Lookback: Grant to Evidence Action’s Deworm the World Program in India (September 2025)

Published: January 2026

Background

In 2013, GiveWell began supporting Evidence Action's Deworm the World program in India, which provides technical assistance to the Ministries of Health and Education to increase coverage of National Deworming Day (NDD) in targeted states.1 Since then, we have directed approximately ~$10 million to support this program across multiple Indian states.2

We made these grants because we thought Evidence Action's technical assistance was a cost-effective way to reduce the burden of parasitic worm infections among school-age children, and that many states would be unlikely to achieve high deworming coverage without this support.3
Evidence Action's technical assistance includes:4

• Monitoring procurement contracts for albendazole to prevent supply disruptions
• Training teachers and health workers in drug administration
• Conducting independent coverage surveys to inform targeting of follow-up activities
• Providing general logistical support to ensure schools and districts don't fall through the cracks during implementation.

Evidence Action started working in different states at different times. Below is a map outlining their activities in India, some of which were supported without GiveWell’s support:5

Would we have made these grants again, knowing what we know now?

Yes, though with important caveats about our confidence level. Our assessment suggests these grants likely achieved cost-effectiveness around our current funding threshold (~10x our benchmark) (more) based on reviewing updated coverage estimates (more). We also received positive feedback from government stakeholders about Evidence Action's contributions (more). However, significant data quality issues limit our confidence in this assessment (more).

If we were making similar grants today, we would invest more in baseline data collection and independent verification systems from the outset. (more)

How did implementation go?

Coverage and prevalence estimates

Coverage

We relied on two sources to estimate how deworming coverage changed after we made these grants: independent surveys commissioned by Evidence Action and coverage estimates published by the government of India.

We don’t have baseline coverage estimates from the independent surveys, which makes it hard to know how much coverage changed before vs. after the TA engagement. We discuss why we didn’t fund baseline data collection here.

For five states where Evidence Action has worked,6 we’ve seen government coverage estimates from before vs. after TA. This data suggests that NDD coverage improved after Evidence Action’s TA started. If we aggregate across states (weighted by population), coverage looks to have increased from ~55% (before TA started) to ~80% three years after.7

Naively, this data suggests Evidence Action’s TA was associated with a 25pp increase in coverage after three years. If we plug this into a retrospective cost-effectiveness model, we get a cost-effectiveness estimate of ~25x,8 well above our current cost-effectiveness threshold.

However, this doesn’t take into account questions about counterfactual coverage and measures of deworming coverage:

• Counterfactual coverage: First, this interpretation assumes coverage would have stagnated in the absence of Evidence Action’s support, which would be an incorrect assumption if coverage was increasing anyway. Importantly, Evidence Action’s TA was not randomized, as they deliberately targeted states with low baseline coverage and high prevalence of worm infections to begin with.9 Coverage in these states may have improved faster in the absence of TA, as the government may have devoted more resources or effort in ensuring these states catch up.
• Measurement: To validate government reported measures of coverage, Evidence Action also commissioned coverage surveys where independent survey companies visited a randomly selected sample of schools just after deworming day. During these visits, enumerators will look at NDD registers, but also interview the headteacher and three randomly selected children, to back check whether the reported data is accurate. Our understanding is that these data don’t line-up particularly well with government data, although there doesn’t appear to be any directional bias (i.e., one source estimating consistently higher coverage than the other).10 We are unable to share this data publicly.

Prevalence

The Indian Council of Medical Research (ICMR) - National Institute of Epidemiology (NIE) and the Post-Graduate Institute of Medical Education and Research (PGIMER) in Chandigarh, supported by Evidence Action, have collected data on the prevalence (percentage of children with eggs in stool) and intensity (eggs per gram of stool) of worm infections before and three to five years after Evidence Action’s TA support. Separately, the National Centre for Disease Control (NCDC) conducted similar surveys across non-TA and TA states. Both surveys followed the same general approach: stool samples were taken from randomly selected schoolchildren and then assessed for worms and eggs under a microscope.11

We’ve found significant discrepancies in results across states where both the ICMR-NIE / PGIMER and the NCDC have conducted follow-up impact assessment surveys.12 Generally speaking, the ICMR-NIE / PGIMER show significant reductions in the worm burden; whereas the NCDC surveys show much more muted declines.

We are very unsure what to make of this discrepancy, and do not have permission to publish results from either survey. The ICMR-NIE / PGIMER surveys (supported by Evidence Action) have produced three peer-reviewed publications (here, here, and here), which gives us reasonable confidence that they followed good protocol. We’ve not been able to interrogate the methodology behind the NCDC estimates as these are not public.

Discussions with local stakeholders

We visited this program in India September 2 to September 6, 2024. We spent three days in Delhi speaking to Evidence Action staff involved with the running of the program, and two days in Uttarakhand (one of the states where we support TA), where we spoke with officials at various levels of the Indian government (block, district, and state).13 Each of these officials was directly involved in NDD, and at the Uttarakhand Department of Health we spoke to the civil servant directly responsible for its implementation.

Generally, the officials we spoke to spoke positively about Evidence Action’s support, and mentioned several bottlenecks that they thought their TA had helped alleviate:14

• Procurement: Evidence Action staff calculate when albendazole15 procurement contracts are due to expire and flag to government officials when they need to start the tendering process. Officials told us that this type of support made it less likely for expiring contracts to fall through the cracks.
• Logistics: Evidence Action helped develop the standard-operating procedures for NDD and runs training of block and district health and education officers, and “nodal” teachers, who are responsible for overseeing deworming on the day itself. They also support in designing informational materials (e.g., posters about NDD) and are on-hand to support last-minute logistics in the lead up to the day itself (e.g., flagging if particular schools haven’t received deworming pills yet). A state official we spoke to cited logistics support as something particularly helpful. This official said that they have 20 different health programs in their portfolio, so having Evidence Action on hand to support the logistics of NDD frees up her time to focus on other programs.
• Coordination: Evidence Action organizes biannual meetings between the Department of Health and Education, and officials said these wouldn't happen without Evidence Action. Officials said these were helpful for proactively flagging potential implementation issues, as school-based deworming requires close coordination between the Department of Health (who procure the drugs) and Education (as drugs are distributed via schools)
• Monitoring: According to state officials in Uttarakhand, the Indian government relies on Evidence Action’s monitoring data for each NDD. After NDD, the government requests real-time coverage data from Evidence Action, which informs supply chain decisions for mop-up day a week later (where deworming pills are given to children that were missed the first time around).

These bottlenecks feel plausible to us, and we view these reports as a positive update on these grants. However, we’re cautious about putting too much weight on this as government stakeholders may be incentivized to report positively, especially because Evidence Action staff were present for these conversations. We’re also not sure exactly how these activities map to increases in counterfactual coverage – for instance, we’re not sure why the back-calculation of procurement contracts and the arrangement of cross-departmental meetings wouldn’t happen anyway, without Evidence Action’s support.

When we asked Evidence Action about why this wouldn’t happen anyway, the reasons they gave for this included:

• Capacity constraints: Government officials responsible for overseeing National Deworming Day have many (~20 in Uttarakhand) health programs in their portfolio. Capacity is often stretched thin, and having external consultants with a mandate to focus on just one program can help prevent things slipping through the cracks
• Complicated reporting lines: Deworming programs awkwardly straddle the Ministry of Education and Ministry of Health. When programs don’t fit neatly into existing silos, accountability and reporting can get muddied, and it can take an external push (e.g., a TA engagement) to jumpstart programming and institutionalize protocols

These bottlenecks seem plausible to us, and the capacity constraints story was generally corroborated by government stakeholders we spoke to. However, these bottlenecks also seem harder to interrogate than a more straightforward supply side story, which is typically the key coverage constraint our ‘direct delivery’ grants alleviate. For example: we think some people don’t get bed nets because National Malaria Control Programs are funding constrained, and so by funding the procurement and distribution of nets, the Against Malaria Foundation can get these to people who would have not otherwise received them. This type of constraint feels more obvious and intuitive to us than the operational constraints Evidence Action claim to have helped alleviate via these grants.

How cost-effective do we think these grants were?

Based on a retrospective cost-effectiveness model, we think these grants were plausibly around our current cost-effectiveness threshold (10x), though with wide uncertainty intervals.

• Effect of program on coverage: We put some weight on data reported by Evidence Action and some weight on data reported by the government
• How coverage would have changed: We make rough forecasts about how deworming coverage would have changed in the absence of TA. To estimate counterfactual impact, we take the ‘wedge’ between (i) our estimated effect of the program on coverage and (ii) how we expect coverage would have changed otherwise. These assumptions are illustrated below.
• Program costs: We based cost estimates on our latest grant to Evidence Action’s Deworm the World program, which earmarked ~$1m to support TA in Bihar and Uttar Pradesh

Did we set ourselves up to learn?

We think these are a few places we could have better set ourselves up to learn:

• Baseline coverage surveys: Though we do have some baseline coverage estimates from the government, we didn’t earmark funding for Evidence Action to commission baseline coverage surveys for these grants, which made them harder to evaluate. Our support for most states started between 2015 and 2018,18 when GiveWell was a much smaller funder.19 At the time, we were more focused on maximizing the immediate impact of our grantmaking vs. investing in learning opportunities to improve our grantmaking in future. As we’ve grown, we’ve started to place more emphasis on monitoring and evaluation, and expect to continue to do so. One way we plan to do this is prioritizing baseline data collection.
• Conversations with key stakeholders: In hindsight, we think it would also have been better if we’d met with government stakeholders alone, rather than with Evidence Action staff present. One of the reasons Evidence Action staff joined for these reasons was due to language barriers – many of the government officials we spoke to didn’t speak English (especially at the district and block level). In the future, we’ll consider bringing independent interpreters to these meetings.
• Independent verification: We didn't establish independent data collection systems to validate both government and grantee-reported data, leading to the current situation where we have conflicting information with no clear way to resolve discrepancies.

How calibrated were our forecasts?

We didn’t make forecasts for our early grants to Evidence Action’s Deworm the World program. We made four forecasts attached to the 2023 grant; only one of which has been realized.

Sources

• 1

  You can find all of our content on Evidence Action’s Deworm the World program on this page, including previous grants, our evaluation of the program, and our evaluation of Evidence Action.

• 2

  You can find a list of our previous grant to Evidence Action’s Deworm the World program in India here.

• 3

  You can read our latest report on our evaluation of the program, from 2022, here.

• 4

  Source: Our report on Evidence Action’s Deworm the World program.

• 5

  Source: Evidence Action’s monitoring data, 2025 (unpublished)

• 6

  We don’t have this for all states where Evidence Action has worked because: (i) we only have coverage estimates going back to 2016; (ii) Evidence Action has been working in some states (e.g., Bihar) before this.

• 7

  Source: GiveWell’s analysis of deworming technical assistance data from India (unpublished)

• 8

  We have set up our retrospective CEA of Deworm the World with an option for what assumptions you want to make for the counterfactual increase in NDD coverage in the absence of Evidence Action’s technical assistance.

  To estimate the cost-effectiveness in a scenario where all of the increase in coverage since the program was implemented was due to Evidence Action’s technical assistance, you can set this value to ‘p100’. We also included options for our own best guess, a 75th percentile value representing an optimistic scenario for the program’s impact, and a 25th percentile value representing a pessimistic scenario for the program’s impact.

• 9

  We discuss Evidence Action’s process for selecting states in our 2022 intervention report for Deworm the World: “Prevalence surveys. Before Deworm the World helps launch a deworming program in a new area, it evaluates whether the prevalence of worm infections is sufficient to justify an MDA for the school-age population.19 If no prevalence surveys have been conducted recently, it generally commissions one.20 The results of prevalence surveys are used to determine the appropriate treatment strategy (in particular, MDA frequency) for a given location.21 It also conducts follow-up prevalence surveys periodically, so that it can track the impact of the MDAs and refine treatment strategies as needed, in accordance with WHO guidelines.22 Deworm the World generally contracts out work on prevalence surveys.23”

• 10

  Source: GiveWell’s analysis of deworming technical assistance data from India (unpublished)

• 11

  Based on conversations with Evidence Action staff. There is nothing published on the methodology of these surveys.

• 12

  Source: GiveWell’s analysis of deworming technical assistance data from India (unpublished)

• 13

  India is organized into states/union territories which are broken into districts then sub-districts then blocks then villages/urban wards. A state (e.g., Uttarakhand) is the primary sub-national government unit. A district is the next tier, overseen by a District Magistrate/Collector. Below that, states use various sub-district terms (e.g., tehsil/taluk), and—especially for health, education, and rural development delivery—community development “blocks” (administrative clusters of villages/Gram Panchayats) led by a Block Development Officer. Programs often plan and supervise implementation at the block level and aggregate reporting at the district and state levels. Source: The Indian government’s census website

• 14

  Source: Conversations with government officials regarding Evidence Action’s Deworm the World program September 2-6, 2024 (unpublished)

• 15

  Albendazole is one of two drugs recommended by the WHO for deworming in children: “Preventive chemotherapy (deworming), using annual or biannual single-dose albendazole (400 mg) or mebendazole (500 mg)b is recommended as a public health intervention for all young children 12–23 months of age, preschool children 1–4 years of age, and school-age children 5–12 years of age (in some settings up to 14 years of age) living in areas where the baseline prevalence of any soil-transmitted infection is 20% or more among children, in order to reduce the worm burden of soil-transmitted helminth infection.”

• 16See our methods for this analysis here.
  
• 17As part of our retrospective CEA for this lookback we’ve modeled several scenarios for the counterfactual increases in NDD coverage which would have occurred without Evidence Action’s assistance. These include for our own best guess, a 75th percentile value representing an optimistic scenario for the program’s impact, a 25th percentile value representing a pessimistic scenario for the program’s impact, and 100th percentile scenario where 100% of the changes in coverage since the program began are attributable to Evidence Action’s work.
  
• 18

  You can see the schedule of our rollouts here.

• 19

  GiveWell raised between $120 million and $159 million USD between 2015 and 2018. In our most recent metrics year, 2024, we raised approximately $415 million USD. Source: GiveWell’s 2024 Metrics Report.