Unlimit Health – November 2021 version

Unlimit Health, formerly known as the SCI Foundation, formerly known as the Schistosomiasis Control Initiative, was one of GiveWell’s top-rated charities from 2011 to 2022. We updated our criteria for top charities in August 2022 and due to these changes, Unlimit Health is no longer one of our top charities.

This does not reflect an update to our view of Unlimit Health. The change was motivated by our desire to clarify our recommendations to donors, not by any shift in our thinking about Unlimit Health. More information is available in this blog post.

We are no longer accepting donations designated for Unlimit Health. You can support Unlimit Health by donating directly here. If you would like to support GiveWell's grantmaking you can do so here—we plan to continue to support deworming through the All Grants Fund.

We are no longer maintaining the review of Unlimit Health below. Please visit their website here to learn more or donate.

Published: November 2021; Last Updated: April 2023

Summary

What do they do? Unlimit Health (https://unlimithealth.org/) works with governments in sub-Saharan Africa to create or scale up programs that treat schistosomiasis and soil-transmitted helminthiasis (STH) ("deworming"). Unlimit Health's role has primarily been to identify country recipients, provide funding to governments for government-implemented programs, provide advisory support, and conduct monitoring and evaluation on the process and outcomes of the programs. (More)

Does it work? We believe that there is strong evidence that administration of deworming drugs reduces worm loads, but weaker evidence on the causal relationship between reducing worm loads and improved life outcomes; we consider deworming a priority program given the possibility of strong benefits at low cost. Unlimit Health has conducted studies to monitor (a non-representative sample of) about 60% of the treatments it has funded since 2014 (70% in the most recently completed cycle) to determine whether its programs have reached a large proportion of children targeted. These studies have generally found positive results, but have some methodological limitations. (More)

What do you get for your dollar? Our best guess is that deworming is generally highly cost-effective. We estimate that children are dewormed for a total of around $1.03 per child in programs supported by Unlimit Health. This figure relies on several difficult-to-estimate inputs including how to account for (a) donated drugs and (b) in-kind contributions from governments with which Unlimit Health works. Excluding drugs and government contributions, we estimate that Unlimit Health's cost per treatment is $0.40. The number of lives significantly improved by deworming is also a function of a number of difficult-to-estimate factors, which we discuss in detail in a separate report. (More)

Is there room for more funding? Unlimit Health expects that it could absorb around $85 million—prior to receiving any grants that GiveWell specifically makes or recommends—to support its deworming portfolio in 2022-24. (More)

Unlimit Health is recommended because of its:

  • Focus on a program with a strong track record and excellent cost-effectiveness. (More)
  • Track record – Unlimit Health has repeatedly demonstrated success at starting and expanding national deworming programs.
  • Standout transparency – it has shared significant, detailed information about its programs with us.

Major open questions:

  • The evidence we have seen on Unlimit Health's track record of reaching those it has targeted for treatment is somewhat limited. Unlimit Health has conducted studies to monitor (a non-random sample of) about 60% of the treatments it has funded since 2014, to determine whether its programs have reached a large proportion of children targeted. We are somewhat uncertain about the results we have seen because of the methodological limitations of the studies.
  • Unlimit Health may use additional funding to, in part, treat adults in areas with moderate or high risk of schistosomiasis infection, as part of a strategy to stop transmission of the parasite. While this strategy has the potential to be cost-effective by decreasing infection rates in children over the long-term, we have not seen sufficient evidence to support it and do not currently believe it meets our criteria for evidence and cost-effectiveness.

Table of Contents

Our review process

We began reviewing Unlimit Health in 2009. Our review has consisted of:

  • Extensive communications with Unlimit Health staff, including leadership, program managers, and monitoring and evaluation and finance staff.
  • Reviewing documents Unlimit Health sent in response to our queries.
  • Visiting a national schistosomiasis control program meeting and demonstration mass drug administration in Malawi in October 2011 (notes and photos from this visit here).

All content on Unlimit Health, including past reviews, updates, blog posts, and conversation notes, is available here. We have also published a page with additional, detailed information on Unlimit Health to supplement some of the sections below.

What do they do?

Unlimit Health works with governments in sub-Saharan Africa to create or scale up mass drug administration programs (MDAs) for neglected tropical diseases (NTDs), particularly schistosomiasis and soil-transmitted helminthiasis (STH), in school-aged children and other groups determined to be at high risk.1 Unlimit Health's role has primarily been to identify country recipients, provide funding to governments for government-implemented programs, provide advisory support, and conduct research on the process and outcomes of the programs.

Unlimit Health's model involves both (a) employing staff for program management, technical assistance, and capacity building and (b) funding governments to carry out infection mapping and treatment programs.

Unlimit Health’s role in mass drug administration programs

Unlimit Health's role in mass drug administrations (MDAs) in general is to:2

  • Advocate for the benefits of MDAs to government officials.
  • Assist with planning and budgeting.
  • Deliver funding and drugs to governments.
  • Provide financial management and technical support.
  • Develop procedures for monitoring and evaluation, analyze data, and write reports.

In 2014, we spoke with four of Unlimit Health's program managers to better understand Unlimit Health's role in four countries. These conversations were mostly consistent with our general understanding of Unlimit Health's work. We selected Côte d'Ivoire, Ethiopia, and Mozambique because Unlimit Health has spent significant unrestricted funds, the type of funding GiveWell has recommended, in these countries (more below). Unlimit Health selected Uganda as the fourth case study. We have summarized Unlimit Health’s work in these countries on a separate page with additional information on Unlimit Health.

Major funding sources for Unlimit Health’s work

In Unlimit Health's April 2020-March 2021 budget year, the major sources of its funding were (a) unrestricted funds—53% of spending3 —and (b) restricted funding from the Accelerating the Sustainable Control and Elimination of Neglected Tropical Diseases (Ascend) program, which was funded by the U.K. Foreign, Commonwealth and Development Office (FCDO). Unlimit Health received this funding through Sightsavers, which received FCDO funding to treat NTDs in West and Central Africa from 2019 to August 20214 and funded Unlimit Health as a subcontractor for this work in Côte d'Ivoire, DRC, Liberia, and Niger (there was no expenditure in Liberia in the April 2020-March 2021 budget year).5 Unlimit Health also received a number of smaller grants from other funders.6

In Unlimit Health's April 2020-March 2021 budget year, we estimate that roughly 74% of the unrestricted funding that SCI received was due to GiveWell’s recommendation of Unlimit Health.7

More detail on Unlimit Health’s current and past major sources of funding is available on a separate page with additional information about Unlimit Health.

Breakdown of Unlimit Health’s spending

Spending breakdown by country

April 2020 to March 2021 expenditures by country (in millions USD)8

Restricted Unrestricted Total % of total
Ethiopia $0.2 $0.1 $0.3 2%
DRC $2.1 - $2.1 18%
Malawi - $1.9 $1.9 16%
Madagascar - $0.2 $0.2 2%
Tanzania - $1.2 $1.2 10%
Uganda - $0.3 $0.3 2%
Côte d'Ivoire $2.0 - $2.0 17%
Niger $0.1 - $0.1 1%
Zanzibar - $0.2 $0.2 1%
Burundi - $0.4 $0.4 3%
Other program costs9 - $0.2 $0.2 2%
Management costs10 $1.1 $1.9 $3.0 26%
Total $5.4 $6.3 $11.8 100%

For breakdowns of Unlimit Health's spending by country in previous years, see our past reviews of Unlimit Health in this footnote.11

Spending breakdown within country programs

In 2015, Unlimit Health began to use a system of country cashbooks, which compare monthly in-country actual spending to budgets.12 Below we have summarized the data from the country cashbooks covering April 2016 to March 2017 for Tanzania (excluding Zanzibar), Uganda, Zanzibar, Niger, Malawi, Côte d'Ivoire, and Burundi.13 We also received 2016-2017 cashbooks from Ethiopia, Madagascar, and Democratic Republic of the Congo (DRC). We have excluded those countries from the overall analysis for reasons that are explained in GiveWell's analysis of SCI cashbooks 2016-2017, "Summary" tab, row 18.

Unlimit Health also sent us data from its 2015-16 budget year, which is summarized here. The 2015-2016 documents included data from Niger, Côte d'Ivoire, Democratic Republic of the Congo, Malawi, Tanzania, and Uganda.14

Notes about the data in these cashbooks:

  • Activity categories in the country cashbooks sometimes appear to be loosely defined and overlapping. For example, expenditure on drug distribution materials is sometimes classified as "drug distribution" and sometimes as "drug logistics."15
  • There are some differences between expenses recorded in country cashbooks and expenses recorded in SCI budget vs. actuals 2015-16 Redacted and SCI Budget vs Actuals April 2016-March 2017 Redacted. For example, for Malawi, the total expenditure recorded for the 2015-16 budget year in both the country cashbook and SCI budget vs. actuals 2015-16 Redacted was $0.8 million, but total expenditure for Tanzania differed substantially in the two sources.16 Unlimit Health notes that the discrepancy for Tanzania is due to Unlimit Health recording the expenditure when the funds were sent to the country, while the cashbooks record when the funds are spent in-country.17 It is our understanding that Unlimit Health planned to change this system.
  • The total amount of spending recorded in the seven country cashbooks we have included in the table below is $3.5 million, approximately one-third of the $10.9 million in total spending within country programs in Unlimit Health's 2016-17 budget year.18

    In-country Unlimit Health spending in seven countries, April 2016 – March 201719

    Activity % of total spending Range across countries Description
    Drug distribution training 34% 0% to 80% Per diem payments for teachers and officers during MDA training, accommodation and meals
    Drug distribution 32% 0% to 65% Per diem payments for teachers and officers, fuel costs, communications costs, drug distribution materials (dose poles, registers, etc.)
    Country management 8% 1% to 34% Per diem payments and salaries for national NTD program staff, fuel costs, communications
    Social mobilization 5% 0% to 14% Per diem payments, advertising, fuel costs
    Drug distribution supervision 4% 0% to 10% Per diem payments for MDA supervisors, fuel costs
    Drug distribution registration 4% 0% to 7% Per diem payments for teachers, drug distribution materials
    Strategic planning 3% 0% to 13% Expenses related to strategic planning meetings, including per diem payments, travel costs, and accommodation and meals
    Drug logistics 3% 0% to 8% Drug distribution materials (dose poles, registers, etc.), fuel costs
    Monitoring & evaluation 2% 0% to 10% Per diem payments for teachers, MDA report writers, and drivers, fuel costs
    Mapping 2% 0% to 26% Per diem payments for teachers, travel costs, laboratory supplies
    Global management 2% 0% to 37% Travel costs to an international conference, miscellaneous items (e.g. laboratory supplies, per diem payments for teachers)
    Advocacy 1% 0% to 2% Per diem payments, fuel costs, communications

    Unlimit Health has also shared country cashbooks from its 2017-18 and 2018-19 budget years. We have not reviewed these cashbooks in depth.

    Does it work?

    This section was last updated in November 2020. The information that Unlimit Health has provided since that date is not yet reflected in this section.

    We believe that there is strong evidence that administration of deworming drugs reduces worm loads but weaker evidence on the causal relationship between reducing worm loads and improved life outcomes; we consider deworming a priority program given the possibility of strong benefits at low cost.

    To evaluate Unlimit Health's track record at executing programs, we have reviewed:

    • Coverage surveys from fourteen of the countries Unlimit Health has worked in, including many of the countries where Unlimit Health's work has been focused in the past five years. These household surveys estimate what percentage of individuals who were targeted for treatment actually received treatment. Overall, coverage of school-aged children was above 75% (the WHO-recommended minimum threshold) in most of the districts surveyed by Unlimit Health. We note some limitations of these surveys below.
    • Prevalence and intensity studies of infection over time from seven of the countries Unlimit Health has worked in over the past few years. Most of the surveys show improvements following Unlimit Health treatment programs. These surveys have a number of limitations and represent a small proportion of the total number of treatments delivered in Unlimit Health's programs.
    • Academic papers that might reflect the treatment coverage achieved by Unlimit Health's programs by directly measuring deworming drug uptake or by measuring worm prevalence in countries where Unlimit Health has worked. The papers did not provide a clear case for or against programs being executed well. We discuss this analysis on a separate page with additional information about Unlimit Health.

    We have now reviewed some recent monitoring results from approximately 90% of the countries in which Unlimit Health works. For those countries from which we have seen monitoring results, we have often seen one or two years of results from the past several years. In total, Unlimit Health has conducted studies to monitor (a non-representative sample of) about 60% of the treatments it funded from 2014-17 to determine whether its programs have reached a large proportion of children targeted. We are somewhat uncertain about the results we have seen because of methodological limitations of the studies.

    In this section, we also discuss how the disease burden in the areas Unlimit Health works in compares to the places where the independent studies that form the evidence base for the impact of deworming were conducted. While Unlimit Health's programs generally target areas that require mass treatment according to World Health Organization (WHO) guidelines, the disease burden in Unlimit Health areas is on average lower than in the study areas, so our expectation is that the average impact per child treated is lower in Unlimit Health areas. We adjust our cost-effectiveness analysis accordingly.

    Is there independent evidence that the program is effective?

    Unlimit Health supports mass school-based deworming programs, the independent evidence for which we discuss extensively in our intervention report on deworming programs. In short, we believe that there is strong evidence that administration of the drugs reduces worm loads but weaker evidence on the causal relationship between reducing worm loads and improved life outcomes; we consider deworming a priority program given the possibility of strong benefits at low cost.

    There are some important differences between the type and severity of worm infections in the places Unlimit Health works and the places where the key studies on improved life outcomes from deworming took place, which we discuss below.

    Are deworming pills delivered to and ingested by recipients?

    Coverage surveys

    We have reviewed results from coverage surveys that Unlimit Health has conducted, or worked with partners to conduct, in:

    • Côte d'Ivoire: 2014, 2016, and 2018
    • Malawi: 2012, 2014, 2016, and 2017
    • Uganda: 2014 and 2017
    • Mozambique: 2015 and 2016
    • Zanzibar: 2015 and 2018
    • Zambia: 2015
    • Ethiopia: 2015, 2016, 2017, and 2018
    • Madagascar: 2016 and 2017
    • Burundi: 2017
    • Liberia: 2017 and 2018
    • Mauritania: 2017 and 2018
    • Niger: 2017 and 2018
    • Nigeria: 2017
    • Tanzania: 2017 and 2018
    • Democratic Republic of the Congo: 2018

    We summarize the methods and results of the above surveys in this spreadsheet and discuss themes in the sections that follow. We have also received coverage surveys that Unlimit Health has conducted, or worked with partners to conduct, in Burundi (2019), Ethiopia (2019), Malawi (2018 and 2019), Uganda (2018 and 2019), and Zambia (2018). We have not reviewed these surveys in detail, and as such they are not covered in the discussion that follows.

    Methods

    In each of the surveys, surveyors visit a sample of households and ask children, or in some cases their parents on their behalf, whether they received treatment in the most recent MDA. Villages and households for these surveys are generally selected randomly or quasi-randomly.20 Other survey questions, such as questions about age, gender, where the respondent received the treatment, and why the respondent did not take the drug(s), are often included as well.21 The main results are reported as "survey coverage" figures (the number of school-aged children interviewed who ingested the drug(s) divided by the total number of school-aged children interviewed), and are intended to check the accuracy of governments' "reported coverage" figures (the number of treatments delivered to school-aged children according to government administrative data divided by the estimated number of eligible school-aged children in the area).22

    The methodology used in Unlimit Health's coverage surveys has differed somewhat across surveys. We have summarized the details of the methodologies used in the surveys in this spreadsheet ("Methods" sheet).

    "Survey coverage" and "reported coverage" estimates of the proportion of school-aged children that ingested deworming pills in Unlimit Health programs can differ substantially.23 We generally believe that coverage surveys provide more reliable information on the proportion of school-aged children that received and ingested deworming pills than "reported coverage" figures calculated by governments of countries with Unlimit Health-supported programs (justifications in footnote).24 However, we note some limitations to Unlimit Health's coverage surveys:

    • Accuracy of survey responses: We have a few concerns about the accuracy of responses from school-aged children, particularly young children (e.g., 5-6 year olds), on whether they received deworming treatments. Some coverage surveys include "verification questions" (e.g., asking children if they recognize pills), but we have not yet seen evidence from these questions that raises our general confidence in the accuracy of Unlimit Health's coverage surveys.
      • Length of time between MDA and survey: This varies between less than a month and nine months in the coverage surveys we have seen from Unlimit Health. Intuitively, the more time that passes, the less likely children are to remember accurately and the more likely they are to confuse past MDAs. Ethiopia (2018) had the shortest interval, at 0-2 months, and Liberia (2018) had the longest, at 3-9 months, of the surveys for which we have seen information about the length of time between the MDA occurring and the survey occurring. Other surveys were generally carried out 2-4 months after the MDA.25
      • Verification questions: In Côte d'Ivoire, Malawi, Uganda, Zanzibar, Zambia, Ethiopia, Madagascar, Mauritania, Liberia, Burundi, Niger, Nigeria, the Democratic Republic of the Congo, and Tanzania, surveyors asked some verification questions in at least one of the coverage surveys that we have seen, such as whether respondents recognized pills or dose poles presented by the interviewers, what they thought of the pills (praziquantel is very large and tastes bitter), and how many pills they took.26
        • In Mozambique (2015), respondents were asked whether they recognized the dose pole used in schistosomiasis MDAs. However, in that survey parents were surveyed on their children's behalf27 and most children (79%) received drugs at school,28 presumably when parents were not present. We do not know how to interpret the result that a very high percentage of parents (median 90%, ranging 61-94% across provinces) reported recognizing the dose pole.29 Unlimit Health hypothesized that parents may either recognize the dose pole from publicity efforts prior to the MDA or remember a similar dose pole from previous MDAs for lymphatic filariasis.30 In either case, this may indicate that the coverage survey was not measuring actual delivery of drugs to children in the 2015 MDA. The Mozambique (2016) coverage survey does not indicate that any verification questions were asked.31
        • Recognition of pills and dose poles was low in some coverage surveys. This is surprising given the high survey coverage rates in those surveys, and reduces our confidence in the accuracy of the survey coverage rates. For example, from the most recent round of coverage surveys we have seen, from 2017 and 2018, children seem to have particularly low recognition in Mauritania (2017) and Ethiopia (2017). In Mauritania (2017), recognition of praziquantel pills is approximately 10% or less in two out of three districts, and yet praziquantel survey coverage rates are greater than 80% in both districts.32 In Ethiopia (2017), recognition of both praziquantel and albendazole pills is approximately 40% or less in at least three out of four regions, and yet survey coverage rates for both treatments are greater than 75% in all regions.33
    • Supervision and auditing of surveys: Many of Unlimit Health's reports on coverage surveys mention that teams of surveyors were overseen by supervisors. It is our understanding that supervisors provide teams with guidance on the logistics of implementing coverage surveys34 but do not implement any specific data quality or auditing procedures, such as re-surveying a sample of households to check the accuracy of the data collected.35 Unlimit Health told us that its U.K.-based staff review data daily as it is entered, looking for missing or concerning data and correcting errors.36

    Additionally, there are limitations that apply only to some of Unlimit Health's coverage surveys:37

    • Selection of geographic target area: All of the surveys were limited to specific geographic areas (such as districts). In Uganda (2014), Zambia (2015), Côte d'Ivoire (2014, 2015, and 2017), Ethiopia (2015), Madagascar (2016), Mauritania (2016), Nigeria (2017), Tanzania (2017 and 2018), Liberia (2017 and 2018), Zanzibar (2017), the Democratic Republic of the Congo (2018), and Niger (2018), these were selected randomly or nearly randomly. In Malawi (2012, 2014, 2016, and 2017), Mozambique (2015), Ethiopia (2016, 2017, and 2018), Burundi (2017a and 2017b), Niger (2017), and Uganda (2017), the districts were purposefully selected and not intended to be nationally representative. In Mauritania (2018), one district was randomly selected while the remaining two districts were purposefully selected. The selection procedures for Zanzibar (2015), Mozambique (2016), and Madagascar (2017) were not given in the documents we have seen on the surveys or were unclear.38 We believe the criteria used for the non-random sampling of districts in Ethiopia (2016 and 2017) and Malawi (2017) are particularly likely to lead to bias in the coverage results, though we are uncertain how many districts were excluded as a result of the criteria used.39 Villages and households are randomly selected within each geographic target area.40
    • Independence from the government: In Mozambique (2015 and 2016), the survey was carried out by government health staff, who may have had an incentive to bias the results. Unlimit Health told us, "[M]ost of the interviews in one district were done by the other district officers with no connection with the district."41 We believe the Ethiopia (2016, 2017, and 2018), Mauritania (2017), and Democratic Republic of the Congo (2018) surveys were also carried out by government health workers. Information we received on the Malawi (2012, 2014, 2016, and 2017), Uganda (2017), and Madagascar (2017) distributions note that the surveyors were independent of the government. Students or staff from universities or research institutions conducted the surveys in Uganda (2014), Zanzibar (2015 and 2018), Zambia (2015), Côte d'Ivoire (2014, 2016, and 2018), Burundi (2017a and 2017b), Niger (2017 and 2018), Mauritania (2018), and Tanzania (2018).42 It is unclear to us who conducted the Ethiopia (2015), Madagascar (2016), Nigeria (2017), Tanzania (2017), and Liberia (2017 and 2018) surveys.
    • Whether parents or children were interviewed: In Mozambique (2015 and 2016), parents were interviewed about whether their children took the drugs. In both Côte d'Ivoire and Malawi (both 2012 and 2014 surveys), if children in a household were not available, then their parents were interviewed about whether the children had received deworming drugs. Unlimit Health made different choices about whether to include these responses in the results, which slightly inflated the results overall.43 Unlimit Health told us that parents were not asked to answer on behalf of their children in Uganda (2014), Zanzibar (2015), Ethiopia (2015), and Madagascar (2016), and that going forward, Unlimit Health no longer plans to have parents answer on the behalf of children.44 We have seen no further mention of this in most of the reports for the 2017 and 2018 coverage surveys. The Ethiopia (2016 and 2018) and Tanzania (2018) coverage survey protocols indicate that caretakers could be asked whether their children received deworming drugs but only for non-school-based MDAs.45
    Results

    Results from the coverage surveys we have seen are available here. For both coverage of praziquantel (to treat schistosomiasis) and coverage of albendazole (to treat STH), the median point estimate for coverage achieved by Unlimit Health, according to coverage surveys, was 85%.

    The fact that the surveys identified low coverage in several cases increases our confidence in their reliability. Given the smaller sample size, government involvement in the survey, and question about parents recognizing the dose pole noted above, we are more skeptical about the results from Mozambique (2015) than those from other surveys. We have low confidence in the results from Mauritania (2017) because of government involvement, uncertainty about how districts were selected and children's low recognition of the pills. We also have low confidence in the results from Mozambique (2016) and Ethiopia (2016 and 2017) because of government involvement and uncertainty/concerns about how districts were selected. Additionally, we have limited confidence in the results of Côte d'Ivoire (2016) due to concerns about the reliability of the data.46

    Prevalence and intensity studies

    Unlimit Health has conducted surveys to track changes in schistosomiasis and STH prevalence and intensity rates following Unlimit Health-supported treatment programs. In each of these studies, Unlimit Health tracked infection rates at the same schools ("sentinel sites") each year. In general, prevalence and intensity of the parasites decreased over time in each of the countries studied. We note several methodological limitations of these surveys below.

    Below, we discuss results from studies of schistosomiasis and STH prevalence and intensity from nine countries: Niger (2004-2006),47 Burundi (2007-2010 and 2007/8-2017),48 Liberia (2012-2013),49 Malawi (2012-2017), Madagascar (2015-2017), Tanzania (2016-2017), Côte d'Ivoire (2013-2016), Ethiopia (unclear when baseline took place; follow-up in 2016/2017), and the Democratic Republic of the Congo (2018).50 We summarize the methods and results in this spreadsheet and discuss themes in the sections that follow. We have also seen impact surveys from Liberia (2018), Côte d'Ivoire (2017 and 2018), Zanzibar (2018), Zambia (2018), Madagascar (2018), Burundi (2018), and Ethiopia (2018) and reassessment reports from Malawi (2017 and 2018), Tanzania (2018), and Côte d'Ivoire (2019) We have not reviewed these surveys in detail, and as such they are not covered in the discussion that follows.

    Results from prevalence and intensity studies

    The studies in Niger, Malawi (2012-17), Burundi (2007-10) and Liberia (2012-13) were originally designed as cohort studies, in which the same individuals are repeatedly surveyed over time.51 In the 2015 Malawi and 2013 Liberia studies, Unlimit Health switched to a cross-sectional sample, where random children from the same schools were surveyed, rather than the same individuals.52 The studies in Burundi (2007/8-2017), Madagascar (2015-17), Tanzania (2016-17), Côte d'Ivoire (2013-16), Ethiopia (2016/17), and the Democratic Republic of the Congo (2018) also sample a random cross-section of children from the same schools in the baseline and follow-up surveys.53 There is no control group for these studies due to the ethical implications of withholding treatment from infected children.54

    Results from the impact surveys we have seen are available here. Prevalence and intensity for the two main types of parasites that cause schistosomiasis, S. haematobium and S. mansoni, decreased over time in each of the countries studied other than Côte d'Ivoire, in which there was a small increase in prevalence for both types of infection, and Tanzania, in which there was an increase in prevalence and intensity for S. mansoni.55 The prevalence and intensity of S. mansoni increased between the second and third follow-up surveys in Burundi (2007/8-2017), but remained significantly lower than in the baseline.56

    The prevalence and intensity of two of the three soil-transmitted helminths, hookworm and ascaris, also decreased over time in most of the countries studied.57

    Though it is possible that other factors besides the treatment program caused these declines (such as improved sanitation infrastructure), the fact that the declines occurred over a short period following treatment strongly suggests that treatment caused or contributed to it.

    The prevalence and intensity of the third soil-transmitted helminth, trichuris, was generally low and the changes over time were more mixed. Prevalence and intensity decreased in approximately the same number of countries as it increased.58

    In some countries, there were large differences in the change in prevalence across schools, with some schools experiencing changes in the opposite direction to the average trend for the country as a whole. In Côte d'Ivoire, despite little change in the prevalence of S. mansoni or S. haematobium on average, several schools experienced increases in the region of 10%.59 In Madagascar, while the prevalence of S. haematobium infection and heavy-intensity infection decreased overall, there were large differences across schools.60 In Ethiopia, the prevalence of S. mansoni increased in nearly one third of schools, despite an overall reduction.61

    Some of the studies also report results for other indicators of disease, such as anemia. We omit discussion of these other indicators because they are more likely to be influenced by external factors than are prevalence and intensity (see our previous review of Unlimit Health for discussion of these indicators and SCI Liberia impact survey dashboard 2012-13 for results from Liberia).

    Limitations of the prevalence and intensity study data include:

    • Monitoring of selected locations in some countries. It appears that, in the Niger and Burundi pilot studies, locations included in the study were selectively chosen rather than selected to be a representative sample of treated areas. In the Côte d'Ivoire, Ethiopia, Liberia, Madagascar, Malawi and Tanzania studies, schools were sampled randomly from locations at medium and high risk of infection in line with a stratified statistical approach. We are uncertain how locations in the Burundi (2007/8-2017) study were chosen.62
    • Attrition of schools from the sample. In Liberia, nearly one-third of the schools surveyed during the baseline year could not be re-visited in the follow-up year (which Unlimit Health attributes to inaccurate school identity numbers, school closures, and/or inaccurate recording of GPS coordinates), and were replaced by nearby schools.63 In Madagascar, two out of twenty-nine locations could not be revisited and were not replaced.64
    • Low follow-up rates in cohort studies. Follow-up rates were low in two of the three studies using a cohort model for follow-up (89% at the first year follow-up in Niger, 33%-50% in the pilot survey and 53%-80% in the other schools survey in Burundi (2007-2010), and 52% in the first follow-up in Malawi).65 To be included in follow-up surveys, children must be present in school when the surveys are done.66 If those who are present in school are less likely to be infected than those who are not present, this could lead to overstating the impact of the program. The connection between infection status and absenteeism could be a direct relationship (infection could cause absenteeism) or an indirect one (a third factor, such as poverty, could cause both higher levels of infection—perhaps through poor sanitation infrastructure—and absenteeism). As previously noted, Unlimit Health has moved away from the use of cohort studies.
    • Different methods of measurement for the prevalence of S. mansoni lead to different prevalence rates in Burundi (2007/8-2017). In Burundi (2007/8-2017), both the rapid urine-based Circulating Cathodic Antigen (CCA) and Kato-Katz methods are used to measure the prevalence of S. mansoni. Prevalence using the CCA method is always higher and for one school was nearly fifty times higher. It is unclear which method is used in the reported results, or whether the same method is used in the baseline and each follow-up survey.67
    • The procedure used to sample children changed between the baseline and follow-up surveys in Malawi. In the second, third and fourth follow-up surveys in Malawi, children were sampled by age, not by grade as in the baseline. It is unclear whether the age distribution of children in these follow-ups is fully comparable to the baseline, and so whether prevalence rates are comparable.68
    • Small sample size in the Democratic Republic of the Congo. The survey in the Democratic Republic of the Congo (2018) included only three schools.69

    Are Unlimit Health's monitoring results representative of its work overall?

    We have received monitoring results that account for about 79% of programmatic spending from 2018-202070 and 55% of the treatments that Unlimit Health delivered in 2014-2017.71 There is one large country program, Sudan, for which we have not seen any monitoring results.72

    Details in this spreadsheet.

    Unlimit Health told us that it is sometimes unable to share results because programmatic data about populations is owned by governments and therefore, in line with data sharing agreements between Unlimit Health and the countries, it is necessary to gain permission from the governments before data can be shared.73

    Since we have not seen coverage surveys from all Unlimit Health-supported campaigns,74 the results we have seen could overstate Unlimit Health's impact if coverage surveys are more likely to be skipped or the results withheld in countries with lower coverage rates. There are a couple of reasons this might be the case:

    • Country programs that have more capacity and experience are likely to be those that both carry out high-quality distributions and complete all the steps necessary for coverage surveys to be implemented and the results shared with Unlimit Health (and thus with GiveWell).
    • Coverage surveys are more likely to be skipped or results withheld if implementers recognize that the surveys are likely to show low coverage results and reflect poorly on them. We have no evidence that this has occurred for Unlimit Health-supported programs, and note this only as a general possibility.

    What is the likely impact per treatment in Unlimit Health's programs compared with the independent studies on the impact of deworming?

    In general, mass deworming programs treat everyone in a targeted demographic, regardless of whether each individual is infected (more). Because of this, the benefits (and therefore the cost-effectiveness) of a program are highly dependent on the baseline prevalence of worm infections.

    In this section, we discuss how the disease burden in the areas Unlimit Health works in compares to the places where the independent studies that form the evidence base for the impact of deworming were conducted. While it is our understanding that Unlimit Health's programs generally target areas that require mass treatment according to WHO guidelines, the disease burden in Unlimit Health areas is on average lower than in the study areas, so our expectation is that the impact per child treated is lower in Unlimit Health areas. We adjust our cost-effectiveness estimate (more below) accordingly.

    We have seen baseline data on the prevalence and intensity of schistosomiasis and STH infections for countries that account for about three-quarters of the treatments Unlimit Health has delivered in recent years. Schistosomiasis and STH prevalence and intensity in these countries was generally fairly low compared to the studies providing the best evidence for the benefits of deworming (Croke 2014 and Miguel and Kremer 2004).

    Baseline data was collected in schools that had been selected for prevalence and intensity studies. The baseline reports use methodologies that seem similar to the other Unlimit Health prevalence and intensity studies discussed above. With the exception of the study discussed above from Malawi, we have not fully vetted the methodology used in these studies.

    In Malawi an error in data collection may have resulted in prevalence being underestimated.75 In Zanzibar, treatment has been ongoing,76 so the study does not reflect pre-treatment conditions.

    Detailed results and sources are available in this spreadsheet.

    Are there any negative or offsetting impacts?

    We discuss several possible considerations but do not see significant concerns.

    Administering deworming drugs seems to be a relatively straightforward program.77 However, there are potential issues that could reduce the effectiveness of some treatments, such as:

    • Drug quality: For example, if drugs are not stored properly, they may lose effectiveness or expire.
    • Dosage: If the incorrect dosage is given, the drugs may not have the intended effect and/or children may experience additional side effects.
    • Replacement of government funding: We have limited information about whether governments would pay for the parts of the program paid for by Unlimit Health in its absence. We also have little information about how governments would use the resources they put toward deworming if they did not choose to implement deworming programs.
    • Diversion of skilled labor: Drug distribution occurs only once every year or two and is conducted by volunteers in communities or teachers in schools. Given the limited time and skill demands of mass drug distribution, we are not highly concerned about distorted incentives for skilled professionals. Planning for the program can take senior government staff time. While we have limited information on what these staff would spend their time on in the absence of deworming programs, we would guess that they would support other education or health initiatives.
    • Adverse effects and unintended consequences of taking deworming drugs: Our understanding is that expected side effects are minimal and there is little reason to be concerned about drug resistance in the near team (more information from our report on deworming). We are somewhat more concerned about potential side effects during integrated NTD MDAs, since multiple drugs are taken within a short time period. However, it is our understanding that organizations follow protocols to space out the treatments so as to avoid adverse effects.
    • Popular discontent: We have heard a couple of accounts of discontent in response to Unlimit Health's mass drug administration campaigns, including one case that led to riots.78 Unlimit Health notes that following episodes of popular discontent, it has worked with governments to improve public education about the programs.79 Additionally, during deworming activities supported by Evidence Action's Deworm the World Initiative in Ogun State, Nigeria in December 2017, rumors of students collapsing reportedly generated panic that led some parents to take their children out of school; the Ogun State government denied that any students collapsed.80

    What do you get for your dollar?

    This section was last updated in November 2020. The information that Unlimit Health has provided since that date is not yet reflected in this section.

    We estimate that on average the total cost of a schistosomiasis treatment (which is often combined with STH treatment) delivered in Unlimit Health's programs is $1.03. Excluding the cost of drugs (which are often donated) and in-kind government contributions to the programs, we estimate that Unlimit Health's cost per treatment is $0.40.

    We make a number of assumptions and judgments in developing our estimate. Our process could introduce errors that overstate or understate the actual cost, and there are some significant sources of uncertainty. More on our approach below.

    Note that our estimate of the number of lives significantly improved by Unlimit Health's programs is a function of a number of difficult-to-estimate factors. We discuss how the cost per treatment figure relates to how much it costs to improve a child's health and development in our report on mass treatment programs for schistosomiasis and STH. We incorporate our estimates into a cost-effectiveness model which is available here.

    Unlimit Health's estimates

    In October 2017, Unlimit Health estimated that it would cost £0.18 on average per additional treatment delivered to a school-aged child, or about $0.24 USD at the exchange rate at the time. It estimated £0.27 per child in the first year of a new country program.81 For community-based distributions (used when adults are also targeted for treatment), Unlimit Health also estimated £0.27 per person in delivery costs, but noted that it would need to purchase praziquantel for adults at a cost of £0.24 per treatment. Praziquantel is currently donated by a pharmaceutical company for school-aged children but not for adults.82

    Our approach

    Our general approach to calculating the cost per treatment is to identify comparable cost and treatment data and take the ratio. We prefer to have a broadly representative selection of treatments in order to mitigate possible distortions, such as using data from a new program, which may incur costs from advocacy, mapping, etc. before it has delivered any treatments.

    It is our understanding that Unlimit Health generally intends to treat for STH in all places where it treats for schistosomiasis, so the treatments Unlimit Health reports can generally be interpreted as combination schistosomiasis and STH treatments,83 though we are aware of several cases in which schistosomiasis-only treatments were delivered either by design or due to problems with implementation, and of some cases where Unlimit Health delivered STH-only treatments (Unlimit Health told us that STH-only treatments are not counted in its treatment numbers).84

    To get the total cost, we attempt to include all partners (not just Unlimit Health), such that our cost per treatment represents everything required to deliver the treatments.85 In particular, we include these categories of costs:

    • Unlimit Health’s funding to country programs (e.g., to fund drug delivery).
    • Unlimit Health headquarters' costs (e.g., for management and technical salaries), including an estimate of costs paid by Imperial College (e.g., office space and some legal and administrative expenses).
    • Cost of drugs. We include the full market cost of all praziquantel that is needed to deliver the treatments, regardless of whether Unlimit Health purchased it or used donated drugs. It is our understanding that DFID funds praziquantel for some countries and that in recent years Unlimit Health has not purchased drugs beyond what is funded by DFID and donated by a pharmaceutical company.
    • Costs incurred by the government implementing the program (e.g., for staff salaries when working on treatment programs).

    We start with this total cost figure and apply adjustments in our cost-effectiveness analysis to account for cases where we believe the charity's funds have caused other actors to shift funds from a less cost-effective use to a more cost-effective use ("leverage") or from a more cost-effective use to a less cost-effective use ("funging"). (Further discussion in this blog post.)

    Unlimit Health notes that cost per treatment calculations should include sensitivity analysis86 —i.e., analysis on the degree to which the cost per treatment varies when various assumptions vary. We have not yet completed such an analysis.

    Our analysis

    We analyzed several sources of data, which cover different country programs and time periods between April 2015 and March 2019, and developed several different cost per treatment estimates based on the inclusion or exclusion of different types of costs. Full details in this spreadsheet. Our estimates are:

    • Unlimit Health’s cost per schistosomiasis treatment, including government costs and drug costs: $1.03,
    • Unlimit Health's cost per schistosomiasis treatment, including drug costs, but excluding government costs: $0.72,
    • Unlimit Health's cost per schistosomiasis treatment, including government costs, but excluding drug costs: $0.71, and
    • Unlimit Health's cost per schistosomiasis treatment, excluding drug costs and government costs: $0.40.

    Shortcomings of our analysis

    While we believe the estimates described above are reasonable, we want to highlight specific reasons to interpret them with caution.

    • We rely on reported treatment data. Our understanding is that these data can often be inaccurate. We have discounted the number of treatments (by 5%) based on the differences between reported treatment rates and treatment rates found in the coverage surveys discussed above (see footnote for why this is an imperfect comparison).87
    • We rely on an estimate that 30% of overall program costs are attributable to the government. We derived this estimate from an analysis of a single program in Niger (this footnote elaborates on this estimate).88
    • We have learned about some cases where Unlimit Health reported treatments for a program where it played a limited role.89 We have adjusted for the cases we know of, but are not confident that we have learned about all such cases.
    • We do not have data that indicate what proportion of drugs are wasted. We expect that in some cases drugs are purchased or donated but expire before use. We do not know how common this is. In our analysis, we have assumed that 10% of drugs are wasted.
    • We do not have data on the expenses Imperial College incurs to support Unlimit Health. Based on a conversation with Unlimit Health, we have roughly estimated these expenses as 10% of Unlimit Health's expenses (excluding drugs and government contributions).90
    • In DRC, Unlimit Health is one of several donors contributing to an integrated neglected tropical disease (NTD) program with pooled funding.91 We are uncertain about how many deworming treatments to attribute to Unlimit Health's contribution to the program. Our understanding is that the treatment numbers Unlimit Health has reported to us in recent years represent the total number of praziquantel treatments delivered in all of DRC, which we would guess would not be appropriate to attribute to Unlimit Health's contributions alone.92 In our 2020 cost per child dewormed analysis, as a rough estimation we have assumed that Unlimit Health's cost per child dewormed in DRC is the same as Sightsavers' cost per child dewormed in DRC.93 We note that Unlimit Health has provided data on total numbers of treatments delivered in DRC (including deworming treatments and treatments for other NTDs) for 2018-19—we have not yet reviewed this data closely, but we may do so for a future update to our cost per child dewormed per year analysis.94
    • Due to limited available information, we have excluded some countries Unlimit Health supported in some years from our cost per person dewormed per year analysis.95

    Is there room for more funding?

    We conduct "room for more funding" analysis to understand what portion of Unlimit Health's ideal future budget it will be unable to support with the funding it has or should expect to have available. We may then choose to either make or recommend grants to support those unfunded activities. Our most recent analysis finds that Unlimit Health expects that it could absorb around $85 million—prior to receiving any grants that GiveWell specifically makes or recommends—to support its deworming portfolio in 2022-24.

    Room for more funding analysis

    In general, we assess top charities' funding needs over a three-year period.96 We ask top charities to report their ideal budgets over the next three years, along with information about their current available funding and funding pipeline. The difference between a charity's three-year budget and the funding we project that it will have available to support that budget is the charity's "room for more funding."

    The main components of our room for more funding analyses are:

    • Available funding. We ask top charities to report how much funding they currently hold in the bank, including in reserves,97 and how much of this funding is committed or expected to be spent on specific future activities. The difference between these figures is the amount available to allocate to the charity's unfunded spending opportunities.
    • Expected funding. We project the amount of additional funding that top charities will receive to support their work over the next three years. These projections represent our best guesses based on top charities' past revenue and our understanding of their funding pipelines. They typically include funding currently held by GiveWell to be granted to the top charity, projected funding due to being a GiveWell top charity,98 and, if the top charity is part of a larger organization, projected unrestricted funding from that parent organization. They exclude any funding we may specifically recommend to the top charity subsequent to the analysis. We add this projected funding to the amount available to allocate to the charity's unfunded spending opportunities.
    • Spending opportunities. We ask top charities to report their ideal budgets in each of the next three years and to provide details on the specific spending opportunities included in these budgets. These opportunities are typically presented as one program year in a specific implementation geography (for example, deworming in Malawi in 2023), and they can represent either an extension of the top charity's previous support to a geography or an expansion of support to a new geography. We ask top charities to report the order in which they would prioritize funding these opportunities, which helps us to understand how available and expected funding will be allocated and what the marginal impact of additional funding beyond that amount would be.

    A charity's room for more funding represents the total budget for the charity's spending opportunities, less its available and expected funding. For example, if a charity proposes spending $50 million over the next three years and holds $10 million in uncommitted funding, and we project that it will receive an additional $15 million in revenue over the next three years, that charity's room for more funding is $25 million. (Note that a charity's total room for more funding figure includes funding gaps at all levels of cost-effectiveness—see below.) Our most recent analysis of Unlimit Health's room for more funding can be found in this spreadsheet.

    Grant investigation process

    Room for more funding analysis is a key part of our grant investigation process. We periodically request the information described above from top charities and update our room for more funding analyses. Our default is to update each top charity's room for more funding analysis annually, though we may choose to do so more or less frequently. The cadence on which we conduct updates depends largely on how often we grant funding to a top charity99 and how much we expect that charity's funding and budgets to have changed since our most recent funding decision.100 We have typically updated our analysis of Unlimit Health's room for more funding on an annual basis. Our most recent analysis of Unlimit Health's room for more funding can be found in this spreadsheet.

    After completing such an update, we may then choose to investigate potential grants to support the spending opportunities that we do not expect to be funded with the charity's available and expected funding, which we refer to as "funding gaps." The principles we follow in deciding whether or not to fill a funding gap are described on this page.

    The first of those principles is to put significant weight on our cost-effectiveness estimates. We use GiveDirectly's unconditional cash transfers as a benchmark for comparing the cost-effectiveness of different funding gaps, which we describe in multiples of "cash." Thus, if we estimate that a funding gap is "10x cash," this means we estimate it to be ten times as cost-effective as unconditional cash transfers. As of this writing, we have typically funded opportunities that meet or exceed a relatively high bar: 8x cash, or eight (or more) times as cost-effective as GiveDirectly's unconditional cash transfers. (Note that a charity's total room for more funding figure includes funding gaps at all levels of cost-effectiveness.)

    If we decide to fill a funding gap, we either make a grant from our Top Charities Fund101 or recommend that another funder—typically Open Philanthropy102 —makes a grant. This page lists all grants made or recommended by GiveWell. Typically, when GiveWell donors make a donation to a top charity,103 we don't expect that donation to be directed to a specific funding gap, but rather to contribute to supporting the overall portfolio of opportunities included within a charity's room for more funding.

    Unlimit Health's room for more funding

    Our most recent analysis of Unlimit Health's room for more funding can be found in this spreadsheet. Our analysis shows that Unlimit Health expects that it could absorb around $85 million—prior to receiving any grants that GiveWell specifically makes or recommends—to support its deworming portfolio in 2022-24.

    We will consider making or recommending grants to fill each of these funding gaps. As of November 2021, we expect to have sufficient funding at our discretion to fill all funding gaps we identify among our top charities that meet our current cost-effectiveness bar of 8x cash or better; we also expect to fill some funding gaps in the 5-8x cash range (for more details, see this blog post). For GiveWell donors who want to support the highest-priority funding needs among our top charities, we recommend donating to the Top Charities Fund.

    Unlimit Health as an organization

    We use qualitative assessments of our top charities to inform our funding recommendations. See this page for more information about this process and for our qualitative assessment of Unlimit Health as an organization.

    Sources

    Document Source
    Alan Fenwick, email to GiveWell, November 3, 2015 Unpublished
    Alan Fenwick, email to GiveWell, October 1, 2015 Unpublished
    Alan Fenwick, email to GiveWell, October 15, 2015 Unpublished
    Alan Fenwick, email to GiveWell, October 29, 2015 Unpublished
    Alan Fenwick, email to GiveWell, September 28, 2015 Source
    Alan Fenwick, SCI Director, conversation with GiveWell, October 14, 2014 Unpublished
    Alan Fenwick, SCI Director, conversation with GiveWell, October 15, 2014 Unpublished
    Alan Fenwick, SCI Director, email to GiveWell, November 24, 2014 Unpublished
    Alan Fenwick, SCI Director, phone conversation with GiveWell, February 16, 2011 Unpublished
    Alan Fenwick, SCI Director, phone conversation with GiveWell, September 15, 2011 Unpublished
    Alderman et al. 2006 Source (archive)
    Allen and Parker 2011 Source (archive)
    Allen and Parker 2012 Source (archive)
    Allen and Parker 2016 Source (archive)
    Anna Phillips, SCI Country Program Manager for Burkina Faso and Niger, email to GiveWell, October 13, 2011 Source
    Benjamin Styles, SCI Senior Biostatistician, phone conversation with GiveWell, August 12, 2011 Unpublished
    Bleakley 2007 Source (archive)
    Brooker et al. 2005 Source (archive)
    Burundi coverage survey dashboard, 2017-2018 Source
    Burundi coverage survey protocol, 2017-2018 Source
    Burundi coverage survey recommendations report, 2017-2018 Source
    Chaula and Tarimo 2014 Source (archive)
    Conversation with Dr. Fiona Fleming, August 1, 2017 Unpublished
    Côte d'Ivoire impact survey dashboard, 2017 Source
    Côte d'Ivoire impact survey protocol, 2017 Source
    Côte d'Ivoire impact survey recommendations report, 2017 Source
    Croke 2014 Source (archive)
    Crown Agents Total ICOSA Procurement Spend (updated 2014) Unpublished
    DFID glossary Source (archive)
    Dr. Wendy Harrison and Najwa Al Abdallah, conversation with GiveWell, May 4, 2016 Unpublished
    Dr. Wendy Harrison and Najwa Al Abdallah, conversation with GiveWell, September 27, 2016 Unpublished
    END Fund, conversation with GiveWell, October 17, 2016 Unpublished
    Ethiopia impact survey protocol, 2017 Source
    Fenwick et al. 2009 Source (archive)
    Fiona Fleming, conversation with GiveWell, November 5, 2015 Unpublished
    Fiona Fleming, conversation with GiveWell, September 19, 2016 Unpublished
    Fiona Fleming, email to GiveWell, March 3, 2016 Unpublished
    Fiona Fleming, email to GiveWell, November 5, 2015 Unpublished
    Fiona Fleming, email to GiveWell, October 11, 2016 Unpublished
    Fiona Fleming, email to GiveWell, September 18, 2015 Unpublished
    Fiona Fleming, SCI Senior Monitoring & Evaluation Manager, conversation with GiveWell, October 14, 2014 Unpublished
    Fiona Fleming, SCI Senior Monitoring & Evaluation Manager, conversation with GiveWell, September 21, 2015 Unpublished
    Fiona Fleming, SCI Senior Monitoring & Evaluation Manager, email to GiveWell, November 9, 2014 Unpublished
    Gates Foundation, Imperial College London (June 2002) Source
    GiveWell summary of SCI finances (October 2014) Source
    GiveWell, Availability of SCI Monitoring Results, 2022 Source
    GiveWell, Room for more funding analysis for SCI, 2021 Source
    GiveWell, SCI Expenditure, April 2020-March 2021 Source
    GiveWell's analysis of SCI 2017-18 budget May 2017 Redacted Source
    GiveWell's analysis of SCI 2018-19 budget May 2018 Redacted Source
    GiveWell's analysis of SCI budget vs. actuals 2015-16 Redacted Source
    GiveWell's analysis of SCI budget vs. actuals 2016-2017 Redacted Unpublished
    GiveWell's analysis of SCI cashbook summary 2015-16 Source
    GiveWell's analysis of SCI cashbooks 2016-2017 Unpublished
    GiveWell's analysis of SCI spending under DFID grant Source
    GiveWell's analysis of SCI spending under DFID grant (updated 2015) Source
    GiveWell's non-verbatim summary of a conversation with Alan Fenwick and Najwa Al Abdallah, September 14, 2015 Source
    GiveWell's non-verbatim summary of a conversation with Alan Fenwick, SCI Director, July 31, 2015 Source
    GiveWell's non-verbatim summary of a conversation with Alan Fenwick, SCI Director, June 17, 2010 Source
    GiveWell's non-verbatim summary of a conversation with Alan Fenwick, SCI Director, October 2, 2015 Source
    GiveWell's non-verbatim summary of a conversation with Blandine Labry, December 15, 2015 Source
    GiveWell's non-verbatim summary of a conversation with Dr. Wendy Harrison and Najwa al Abdallah, February 17, 2016 Source
    GiveWell's non-verbatim summary of a conversation with Dr. Wendy Harrison, Najwa Al Abdallah, and Dr. Lynsey Blair, April 6, 2016 Source
    GiveWell's non-verbatim summary of a conversation with Giuseppina Ortu on June 20, 2014 Source
    GiveWell's non-verbatim summary of a conversation with Lynsey Blair, October 16, 2014 Source
    GiveWell's non-verbatim summary of a conversation with Michael French, October 15, 2014 Source
    GiveWell's non-verbatim summary of a conversation with Oumer Shafi, November 4, 2014 Source
    GiveWell's non-verbatim summary of a conversation with Sarah Nogaro, October 16, 2014 Source
    GiveWell's non-verbatim summary of a conversation with Wendy Harrison and Najwa Al Abdallah on September 8, 2015 Source
    GiveWell's non-verbatim summary of a conversation with Yolisa Nalule, October 14, 2014 Source
    GiveWell's notes from visit to Malawi on October 17-19, 2011 Source
    GiveWell’s non-verbatim summary of conversations with Grace Hollister on September 21 and October 1, 2015 Unpublished
    Kabatereine et al. 2001 Source (archive)
    Kabatereine et al. 2006 Source (archive)
    Kabatereine et al. 2007 Source (archive)
    Kieran Bird, SCI Finance Manager, conversation with GiveWell, October 16, 2014 Unpublished
    Kieran Bird, SCI Finance Manager, email to GiveWell, November 4, 2014 Unpublished
    Knopp et al. 2009 Source (archive)
    Knopp et al. 2013 Source (archive)
    Koukounari 2011 Source
    Koukounari et al. 2007 Source (archive)
    Leslie et al. 2011 Source (archive)
    Liberia coverage survey dashboard, 2018 Source
    Liberia coverage survey protocol, 2018 Source
    Liberia coverage survey recommendations report, 2018 Source
    Liberia impact survey dashboard, 2018 Source
    LSTM Mozambique trip report (May 2015) Unpublished
    Malawi coverage survey dashboard, 2017 Source
    Malawi coverage survey protocol, 2017 Source
    Malawi coverage survey recommendations report, 2017 Source
    Malawi impact survey recommendations report, 2016 Source
    Malawi reassessment survey dashboard, 2017 Source
    Malawi reassessment survey protocol, 2017 Source
    Malawi reassessment survey protocol, 2018 Source
    Malawi reassessment survey recommendations report, 2017 Source
    Malawi reassessment survey recommendations report, 2018 Source
    Mapping of Schistosomiasis and Soil-transmitted helminthiasis in Yemen Source
    Mazigo et al. 2012 Source (archive)
    Michelle Clements, SCI Senior Biostatistician, conversation with GiveWell, October 15, 2014 Unpublished
    Miguel and Kremer 2004 Source (archive)
    Muhumuza et al. 2009 Source (archive)
    Muhumuza et al. 2013 Source (archive)
    Muhumuza et al. 2014 Source (archive)
    Najwa Al Abdallah, Alan Fenwick, and Wendy Harrison, conversation with GiveWell, October 11, 2016 Unpublished
    Najwa Al Abdallah, conversation with GiveWell, October 22, 2015 Unpublished
    Najwa Al Abdallah, Dr. Lynsey Blair, and Dr. Wendy Harrison, conversation with GiveWell, September 27, 2016 Unpublished
    Najwa Al Abdallah, email to GiveWell, August 15, 2018 Unpublished
    Najwa Al Abdallah, email to GiveWell, May 18, 2016 Unpublished
    Najwa Al Abdallah, email to GiveWell, October 14, 2016 Unpublished
    Najwa Al Abdallah, email to GiveWell, October 19, 2016 Unpublished
    Najwa Al Abdallah, email to GiveWell, September 22, 2016 Unpublished
    Nigeria NTD stakeholders meeting and potential SCI involvement scoping document Source
    Nigeria stakeholders meeting summary (May 2015) Source
    Nigeria trip report (June 2015) Source
    Nigerian Tribune, "Panic in Ogun schools over deworm exercise," December 2017 Source (archive)
    OFX, Historical Exchange Rates Source (archive)
    Parker and Allen 2011 Source (archive)
    Parker and Allen 2014 Source (archive)
    Parker, Allen, and Hastings 2007 Source (archive)
    Pinot de Moira et al. 2010 Source (archive)
    Rudge et al. 2008 Source (archive)
    Scheich et al., 2012 Source (archive)
    Schematic of SCIs Income Sources and Imperial Account structure Apr 2016 Source
    Schistosomiasis Control Initiative, conversation with GiveWell, October 7, 2016 Unpublished
    Schistosomiasis Control Initiative, conversation with GiveWell, September 6, 2016 Unpublished
    SCI 3 Year Budget September 2017 Redacted Source
    SCI Account summary (May 2011) Source
    SCI advisory board financial report (June 2013) Source
    SCI advisory board financial report (June 2014) Source
    SCI allocation table 2016-2017 Source
    SCI board financial details (June 2014) Unpublished
    SCI Board management accounts (April 2010) Source
    SCI budget 3 options October 2016 Redacted Source
    SCI Budget vs Actuals April 2016-March 2017 Redacted Source
    SCI Budget vs Actuals April 2017-March 2018 Source
    SCI Budget vs Actuals April 2018-March 2019 Source
    SCI budget vs. actuals 2015-16 Redacted Source
    SCI budget vs. actuals April-July 2016 Redacted Source
    SCI Budget vs. actuals through June 2018, Redacted Source
    SCI Burundi coverage survey draft results 2017-18 (French) Unpublished
    SCI Burundi coverage survey protocol 2017 Unpublished
    SCI Burundi impact survey protocol 2017 Unpublished
    SCI Burundi impact survey recommendations report 2017 Source
    SCI Burundi June 2014 Open Day poster Source
    SCI Burundi: Impact Source (archive)
    SCI cashbook Burundi 2016-17 Unpublished
    SCI cashbook Côte d'Ivoire 2015-16 Unpublished
    SCI cashbook Côte d'Ivoire 2016-17 Unpublished
    SCI cashbook DRC 2015-16 Unpublished
    SCI cashbook DRC 2016-17 Unpublished
    SCI cashbook Ethiopia 2016-17 Unpublished
    SCI cashbook Madagascar 2016-17 Unpublished
    SCI cashbook Malawi 2015-16 Unpublished
    SCI cashbook Malawi 2016-17 Unpublished
    SCI cashbook Niger 2015-16 Unpublished
    SCI cashbook Niger 2016-17 Unpublished
    SCI cashbook summary 2015-16 Source
    SCI cashbook Tanzania 2015-16 Unpublished
    SCI cashbook Tanzania 2016-17 Unpublished
    SCI cashbook Uganda 2015-16 Unpublished
    SCI cashbook Uganda 2016-17 Unpublished
    SCI cashbook Zanzibar 2016-17 Unpublished
    SCI CNTD spending data (FY2-FY4) Source
    SCI Commitments July 2017 to March 2018 Redacted Source
    SCI contribution to the global effort to control and eliminate schistosomiasis Source
    SCI Côte d'Ivoire coverage survey 2014 Unpublished
    SCI Côte d'Ivoire coverage survey dashboard 2016 Unpublished
    SCI Côte d'Ivoire coverage survey protocol 2016 Unpublished
    SCI Côte d'Ivoire coverage survey protocol 2018 Unpublished
    SCI Côte d'Ivoire coverage survey recommendations report 2016 Source
    SCI Côte d'Ivoire coverage survey recommendations report 2018 Source
    SCI Côte d'Ivoire impact survey protocol 2016 Unpublished
    SCI Côte d'Ivoire impact survey recommendations report 2016 (Phase 1 schools follow-up and Phase 2 schools baseline) Source
    SCI Côte d'Ivoire panel study baseline report Unpublished
    SCI draft budget 2015-2016 Source
    SCI draft financial statements for 2013/14 and 2014/15 Source
    SCI draft reserves policy (September 2015) Source
    SCI DRC baseline impact survey 2015 Unpublished
    SCI DRC impact survey recommendations report, 2017-2018 Unpublished
    SCI DRC neglected tropical disease treatment data 2016-17 Source
    SCI DRC summary budgets and treatment numbers 2016-17 Source
    SCI Ethiopia annual treatments 2016 Source
    SCI Ethiopia coverage survey 2015 Unpublished
    SCI Ethiopia coverage survey dashboard 2016 Source
    SCI Ethiopia coverage survey protocol 2017 Source
    SCI Ethiopia coverage survey recommendations report 2017 Source
    SCI Ethiopia impact survey baseline report 2015-16 Unpublished
    SCI Ethiopia impact survey protocol 2016-17 Unpublished
    SCI Ethiopia impact survey recommendations report 2016-17 Source
    SCI Ethiopia mapping of SCH and STH 2014 Source
    SCI Ethiopia mapping surveys 2013-15 Unpublished
    SCI Ethiopia treatment campaign summary report (April 2015) Source
    SCI financial statement 2013/14 and 2014/15 (revised October 2015) Source
    SCI Foundation, 2017-19 consolidated treatment numbers Source
    SCI Foundation, 2018-19 DRC NTD treatments Source
    SCI Foundation, 2018-19 treatment numbers Source
    SCI Foundation, budget vs. actuals 2018-19 Source
    SCI Foundation, Financial statements April 2019-March 2020 Source
    SCI Foundation, Financial statements April 2020-March 2021 Source
    SCI Full 5 year budget (July 2018), Redacted Source
    SCI fundraising targets (November 2014) Source
    SCI Gates Foundation final report (January 2011) Source
    SCI GiveWell Funding Scenario Notes and Assumptions Source
    SCI GiveWell Funding Scenarios 2018-2021 Source
    SCI GiveWell income reconciliation 2016 Unpublished
    SCI Global treatment numbers 2015-16 Source
    SCI IC Trust statement July 2016 Redacted Source
    SCI IC Trust summary (September 2011) Source
    SCI ICOSA Mid-Year Report 2014 Source
    SCI impact and coverage survey plans Source
    SCI Imperial initiative to protect children from tropical disease awarded ₤25m government backing Source (archive)
    SCI L-account 2016 Unpublished
    SCI L-account April-October 2016 Unpublished
    SCI Liberia coverage survey dashboard 2017 Unpublished
    SCI Liberia coverage survey protocol 2017 Unpublished
    SCI Liberia coverage survey recommendations report 2017 Source
    SCI Liberia impact survey dashboard 2012-13 Unpublished
    SCI Liberia impact survey follow up recommendations report 2013 Unpublished
    SCI Liberia panel study baseline report Unpublished
    SCI M&E survey schedule March 2017 Source
    SCI M&E survey schedule September 2016 Source
    SCI M&E timeline (May 2015) Source
    SCI Madagascar baseline impact survey 2015 Unpublished
    SCI Madagascar coverage survey dashboard 2016 Source
    SCI Madagascar coverage survey dashboard 2017 Source
    SCI Madagascar coverage survey protocol 2017 Unpublished
    SCI Madagascar coverage survey recommendation report 2016 Unpublished
    SCI Madagascar coverage survey recommendations report 2017 Source
    SCI Madagascar coverage survey recommendations report-EN-draft Source
    SCI Madagascar impact survey protocol 2017 Unpublished
    SCI Madagascar impact survey recommendations report 2017 Source
    SCI Malawi coverage survey 2012 Source
    SCI Malawi coverage survey 2014 Source
    SCI Malawi coverage survey dashboard 2017 Unpublished
    SCI Malawi Coverage Survey Protocol 2016 Unpublished
    SCI Malawi coverage survey protocol 2017 Unpublished
    SCI Malawi impact study – second follow up Source
    SCI Malawi impact survey dashboard 2016 Unpublished
    SCI Malawi impact survey protocol 2017 Unpublished
    SCI Malawi impact survey recommendations report 2017 Source
    SCI Malawi panel study Source
    SCI Malawi spending data (November 2011 to August 2013) Source
    SCI Mauritania coverage survey protocol 2017 Unpublished
    SCI Mauritania coverage survey recommendations report 2017 Source
    SCI Mozambique coverage survey 2015 Source
    SCI Mozambique coverage survey 2016 Unpublished
    SCI Mozambique coverage survey presentation 2016 Unpublished
    SCI Mozambique PRZ LQAS Coverage Survey Northern Provinces 2016 Unpublished
    SCI Neglected tropical diseases in Mozambique Unpublished
    SCI Niger coverage survey dashboard 2017 Unpublished
    SCI Niger coverage survey protocol 2017 Unpublished
    SCI Niger coverage survey recommendations report 2017 Source
    SCI Niger panel study 2011 Unpublished
    SCI Niger spending data (October 2011 to May 2013) Source
    SCI Nigeria coverage survey protocol 2017 (with Sightsavers) Unpublished
    SCI Nigeria coverage survey report 2017 (with Sightsavers) Unpublished
    SCI planned SCH treatment numbers by country by year (October 2015) Source
    SCI Proposal by SCI, Imperial College to manage the Program for Integrated Control of Neglected Tropical Diseases in Côte d'Ivoire Unpublished
    SCI report to DFID (October 2013) Source
    SCI report to DFID (September 2015) Source
    SCI report to GiveWell (September 2013) Unpublished
    SCI report to GiveWell (September 2014) Source
    SCI responses to GiveWell questions on financial statements (October 2015) Source
    SCI Revised 5 year budget for GiveWell (August 2018), Redacted Source
    SCI Rwanda June 2014 Open Day Poster Source
    SCI Rwanda: Strategy Source (archive)
    SCI Summary sheet of treatments instigated and overseen by SCI Source
    SCI supporting documents matrix (September 2015) Source
    SCI Tanzania coverage survey national technical report 2017 Unpublished
    SCI Tanzania coverage survey protocol 2016-17 draft Unpublished
    SCI Tanzania impact survey baseline report 2016 Unpublished
    SCI Tanzania impact survey protocol 2017 Unpublished
    SCI Tanzania impact survey recommendations report 2017 Unpublished
    SCI Tanzania spending data (March 2011 to July 2013) Source
    SCI treatment data 2014-16 Source
    SCI treatment gap forecast 2016 Source
    SCI treatment numbers (October 2014) Source
    SCI Treatment numbers 2010-2017 Source
    SCI Treatment numbers 2010-2017 and targets 17-18 all countries (August 2018) Source
    SCI Uganda coverage survey 2014 Source
    SCI Uganda coverage survey protocol 2017 Unpublished
    SCI Uganda coverage survey recommendations report 2017 Source
    SCI Uganda panel study baseline report Unpublished
    SCI Uganda spending data (September 2011 to August 2013) Source
    SCI Unrestricted income update (July 2018) Source
    SCI Zambia coverage survey 2015 Source
    SCI Zambia panel study baseline report Unpublished
    SCI Zanzibar coverage survey 2015 Source
    Sightsavers, "Ascend: fighting disease in West and Central Africa" Source
    Standley et al. 2009 Source (archive)
    Standley et al. 2010 Source (archive)
    Stothard et al. 2009 Source
    Stothard et al. 2013 Source
    Styles 2011 Source
    Sudan annual workplan (April 2015 to March 2016) Source
    Sudan annual workplan for WHO (2015) Source
    Sudan campaign photos Source
    Sudan cash book Source
    Sudan joint request for selected PC medicines Source
    Sudan NTD concept paper (2015-2018) Source
    Sudan PZQ and ALB treatments by locality (2015) Source
    Summary Technical Report: Schistosomiasis Control in Yemen (July 2014) Source
    Tohon et al. 2008 Source (archive)
    Top 20 countries, estimated schistosomiasis infections Source
    Touré et al. 2008 Source (archive)
    Utroska et al. 1989 Source (archive)
    Wendy Harrison and Sarah Knowles, SCI Managing Director and Biostatistician, conversations with GiveWell, April 9 and 14, 2014 Source
    Wendy Harrison, email to GiveWell, October 11, 2016 Unpublished
    Wendy Harrison, email to GiveWell, October 30, 2018 Unpublished
    Wendy Harrison, email to GiveWell, September 8, 2015 Unpublished
    Wendy Harrison, SCI Managing Director, email to GiveWell, March 4, 2014 Unpublished
    WHO schistosomiasis treatment gap data Unpublished
    WHO STH factsheet Source (archive)
    WHO STH treatment report Source (archive)
    WHO Weekly epidemiological record, 18 December 2015 Source (archive)
    WHO Weekly epidemiological record, 6 March 2015 Source (archive)
    WHO, Summary of global update on preventive chemotherapy implementation in 2015 Source (archive)
    WHO, Summary of global update on preventive chemotherapy implementation in 2016 Source (archive)
    Wikipedia entry for Unguja Source (archive)
    Zanzibar coverage survey dashboard, 2018 Source
    Zanzibar coverage survey protocol, 2018 Source
    Zanzibar coverage survey recommendations report, 2018 Source
    • 1

      "Objectives of SCI

      • To encourage development of sustainable schistosomiasis and STH control programmes in sub-Saharan Africa.
      • In the selected countries: to reach at least 75% of school-aged children (which in most countries would be from 6 to 15-year-old) and other high-risk groups with chemotherapy, namely PZQ and ALB; and thereby reducing prevalence and intensity of schistosomiasis and STH infections; as well as reducing schistosomiasis-related morbidity in high risk groups; and burdens due to STH infections in the targeted populations.
      • To create a demand for sustained schistosomiasis and STH control.
      • To promote access to anthelmintic drugs and good case management in the regular health system.
      • To develop a rigorous monitoring and evaluation plan which will generate the information required to determine whether or not the objectives have been met."

      Fenwick et al. 2009, Pg. 3.

    • 2

    • 3

      See this spreadsheet, sheet "Country breakdown in USD," cell C17.

    • 4

      See here for details.

    • 5

      See this spreadsheet, sheet "Source: Program Costs," rows "Cote d'Ivoire," "DRC," "Liberia," and "Niger."

    • 6

      See this spreadsheet, sheet "Source: Program Costs," columns O-Q.

    • 7

      See this spreadsheet, sheet "Calculation: Unrestricted revenue" for calculation.

    • 8

    • 9

      For details on what is included in this cost category, see cell note in cell A14 of this spreadsheet, sheet "Country Breakdown in USD."

    • 10

      For details on what is included in this cost category, see cell note in cell A15 of this spreadsheet, sheet "Country Breakdown in USD."

    • 11

    • 12

      "In 2015, SCI implemented the use of cashbooks by its country programs to report monthly spending. Countries where SCI receives funding from the Department for International Development (DFID) implemented this system first, and it has now been rolled out to all of SCI's country programs. A few countries are currently delayed in submitting cashbooks to SCI because of limited capacity." @GiveWell's non-verbatim summary of a conversation with Dr. Wendy Harrison and Najwa Al Abdallah, February 17, 2016@.

    • 13
    • Cashbooks:
      • SCI cashbook Tanzania 2016-17.
      • SCI cashbook Uganda 2016-17.
      • SCI cashbook Zanzibar 2016-17.
      • SCI cashbook Niger 2016-17.
      • SCI cashbook Malawi 2016-17.
      • SCI cashbook Côte d'Ivoire 2016-17.
      • SCI cashbook Burundi 2016-17.
    • We also received 2016-2017 cashbooks from Ethiopia, Madagascar, and Democratic Republic of the Congo (DRC). We have excluded those countries from the overall analysis for reasons that are explained in GiveWell's analysis of SCI cashbooks 2016-2017, "Summary" tab, row 18.
      • SCI cashbook Ethiopia 2016-17
      • SCI cashbook Madagascar 2016-17
      • SCI cashbook DRC 2016-17

    • 14
    • SCI cashbook Niger 2015-16.
    • SCI cashbook Côte d'Ivoire 2015-16.
    • SCI cashbook DRC 2015-16.
    • SCI cashbook Malawi 2015-16.
    • SCI cashbook Tanzania 2015-16.
    • SCI cashbook Uganda 2015-16.
    • SCI cashbook summary 2015-16.

    • 15

      Unlimit Health notes, "SCI carries out capacity development of and actively trains in-country accountants on the activity terminology and types of expenses that fall under each which are clearly defined e.g Drug distribution is for expenses related to conducting MDA at a site while drug logistics refer to the transportation of drugs from the source to MDA sites." Comment provided in response to our draft of this page in June 2016.

    • 16

    • 17

      "The reason for the differences: SCI BvA is prepared based on imperial college information system (ICIS) -expenditure is recorded when a transfer is made to a country while the cash book record when the money is being spent. Example of Tanzania 2 transfers were made: on 16 Dec. 2015 for the amount GBP 344,018 from DFID and GBP 406,696 from unrestricted. In the cash books Tanzania didn’t spend all the money received in Dec. 2015 as the MDA was delayed and took place during Jan-March 2016." Comment provided in response to a draft of this page in June 2016.

    • 18

    • 19
      • GiveWell's analysis of SCI cashbooks 2016-2017 "Summary" sheet.
      • Descriptions for these activities drawn from line items in SCI cashbooks from 2015-2016 and 2016-2017.

    • 20
      • For a full description of the methodology used in each survey (and sources for the statements below), see this spreadsheet, "Methods" sheet.
        • Unlimit Health has told us (in reports on the coverage surveys or through personal communication) that villages are selected randomly within districts in which the coverage survey is implemented (often with adjustments to account for differing population sizes of villages).
        • Unlimit Health has told us (in reports on the coverage surveys or through personal communication) that households are either chosen through random selection from a village register or through a "random walk" method. A description of a "random walk" method for choosing households:
          • "Households were selected using the random walk procedure. A central point in the village was designated and a bottle was spun to randomly select a direction of walk. All households along the direction of walk were counted. A sampling fraction was calculated and the households selected." SCI Madagascar coverage survey recommendation report 2016, Pgs 4-5.
        • Note that the selection process for villages and households is not fully clear to us for all coverage surveys we have seen. Also note that coverage surveys are implemented in a selection of districts covered by the MDA program, and the selection of districts for coverage surveys is not always random or intended to be representative of the MDA program as a whole. See this spreadsheet, "Methods" sheet, for sources and details.

    • 21

      For a full description of the methodology used in each survey, see this spreadsheet, "Methods" sheet and the primary sources linked in the "Report" column.

    • 22
      • "The primary objectives of this coverage survey were to:
        1. Quantify and validate PZQ and MBD treatment coverage for SCH and STH, respectively;
        2. Assess coverage rates disaggregated by school attendance and gender for SAC;
        3. Collect information on why targeted eligible individuals did not receive or accept treatment.

        Reported coverage was defined as, number of SAC ingesting drugs / eligible SAC population x 100
        Survey coverage defined as, number of SAC interviewed that ingested the drug / total number of interviewed
        SAC x 100" SCI Madagascar coverage survey recommendation report 2016, Pg. 4.

      • Our understanding (formed over several conversations with Unlimit Health) is that the governments of countries with Unlimit Health-supported programs calculate "reported coverage" figures. The numerator of these figures (the number of treatments delivered to school-aged children according to government administrative data) is calculated by aggregating data from each school in the program on the number of reported treatments delivered. The denominator of this figure (estimated number of eligible school-aged children in the area) is calculated by referencing the most recent government census in the area, which may be adjusted to estimate the effect of population growth since the last census, or by using school enrollment data.

    • 23

      For a comparison between survey coverage and reported coverage rates, see this spreadsheet, “Results” sheet.

    • 24
      • It seems plausible to us that there may be an incentive for schools to over-report the number of treatments delivered, or for districts or regions to over-report aggregated treatment figures.
      • We have not seen calculations used by governments to find the denominator of the reported coverage calculation (estimated number of eligible school-aged children in the area), but it seems that these estimates are sometimes substantially inaccurate. For example, note that in this spreadsheet, "Results" sheet, some reported coverage estimates are over 100% (which means that the aggregated figures for reported treatment of school-aged children are greater than the government's estimate of the total number of school-aged children in the area).
      • Several people have told us that it is difficult to get accurate government administrative data and that data is often missing from some portion of schools/clinics/etc.

    • 25

      See this spreadsheet, "Methods" sheet, column D.

    • 26

      See this spreadsheet, "Methods" sheet, column T as well as the sources in the "Report" column.

    • 27

      "The actual sample included caretakers of 578 children." SCI Mozambique coverage survey 2015, Pg. 5. Note that we have only received permission to publish a summary of this report. Quotation is from the full, unpublished report.

    • 28

      SCI Mozambique coverage survey 2015, Pg. 23. Note that we have only received permission to publish a summary of this report. Quotation is from the full, unpublished report.

    • 29

      SCI Mozambique coverage survey 2015, Pg. 24. Note that we have only received permission to publish a summary of this report. Quotation is from the full, unpublished report.

    • 30

      "Pole doses were used as a proxy for the medication – rather than showing the medication we showed them the poles. Only parents (not teachers) can authorize medication to kids in Mozambique and they need to be involved and informed during the social mobilization which is conducted usually by the district officers and the activists, in the week preceding the distribution of the medication. The activists use posters and poles when they inform the community and the poles are very recognizable. The poles are also used in the integrated campaigns – and in most of these districts campaigns have been conducted for a number of years. So…they may be recognized as part of a preventive medication campaign rather than related to the PZQ only." Fiona Fleming, conversation with GiveWell, November 5, 2015 quoting from an email from FPSU.

    • 31

      The interview protocol (provided in Portuguese) does not include any verification questions. See SCI Mozambique coverage survey 2016, Pgs 28-29.

    • 32

      See the bottom left figure on Pg 7 of SCI Mauritania coverage survey recommendations report 2017 for the recognition rates, and this spreadsheet, “Results” sheet, for the survey coverage rates.

    • 33

      See the top right figure on Pg 6 of SCI Ethiopia coverage survey recommendations report 2017 for the recognition rates, and this spreadsheet, “Results” sheet, for the survey coverage rates.

    • 34
      • Fiona Fleming, conversation with GiveWell, September 19, 2016
      • Email from Unlimit Health's partner FPSU, quoted to us by Unlimit Health:
        "Further to our conversation, here is a recount of the supervision of the survey:
        1. Initially there were 6 teams with Cabo Delgado and Nampula divided into two. All of them were supervised by Don and myself during 2 days of intense data collection in Nampula Cidade to ensure the areas selected were visited, the random selection of houses and respondent and the correct administration of the questionnaire plus the team organization and supervision.
        2. Teams were then reduced to 4 as it made more sense to complete districts before traveling to the next.
        3. Each team had one or two supervisors, one being the provincial NTD or M&E staff in the case of Zambezia, the national supervisor being NTD or other MISAU Departments (Cabo Delgado, Zambezia). The district officers worked in teams so that most of the interviews in one district were done by the other district officers with no connection with the district. In fact in most cases the district officer of that particular district was solving logistical and administrative issues which are quite intense in Mozambique (permission by the provincial delegate, permission by the district and the locality chief letters of introduction all signed and stamped, motorcycles rental and bills etc).
        4. Nampula was identified as the most complex area and the team was supervised by me during the completion of Ilha and Mossuril districts plus Cidade Nampula while Don supervised District Nampula and again the finalization of Cidade Nampula which was pretty complicated. In other areas the survey was easier but in Chinde Islands where Dr Xose went from one island to the next to find the selected village and got lost for a few days because the district has been divided into two and we had to select again the villages. All other supervisors provided daily reports to Don. They also wrote a report of the questionnaires received every day, reviewed and sent.
        5. In most cases, data was sent to the database and reviewed every day. Once the supervisor had sent the questionnaire editing was only allowed to Don Whitson. Some areas did paper based questionnaires and this was again submitted and reviewed."

        Fiona Fleming, email to GiveWell, November 5, 2015.

    • 35

      See notes in this spreadsheet, "Methods" sheet, column N, "data quality control."

    • 36
      • Conversation with Unlimit Health staff, April 3, 2019
      • This process involves "identifying obvious missing or incorrect data and subsequently rectifying errors in real time." Unlimit Health, comments on review, November 11, 2019.

    • 37

      Citations for all statements in this list can be found in this spreadsheet, "Methods" sheet.

    • 38

      For the Zanzibar (2015) survey, Unlimit Health told us, "Shehias [sub-district administrative units] were stratified by MoH treatment programme and by elimination programme; required sample size selected from each startum [sic]." Fiona Fleming, email to GiveWell, November 5, 2015. We do not know how to interpret this.

    • 39
      • In Ethiopia (2016 and 2017): "Criteria for purposive selection of the Districts include:
        • Treatment was given for both SCH and STH
        • Safety of data collectors while in the district
        • Districts that have been treated within the last 3 months
        • Too high or low reported coverage (either from the current round or historical)",

        SCI Ethiopia coverage survey protocol 2017, Pg 10.

      • In Malawi (2017): "Districts were purposively selected for this survey, four of which were surveyed in 2016, based on safety and whether they were surveyed in previously." "...in the Southern districts of Malawi, health workers, local leaders and medical services were attacked in villages due to superstition surrounding witchcraft. It was decided for the security of survey teams that the Southern districts would not be visited", SCI Malawi coverage survey protocol 2017, p.12.

    • 40

      See this spreadsheet, “Methods” sheet, column H (“How villages chosen”) and column J (“How households chosen”), for details.

    • 41

      Fiona Fleming, email to GiveWell, November 5, 2015 (quoting from email from FPSU, which runs the program in Mozambique).

    • 42
      • See this spreadsheet, “Methods” sheet, column L, “Who conducted the survey,” for details.
      • For Malawi (2012), Uganda (2014), Zanzibar (2015), and Zambia (2015), see Fiona Fleming, email to GiveWell, November 5, 2015.

    • 43
      • For Côte d'Ivoire, coverage reported by parents was lower than coverage reported by children. In Malawi, coverage reported by parents was similar to coverage reported by children. In Côte d'Ivoire, parents' answers were excluded from the reported results: "Calculation of validated coverage rates initially included answers from both proxy and direct interviews across all 4 districts. As shown in Figure 2, coverage rates calculated based on direct interviews were higher than those which included responses given by proxy (p < 0.001), perhaps due to the parent erring on the side of caution when giving their answer. As direct interviews are believed to be more robust, we omitted data from proxy interviews when calculating final coverage rates (Table 4)." SCI Côte d'Ivoire coverage survey 2014, Pg. 10.
      • In Malawi, parents' answers were included in reported results. Proxy results were similar to results from direct interviewing in Malawi, and were included in the headline analysis. SCI Malawi coverage survey 2012, Pg. 11, Figure 4.
      • About 9% of the responses in Côte d'Ivoire were from parents. SCI Côte d'Ivoire coverage survey 2014, Pg. 9, Table 3.
        • Number of yes/no answers obtained in person (summed across four districts): 2178.
        • Number of proxy yes/no answers obtained (summed across four districts): 228.
        • 228 / (228 + 2178) = ~.09.

    • 44
      • Fiona Fleming, email to GiveWell, November 5, 2015.
      • Fiona Fleming, conversation with GiveWell, September 19, 2016.

    • 45

      See this spreadsheet, “Methods” sheet, column R, “Parents answer for children?,” for details.

    • 46

      "Some of the data for the Côte d'Ivoire coverage survey conducted in 2016 may not
      be reliable for the following reasons.
      Possibility of other NTD MDAs
      Unlimit Health found that interviewees were more likely to report receiving ALB than PZQ, even though the two drugs were both meant to be administered together in the latest MDA. It is possible that a treatment for lymphatic filariasis (LF) may have been administered between the time of the PZQ and ALB treatment and the time of the survey. This would affect the interviewees' answers because ALB is also used to treat LF. It is also possible that the amount of drugs available was insufficient, so people received only ALB and not PZQ. Unlimit Health plans to investigate this by asking further questions of the national program and of the district officials to find out whether other NTD MDAs may have happened during that time period. This will help Unlimit Health know how reliable their data are. Unlimit Health believes that the data from one of the three districts surveyed, Prikro, does not suffer from this issue.
      District-specific problems
      Bangolo
      Unlimit Health was told before the survey that Bangolo was one of the treated districts, and conducted its survey there based on this information. However, Unlimit Health later found out that Bangolo had not been a part of the latest MDA. This leads Unlimit Health to believe that interviewees in this district may have been responding about some treatment other than the PZQ and ALB.
      Aboisso
      In Aboisso, Unlimit Health had difficulty finding people to interview, since many people there live in camps such as internally displaced people's camps. Since the district lacks a regular community structure, there were fewer people available to interview than usual. This resulted in very large confidence intervals for the district."
      Conversation with Dr. Fiona Fleming, August 1, 2017.

    • 47

      Tohon et al. 2008. In 2014, Unlimit Health sent us a more recent report, SCI Niger panel study 2011. SCI Niger panel study 2011 is our only source of data on the second to fifth year followups. This analysis excludes three of the eight schools that were originally in the study as well as the 56% of participants lost to follow-up in the remaining five schools. Participants who were followed up for some years but lost to follow-up by the end of the study are completely excluded rather than being included in a separate analysis of the earlier follow-ups. The 3 schools were apparently dropped because they were not surveyed in some years:

      • "Of these eight sentinel sites, five had data available in all six years of the study up to 2010." SCI Niger panel study 2011, Pg. 19.
      • "455 out of 1024 (44%) children recruited at baseline from November 2004 to April 2005 were successfully followed up for the full duration of the 6-year sentinel site monitoring." SCI Niger panel study 2011, Pg. 25.

      For this reason, we feel that the results from SCI Niger panel study 2011 are hard to interpret, and we don't present them here, relying instead on Tohon et al. 2008.

    • 48
      • We have seen two reports on the Burundi (2007-2010) study, Styles 2011 and Koukounari 2011, which use different methods of analysis and included different numbers of participants. Which individuals they included in their analyses is not always clear, however the fact they report similar results for prevalence provides some evidence that the results are not highly dependent on these choices. (Both Dr. Koukounari and Dr. Styles are statisticians who formerly worked at Unlimit Health.)
        • Pilot study: Styles 2011 and Koukounari 2011 have strengths and weaknesses in defining the sample for the pilot study. Styles 2011 includes a much larger number of individuals (Styles 2011, Pg. 3, Figure 1b compared to 710 in Koukounari 2011, Pg. 7, Figure 2). However, it is unclear whether the individuals in the follow-up were all in the original cohort or whether some were added to the sample later. The two reports indicate different numbers of students included in the study and retraced at each followup. We have seen two explanations for the differences:
          (1) "Dr. Koukounari only included students who were in first grade during the first year of the study and who were successfully surveyed every year of the study. In addition to the children counted by Dr. Koukounari, Dr. Styles included students who entered first grade and were added into the study in subsequent years, as well as students who were missing data from some years. Each of these strategies for data analysis has benefits and drawbacks. SCI initially planned to do a cross-sectional evaluation of sixth grade students every year, because each year the current sixth grade class would have received more rounds of treatment over the course of elementary school than the previous year. SCI did not complete this plan, but Dr. Koukounari included the data from the sixth grade students in the baseline data. Dr. Styles did not include this data." GiveWell's non-verbatim summary of a conversation with Giuseppina Ortu on June 20, 2014.
          (2) "In addition to the longitudinal studies at each follow-up newly recruited children were added to these surveys. At 1st follow-up (2008) 2288 newly recruited children were added to these surveys with range age: 6-21 years old and median age: 12 years old. Of these 2288, only 210 i.e. (9.18 %), were of age 6 and eligible to be included in the specific cross sectional data analysis. At 2nd follow-up (2009) 2311 newly recruited children were added to these surveys with range age: 5-20 years old and median age: 11 years old. Of these 2311, only 160 i.e. (6.92 %), were of age 6 and eligible to be included in this specific data analysis. Finally, at 3rd follow-up (2010) 2224 newly recruited children were added to these surveys with range age: 6-20 years old and median age: 12 years old. Of these 2224 only 189 i.e. (8.50 %) were of age 6 and eligible to be included in this specific data analysis." Koukounari 2011, Pg. 6. Koukounari 2011 notes that it only includes the 20% of participants in the pilot study who were tracked through all follow-ups: "Finally at 3rd follow-up (2010) there were 713 children successfully followed-up (i.e. follow-up rate=19.71%). Longitudinal analyses for the 4 years are presented in the next pages for these 713 children." Koukounari 2011, Pg. 6.
        • Other schools: Koukounari 2011 claims that 5,700 participants were recruited at baseline: "At baseline (2008) there were recruited 5700 children while the follow-up rate one year later was 53.42 % (3045/5700)." Koukounari 2011, Pg. 16. Styles 2011 Pg. 13, Table 8 claims there were 3,781 participants recruited at baseline. Unlimit Health later told us that this may be because Koukounari 2011 included data from a group of students who were surveyed for a concurrent cross-sectional study, which was not completed: "SCI initially planned to do a cross-sectional evaluation of sixth grade students every year, because each year the current sixth grade class would have received more rounds of treatment over the course of elementary school than the previous year. SCI did not complete this plan, but Dr. Koukounari included the data from the sixth grade students in the baseline data. Dr. Styles did not include this data." GiveWell's non-verbatim summary of a conversation with Giuseppina Ortu on June 20, 2014. Both studies include a similar number of participants at the first follow up: "At baseline (2008) there were recruited 5700 children while the follow-up rate one year later was 53.42 % (3045/5700)." Koukounari 2011, Pg. 16; Styles 2011 Pg. 13, Table 8 shows about 3030 students were included in the first year follow-up analysis.
      • Details on the 2017 survey, evaluating the period 2007/8-2017, can be found in this spreadsheet, "Methods" sheet.

    • 49
      • We have seen reports on a baseline impact survey from 2012 (SCI Liberia panel study baseline report) and one follow-up survey from 2013 (SCI Liberia impact survey follow up recommendations report 2013) and (SCI Liberia impact survey dashboard 2012-13).
      • Implementation of the follow-up survey as described in SCI Liberia impact survey follow up recommendations report 2013:
        • "The survey aimed to revisit all of the schools that were visited in 2012 for the baseline survey to determine changes in prevalence and intensity over time. The schools were randomly selected at baseline and the sampled children within each school were also randomly selected at baseline, with stratification by gender and grade." Pg 3.
        • "There were some difficulties in matching pupils to schools, and also matching schools between years. This was because the school identity numbers became muddled making matching not possible. This was fixed by going back to the paper records to determine the data associated with each school." Pg 4.
        • "Eleven of the baseline schools (with code 1, 7, 9, 16, 17, 18, 20, 23, 24, 25, 28) could not be re-visited in 2013 and were substituted by schools as close as possible. Three of these new schools (code 39, 41, 48) were deemed to be too far away from the corresponding baseline schools to serve as their follow up and were added to the analysis as independent new schools." Pg 4.
        • "The original protocol was originally designed as a ‘cohort’ study where the same children are surveyed repeatedly across a number of years. Issues with identification numbers meant that children could not be matched between the years and consequently we analysed the data as a cross-sectional study. Any future surveys of these schools in Liberia will be cross-sectional and SCI has moved away from cohort studies across all its programmes. In addition, as time that has passed since the original baseline study in 2012 many of the original children surveyed will no longer be at school." Pg 4.

    • 50
      • Details on the methods for the surveys in Côte d'Ivoire, Tanzania, Burundi, Madagascar, Ethiopia, Malawi, and the Democratic Republic of the Congo can be found in the "Methods" sheet of this spreadsheet.
      • Note that we have excluded studies from Uganda (Kabatereine et al. 2007) and Burkina Faso (Touré et al. 2008 and Koukounari et al. 2007) because participants in these studies received separate, more intensive treatment than other children in those countries (discussed in blog posts in 2013 and 2014). Therefore, we believe that the results from Uganda and Burkina Faso do not reflect the quality of the national programs which were supported by Unlimit Health.
      • It is our understanding that, in the studies that we have included, study participants received treatment in the same manner as other children in the country, and thus that those studies reflect the performance of the national MDAs:
        GW: "In our current review, when discussing prevalence and intensity sentinel site surveys, we write, 'It is our understanding that, in the Niger, Burundi, and Malawi studies, study participants received treatment in the same manner as other children in the country, and thus that those studies reflect the performance of the national MDAs.' Would it be accurate to include the Liberia study in the statement above?
        Fiona Fleming: "Yes it would, participants in all these surveys receive treatment as part of the national programme and not at the time of the survey. Participants only received treatment in surveys in Uganda in the first 2 years back in 2003 and no surveys have treated participants since that time."
        Fiona Fleming, email to GiveWell, October 11, 2016. This is also noted in several specific studies; see the "Methods" sheet of this spreadsheet.

    • 51
      • Niger: “Praziquantel (using dose-pole corresponding to 40 mg/kg) and Albendazole (400 mg) were given to the target population regardless of infection status, during the mass drug administration campaign that took place 3–4 weeks after the surveys were conducted.” Tohon et al. 2008, Pg. 3. “A total of 89% of the initial sample group were re-examined one year after baseline data collection and the first round of treatment with praziquantel and albendazole.” Tohon et al. 2008, Pg. 4.
      • Burundi pilot: “At baseline (2007) there were recruited 3616 children. At 2008 the 1st follow-up took place where 1188 children were retraced since baseline (i.e. follow-up rate=32.85 %). At 2nd follow-up (2009) there were 1004 children successfully followed up since baseline (i.e. follow-up rate=27.77%). Finally at 3rd follow-up (2010) there were 713 children successfully followed-up (i.e. follow-up rate=19.71%). Longitudinal analyses for the 4 years are presented in the next pages for these 713 children.” Koukounari 2011, Pg. 6.
      • Burundi other schools: “At baseline (2008) there were recruited 5700 children while the follow-up rate one year later was 53.42% (3045/5700).” Koukounari 2011, Pg. 15.
      • Malawi: “A longitudinal survey design requires baseline data collection from schools prior to the initiation of large-scale distribution of praziquantel and albendazole or mebendazole through the school-based platform. Follow up surveys will be conducted immediately prior to subsequent rounds of treatment for the life of the programme to monitor the impact of the health intervention.” SCI Malawi panel study, Pg. 3. “During the baseline survey, cohorts of 125 children from Standards 1, 2 and 3 (aged approximately 6, 7 and 8 years) were randomly selected in each of the schools and enrolled into the study. [...] This group of selected children, now in standards 2, 3, and 4, as well as a new group of 40 Standard 1 children, were re-tested to measure the same indicators during the 1st follow-up.” SCI Malawi panel study, Pg. 4.
      • Liberia: "The original protocol was originally designed as a ‘cohort’ study where the same children are surveyed repeatedly across a number of years. Issues with identification numbers meant that children could not be matched between the years and consequently we analysed the data as a cross-sectional study. Any future surveys of these schools in Liberia will be cross-sectional and SCI has moved away from cohort studies across all its programmes." SCI Liberia impact survey follow up recommendations report 2013, Pg 4.
      • For all four studies, the methodology does not discuss a control group, and with context, it is sufficiently clear that there was not one.

    • 52
      • "This survey year saw a switch from a longitudinal survey design to a cross-sectional design which occurred following internal SCI reviews of the data and issues arriving from field surveys. The change in survey design led to the ages of children included in the study to be altered slightly to allow for like for like comparison over time and to capture those with the highest burden of infection." SCI Malawi impact study – second follow up, Pg. 3.
      • "The original protocol was originally designed as a ‘cohort’ study where the same children are surveyed repeatedly across a number of years. Issues with identification numbers meant that children could not be matched between the years and consequently we analysed the data as a cross-sectional study. Any future surveys of these schools in Liberia will be cross-sectional and SCI has moved away from cohort studies across all its programmes." SCI Liberia impact survey follow up recommendations report 2013, Pg 4.

    • 53

      See columns K and O in the "Methods" sheet of this spreadsheet.

    • 54

      Unlimit Health notes that because children in a control group would be tested for infection, ethical guidelines would require that those found to be infected receive treatment and thus would no longer serve as controls.

    • 55
      • See the "Results" sheet in this spreadsheet.
      • Unlimit Health reports that the increase in prevalence of S. mansoni in Côte d'Ivoire and Tanzania and the increase in prevalence of S. haematobium in Côte d'Ivoire is not statistically significant, though we have not verified these statistical tests. The statistical significance of the increase in intensity of S. mansoni in Tanzania was not reported.

    • 56

      "For S. mansoni both prevalence and intensity increased between FU2 and FU3. All differences are significant." SCI Burundi impact survey recommendations report 2017, Pg 6. See the notes in the cells for the Burundi 2017 study in the “Results” sheet of this spreadsheet.

    • 57

      See the "Results" sheet in this spreadsheet.

    • 58

      See the "Results" sheet in this spreadsheet.

    • 59
      • See SCI Côte d'Ivoire impact survey recommendations report 2016 (Phase 1 schools follow-up and Phase 2 schools baseline).
      • "There was an insignificant increase in overall prevalence of S. mansoni infections. However, there were big differences between schools. The biggest increase in S. mansoni prevalence was registered for schools 2 (+8.9%), 9 (+9.3%), and 20 (+11.5%)", p.2.
      • "The overall prevalence in S. haematobium increased insignificantly although there were big differences between schools. The biggest increase was registered at the schools 3 (+13.1%), and 11 (+11.3%)", p.2.

    • 60

      "The prevalence of S. haematobium infections and heavy intensity infection decreased overall. However, there is significant heterogeneity between schools. In many schools prevalence is still high following two rounds of MDA", SCI Madagascar impact survey recommendations report 2017, p.3.

    • 61

      "Increase in prevalence of S. mansoni (42 out of 146 schools) and heavy intensity infection (5 out of 146 schools) despite overall reduction in both", SCI Ethiopia impact survey recommendations report 2016-17, p.3.

    • 62
      • Niger:
        • "Eight villages located in schistosomiasis endemic regions were randomly selected to represent the two main transmission patterns in Niger: six villages located near permanent (Tabalak, Kokorou) or semi-permanent (Kaou, Mozague, Rouafi, and Sabon Birni) ponds and two (Saga Fondo, Sanguile) located along the Niger River. The villages represented the south-western region (Tillabéry) and the central-northern region (Tahoua) of the country, with four villages from each region. One village is located in the Sudanian climatic zone and the seven others are in the Sahelian climatic zone." Tohon et al. 2008, Pg. 2.
        • Unlimit Health told us that these locations "are not representative of the treatment population as a whole. They were selected to indicate the impact of treatment in schools with varying prevalence and intensity of both [types of schistosomiasis]." Anna Phillips, SCI Country Program Manager for Burkina Faso and Niger, email to GiveWell, October 13, 2011.
      • Burundi pilot survey: “More precisely, the 12 schools were chosen based on 3 zones-believed at the time that they would have the majority of NTDs. 4 schools were selected randomly so that they represent the ‘STHs +Schisto +oncho’ zone (these were Musenyi, Nyamibu, Munyika, Rukinga); then another 4 schools were selected randomly so that they represent the ‘STHs +oncho’ zone (Mirombero, Kizuga, Ruzibira, Mudende) and finally 4 schools were selected randomly so that they represent STHs only endemic areas (Gatwe, Ruko, Condi, Gitobo). Such decisions were based on available historic data. Thus, SCI Programme Manager advised not to stratify the statistical analysis by province and so such results (i.e. stratification by province) are not presented anywhere in this report.” Koukounari 2011 Pg. 6.
      • Burundi other schools: it appears that schools were selected to be representative, though this is not fully clear in the reports we have seen.
        • Styles 2011 says, "The selection of schools was done randomly from the non-pilot provinces; taking into account 11 separate ecological zones." Pg. 13.
        • Koukounari 2011 says, “At baseline (2008) there were recruited 5700 children while the follow-up rate one year later was 53.42% (3045/5700). For these set of studies as they were designed to cover almost all of the country, it is worthwhile to also examine stratifications of analyses by district and such results are also presented in the following subsections. However, in most of the districts the children were coming only from 1 school (see relevant graphs for district whenever n<200; when this is the case then this is only 1 school per district and thus results should be treated there with caution and programmatic decisions to be taken with reservations). Whenever/wherever this is the case, results should be interpreted with caution as just 1 school would be quite ‘risky’ to represent inference/decisions for a whole district.” Pg. 16.
      • Malawi: The schools seem to have been selected in a way that makes them representative of districts with moderate to high prevalence, which are those districts that receive annual treatment (low prevalence districts receive more limited treatment). The schools were selected only from districts found to have moderate schistosomiasis prevalence in Unlimit Health's mapping: "Method of sentinel site selection: SCI’s protocol is to monitor only in those districts where prevalence of schistosomiasis is moderate or high i.e. SCI does not monitor in non-endemic or low prevalence districts where a full control program is not implemented. All districts except Mzuzu City surveyed in the mapping in February 2012 were determined to have moderate prevalence of S. haematobium, and consequently all districts except Mzuzu City were included in the selection of sites to be monitored for this species (see Table 1). S. mansoni infection was more focal and only present at moderate prevalence in Chiradzulu, Blantrye Rural, Lilongwe City and Lilongwe Rural East. Following district stratification by S. mansoni infection, such that the number of schools selected for S. mansoni monitoring, reflected the frequency of moderate risk areas in the monitoring areas, 22 schools were selected that would be monitor S. haematobium infection with a subset of 9 schools which also monitor S. mansoni infection. Due to the low prevalence of STH’s, STH infection was only monitored in those schools where the Kato-Katz slides were already prepared for S. mansoni." SCI Malawi panel study, Pg. 4; Unlimit Health's senior biostatistician told us that the sampling method would produce a sample representative of the treated districts. Michelle Clements, SCI Senior Biostatistician, conversation with GiveWell, October 15, 2014.
      • Details on the sampling procedure for all other studies are given in the "Methods" sheet in this spreadsheet.

    • 63
      • "Baseline Impact M&E surveys took place in 38 schools in three provinces of Liberia (Bong, Nimba, Lofa) in Nov-Dec 2012, as part of the ICOSA programme." SCI Liberia panel study baseline report, Pg 3.
      • SCI Liberia impact survey follow up recommendations report 2013:
        • "Eleven of the baseline schools (with code 1, 7, 9, 16, 17, 18, 20, 23, 24, 25, 28) could not be re-visited in 2013 and were substituted by schools as close as possible. Three of these new schools (code 39, 41, 48) were deemed to be too far away from the corresponding baseline schools to serve as their follow up and were added to the analysis as independent new schools." Pg 4.
        • Survey recommendations table, Pg 6:
          • "School identity numbers were not always correct."
          • "Many of the GPS coordinates were not recorded correctly."
          • "Schools closed between years of data collection."

    • 64

      See the "Methods" sheet, cells Q10 and R10, in this spreadsheet.

    • 65
      • Niger: "A total of 89% of the initial sample group were re-examined one year after baseline data collection and the first round of treatment with praziquantel and albendazole." Tohon et al. 2008, Pg. 4.
      • Burundi: Pilot schools: 33%: “Without taking into consideration the parasitological exams, at baseline (2007) there were recruited 3616 children. At 2008 the 1st follow-up took place where 1188 children were retraced since baseline (i.e. follow-up rate=32.85 %).” Koukounari 2011, Pg. 16. 50%: Styles 2011 Pg. 1, Table 1. Other schools: 53%: “At baseline (2008) there were recruited 5700 children while the follow-up rate one year later was 53.42 % (3045/5700).” Koukounari 2011, Pg. 15. 80%: Styles 2011 Pg. 13, Table 8. The discrepancy between the populations included in Koukounari 2011 and Styles 2011 is described in more detail below.
      • Malawi: "Overall, the drop-out rate was higher than expected for both species of schistosomiasis. 48% of those pupils monitored for S. haematobium dropped out the study between baseline and follow-up, and 64% of those pupils monitored for S. mansoni dropped out of the study." SCI Malawi panel study, Pg. 17. Since the most meaningful results from Malawi were for Schistosoma haemotobium, we focus on the follow-up rate for that species.
      • Liberia: Although the Liberia study was originally designed as a cohort study, the survey implementers switched to a cross-sectional methodology at the first follow-up: "The original protocol was originally designed as a ‘cohort’ study where the same children are surveyed repeatedly across a number of years. Issues with identification numbers meant that children could not be matched between the years and consequently we analysed the data as a cross-sectional study." SCI Liberia impact survey follow up recommendations report 2013.

    • 66

      Benjamin Styles, SCI Senior Biostatistician, phone conversation with GiveWell, August 12, 2011.

    • 67

      "Prevalence with CCA is always higher (or equal if 0) than with KK. Prevalence using CCA with trace positive is between 3 and nearly 50 times larger than with KK." SCI Burundi impact survey recommendations report 2017, p.7.

    • 68

      For further details, see the "Methods" sheet, cell S12 in this spreadsheet.

    • 69
      • “This survey included baseline results for 18 schools in Tshopo, Ubangi Nord and Uele co-ordinations (Phase III) and follow-up year 2 (FU2) results for 1 schools in Kongo Central and 2 schools in Kasai Mbujimayi (Phase I).”
      • “The number of schools followed-up Phase 1 FU2 n=3 is too small to determine whether the results are representative, or of statistical significance, in evaluating programme impact and informing programmatic actions.”

      SCI DRC impact survey recommendations report, 2017-2018, Pg. 2.

    • 70

      We have not seen treatment numbers from 2018-2020, so we are using spending to estimate the proportion of monitoring we have seen. See this spreadsheet, sheet "SCI monitoring," row "Amount of spending with coverage or impact survey, 2018-2020."

    • 71

      See this spreadsheet, sheet "SCI monitoring," row "Number of treatments in country-years with coverage survey, 2014-2017."

    • 72

      See here.

      Unlimit Health told us that it is unable to visit Sudan because of safety concerns and political unrest, and the country staff requires in-country support from Unlimit Health in order to implement a coverage survey for the first time. Conversation with Unlimit Health staff, April 3, 2019.

    • 73

      Wendy Harrison and Sarah Knowles, SCI Managing Director and Biostatistician, conversations with GiveWell, April 9 and 14, 2014.

      For example, for Mozambique, Unlimit Health shared a report from a consultant who visited the country in May 2015 to assist with data cleaning and analysis for prevalence data from 2012, 2013, and 2014. The report notes major problems with this data and the refusal of the government to allow Unlimit Health and other international partners to have access to the data outside of Mozambique.LSTM Mozambique trip report (May 2015)

      • "With authorization of Dra Olga, I am sharing with you the EXCEL file with results (not the databases), and she reminds me that the data belongs to MISAU." Pg. 2.
      • "The Team reviewed the results for 2012-2013-2014 and they confirmed that only 2 sentinel sites were strictly comparable for 2013-2014 (Mecula and Mandiba) and none of the Sentinel Sites for 2015 are going to be comparable to 2014. They decided to review the location of the new sentinel sites, to be able to have a better and stronger comparisons." Pg. 2.

    • 74

      See this spreadsheet, sheet "SCI monitoring," column G "Coverage survey?"

    • 75

      "In Malawi, urine volumes were not accurately recorded thus it is possible that data is indicating lower overall prevalence in sentinel sites. ICOSA will be undertaking further data analysis to quantify underestimates using mapping data from 2012 and baseline data in appropriate districts." SCI report to DFID (October 2013), Pg. 15.

    • 76

      "The NTD programme in Zanzibar has recently completed the 3rd round of MDA." SCI report to DFID (October 2013), Pg. 17.

    • 77
      • Our intervention report discusses this briefly.
      • Other conversations and observations have reinforced our impression that administering deworming drugs is fairly straightforward.
      • The WHO factsheet on STH: "The WHO recommended medicines – albendazole (400 mg) and mebendazole (500 mg) – are effective, inexpensive and easy to administer by non-medical personnel (e.g. teachers)." WHO STH factsheet.

    • 78
      • "In Tanzania matters came to a head in places around Morogoro in 2008. Distribution in schools of tablets for schistosomiasis and soil-transmitted helminths provoked riots, which had to be contained by armed police. It became a significant national incident, and one of the consequences has been the delay in Tanzania adopting a fully integrated NTD programme, and the scaling back some existing drug distributions." Allen and Parker 2011, Pg. 109.
      • "From these reports a number of problems with the MDA were raised which included fear of side effects from the tablets, particularly following the mass hysteria and death in Blantyre and Rumphi respectively and may explain some of the geographic heterogeneity seen. Furthermore most districts reported that MDA occurred after standard 8 students had finished exams and left school, and due to having inadequate resources for drug distribution...The side-effects incident in Blantyre and death in Rumphi had a large effect on districts and with many district reports stating that after the incidence many families refused to participate." SCI Malawi coverage survey 2012 Pgs 5, 21.

    • 79

      Fiona Fleming, conversation with GiveWell, November 5, 2015.

    • 80

      "There was confusion on Wednesday in some public primary and secondary schools in Ogun State, over the administration of anti-worm tablets. Nigerian Tribune gathered that some students reportedly collapsed in the cause of administering the tablets on them. This resulted into rumour that spread like wildfire across the length and breadth of the state, as parents stormed various school to withdraw their wards. When the Nigerian Tribune visited Egba High School, Asero and Asero High School both in Abeokuta South Local Government Area of the state, some parents were sighted at the school gate, who had come to confirm the incident and probably withdraw their wards. There was calmness in both schools as students in the Senior Secondary Classes were said to be preparing for their examinations. Meanwhile, the Ogun State Government through the State Commissioner for Health, Dr Babatunde Ipaye, has denied any case as a result of the anti-worm drug. Ipaye in a statement made available to the Nigerian Tribune in Abeokuta, said that no pupil or student to the best of his knowledge had reacted to the drug in the state. He explained that the exercise was done by his Ministry in collaboration with Evidence Action." Nigerian Tribune, "Panic in Ogun schools over deworm exercise," December 2017

    • 81
      • In October 2018, Unlimit Health described its calculations:
      • "The financial reports on in-country expenditure and reported treatment data allow SCI to perform cost analyses on the cost per treatment by year and by country. The £0.18 is derived from average of the country cost per treatment for a school-age child over several financial years. The £0.27 for community-based treatments is based on the same overall dataset, but with the cost data disaggregated by community-based delivery.

        "The £0.27 per child in the first year of a new country program is also derived from cost analyses on the same financial data when looking at the first year of MDA across multiple countries. The cost is higher due to baseline prevalence mapping in the country to determine treatment strategy, as well as the purchase of capital goods and because countries tend to be less efficient in the first year of implementation. Subsequently efficiency increases." Wendy Harrison, email to GiveWell, October 30, 2018

    • 82
      • For cost per school-aged child see, for example, cell D12 on sheet "Ideal Scenario 2018-2021."
      • For cost in community-based distributions, see, for example, cell D10 on sheet "Ideal Scenario 2018-2021."
      • For cost to purchase drugs for adults, see, for example, cell E10 on sheet "Ideal Scenario 2018-2021" showing that 3 tablets are needed per adult and cell F10 showing a cost per tablet of £0.08.
      • For cost the first year of a new country program, see, for example, cell D33 on sheet "Expansion_Scenario 2018-2021."

      SCI GiveWell Funding Scenarios 2018-2021.

    • 83

      This understanding is from undocumented conversations with Unlimit Health from early in our investigation of Unlimit Health.

    • 84
      • "Ordering of ALB is carried out by the Ministry Of Health Programme Manager for Lymphatic Filariasis (LF) Elimination. Previously the ALB had been used for the LF program and stocks had not been replenished in time for the SCH campaign which contributed to the low ALB coverage. Treatments were also carried out in conjunction with the child health days which may have caused confusion with who was eligible for treatment and prioritising the younger children. Furthermore due to the complexities of distributing PZQ compared to ALB there may have been more focus during the training and distribution on reporting and dispensing PZQ." SCI Malawi coverage survey 2012, Pgs 20-21.
      • "Miscommunication between the national Program Managers for lymphatic filariasis and schistosomiasis led many districts to believe that the ALB which they received should have been made available for the MDA but was not used." SCI Malawi coverage survey 2014, Pg. 37.
      • "Unlike in Sub-Saharan Africa, where SCI distributes praziquantel for schistosomiasis and albendazole for STH in equal amounts, Sudan has many areas where schistosomiasis is a problem but STH are not, and SCI distributes more praziquantel than albendazole. However, a small number of areas including Kassala and eastern Sennar have STH and no schistosomiasis. In these areas, the Ministry distributes more albendazole than praziquantel." GiveWell's non-verbatim summary of a conversation with Alan Fenwick and Najwa Al Abdallah, September 14, 2015.
      • "Albendazole (ALB) was also distributed in some of the 30 districts under different partners, however ICOSA programme did not support its procurement, distribution or data collection in 2013/2014." SCI Uganda coverage survey 2014, Pg. 4.
      • Schistosomiasis Control Initiative, conversation with GiveWell, September 6, 2016:
        • In Ethiopia, the Unlimit Health-supported program has delivered STH-only treatments in some districts where schistosomiasis prevalence was low.
        • Unlimit Health told us that its reported treatment numbers were total numbers of schistosomiasis treatments.

    • 85

      We explain why we take this approach in this blog post.

    • 86

      Alan Fenwick, SCI Director, email to GiveWell, November 24, 2014.

    • 87

      See this spreadsheet for details. Note that this is not an ideal comparison because reported coverage rates are calculated from reported treatment numbers and an assumed target population. We would like to adjust the reported treatment numbers, rather than the reported coverage rate, but don't have the data to do so. It is possible that the difference in the reported coverage rate and the coverage rate from the coverage surveys is due to errors in the numbers used for the target population rather than the reported treatment numbers.

    • 88

      We used Leslie et al. 2011, a study of the costs of a Unlimit Health-funded deworming program in four districts of Niger in 2004-2006, to estimate non-Unlimit Health contributions to Unlimit Health's deworming programs. Three of the authors of the study were affiliated with Unlimit Health. The study aimed to account for all costs of the program, including costs funded by the government and non-financial costs such as the value of volunteers' time:

      “This was a retrospective study which covered a two year period from April 2004 to May 2006, including the first and second years of MDA and related programme activities in four health districts. All data on first year costs at national, regional, district, and sub district levels were taken from the PNLBG accounts and receipts and records of staff missions or activities. Second year cost data for national and regional level activities were taken from receipts. District and sub district, school and community MDA resource use data for 2005 were collected in June 2006 through a retrospective survey…
      The main cost elements include: the programme specific expenditure; the opportunity cost or value of government contributions related to in-kind costs of using local government staff and vehicles and the value of CDD’s time (taken as the daily agricultural labour rate); and the international costs of programme co-ordination, reporting and technical support." Leslie et al. 2011, Pgs. 2-3.

      The study is of a single country, looked at a program that was carried out years ago, and the program may differ in some ways from current programs, but overall it is of high quality and provides us with a sense for the portion of resources contributed by Unlimit Health versus non-Unlimit Health parties.

      Two examples of how the area where the study was conducted may not be representative of all areas in which Unlimit Health works:

      • Due to low school enrollment rates, a substantial portion of the program was through community distribution. Current Unlimit Health programs focus on school-based distribution. "The primary school net enrolment rate (NER) in 2004 in Niger was 41%... To achieve high treatment coverage in targeted school age children and at risk adults two treatment strategies, school-based and community-based distribution, were established." Leslie et al. 2011, Pg. 2.
      • "The cost per treatment and prevalence figures relate to the study sample of four districts located in the Niger River Valley. This was and is an area of high disease prevalence and high population density relative to other parts of the country. The costs per person treated may be higher in lower density and more remote areas." Leslie et al. 2011, Pg. 8.

      Non-Unlimit Health costs were 18% of the total cost of the program and 33% of the cost of school-based deworming (the program also included community-based deworming).

      • ”Programme cost: 75%
      • Government cost: 18%
      • International tech. support: 7%”

      Leslie et al. 2011, Pg. 5, Table 2.

      It is our understanding from the paper and our past conversations with Unlimit Health that "programme expenditure" was fully funded by Unlimit Health. We believe that "international tech. support" refers to Unlimit Health staff time and travel costs; we're somewhat less confident in this than in our understanding of “programme expenditure.” Government costs are "related to in-kind costs of using local government staff and vehicles and the value of CDD’s time (taken as the daily agricultural labour rate)." Leslie et al. 2011, Pg. 3.

      Calculating non-Unlimit Health costs of school-based delivery:

      • The average cost/treatment in the study was $0.58: “The total economic cost per treatment was $0.58. This includes programme, government and international costs.” Leslie et al. 2011, Author Summary. At 7% of the total cost, international tech. support accounts for $0.04/treatment.
      • “The full economic delivery cost of school based treatment in 2005/06 was $0.76, and community treatment was $0.46. If only programme costs are included these figures are $0.47 and $0.41 respectively.” Leslie et al. 2011, Pgs 7-8.
      • Therefore, non-program costs (government and international tech. support) are $0.29 ($0.76 - $0.47) of the $0.76 cost of each school based treatment. Since $0.04 is international tech. support, that leaves $0.25 of government costs, or 33% of the total cost.

      It is our understanding that in recent programs Unlimit Health has continued to do some community-based deworming but that most of its treatments are delivered through schools. Therefore, we conservatively estimate that non-Unlimit Health actors contribute 30% of the cost of a Unlimit Health deworming program.

    • 89

      For example, Unlimit Health provides technical assistance to the program in Yemen and does not pay other costs. It reported the full number of treatments delivered in that program.

      Najwa Al-Abdallah and Lynsey Blair, conversation with GiveWell, August 11, 2017.

    • 90

      "Imperial College pays the majority of SCI’s overhead costs, including rent and utilities, and offers free services such as legal assistance. Imperial’s legal department prepares all of SCI’s contracts. The college also covers some risks and liabilities that SCI may face. In return, SCI pays Imperial 6% of its funding from DFID, but this does not cover the full cost of the services that SCI receives. The 6% of the DFID funding that is paid to the college is not included in SCI’s budget, but is included in the total size of the DFID grant. SCI may perform an analysis of the costs covered by Imperial College, but this is not currently a high priority." GiveWell's non-verbatim summary of a conversation with Wendy Harrison and Najwa Al Abdallah on September 8, 2015, Pg. 4.

    • 91

    • 92
      • For example, see sheets "SCI 2018-19 treatment numbers" and "SCI 2017-18 Final Treatment No" in our 2020 analysis of SCI's cost per person dewormed per year.
      • "I note that you mention in your request below the sharing of deworming figures only for the SCI assigned health zones. In 2018/19, SCI funding was allocated on the basis of the total number of SCH treatments targeted nationally. If only those deworming treatments in the SCI assigned health zones are considered against that funding, it will influence dramatically the cost of those treatments given that the number will be significantly lower. The NTD Programme have used the assignation of health zones by partner to streamline the overall management of the programme for administrative purposes, thus if SCI funding was being considered against treatment delivery, we would recommend that either of the options above are used, rather than the deworming treatments in SCI assigned health zones only." Lynsey Blair, Head of Programmes, SCI Foundation, email to GiveWell, July 6, 2020 (unpublished)

    • 93
      • See our 2020 cost per child analysis for Sightsavers (sheet "Sightsavers' costs by country and year," "DRC" section).
      • In our 2018 cost per child dewormed analysis for Unlimit Health, we used data on the total costs and treatments across all actors in DRC from 2016-17 to calculate a cost per child dewormed estimate for Unlimit Health in DRC (sheet "Countries with pooled funding"). We changed our analysis method for 2020 because we found that applying the methodology from the 2018 cost per child dewormed analysis to Unlimit Health's 2018-19 spending led to implausibly high numbers of deworming treatments attributed to Unlimit Health in DRC.

    • 94

      SCI Foundation, 2018-19 DRC NTD treatments

    • 95

      See GiveWell analysis of SCI's cost per person dewormed per year [2020], "GW 2018-19 costs and treatments," "GW 2017-18 costs and treatments," "GW 2016-17 costs and treatments," and "GW 2015-16 costs and treatments" sheets for a list of the countries excluded from our analysis in each year and their reasons for exclusion.

    • 96

      For a discussion of why we consider funding a charity's work up to three years in the future, see this blog post.

    • 97

      Some of our top charities have a policy of holding funding reserves. In our room for more funding analyses, we typically include reserved funding as funding available to support program activities. We do this both to ensure consistency across top charities (as not all top charities hold reserves) and to understand the true effect of granting additional funding (i.e. whether additional funding would support undertaking additional program activities versus building or maintaining reserves).

    • 98

      GiveWell maintains a list of all charities that meet our criteria, along with a recommendation to donate to our Top Charities Fund. Some donors give based on our top charity list but do not follow our donation recommendation. In our projections of future funding, we typically count only one year of funding that an organization receives as a result of being on our list of top charities in order to retain the flexibility to change our recommendations in future years.

    • 99

      We update a top charity's room for more funding analysis more frequently if we grant funding to it more frequently.

    • 100

      We update a top charity's room for more funding analysis more frequently if we have reason to believe that its funding and budgets have changed substantially.

    • 101

      For a list of grants we have made from our Top Charities Fund, see this page, section "Past recipients of the Top Charities Fund."

    • 102

      Open Philanthropy, a philanthropic organization with which we work closely, is the largest single funder of our top charities. The vast majority of Open Philanthropy's current giving comes from Good Ventures. We make recommendations to Open Philanthropy each year for how much funding to provide to our top charities and how to allocate that funding among them. An example of these recommendations from November 2020 can be found on this page.

    • 103

      This includes donations made to charities on our checkout page, donations made directly to the organizations, and donations through other third-party organizations that share GiveWell’s recommendations (e.g., One for the World).