Last Mile Health

Our Recommendation

Last Mile Health helped the Liberian government design and launch a national community health worker program, then measured what it did. The intermediate results are strong and comparison-based. In Grand Bassa County, care of childhood illness by a qualified provider rose 60.3 percentage points more in program areas than in areas without the program, while comparison areas stayed flat. Care from a community health worker went from zero to 81.6 percent. In Konobo district, access to qualified care rose 60.1 points for diarrhea, 51.2 for acute respiratory infection, and 30.6 for fever.

Then the organization pre-registered a protocol, published it in advance, surveyed 22,315 mothers about 32,743 children in a full census of remote communities, and used the staggered district-by-district rollout as a stepped-wedge design to test whether the program reduced under-five mortality. The study was powered on the implementers’ own stated expectation of a 25 percent reduction. It detected no reduction at all, alongside large improvements in treatment coverage and a shift in care seeking from informal providers to trained ones.

A donor should hold two things at once. Last Mile Health reaches 52 million people, supports more than 30,000 health workers, and won the 2026 Hilton Humanitarian Prize. It also went looking for proof that children were surviving, did not find it, and published that. We would add one caution: the organization’s own homepage reports 10,140 estimated lives saved of children under five, derived from a model, on the same site as the evaluation that found no measured mortality reduction. Those two numbers should not sit that far apart without explanation.

Last Mile Health's Fierce Certification score is 100/100 points based on our criteria:
✔ Has Ultimate Outcome Goals (20 pts)
✔ Measures Intermediate Outcomes (15 pts)
✔ Measures Ultimate Outcomes (20 pts)
✔ Measures Intermediate and/or Ultimate Counterfactual (20 pts)
✔ Shows Continual Learning & Adaptation (25 pts)

The Social Problem

Last Mile Health states the problem in a sentence on its homepage: “Illness is universal, healthcare is not.”

In rural Liberia and comparable settings, the determining factor in whether a sick child receives treatment is distance. The national program defines its target population as everyone living five kilometers or farther from a health facility. Those communities sit hours from the nearest clinic over terrain that becomes impassable in the rainy season. When a child develops malaria, diarrhea, or pneumonia, the three conditions estimated to cause about half of under-five deaths in Liberia, the family either makes a long journey or does not. Many do not, and the child is instead treated by whoever is nearby, qualified or not.

The conventional answer, community health volunteers, has its own failure mode. Unpaid volunteers with minimal training, no supervision, and no reliable drug supply drop out when their families need feeding. Programs that look identical on paper deliver nothing. Last Mile Health’s own baseline mapping of Grand Bassa County identified 1,733 remote communities containing 23,702 households.

The deeper problem is that no government had a financed national model. Community health remained a landscape of pilots.

The Solution

Last Mile Health does not run its own parallel health system. It supports governments to build and finance national community health worker programs, which is the choice that defines everything else about the organization.

The model rests on a specific design standard, developed in its Konobo pilot in Grand Gedeh County in 2012 and carried into Liberia’s national program. The World Health Organization noted that several features informed the national design: contracts and cash payments rather than volunteering, a worker-to-population ratio of 1:350, explicit targeting of remote communities, and field-based supervision.

Community health assistants are recruited from the communities they serve, receive a monthly stipend of $70, and are supervised by Community Health Services Supervisors who are mostly nurses, physician assistants, or midwives. The design calls for two supervision visits per assistant per month. At the time of the Grand Bassa evaluation there were 100 assistants and 10 supervisors in the county.

The work now runs along three lines:

Strengthen health systems. Policy, financing, and supervision architecture with Ministries of Health, plus Africa Frontline First, co-founded with the Community Health Impact Coalition and the Financing Alliance for Health.

Train health workers. Curriculum, digital tools, and knowledge assessment, including clinical vignettes built into routine supervision.

Deliver healthcare. Integrated community case management of childhood illness, malaria testing and treatment, vaccination outreach, and perinatal care support.

Last Mile Health co-designed Liberia’s National Community Health Assistant Program, launched in 2016, which extends primary healthcare to 1.2 million people living at Liberia’s last mile. That program is the source of nearly all the evidence in this review, and it is one country within a four-country footprint.

Key Outputs

• Reach: 52 million people across Ethiopia, Liberia, Malawi, and Sierra Leone, reached by community health workers supported by Last Mile Health.
• Workforce: More than 30,000 community and frontline health workers supported, including 4,118 community health workers in Liberia.
• Treatment volume: 650,439 total treatments for children under five in Liberia since July 2023, including 354,097 malaria treatments.
• Malaria vaccination: 90,805 malaria vaccine doses provided during outreach in Liberia since the national campaign launched in 2024, with community health supervisors administering and 4,118 community health workers conducting education, demand generation, and tracking of children through the full four-dose schedule.
• Supervision quality: 82% of Liberian community health workers received at least two supervision visits in the previous four weeks, which is the program’s design standard; four countries have strengthened supervision systems with Last Mile Health support.
• Facility births: 87% of all births in managed counties in Liberia took place in health facilities.
• Countries: Liberia, Ethiopia, Malawi, Sierra Leone, plus global policy work.
• Recognition: 2026 Conrad N. Hilton Humanitarian Prize.

Key Intermediate Outcomes

Childhood illness treatment coverage (counterfactual). The Grand Bassa County evaluation used a difference-in-differences analysis of population-based survey data covering 1,291 illness episodes, comparing areas where the national program had been implemented against areas where it had not (PLOS Global Public Health).

Care of childhood illness by a qualified provider increased 60.3 percentage points more in program areas. Treatment coverage rose dramatically in intervention areas while remaining approximately constant in comparison areas. Care from a community health worker rose from zero to 81.6 percent, and care from unqualified providers dropped. The paper also examined equity by distance from facility, maternal education, and household wealth.

Konobo district (counterfactual). The earlier controlled before-and-after evaluation, published in the Bulletin of the World Health Organization, found the regression-adjusted percentage of children receiving care from formal providers increased by 60.1 percentage points for diarrhea (95% CI 51.6 to 68.7), 51.2 for acute respiratory infection (95% CI 37.9 to 64.5), and 30.6 for fever (95% CI 20.5 to 40.7).

Facility delivery rose while vaccination fell (before and after). The mortality evaluation measured both. Facility-based delivery rose from 41.0% to 57.1% (p < 0.001). Full vaccination among children aged 12 to 23 months fell from 16.2% in 2018 to 3.5% in 2022 (p < 0.001), a decline across every individual vaccine. The authors cannot fully explain the vaccination fall and point to likely COVID-19 disruption. At the time of the study, neither community health assistants nor their supervisors delivered vaccination in communities; they promoted and referred only. Policy has since changed to let supervisors vaccinate during supervision visits.

Vaccine refusals where community health workers educate (no counterfactual). In communities where community health workers provide malaria vaccine education, only 4% report any vaccine refusals. This is reported by Gavi and VillageReach, both citing Last Mile Health, with no comparison communities, so it describes refusal rates where the program operates rather than an effect of the program. We could not locate the figure in Last Mile Health’s own materials.

Health worker competence and supervision (no counterfactual). Clinical vignettes were integrated into routine programmatic supervision to assess community health worker knowledge of integrated community case management, with the method and results published in Global Health: Science and Practice. Separately, 82% of Liberian workers received at least two supervision visits in the previous four weeks, meaning roughly one in five fell short of the program’s twice-monthly design standard. The mortality evaluation names that shortfall among its reasons for the null.

What the authors said about the limits of their own result. In the Grand Bassa paper, the team wrote that they lacked statistical power to directly assess changes in childhood mortality, and that they were executing a subsequent study in 2022 powered to assess child mortality. They then executed it.

Key Ultimate Outcomes

Under-five mortality (counterfactual, null). This is the measurement that earns the perfect score.

Last Mile Health and the Liberian Ministry of Health designed a quasi-experimental stepped-wedge evaluation taking advantage of the fact that the national program rolled out in phases across seven of Grand Bassa County’s eight districts between 2018 and 2022, two districts in September 2018, two more in February 2019, one in April 2021, and two in June 2022. One district was excluded because security threats to program staff prevented both implementation and enumeration.

The design used mixed-effects Cox proportional hazards models with a random intercept for community. The impact assessment survey was a census of all remote communities in seven of the county’s eight districts, reaching 22,315 mothers reporting on 32,743 children under five, with births and child deaths recorded from women aged 15 to 49 and a locally developed landmark event calendar to anchor recall. The study was powered on the program implementers’ own expectation of a 25 percent reduction in under-five mortality. The protocol was registered as NCT05123378 and published in advance in PLOS ONE.

The result: no reduction in under-five mortality. Unadjusted rates were 56.3 deaths per 1,000 live births before the program and 62.0 under it; adjusted, 55.5 and 57.0, with a hazard ratio of 1.03 (95% confidence interval 0.82 to 1.31). Neonatal, post-neonatal, infant, and child mortality all showed the same absence of effect. Models accounting for time-varying effects found a small non-significant trend toward improvement at two and three years of exposure, which the authors report and decline to claim.

Three caveats. The study examines early implementation over a period that included the COVID-19 pandemic. Fidelity of implementation was explicitly not a component of the impact study and is the subject of separate ongoing evaluation. And a null in seven districts of one county over four years does not establish that community health worker programs do not save lives. But it does mean the claim is not yet demonstrated here.

The modeled figure. Last Mile Health’s homepage reports 10,140 estimated lives saved of children under five, footnoted to its under-five lives saved model. Under our standard, modeled impact is not measured impact and is excluded from the table.

What remains unmeasured. Maternal and newborn mortality, prolonged illness and lasting harm, and any measure of community security or wellbeing.

Continual Learning & Adaptation

It said what it could not yet measure, then measured it. The 2022 Grand Bassa paper stated plainly that it lacked power to assess mortality and announced the follow-up study. The follow-up happened.

It pre-registered, then published the null. The mortality evaluation was registered on ClinicalTrials.gov and its protocol appeared in a peer-reviewed journal before results were known, which removes the option of quietly reframing a disappointing finding. The null was then published, and is the subject of a public blog post, “2018-2022 Liberia impact evaluation: Findings and lessons learned.” More unusual still, the study was powered on the implementers’ own stated expectation of a 25 percent reduction, which means the organization published a number it expected to hit and then published the result when it did not.

It diagnosed its own null in print. The evaluation names probable reasons the program did not move mortality: supervision visits below designed levels, neonatal deaths requiring facility care that community health workers cannot provide, nutrition services outside the original design, no community vaccination delivery, and no mechanism to improve facility quality. It then lists the policy revisions responding to each.

It named its own conflict of interest. The limitations section acknowledges that most authors were affiliated with implementing organizations, defends embedded research on its merits, and notes that the concern is mitigated when the result being reported is null.

It published an unflattering gender assessment. After five years of national implementation, only 17% of community health workers were female against 70% of the global health workforce, with the assessment documenting the cultural and selection mechanisms producing that gap.

It built measurement into supervision rather than bolting it on. Clinical vignettes assessing worker knowledge run through routine supportive supervision visits.

It separated impact evaluation from fidelity evaluation deliberately, keeping a parallel program evaluation measuring dose, coverage, and contextual factors including health facility readiness and the pandemic, which is the right structure for interpreting a null.

It supports an ongoing fidelity system rather than a one-off audit. Supervision visit frequency and components are tracked through the Ministry of Health’s Implementation Fidelity Initiative, under which supervisors attached to 25 percent of health facilities are randomly selected for assessment every three months.

It acted on its own diagnosis. Community vaccination delivery, which the evaluation names as one reason the program did not move mortality, was absent during the study period. Policy has since changed to permit supervisors to vaccinate during supervision visits.

Further Reading

• Last Mile Health homepage (lastmilehealth.org), Impact, Research, and Reports and Financials
• “2018-2022 Liberia impact evaluation: Findings and lessons learned”
• Obaje, H., White, E. E., Dennis, M. L., Jockers, D., Bärnighausen, T., Kenny, A., Kraemer, J. D., et al. “Early impact of the Liberian National Community Health Assistant Program on under-five mortality in Grand Bassa County: A non-randomized stepped wedge evaluation.” PLOS Global Public Health, September 28, 2026. DOI 10.1371/journal.pgph.0005681. The stepped-wedge mortality evaluation. Source of the null finding, the hazard ratios, the facility delivery and vaccination figures, the fidelity initiative description, and the authors’ diagnosis of the result. (link)
• Jockers, D., Ngafuan, R., Baernighausen, T., Kessley, A., White, E. E., Kenny, A., Kraemer, J., Geedeh, J., Rozelle, J., Holmes, L., Obaje, H., Wheh, S., Pedersen, J., Siedner, M. J., Mendin, S., Subah, M., and Hirschhorn, L. R. “Under-five mortality before and after implementation of the Liberia National Community Health Assistant (NCHA) program: A study protocol.” PLOS ONE, January 2024. DOI 10.1371/journal.pone.0272172. The pre-registered protocol, published before results were known. Source of the 25 percent expected reduction used to power the study. (link)
• Trial registration NCT05123378 (link)
• White, E., Mendin, S., Kolubah, F. R., Karlay, R., Grant, B., Jacobs, G. P., Subah, M., Siedner, M. J., Kraemer, J. D., and Hirschhorn, L. R. “Impact of the Liberian National Community Health Assistant Program on childhood illness care in Grand Bassa County, Liberia.” PLOS Global Public Health, 2022, 2(6): e0000668. DOI 10.1371/journal.pgph.0000668. Source of the 60.3 percentage point difference-in-differences result, the zero to 81.6 percent community health worker care figure, the $70 monthly stipend, and the twice-monthly supervision design standard.
• Luckow, P. W., Kenny, A., White, E., Ballard, M., Dorr, L., Erlandson, K., et al. “Implementation research on community health workers’ provision of maternal and child health services in rural Liberia.” Bulletin of the World Health Organization, 2017, 95(2): 113-120. DOI 10.2471/BLT.16.175513. The Konobo study. Source of the 60.1, 51.2, and 30.6 percentage point increases.
• White, E. E., Downey, J., Sathananthan, V., et al. “A Community Health Worker Intervention to Increase Childhood Disease Treatment Coverage in Rural Liberia: A Controlled Before-and-After Evaluation.” American Journal of Public Health, 2018, 108(9): 1252-9. DOI 10.2105/AJPH.2018.304555. Rivercess County, with half the county receiving community health workers in August 2015 and the other half serving as comparison until July 2016.
• Downey, J., et al. “Measuring Knowledge of Community Health Workers at the Last Mile in Liberia: Feasibility and Results of Clinical Vignette Assessments.” Global Health: Science and Practice, 2021
• VillageReach and Gavi, “How community health workers are improving vaccine delivery in Africa,” 2026. Source of the 4% vaccine refusal figure
• “Towards a More Gender-Equal Community Health Worker Program: Key Takeaways from the Gender Assessment of Liberia’s National Program”
• Last Mile Health under-five lives saved model
• Africa Frontline First
• Conrad N. Hilton Foundation, 2026 Humanitarian Prize laureate

• Share
• Tweet