cancer treatment and heart operations "life improving" interventions where most of the benefit comes from things like fistular surgery, sight-saving medication and surgery and fixing retracted testicles.
The problem is though where do you stop considering something a single "intervention"? Treatment for malaria, pneumonia and sickle cell could be cost effective but is already a standard part of almost any health systems, and its more about making sure people get the treatment quickly and cheaply or for free.
I also like your idea of "networks of people we can trust" to support individuals as well, which could have the advantage of building EA community and resisting value drift as well. Keep in mind though these kinds of people might also be busy with other work. I personally love to be contacted with questions about this kind of thing so am always happy to give my poorly-to-moderately-informed opinion on medical stuff.
Thank you so much for doing this and writing about it. I've been thinking about this post for weeks.
It feels extremely weird to quibble about numbers in situations like this.[1] As you write, a person's life and health are worth so much more than anything else we could buy with a few thousand dollars, we should be willing to give a thousand suns. The difference between these two is always a reminder of how unfair things are.
That said, keeping in mind that I'm not at all an expert in cost-effectiveness estimates, here are some considerations for cost-effectiveness that might be worth considering.
I think the "bar" shouldn't be the cost per life saved of the average GiveWell intervention ($5.5k) but the marginal cost of ~$7.5k. And of course we should keep in mind that these numbers are extremely rough with huge uncertainties, so I don't think any decision should hinge on the difference between $7.5k and $5k.
On the other hand, GiveWell's interventions have many other positive effects. For example, ~35% of the value of AMF's program is estimated to come from non-life-saving development benefits, which would compensate for the higher bar.
Considering both the "chance of help without intervention" and the "cost for the hospital" seems to be excessively conservative. I would imagine that the help from someone else would still incur some cost for the hospital and the people helping. So, I would only consider one of them. This seems especially relevant in the malnutrition case.
I also wonder about the hospital's opportunity cost. I might be showing my ignorance here, but would the burns unit in your first example and the cardiopulmonary bypass machine in your last example be doing non-life-saving work without this intervention? If that's the case, I think you could (very) heavily discount the economic cost for them. Otherwise, the economic cost could be modelled as similar to a donation to the hospital.
Lastly, the estimates for the relative quality of life in case of success might sadly be too high. Possibly ignorantly, I would expect these individuals to live in a worse environment with fewer resources and a worse support network compared to the population average; they might have a lower life expectancy and more future risks.
That said, I tried to do a BOTEC and the results were in the same ballpark (but highly sensitive to e.g. the chance of survival without help).
Again, thank you so much for all the work you're doing.
I've been sitting on this comment since I read this post, since it feels so insensitive to nitpick numbers in response to such a personal and "real" post. I can't even imagine what it must feel like to act on these numbers.
Amazing response and thanks so much for looking more closely at the numbers, I was hoping someone would!. Absolutely love that you made your own spreadsheet, great to get a sanity check at least. To respond individually to your fantastic points.
I hadn't considered the 7.5k bar I agree that may make more sense. I'm not sure though that if this was going to be done at any scale, "high uncertainties" should lead to a fuzzier bar - I struggle to see how that works practically. I'm not sure how any cost-effectiveness seeking intervention works without a fairly solid theoretical cut-off? a As a side note as well, as cost-effective interventions go, helping individuals would perhaps be at the lower end of uncertainty
This is a good point about the 35% of non life-saving benefits for AMF. Obviously here there are also many non life-saving events with similar pathways to AMF, but I'm not sure how that compares
Thanks for this comment on excessive conservatism, I think I agree nice one!
Yes how to include other non-donation costs is a real head scratcher and I struggle to get my head around.. For burns and Heart surgery, much of the work the hospitals are doing is indeed life saving - more so for the heart surgery than the burns. Even the non-life saving burns work has potentially huge impacts on the patients' lives.
This is an excellent point, I think I was probably being overly optimistic here and would downgrade these in future.
Thanks so much again for your work as well. As much as thinking about helping individual people might speak to us and hit us emotionally, my guess would be that your work developing software at Giving What We Can may well be a more cost-effective way of helping people.
Loved this post. Like sawyer wrote - it made me emotional and made me think, and feels like a great example of what EA should be.
There actually is a non-profit I'm aware of (no affiliation) that hits a lot of the criteria mentioned in the comments - https://saveachildsheart.org/, they treat life-threatening heart disease in developing countries, often by paying for transportation to Israel where the children receive pro-bono treatment from a hospital the nonprofit has a partnership with. From a (very) quick look at their financial statements and annual report, it looks like it costs them around ~$6,300 to save a life, although that number could be significantly off in either direction (by looking through the annual report, it looks like the nonprofit is not especially focused on the most cost-effective parts of its programming, and does many activities that look like PR, which is probably morally good if it allows them to scale. On the other hand, it's not clear from the AR what the severity of the disease is in the children treated, and what share of their treatments are actually life saving).
Your post, and nonprofits like this, make me think of something EA often misses from it's birds-eye approach to solutions - leverage. Both you and saveachildsheart use their leverage (your proximity, their partnership with a first world medical institution) to be impressively cost-effective, but leverage is hard to spot in a-priori spreadsheets.
Could you say a bit more about the ~$6,300 figure? I have 547 lives saved from the annual report (p. 5) and about $9.5MM USD in expenses from the financial statements. Admittedly, most of this is related to the establishment of a "Children's Hospital at Wolfson" -- but it's not clear to me that these costs should be excluded. I suppose that the organization is doing its current work without said hospital existing yet, but the presence and magnitude of that expenditure makes me wonder -- at a minimum -- whether they have room for more funding at ~$6,300.
By rough analogy, it wouldn't be appropriate for an organization to fundraise separately for bednets and for distribution costs, and quote a cost-effectiveness figure to distribution-cost donors of (distribution costs / total impact).
Thanks for this source, ezrah! These 6,3k seem to possibly be a bit misleading to me. Without taking a closer look, it sounds like the costs of the treatment (currently covered by the hospital's CSR funds or donations?) are hidden from the total costs as someone had the great idea of splitting the true costs into two different donation opportunities, both of which on each side suggest that "if you give to this opportunity, the other side is covered". But given the high-quality medical facility/treatment the children receive, the true cost of donations consumed per patient are probably much higher. Am I mistaken here?
From what I understand, the per-patient treatments costs are both quite low and are given pro-bono, so given how GiveWell understands leverage (which @Mo Putera pointed out in the response below), they should be strongly discounted from the costs. The question of how to incorporate the infrastructure costs, ie - the hospital, staff training, etc - that enable the program to operate, is quite interesting, and I honestly don't have a great idea how that fits into the model.
something EA often misses from it's birds-eye approach to solutions - leverage.
I'd be curious as to what you mean here, since my impression was always that EA discourse heavily emphasises leverage – e.g. in the SPC framework for cause prioritisation, in career advice by 80,000 Hours and Probably Good, in GiveWell's reasoning (for instance here is how GW's spreadsheet adjusts for leverage in evaluating AMF).
Great question! I realize that I really wasn't clear, and that it probably does exist more in EA than my instinctive impression (also - great links, I hadn't been familiar with all of them).
What I meant by leverage was more along the lines of "the value of insider's perspective and the ability to leverage individual networks and skill sets". In these cases, Nick was able to identify potential cost-effective ways to save lives because of both his training and location, and SACH is able to similarly have a cost-effective program because of their close connections with a hospital. I have a few other examples as well, such as NALA's WASH on Wheel's program (which essentially trains a team a plumbers and provides access to clean water to hundreds of program, leveraging the existing infrastructure), and anecdotes I've heard about people on the ground being able to provide crucial solutions during the current Israel-Hamas crisis.
I have a sense that the classic EA (and I could very much be strawmanning here) thinks along the lines of: big problems, good solutions, niche area - but doesn't think about who is best placed to identify or implement even better solutions that can come up because the world is messy.
After thinking about it, the "leverage" I'm referring to is probably more common than I thought, but maybe not so very well defined.
Thanks for your post! As mentioned in other comments, one central factor to understand how cost-effective your procedure appears, could be the cost of having a medical professional as you are (or a person with the minimum required medical expertise and local context knowledge) spend their time roaming, (incentivizing people to meet them, this one was easy for you and probably always is), meeting people and identifying patients to support. Could you roughly estimate these costs per case if you were to do that as your main occupation?
The second part of it, paying for their treatment, seems like something that could be rather easy (much uncertainty here ofc) to do when embedded into the right framing. As pointed out by Jason and others in the comments, these are patients with very tangible, heart-warming stories and very obvious direct impact-connection to a donation, which makes them suitable for successful fundraising within the broader non-EA target-groups.
This is of course without evaluating the potentially quickly arising adverse effects such as the deterioration of informal institutions of local fundraising among families and friends due to them pointing potential patients to you etc..
I haven't got this far in terms of estimating the costs of doing it at a bigger scale - after this trial I was probably a little less enthusiastic about the idea than before I started to be honest so I probably won't go ahead and do that right now anyway.
I don't think I'm a person with the minimum required medical knowledge at all though, it would be harder for others but doable.
The adverse effects you speak of is certainly a real risk but I don't think one of the biggest factors. I think the approach would be to only accept poor subsistance farmers from rural areas who were super unlikely to be able to afford treatment counterfactually - my instinct would be to just exclude anyone who lived in any kind of city or even township. I think helping individuals like this could only remain cost-effective if people were identified by our agents, not people "applying" or coming to ask for help or anything like that.
Thank you for writing and doing this. It moved me to read how putting cold calculations next to your warm personal connection with them is inhumanly strange and absurd.
To me it means that I discovered a new priviledge of living in a rich country. Where we can abstract donating behaviour from the decision process of who gets the help.
I kind of look at the EA-principles as a good way of sharpening my intuitions of impact in a far away country. But for you its different: as a medical professional, your intuition of where the most good can be done will be spot on, in comparison to me atleast.
While it's interesting to see the calculations, I would think it's more healthy in your context to stick to your medical intuition in combination to how likely is it that this person will die if you don't help financially.
I've got a slight pushback here. I don't think my intuition of where the most good can be done is actually all that good at all, even now after thinking more deeply from an effective-altruism perspectives. It often surprises me when I crunch the numbers that what I intuitively think might be best doesn't stack up. Others intuition might be far better though.
When its a doctors' own patient though, part of their role might be (arguably) to advocate for that patient cost-effectiveness be darned. Not sure what I think about that though.
A way your decisions are underrated is that this charity, if it existed, would possibly be much easier to fundraise for than GiveWell. Like rather than talking about bednets you'd have pictures of actual children. Perhaps typical donors would give to that competitively.
Have you considered writing up these as manifund impact grants. I can imagine some people might buy having saved some fraction of a child and then you'd have more money to spend. Likewise if you saw promising opportunities you could put them on there.
Finally I find it pretty tragic that you might do less of this for white saviour reasons. White saviour or no, you helped Sunday a lot. I imagine her and her family are not worried about white saviourism. What's your experience of the people you know resenting and your wife's help?
Thanks for the manifund idea, but to be honest in the short term at least I'm focused on OneDay Health and am not looking to either do this systematically or set up a charity around this at this stage (although the encouragement has been great and I'd be open to it in future). I also think if someone was going to start a charity around this, as a few people have suggested it might be fairly straightforward to target non-EA donors which I believe where possible is better than supping from the limited EA money pots.
The white saviour thing is complex but I do think its worth thinking deeply about the potential harms of any intervention (reputational harms, disempowering local social entrepreneurs from filling this kind of space, disrupting local systems of mutual support etc.). There are always down sides to global heal interventions and we often underrate them.
You're absolutely right though in Northern Uganda almost everyone is appreciative of any kind of help. In neighbouring countries though the situation is very different I think because there were far greater negative effects from Colonialism. No-one has ever resented help in 10 years. There are a handful of elite Ugandans in the capital Kampala who would probably resent this kind of thing, but its a very small percentage of the population.
Could this could be made even close to cost-effective and scaleable? If so, I think it has strong potential appeal. Perhaps not so much to hard-core rationalists and EAs, but as a bridge to making mainstream donation more effective. From this perspective, I'm more optimistic about your 'hands oncharity' proposal.
Wow. I don't have a medical background, but I was volunteering in northern Ghana for several months, and had several opportunities to contribute to individual medical care: an (unsuccessful) brain cancer operation for a 9 year old girl (may she rest in peace), malaria and typhoid medication for a few friends who couldn't afford it, and antibiotics (or pain relief?) for a boy who stepped on a nail. And I worried about the same complexities as you did! Thank you for taking the time, effort, and resources to complete this project and share it with us.
I met a group of nurses, who are teaching First Aid to people, the way we do in North America... do you see that as something that could be more cost-effective?
Hey Shelley thanks for the reflections and for doing your best to help those people in Ghana. Teaching first aid is a really interesting idea that I've thought about a little but am unsure about. If you want to have a chat about it sometime feel free to message :).
My prior is that there are a lot of cost effective actions in everyday life, even if you don’t live in Uganda, but that it is hard to scale. The circumstances of your life are probably exposing you to more significant scaling opportunities though, even compare to others living in Uganda.
Thanks Ian that's an interesting reflection, to be honest I hadn't really thought that way before. Can you share the kind of things you think might sometimes be cost effective in everyday life, if you are comfortable with that all good if not!
So, to be clear, it's not like I have a back-of-the-envelope calculation or anything.
The way I see it, charity is hard mainly because it's hard to identify opportunities that scale, and even when we do, most of our efforts are wasted. With Deworm The World, for example, only about half of treated children have any worm infection at all. Targeting charitable interventions is usually not cost-effective because the best beneficiaries can be hard to find. This is even harder if we need the reasoning and evidence to be legible.
But, if we are able to identify targeted cases "by accident" (or, in the course of living life), then we get the benefits of targeting for free, without either the cost of finding beneficiaries or the cost of legible/rigorous impact evaluation.
In the rich world, I think this sort of impact usually comes from behaviors that are free or very low cost to the donor. An example is giving CPR in a public place — it could potentially save a life, for a pretty small opportunity cost, but it wouldn't be worth it to give up your career just to be around in case someone needs CPR. Or a more minor (but also maybe more common) example might be introducing two people who are well positioned to help one another, where the potential connection is discovered incidentally, or by accident.
Thanks Ian that makes perfect sense, really like the CPR and introducing people examples makes a lot of sense. Also like the "free targeting" concept, its true that we are likely to stumble accross cost-effective things along the way :).
Nick, this is one of the best posts I've ever read on the Forum. As you already know, I have huge respect for your commitment to living out your values and I can't wait to read more about your efforts.
Love you Nick. Keep be the person you are. Modernization, in the sense of making the world less darker, was always related to nurses and doctors working in rural and remote areas. It is not only the direct giving but also the spirit of it.
I hope people will donate oneDay Health above the explicit cost effectiveness of it.
This is a crosspost from the new Animal Welfare Alignment Newsletter by Anima International. You can subscribe on Substack if you are interested in following these efforts. Audio reading also available on Substack.
The goals of this post are to:
1. Raise a question I see as crucially important to the goal of aligning AI to animal welfare...
I used AI to fix transcription errors, rerrarange the ideas, and suggest tweaks to the title and some sentences.
Three of the most exciting projects to come out of EA in recent years are, in a vague sense, CEA spinouts:
* Kairos is directly a spinout of CEA and now handles most support for university AI safety groups. Basically everyone I've found who knows them is really excited about what they do
* NEST is an opinionated ideas-fi...
Hello! I'm Justin Portela. I got hired by GWWC to make YouTube videos after AI in Context did such a kickass job.
My channel is using that same cinematic, high-production value beauty to talk about everything in the EA universe that isn't AI.
...
TLDR: My wife and I tried to help three children with severe illness cost-effectively. We paid for their medical care, followed up on what happened and reflected on the process.
"Hearing about her death hurt badly – as it should. But not only had I done what I could, it made sense. Those 300 dollars gave her a chance of not only surviving, but living a long and happy life."
My wife and I live in Gulu, Northern Uganda, where I operate OneDay Health, a social enterprise which provides cost-effective health care in remote rural areas – but this post isn’t about that.
We naturally encounter many situations where we feel compelled to help those around us, We sometimes help with school fees, medical bills and sometimes just cash bailouts. Many of these might not be very cost effective and fall into the “fuzzy giving” category, but by virtue of living in Uganda I suspect these contributions may often hit the unusual jackpot of both feeling good and being cost-effective, earning both “fuzzys” and “utilons” at the same time (sorry Yudkowsky, we mix them).
Which got me thinking, could we target some of our support in a more cost-effective way? As a medical guy in Uganda, I figured we might be in a decent position to try, so we embarked on a mini-experiment to see if we could find a few medical cases which could be super cost-effective to treat, on par with top GiveWell charities.
Thanks Dall E
Our Cost-Effectiveness Bar
We kept it simple here, and deferred to GiveWell’s top-charity bar of $5500 per life saved. [1] I estimated cost effectiveness by dividing the money we would spend on medical care by a series of discount multipliers based on loosely estimated counterfactuals (see calculations below). My approach is deeply flawed as it fails to account for many variables, but allowed me to make a reasonable estimate in about 15 minutes. Feel free to criticize and discuss the method, I'm sure there more accurate and easier ways (@NunoSempere@Vasco Grilo). I didn’t attempt a cost-effectiveness range as it would take longer.
But where to find these sick people who might be cost-effectively helped? We planned to identify poor rural farmers who had life-threatening treatable illnesses, who were unlikely to be treated unless we paid for it. So I asked a couple of our OneDay Health managers to be on the lookout for sick people who presented to our Health Centers, while I also kept an eye out in the health center where I work as a doctor - people who could die without treatment that they couldn’t afford.
Finding people to "cost-effectively" help turned out to be harder than expected. Over a 2 month period we identified three young girls to help, while we decided not to help in another 5 or so cases. These are the stories people we tried to help– feel free to skip the BOTEC sections if you aren’t into that!
Lamunu – age 9
Even for a hardened doctor, Lamunu’s photo was tough to look at. She showed up in our Health center with about 30% of her body burned after falling into a fire. She had spent 10 days in an ill-equipped government hospital and she continued deteriorate. Her father ran out of money and took her home, before arriving one of our health centers (which can't treat burns) as perhaps desperate last-ditch effort. Her burns were infected, she was malnourished and everybody involved knew that she probably wasn’t long for this world.
I happen to live 500 meters away from the best burns unit in Northern Uganda at St. Mary’s hospital Lacor so I thought we might be able to help. We paid for her with herfather to take a 5 hour bus ride to Lacor hospital, and supported the family with hospital fees and some food.
Cost effectiveness BOTEC?
Chance of help without us? 10%. – I couldn’t really conceive how this would have happened, given my specific connection to the burns unit and the poor village she came from
Long term quality of life compared to the average person if survives? 70% - This is very hard to judge, but with the disfiguration and scarring, she would have a hard time socially and she would have moderate physical disability
Chance of survival without help? 10% - Probably conservative, its hard to imagine a scenario where she survived at home, given how quickly she was deteriorating
Chance of survival with help? 25% - Although the situation looked dire, she had already survived a couple of weeks and the condition was treatable. She would need nutritional support, fantastic wound care and skin grafts. In a high-income country hospital with all the resources in the world, I would predict her chance of survival at over 70%. In retrospect her chances were probably better than 25% - the Lacor hospital care was even better than I expected.
Cost of treatment $600 – Of this we paid only 300 dollars (transport from Soroti, hospital bill, some food). The other half is the estimated cost of her treatment bourne by the wonderful Lacor Hospital which looked after her. This hospital is over 50% donor funded, so this 300 dollars reflects their contribution.[2]
The calculation...
$600 ____________ 0.9 x 0.7 x (0.25-0.10) =$6349 estimated to save a life [3]
The result Over the first couple of weeks she recovered surprisingly well, while still remaining very sick. Her wounds were dressed multiple times daily, she was started on a high protein diet and the skin infections cleared. She started to eat more and talk a little. After she stabilized, she was booked for a skin graft surgery which could solve her underlying problem. On the day of surgery she was wheeled into the operating theatre where they discovered she had high fever and so delayed the surgery until her infection cleared.
The infection never cleared, and she tragically died two days later. Hearing about her death hurt badly – as it should. But not only had I done what I could, it made sense. Those 300 extra dollars gave her a chance of not only surviving, but living a long and happy life – at least that’s what I told myself…
Sunday had it tough from the start. Her mother struggled to breastfeed her (surprisingly rare here in Uganda) and she had been feeding her a mixture of what little mil her mother managed, plus cows milk to supplement. This didn’t help at all. Sunday had severe malnutrition, the worst I have seen with a nasty chest infection to boot. Her mother knew she was in a bad state, but didn’t comprehend just how sick she was. Fortunately there was an NGO malnutrition treatment center in nearby Soroti town, which had a good reputation. The mother was so poor that she didn’t even have the money to transport the kid to the center. To my surprise though the malnutrition center was completely free, and recommended they keep the kid for a minimum of 6 weeks until she improved.
Cost-Effectiveness BOTEC
Chance of getting help without us 50% - This is a hard one, as if her mother had realized that she could just go into Soroti town and drop the kid off at the treatment center, she might have managed to raise the transport money eventually, but it seems very unlikely given she hadn’t already for many months.
Survival chance without treatment 60% - In good treatment centers here, severe malnutrition mortality rates are 10-20%, but I doubled this worst case as she was at home, not In a malnutrition center. In retrospect given how dire the situation was, this seems too conservative and maybe 40-50% would have been a better estimate
Survival chance after treatment 80%. - As this was one of the worst cases of malnutrition I’ve seen, I put this probability at the worse end of mortality in malnutrition centers
Quality of life after successful treatment 75% of normal - Although she likely will have some form of developmental delay, most kids with severe malnutrition do surprisingly well long term after successful treatment
Cost Only $300. Of this we contributed $100 to the mother for transport to and from the malnutrition facility a few times and, then made a $100 donation to the treatment center. I’ve estimated the cost of treatment at the center at around $200 in total
The Calculation...
$300 _____________
0.5 x (0.8-0.6) x 0.75 = Estimated $4000 to save a life
The result The treatment center did an amazing job and to cut a long story short she gained weight, got healthier and went home. A few months later she is still growing well and thriving, a great result. Given that malnutrition treatment is already a GiveWell target, this likely being the most cost-effective intervention shouldn’t be a surprise.
Angel – Age 7
Angel didn’t tick the boxes the same way the first two girls did. From minute one she pulled on my heart strings from the moment she walked into my consultation office and if I’m honest I would probably have helped her regardless, cost effectiveness be damned. A bright eyed seven year old girl, born with a 99.99% perfect body. But the tiny defect was a disaster, a mere 5 millimeter membrane which was enough to block her aortic valve in her heart.
Enough to make her heart not work properly Enough to stunt her growth so she was only 15kg at age 7 Enough to ruin a full and fulfilling life
Five years ago in Uganda there was no heart surgery available, so the rich politicians and business people would fly their kids to India to get operations while everyone else had no option.
I was struck by the injustice of it all – in my birth country of New Zealand the operation would have been done 5 years ago. And enraged by the absurdity of the situation – a physical membrane of just a few millimeters standing between a girl and her happy and fulfilling life. At first the Mulago heart institute quoted $4,500 US dollars for the operation, before we bargained them down to $2800 on the basis of Angel’s extreme poverty. This is as close as I’ve ever come to literally bargaining for someone’s life.
My (probably biased) BOTEC
Chance of getting operation without us 10% (probably optimistic)
Probability of survival within 10 years without operation 10% - Her heart failure is worsening, and I’m worried about a timeline of months rather than years.
Chance of operation success and a normal-ish life expectancy afterwards 60% - About 20% of aortic membranes recur after surgery in high income countries, which I arbitrarily increased to 40% given likely poorer outcomes in this case She doesn’t however have many features which increase risk or recurrence. If this operation failed there would be the possibility of another, but I kept the math simple. Her current re-modelled, possibly permanently damaged heart may mean she will have a lower life expectancy and possibly lower quality of life even after a successful operation, but this is very hard to predict and many children have normal lives after operation.
Cost $2800 for the operation, plus $300 for pre-tests and follow up
The claculation.. 3100 _________ = $6888 estimated to save a life. Not <$5000, but close enough. 0.9 x (0.6-0.1)
BUT
The operation hasn’t happened yet. Apparently there’s only one cardiopulmonary bypass machine for kids in Uganda – you can't do heart operations without it and it was broken for 3 months. They’ve just fixed it, but have a huge backlog of kids waiting for surgery, so it might be another 1 to 3 months before Angel gets hers. I hope its soon enough.
Thanks again Dall E
My reflections and mixed feelings
This whole exercise felt absurd and uncomfortable. Even discussing my experience feels off, given the immense suffering of all the sick kids and their families. Despite the obvious emotional roller coaster, the process was probably net-“fuzzy”-positive overall while coupled with enormous discomfort. What right do I have to make these kind of decisions? Why did I do 100x better than these wonderful people in the lottery of life? Also the reality of valuing of one real, hurting, living, desperate. beautiful, hurting, precious, human at only $5500 seemed not just morally wrong but almost abhorrent. I don’t think many would disagree with me that a human life should be or even “is” objectively worth far, far more. Its also easier to be two or three steps removed from a donation. And also easier to switch our donations from World Vision to mosquito nets, than to deny someone you talked to potentially life saving treatment just because the theoretical cost is $10,000 not $5500.
Complexity, problems and downsides
I didn’t account for the value of my time or emotional energy into any of these calculations. Although gained energy and a sharpened sense of purpose through this experiment, if I’m honest there was probably a small net drain on my overall time and energy
Risk of “Savior Complex"[5] formation, where I feel proud or superior to others because of these “good works”. Its interesting that this dangerous mind-state may well be more likely to form helping individuals in a less cost-effective way than it would giving money to highly effective charities
Alternatively there's the chance of negative external perception where I’m accused of being a lone-ranger White Savior. For deciding from my privileged and literally white Ivory tower who might live and die, without consulting local people on what they think about this whole initiative. I’m aware there is a pretty reasonable “Playing God” critique here with a couple of people making important decisions about who to help based on deeply flawed and loosely calculated BOTECs…
Feeling Guilty and Overwhelmed at times – almost the opposite of the Savior Complex. Guilt from denying people help that didn’t meet our bar, and a litle overwhelmed by the pressure and responsibility.
Bad decision making ,due to bias from personal connections – see Angel… Its almost impossible to stay objective when dealing with real people in real time and sometimes in person.
I was surprised how few cases emerged that seemed likely to be super cost-effective. This made this trial take a bit longer than I had hoped and made me less positive about similar projects being viable in the long term.
A Charity which cost-effectively helps individual sick people?
As well as our intention to make our own close-to-home donations more cost-effective, I had a vague idea that helping individuals sick people cost-effectively could become an effective charity. Perhaps it could be a bit like Kiva loans for effective altruists, where those who donated would know who they donated to and even the potential cost-effectiveness of each individual donation. There are many charities which pay directly for cost-effective medical care, for example fistula treatment or cataract surgery, but they are systematic entities which only treat one condition. This approach could pick up on any condition that might be cost-effective to treat, which in theory could increase the pool of people able to be helped.
After trying this on a small scale though, I’m less enthusiastic. First finding cases where the life-saving cost might meet the GiveWell bar was harder than I expected. This shouldn’t be too much of a surprise given that most charities which meet the GiveWell bar focus on prevention, which is usually more cost-effective than cure. Because identifying suitable people is tricky, the admin costs of a large scale charity might be so high as to make the work not cost-effective. Also there is a high risk of a charity along these lines being accused of “Playing God” and of “White Saviorism” given that we would deny most sick kids potentially life-saving treatment. I think this criticisms doesn’t hold much weight as we “Play God” in a similar way every time we choose a coffee over a mosquito net, but the negative perception risk remains.
On the other hand, this was a tiny side project over a couple months in one location and as a tiny trial there's no optimisation of any aspect. There might still be a way to run a cost-effective charity which helps individuals with life-saving care.
There’s no avoiding the strange juxtaposition between helping people we’re connected to, and funding/operating cost-effective organisations. The $3100 we're using to help Angel might save an estimated 0.5 lives, while for the same amount we could treat over 1800 patients at a OneDay Health center and save perhaps a couple of lives. So on one level funding Angel’s heart operation might not make logical sense, but on another level it connects us with our community, helps us love and mourn better and reminds us of the messy injustices in this strange world we're a part of. I hope I'm a better person after this wee endeavour.
We’ll probably keep doing some more cost-effective life-saving giving on a slower, less intentional rolling basis. It might just be MoreGood than not doing it all.
At the time, I calculated this lower at around $4000 and just corrected it now after I realised my mistake, so it seemed more clearly cost-effectiv at the time. However we would probably have helped regardless as it was close enough to the bar.
"Savior Complex" is a Highly recommended HBO 3 part miniseries, raising great questions about the risks of white people trying to help in low income countries. Also interesting from thinking about net positive/net negative in what we do.
420
More posts like this
166
Are our Top Charities saving the same lives each year?
Building off of Jason's comment: Another way to express this is that comparing directly to the $5,500 Givewell bar is only fair for risk-neutral donors (I think?). Most potential donors are not really risk neutral, and would rather spend $5,001 to definitely save one life than $5,000 to have a 10% chance of saving 10 lives. Risk neutrality is a totally defensible position, but so is non-neutrality. It's good to have the option of paying a "premium" for a higher confidence (but lower risk-neutral EV).
Leaving math mode...I love this post. It made me emotional and also made me think, and it feels like a really central example of what EA should be about. I'm very impressed by your resolve here in following through with this plan, and I'm really glad to have people like you in this community.
Thanks so much for the encouragment, really do appreciate it.
Great point I hadn't thought about risk neutrality vs non-neutrality here and that there might be a pool of people even within EA who would rather pay a "premium" for higher confidence. Outside EA my experience has been that perhaps even the majority of people would probably prefer to pay for higher confidence.
I agree with your last sentence, and I think in some versions of this it's the vast majority of people. A lot of charity advertising seems to encourage a false sense of confidence, e.g. "Feed this child for $1," or "adopt this manatee". I think this makes use of a near-universal human bias which probably has a name but which I am not recalling at the moment. For a less deceptive version of this, note how much effort AMF and GiveDirectly seem to have put in into tracking the concrete impact of your specific donation.
Orthogonally, I think most people are willing to pay more for a more legible/direct theory of impact.
"I give $2800, this kid has lifesaving heart surgery" is certainly more legible and direct than a GiveWell-type charity. In the former case, the donor doesn't have to trust GiveWell's methodologies, data gathering abilities, and freedom from bias. I've invested a significant amount of time and thought into getting to my current high level of confidence in GiveWell's analyses, more time than most people are prepared to spend thinking about their charitable donations.
And I think most people -- including myself -- have a prior that projections and analyses of all sorts tend to be overinflated in comparison to reality. How many building projects come in on time and under budget? How many IT projects? The less complex the theory of impact, the less the lightly-researching donor will end up implicitly discounting the organization's claims on account of their background skepticism of projections and analyses.
I think that’s such a good point about risk neutrality - and when looking outside of EA spaces this is important to keep in mind, I think risk neutrality is so accepted in EA spaces it’s sometimes forgotten that outside it is not necessarily so.
Also, just a really moving post. Trying to do the best you can in an emotional yet rational way. I really appreciate this.