Note: This page summarizes the rationale behind a GiveWell grant to Johns Hopkins University. Johns Hopkins staff reviewed this page prior to publication.
In a nutshell
In August 2025, GiveWell recommended a $545,780 grant to the Johns Hopkins CA CODE (Child and Adolescent Causes of Death Estimation) team to produce state-level estimates of causes of death for children ages 0-5 in Nigeria from 2000-2024. We recommended this grant because:
- Nigeria receives substantial GiveWell funding across multiple states and health interventions, and we rely on cause-specific mortality estimates when evaluating these different programs.
- When two major estimation groups (IHME and IGME) have produced comparable estimates in other countries, they've sometimes differed substantially. Having multiple sources of estimation could improve our decision-making.
- The researchers have committed to producing a transparent, accessible explanation of their methodology. This will help us and others vet their estimates.
Our main reservations are that the new estimates could be lower quality than existing ones, and that this grant alone won't explain why different estimates might disagree. (More)
Published: October 2025
Background
At GiveWell we use estimates of disease burden in our evaluations of the cost-effectiveness of different health programs. For example, to assess malaria prevention programs, we need to know how many children die from malaria in areas where we might fund interventions.
Several different research groups produce these estimates: IHME (Institute for Health Metrics and Evaluation), IGME (UN Inter-agency Group for Child Mortality Estimation), and CA CODE (Child and Adolescent Causes of Death Estimation), a research group based at Johns Hopkins University. IHME produces both cause-specific and all-cause child mortality estimates; IGME produces only all-cause child mortality estimates; CA CODE produces cause specific mortality estimates. Only IHME produces state-level cause-specific mortality estimates for Nigeria.
Until recently, we typically relied solely on IHME for all disease burden estimates. However, since completing a project looking into different burden of disease estimates, we couldn’t find strong reasons to defer entirely to IHME, and so we plan to use IHME, IGME, and CA CODE estimates going forward. We generally expect blending estimates to get us closer to the truth vs. relying on one set alone.1
The grant
The CA CODE team at Johns Hopkins produces cause-specific child mortality estimates at the national level.2 This grant will fund them to extend their work to produce state-level estimates for Nigeria's 36 states and Federal Capital Territory.
Over 18 months, starting in fall 2025, the team will:
- Compile and analyze existing data sources on child deaths in Nigeria
- Develop statistical models to estimate causes of death by state
- Conduct a dissemination workshop with the Nigeria National Population Commission3
- Publish final estimates and a detailed methodology report
The research team includes nine researchers, led by Principal Investigator Li Liu.4 All estimates and methodology will be made publicly available.5
The case for the grant
We're making this grant primarily because collecting more information about causes of death could improve our burden of disease estimates in Nigeria, which could improve our funding decisions in our largest grantmaking geography.
Blending burden of disease estimates has already unlocked grant opportunities we wouldn’t have otherwise made. For example, in Chad, CA CODE estimates that malaria causes 14% of child deaths, while IHME estimates 6%.6 When we began incorporating CA CODE estimates into our analysis last year, this difference led us to allocate an additional $29 million to malaria programs there.7
The CA CODE team has agreed to produce a "layperson's summary" explaining their methodology in accessible terms.8 This will make it easier for us and others to vet their estimates. If estimates differ substantially between IHME and CA CODE, this could provide an opportunity to better understand why these differences exist, which is something we’ve struggled with in the past. It’s possible this could serve as an example of transparency for other organizations producing similar estimates.
Risks and reservations
Our main concern is that blending estimates might not improve accuracy if one set of estimates is significantly lower quality. While we generally believe that incorporating multiple perspectives improves decision-making, there's a risk that adding lower-quality estimates could lead to worse funding decisions.
Additionally, while we expect this grant to shed light on CA CODE’s methodology, this grant will not directly compare the methodologies of IHME and CA CODE or explain why IHME and CA CODE estimates might differ. If the estimates disagree substantially, we might need to do additional work to understand which estimates are more reliable for specific uses.
Plans for follow-up
We plan to:
- Work with the research team on the format and content of the layperson's methodology summary
- Review preliminary results at the 15-month mark
- Monitor whether the estimates lead to changes in our funding allocations
Internal forecasts
For this grant, we are recording the following forecasts:
| Confidence | Prediction | By time | Resolution |
|---|---|---|---|
| 80% | CA CODE will deliver preliminary estimates on time. | Oct 2026 | |
| 70% | CA CODE's malaria cause-share estimates will differ from IHME's by >50% in at least one state. | Oct 2026 | |
Our process
This opportunity arose through GiveWell's work on improving burden of disease estimates.9 We evaluated it through conversations with:
- The CA CODE research team
- David Blazes and Laura Lamberti at the Gates Foundation10
- Greg Roth at IHME11
- Nigeria National Population Commission12
We also developed a threshold analysis suggesting the grant would meet our cost-effectiveness bar if it has a 10% chance of changing $8 million in annual funding allocations.13
Sources
- 1
For more on how we estimate disease burden, see this page.
- 2
“Cause-specific mortality estimates produced by the CA CODE project have now been incorporated for the first time, aiming to start a dialogue with countries about their mortality data to improve cause-specific estimates while increasing data transparency and use at the country level.” Villavicencio et al. 2024.
- 3
The Nigeria National Population Commission is the main government agency responsible for mortality estimation.
- 4
Johns Hopkins CA CODE team, budget submission to GiveWell, 2025 (unpublished).
- 5
Johns Hopkins CA CODE team, correspondence with GiveWell, 2025 (unpublished).
- 6
GiveWell, Analysis of IHME and CA CODE estimates for Chad, 2024. CA CODE estimates under-5 malaria mortality about 2.5 times higher than IHME.
- 7
GiveWell, Guidance on Burden, 2025: "For example, in Chad, incorporating CA CODE estimates led us to increase our estimate of malaria mortality by about 70%. This caused us to make a $3 million grant to seasonal malaria chemoprevention (SMC) in Chad and $25 million grant in insecticide-treated nets (ITNs) in Chad that we likely would not have made otherwise."
- 8
Johns Hopkins CA CODE team, communications with GiveWell, 2025 (unpublished).
- 9
Improving burden of disease estimates was a key priority for GiveWell's cross-cutting team in 2025. The main result of this focus was updated guidance for the GiveWell research team to refer to when using burden of disease estimates.
- 10
- David Blazes is the Deputy Director of Enteric and Diarrheal Diseases at the Gates Foundation
- Laura Lamberti is the Deputy Director of Epidemiology & Surveillance Research & Development in MNCNH & Family Planning at the Gates Foundation
- 11
Greg Roth leads IHME's client services group.
- 12
Nigeria National Population Commission, conversation with GiveWell, 2025 (unpublished).
- 13
GiveWell, Threshold analysis for CA CODE grant, 2025. Given annual Nigeria grantmaking of $135 million, an $8 million change represents approximately 6% of allocations.