The GiveWell Blog

The case against disaster relief

(updated on: ) | by Holden

When a natural disaster and humanitarian crisis hits the headlines, many of us (including me) reach straight for our wallets. Emergencies have an easier time getting our attention (and emotional investment) than the chronic health problems that plague the developing world every day. But to hear the Disease Control Priorities report tell it, emergency aid is one of the worst uses of donations, despite being one of the most emotionally compelling.

The full discussion is on pages 1147-1161 of the report. A couple highlights:

The immediate lifesaving response time is much shorter than humanitarian organizations recognize. In a matter of weeks, if not days, the concerns of both the population and authorities shift from search and rescue and trauma care to the rehabilitation of infrastructure (temporary restoration of basic services and reconstruction). In Banda Aceh, Indonesia, after the December 2004 tsunami, victims were eager to return to normalcy while external medical relief workers were still arriving in large numbers.

Even if a donation is made minutes after a disaster, it might not be used in any meaningful way until it’s too late for emergency relief. Another reason to favor organizations with staff already on the ground.

Several specific emergency interventions are criticized for high costs and low or negative effects, including mobile hospitals:

The limited lifesaving usefulness of foreign field hospitals has been discussed. Again, the lessons learned from the Bam earthquake are clear. The international community spent an estimated US$10.5 million to dispatch approximately 10 mobile hospitals, which arrived from two to five days after the impact, long after the last casualty had been evacuated to other Iranian provinces.

And search-and-rescue operations (particularly those not carried out by locals):

Few developing countries have established the technical capacity to search for and attend to victims
trapped in confined spaces in the event of the collapse of multistory buildings. Industrial nations routinely dispatch search
and rescue (SAR) teams. Costs are high and effectiveness is reduced by delayed arrival and quickly diminishing returns.
Following the 1988 earthquake in Armenia, in the former Soviet Union, the U.S. SAR team extracted alive only two victims at a cost of over US$500,000. In Turkey in 1999, 98 percent of the 50,000 people pulled alive from the rubble were salvaged by relatives and neighbors. In Bam in 2003, the absence of high-rise and reinforced concrete buildings ruled out the need for specialized teams. Nevertheless, according to UN statistics, at least US$2.8 million was spent on SAR teams. An alternative solution consists of investing these resources in building the capacity of local or regional SAR teams—the only ones able to be effective within hours—and training local hospitals to dispatch their emergency medical services to the disaster site.

The report is also harsh on in-kind donations, which it says are “not only are of limited use, but often cause serious logistic, economic, and political problems in the recipient country” due to warehousing issues.

The report’s bottom line is that “emergency relief is “one of the least cost-effective health activities,” and no substitute for (a) disaster preparedness (discussed on pgs 1158-9); (b) proven interventions to deal with chronic, everyday health problems.

I should note that this chapter is less thoroughly referenced than most others in the report, although this is likely because emergencies are a bad environment for meticulous study (and so evidence must be informal and observational instead). Having read it, I’m personally hesitant to give to disaster relief again. I’d rather up my donations to projects that aim to strengthen everyday health infrastructure for those in chronic need. I do feel an emotional pull to try to help when disaster strikes, and I feel this pull more strongly in the aftermath of the headline than contemplating it in the abstract – but I also agree with the DCP report’s emphasis on using limited funds as well as possible:

The willingness to spend hundreds of thousand of dollars per victim rescued from a collapsed building in a foreign coun-
try is a credit to the solidarity of the international community, but it also presents an ethical issue when, once the attention has
shifted away, modest funding is unavailable for the mid-term survival of tens of thousands of victims.

Cost-effectiveness is in the eye of the beholder

(updated on: ) | by Holden

Broadly-speaking, we think of “cost-effectiveness” as referring to how much of value is accomplished for a given amount of money. If two interventions are both proven and scalable to similar degrees, the more cost-effective one is a better investment because it allows the same donation to accomplish more of value. However, “value” means different things to different donors: to some, $200 to save a child’s life might be a great deal, while others may prefer $450 to repair a fistula.

We prefer to leave major judgment calls to our donors when practical. This means that rather adopt a single definition of value (such as Disability-Adjusted Life-Years), we hope to find the most cost-effective interventions for several different definitions of value. We can’t cover every possible notion of what’s worth funding, but we hope to be of use to as many donors as possible within the cause of developing-world health (our current focus – we will be moving on to economic empowerment later).

The following are several philosophical goals – i.e., definitions of what results make an intervention “valuable” – that we think will appeal to many donors.

  1. DALYs averted. Although it isn’t our favorite measure of value, the Disability-Adjusted Life-Year is a widely used metric that considers all forms of mortality and morbidity. Some donors may feel most comfortable aiming to avert as many DALYs as possible for their donation.
  2. Economic benefits. Health problems impose an economic burden, not just a moral one. There are sometimes attempts (such as the “benefit:cost ratio” used by the Copenhagen Consensus and some versions of the social return on investment metric) to combine economic and moral benefits into a single figure, measured in dollars.
  3. Life-years saved – for those who put a lot of weight on being alive vs. not alive (and less weight on quality of life). Interventions that focus on infant mortality are likely to be cost-effective in terms of saving life-years.
  4. Lives saved.
  5. Adult lives saved. It is common to value adult lives more than children’s lives. In addition, adults are more likely to have dependents, making their deaths arguably more tragic (in a way that DALYs could capture in theory, but that DALY estimates generally don’t capture in practice).
  6. Cases of extreme suffering prevented/rectified. This goal has several subcategories, for different conceptions of what constitutes extreme suffering. One that jumps to mind is fistula (and other deformities associated with ostracization).
  7. People brought to a normal standard of health and potential productivity. In some ways the opposite of the cause immediately above, in that it focuses on helping those with high potential rather than helping those with high need. (There are different places one could draw the line for “normal health and potential productivity” – is it enough to prevent/cure someone’s blindness, or is it also important that they be adequately nourished and have reasonable job opportunities?)
  8. Unwanted pregnancies averted / population growth slowed. Some donors might see births averted as a negative; others might feel that it is the key to better quality of life and sustainability.

Some of these metrics are highly well-defined and relatively easy to find or form cost-effectiveness estimates on. Others are far less so. We aren’t necessarily going to be conducting a separate search on interventions for each, but we think it’s productive to get as many appealing “goals” on the table as possible before beginning to narrow down the programs we’re focusing on. We hope that we won’t leave out programs whose potential effects are highly appealing to a large/important set of donors, even if they fail a “cost-effectiveness test” based on other metrics.

If you feel that we’re leaving out any important ones, please let us know. (We will also be consulting with our supporters and Advisory Board on this question.)

Donated technological equipment

| by Elie

Business Week writes:

According to the World Health Organization, about half of the imaging equipment sent to developing countries goes unused because local technicians aren’t trained to operate it or lack the necessary spare parts.

(H/t Aman at THDBlog)

Is this possible?

(Note: I tried to find the original WHO source and all I could find was this article stating that “more than half of the medical equipment in developing countries is left unused or broken because it is too complicated or expensive to operate and repair,” which is substantially different.)

Selecting program-based health interventions

| by Holden

We are doing both region-based and program-based research on developing-world health. First I will discuss our program-based research, which will focus on (a) finding particular interventions (and/or clusters of interventions) that appeal to us; (b) finding organizations that implement these interventions “vertically,” i.e., replicating the same basic program across a variety of regions.

For these purposes, we particularly value interventions that are proven, scalable, and cost-effective.

Unlike most foundations, we are seeking explicitly to serve individual donors – donors who don’t have personal connections to particular organizations or expertise in the issues, and who therefore are better suited to expanding what already works (i.e., interventions that are proven and scalable) than to exploring unproven innovations. More on this idea at our FAQ, as well as in this blog post.

We think of an intervention as proven when:

  • It has been previously carried out and carefully, publicly evaluated (often through academic research) in a way that provides strong empirical evidence for its positive impact on people’s lives. (A future post will further discuss our position on the “evaluation hierarchy” and what sorts of evidence we think are necessary under different circumstances.)
  • The conditions under which it has been evaluated match the conditions under which it is likely to be carried out again, in as many relevant ways as possible.

We think of an intervention as scalable when:

  • There is a significant amount of unmet “need” for the intervention, i.e., conditions under which the intervention would be helpful but has not yet been funded.
  • Donations can be used to replicate the intervention in a variety of areas, while recording enough information about its execution to be reasonably confident that it is working as intended.

Defining “cost-effectiveness” is significantly more complex, and will be discussed in the next post.

Region-based vs. program-based approaches to developing-world health

(updated on: ) | by Holden

Deciding where to give involves making major judgment calls: decisions that rest on subjective and otherwise highly debatable claims (such as the decision of which sort of life change to aim for). We have no pretense of being able to make such judgment calls “objectively” or “perfectly. Rather, we try to:

  • Be explicit about which values we are pursuing and which judgment calls we’re making. For example, we’ve already declared our preference for funding proven and repeatable interventions, and we’ve declared our decision to focus on developing-world direct aid for the coming year.
  • When practical, leave major judgment calls to our donors rather than making them ourselves, by recommending a variety of charities that are strong according to different criteria. This approach is especially important for judgment calls that our donors haven’t agreed to in advance (unlike the two listed directly above).

We are currently focusing on health interventions for the developing world (we will research other aspects of developing-world aid later in our process). One of the major judgment calls involved in choosing a health intervention is the decision between a region-based approach and a program-based approach to giving.

Taking a region-based approach means focusing on a particular part of the world; learning as much as possible about the people who live there, the opportunities they have, and the problems they face; and then finding a program that is well-suited to addressing the particular needs of this region. For regions with many interrelated health problems, such a program will likely be one that aims to strengthen the general quality of health care in the region, which will make it possible to address many health issues at once. (One such program is Partners in Health, a recommended charity from our last round of research that focuses on bringing full-service health care systems to poor rural areas.)

Taking a program-based approach means focusing on a particular intervention (or cluster of interventions); learning as much as possible about the conditions under which this intervention has been shown to improve lives in the past; and then finding a program that replicates this intervention across many regions. (This is the approach we took in evaluating Population Services International, which markets materials such as condoms and bednets across the world.)

A region-based approach has the advantage that it is more likely to be well-fitted to the particular people it serves and their needs. A program-based approach, however, may in some cases be a simpler and/or more cost-effective way to address a particular health problem, which means that it may be a more efficient and reliable way of changing lives for the better.

We will be taking both approaches, and presenting the options for donors that correspond to each. In future posts, we will discuss the specifics of (a) our region-based approach, i.e., how we will be choosing one or more countries to focus on; (b) our program-based approach, i.e., how we will be choosing one or more interventions to focus on.

DALYs and disagreement

(updated on: ) | by Holden

Elie thinks that fistula is worse than death. jsalvati disagrees.

I’d rather bring someone to full health than save an infant’s life. Ryan agrees, but Basti does not and Ron Noble emphatically does not.

It’s possible that we would all agree if we knew more about the lives of people in the developing world, or if we just had a long enough to argue about our values. It’s also possible that we wouldn’t. And as long as we disagree, we’ll have different opinions on what the most “cost-effective” interventions are. For example, if it’s true that fistulas can be repaired for $450 each, is this a better or worse use of donations than preventing children’s deaths for $200 each through vaccinations? My answer would be “It depends on the donor.”

Converting disease burdens and intervention benefits into DALYs doesn’t resolve questions like this. Rather, it obfuscates them, by converting the two interventions into the same terms using a single set of philosophical values. If the numbers above ($200/death averted for vaccinations, $450/surgery for fistula) are accurate, they allow different donors to make their own judgment calls, while being informed about their options. But these aren’t the numbers you’ll find in the Disease Control Priorities Project’s summary tables; instead, you’ll see only that surgical services cost an average of $136 per DALY averted (Jamison et al. 2006, Pg 75) and that the vaccinations interventions costs an average of $7 per DALY averted (Jamison et al. 2006, Pg 77).

Some simplification and information loss is necessary in order to compare different options, but reducing everything to a single unit means being able to serve only a single kind of donor. I’d prefer to estimate the effect of different interventions on a variety of “life outcomes” that different donors might value differently. We will discuss this variety more in a future post, but here’s a quick list:

  • Total life-years saved.
  • Adult lives saved (as it is common to value adult lives more than children’s lives).
  • Cases of extreme misery, such as fistula or perhaps severe elephantiasis, averted.
  • People brought to a “normal” level of health, i.e., without any debilitating nutritional or other conditions.

All of these things need to be separately estimated to produce DALY estimates. The DCP report did so with admirable thoroughness and far more people than we have (Jamison et al. 2006, Pgs xxiii-xxxiv). Yet because they published only their DALY estimates (not, with some exceptions, the estimates of different health problems that went into them), they buried a great deal of this work, and produce cost-effective estimates that are useful only if you’re completely on board with all of their values (from how bad each disability is to how to value different years of life). We’re currently trying to get in touch with the authors so we can get access to more of the details; if we don’t, we’ll have to repeat much of their work (with less capacity to do so).

Sources

Previously in series: