The GiveWell Blog

“Straightforward” interventions

As discussed previously, we are looking for (program-based) interventions that are both proven and scalable. The Disease Control Priorities report lists many interventions that are “proven,” in the sense that one or more studies have been done indicating that the program has improved health outcomes in the region. However, showing that a program has worked once – or even several times – is far from showing that it will work again. Not only could the context, circumstances, and clients change, but the people running the program (and the scrutiny they’re under) most likely do change from a studied experiment to a less thoroughly tracked replication.

In examining the case for different interventions (see the list of interventions here), we have noted that some types of programs are inherently more “straightforward” than others, in the sense that a lower burden of proof is necessary to extrapolate from their past effects to their future effects. For example:

  • Many programs focus on educating clients, and their success thus depends on sustained and difficult-to-track behavior change from locals. Even if a sanitation program has successfully reduced diarrhea incidence rates in one region, bringing it to a new region means dealing with a new group of people, who may respond completely differently to the same education techniques. Without thorough and ongoing monitoring and evaluation, we feel it is difficult to be confident that a program along these lines is continuing to work.
  • Other programs, such as DOTS for tuberculosis, focus on improving medical care. They may depend on sustained behavior change from local medical professionals, but not from the population in general. The success of such programs will depend heavily on the existing health care infrastructure, qualifications of health professionals, and quality of training. We feel it is difficult to have confidence in such programs without continually tracking patient outcomes.
  • Other programs, such as vaccinations, do not rely on sustained behavior change from anyone in the local population. A vaccination campaign in a given region can be carried out fairly infrequently (i.e., once a year) as long as the vaccines administered are legitimate and a large number of people can be gathered and treated. Both of these factors are relatively straightforward to track and report.

Below are all the interventions we’re aware of that fit in the last category: interventions that can be carried out relatively infrequently (i.e., every 6-12 months), and for which all necessary behavior change can be directly observed and reported by the people carrying them out. For each intervention, we briefly characterize what information would need to be reported in order to reasonably extrapolate from the intervention’s past, studied effects on health outcomes.

  1. Intervention: vaccination campaigns. Information needed: number and age of people vaccinated; disease incidence/prevalence estimates for the region.
  2. Intervention: Mass drug administration programs, including albendazole for treating helminths and ivermectin for treating onchocerciasis and lymphatic filariasis (more on these conditions in future posts). Information needed: number and age of people treated; disease incidence/prevalence estimates for the region.
  3. Intervention: Vitamin supplementation programs, with nutrients that need be taken only infrequently (particularly vitamin A). Information needed: number and age of people treated; estimates of vitamin A deficiency prevalence for the region.
  4. Intervention: Vitamin fortification programs, such as the iodization of salt. Information needed: amount and circulation of food fortified; estimates of iodine and other deficiency prevalence for the region.
  5. Intervention: One-time surgery programs, along the lines of Interplast’s surgical team trips. Information needed: Condition treated for each client; completeness of surgery (before-and-after pictures would capture this information).

In labeling interventions “straightforward,” we are not claiming that they are easy to carry out or that they will always work. However, all else equal, we find them more promising for our purposes than other program-based interventions. This is largely because our experience to date with developing-world aid has shown that thorough, high-cost monitoring and evaluation (of the sort that could track sustained behavior change, for example) is relatively rare. We believe we are most likely to be able to confidently recommend interventions along the lines of those listed above, for which the necessary burden of monitoring and evaluation is lower.