[This is the second post in a brief series explaining the current GiveWell top charities. You can get all the information in this post on GiveWell’s website, but my blog post is both shorter and less boring.]
Helen Keller International performs Vitamin A supplementation.
Vitamin A deficiency can cause stunting, anemia, blindness, more severe infections, and death. It is particularly harmful to pregnant or lactating people, infants, and children. People in developing countries who don’t eat much meat and don’t eat Vitamin-A-fortified food may suffer from deficiencies. (Vitamin A is also a nutrient of concern for vegans and vegetarians in the developed world, but if you have an ample and diverse diet of fruits and vegetables you probably don’t have to worry about it.) Luckily, your liver can store Vitamin A for several months; if you take a very high dose of Vitamin A once every six months, you won’t suffer from Vitamin A deficiency. It is recommended that children between 6 months and 5 years old get a supplement once every six months.
Helen Keller International provides technical assistance, advocacy, and funding. Technical assistance includes helping countries monitor how many children are getting Vitamin A supplements, running campaigns that educate parents about the importance of Vitamin A supplements, training health workers to give out supplements, and helping governments figure out why their vitamin A supplementation rates are so low and how they can fix them. Advocacy involves convincing governments that they should prioritize mass vitamin A supplementation. Helen Keller International also sometimes provides grants to governments to help them pay for vitamin A supplementation programs.
It is somewhat unclear whether vitamin A supplementation actually works to reduce mortality rates. Normally, this sort of thing is unclear because we don’t have any evidence about it. In the case of vitamin A supplementation, we have two pieces of really good evidence; they just point in opposite directions.
The Cochrane Collaboration, whom you may remember from the previous post as the people who aren’t sure whether flossing your teeth makes your teeth better, performed a meta-analysis that suggests that vitamin A supplementation reduces all-cause mortality by 24%, with a 95% confidence interval ranging from 17% to 31%. The Cochrane Collaboration primarily used studies that were conducted in the 1980s and 1990s.
Not long after the Cochrane Collaboration’s meta-analysis came out, we learned the results of the DEVTA trial. The DEVTA trial is the single largest randomized controlled trial ever conducted, with one million children participating. It estimated that Vitamin A supplementation reduces child mortality rates by 4% and could not rule out the possibility that it did not affect child mortality rates at all. You may notice that 4% is in fact much much smaller than 24%.
What the fuck is going on?
- It might just be random chance. That’s pretty unlikely: the Cochrane Collaborations 95% confidence interval doesn’t overlap at all with DEVTA’s.
- DEVTA might not have treated as high a percentage of the children in the study as claimed. DEVTA claims to have treated 86% of children, but some researchers are skeptical because DEVTA was done very cheaply. DEVTA seems to be using broadly reasonable strategies to get all children to take vitamin A supplements and to figure out how many children actually took it, but their strategies aren’t very well-documented and sometimes they didn’t implement them until halfway through the study. Nevertheless, the percentage of children treated in order to make ‘they didn’t treat enough kids’ plausible as an explanation is so much lower than the percentage of children claimed to have been treated that this is not a very plausible explanation.
- DEVTA might have treated a population with less severe or prevalent vitamin A deficiency. However, the rate of vitamin A deficiency, severe vitamin A deficiency, and complications related to vitamin A deficiency is similar in DEVTA as it is in other studies. It’s more likely they underestimated vitamin A deficiency than that they overestimated it.
- DEVTA’s population might be healthier than other populations. Vitamin A deficiency doesn’t generally kill children directly; it kills them indirectly, by making them more susceptible to infections. If those deaths are being prevented some other way (e.g. measles vaccianations, oral rehydration treatment), treating Vitamin A deficiency saves fewer children’s lives. DEVTA had a lower child mortality rate than most of the studies in the Cochrane review, and generally studies with a lower child mortality rate show a smaller effect from vitamin A supplementation. However, it’s unclear whether measles and diarrhea– the two biggest killers related to Vitamin A– were less common in DEVTA than in other studies.
GiveWell thinks the most likely explanation is the last one. That means that whether Vitamin A supplementation is cost-effective depends on not just how high the Vitamin A deficiency rates are but also how high the child mortality rate is.
The countries Helen Keller International works in typically have lower rates of vitamin A deficiency than in any study of the effects of vitamin A supplementation: they work in countries where 20% of preschool-aged children have vitamin A deficiency, compared to 59% in the Cochrane meta-analysis. Hellen Keller International works in countries where 12 children out of every 1000 die every year; previous studies have found an effect of vitamin A supplementation if more than 10 children out of every 1000 die every year.
However, there are a lot of limitations of this estimate. Helen Keller International typically works in regions, rather than in whole countries, which might have higher or lower child mortality rates than the country as a whole. It seems really unlikely that vitamin A supplementation doesn’t do anything below 10 children out of 1000 dying each year and then suddenly has a big effect as soon as you get to 10; it’s probably a smoother effect that’s harder to analyze. “Child mortality” is a statistic that includes a lot of different things. It’s unclear whether vitamin A supplementation helps with all infectious diseases or a subset, such as measles and diarrhea; it’s pretty clear that vitamin A supplementation has no effect on some other causes of child mortality, like car accidents. If a region has a high child mortality rate because there are a lot of car accidents, vitamin A supplementation might not do anything. More realistically, if vitamin A supplementation reduces deaths by causing children to be less likely to die if they get measles, then if a region has a low rate of measles, it won’t have a big effect from vitamin A supplementation, even if it has a high child mortality rate.
There are not likely to be any negative long-term side effects from vitamin A supplementation. In the short term, less than ten percent of children experience some sort of adverse side effect, such as headaches, nausea, vomiting, irritability, fever, or loose stools. Vitamin A supplementation does not cause vitamin A overdose or increase mortality when given alongside an inactivated vaccine.
Helen Keller International gives a relatively high percentage of target children vitamin A supplements (between 46% and 81%, depending on region).
GiveWell believes Helen Keller International’s grants cause vitamin A supplementation distributions that otherwise would not occur, but does not know whether its technical assistance helps countries to give vitamin A to children who otherwise wouldn’t receive vitamin A. For this reason, GiveWell’s analyses include only the effect of grantmaking, not the effect from technical assistance.
Helen Keller International needs $20.6 million over the next three years.
Vitamin A supplementation is only one program which Helen Keller International implements. GiveWell recommends donating to Vitamin A supplementation and does not recommend donating to their other programs.
One big area of uncertainty comes from the fact that Helen Keller International has been investigated in less detail than other top charities. In general, over time, GiveWell tends to become more uncertain about charities, learn more about their limitations, and have a higher cost-per-life-saved-equivalent number attached to the charity.
Why might you donate to Helen Keller International?
- You want to save the lives of children under 6.
- You want to donate to something that definitely won’t cause significant harm, even if it might not have an effect.
- You’re optimistic about GiveWell’s further investigations finding that Helen Keller International is as effective as we thought, not less effective.
- You think Helen Keller International’s technical assistance is more likely to be effective than GiveWell thinks it is.
- You’re not too concerned about low rates of vitamin A deficiency or about the uncertainties associated with child mortality rates.
> In general, over time, GiveWell tends to become more uncertain about charities,
Did you mean “less uncertain about charities”?
LikeLiked by 1 person
Don’t think so – I think the claim is that GiveWell has historically been overconfident about new charities as it adds them, and then discovered unexpected limitations and walked back on its cost effectiveness estimates (i.e., “become more uncertain” about their effectiveness).
Implication being that it might be reasonable to be concerned about that happening with HKI (since it’s a new charity that GiveWell is bullish on, but has some pending concerns).
LikeLiked by 2 people
Nancy Lebovitz said:
I’m wondering whether the children enrolled in trials might be getting better care on the average than children not enrolled in trials.
LikeLiked by 1 person
>vitamin A supplementation reduces all-cause mortality by 24%, with a 95% confidence interval ranging from 17% to 31%
What does it mean?
Variant A. Using some probabilistic model, p(17% ≤ all-cause mortality reduce ≤ 31% | data) = 95%.
Variant B. Using some probabilistic model, all-cause mortality reduce rate is between 17% and 31% if and only if the probability to get data at least as extreme as we actually have (given the reduce rate) is less than 5%.