> For the complete documentation index, see [llms.txt](https://research.ambitiousimpact.com/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://research.ambitiousimpact.com/bemonc/simulation-based-training-for-emergency-obstetric-and-newborn-care/6-cost-effectiveness-analysis.md).
# 6 Cost-effectiveness analysis
*We modeled the potential cost-effectiveness of a charity delivering a cascade-based LDHF BEmONC training. We estimate that, at scale, this charity could avert 10,000s or 100,000s of DALYs per year, with cost-effectiveness in the range of USD 16–76/DALY for the most cost-effective* [*countries*](#user-content-fn-1)[^1]*.*
[Link to our model](https://docs.google.com/spreadsheets/d/1qzLTQbqSK3VSZki6Y_nDRbJ0ArZuSpKk8ZYE9B19gvQ/edit?gid=1601291114#gid=1601291114)
**This section presents the cost-effectiveness analysis (CEA) for a proposed charity delivering LDHF simulation-based BEmONC training at primary health facilities in low- and middle-income countries.**
We estimate the cost per disability-adjusted life year (DALY) averted at steady-state operations across ten candidate countries, building a bottom-up cost model anchored in published evidence on LDHF implementation in Ghana ([Willcox et al., 2017](https://doi.org/10.1186/s12992-017-0313-x)).
We begin by presenting the headline results before turning to explaining our modeling choices.
## 6.1 Results
**The results of our model suggest high cost-effectiveness relative to our bar, at a high scale reached per year.** Table 7 shows results for all countries modeled[^2].
**Table 7: Cost-effectiveness (CE) at scale by country.**
| Country | Total impact / year (DALYs) | CE (2023 USD/DALY) – including costs of commodities | CE (2023 USD/DALY) – excluding costs of commodities |
| ----------------- | --------------------------- | --------------------------------------------------- | --------------------------------------------------- |
| **Ghana** | 39K | 45 | 33 |
| **Nigeria** | 159K | 37 | 28 |
| **Pakistan** | 234K | 25 | 16 |
| **Ethiopia** | 123K | 45 | 36 |
| **India** | 256K | 99 | 76 |
| **Chad** | 20K | 106 | 99 |
| **Guinea-Bissau** | 3.4K | 164 | 156 |
| **Sierra Leone** | 15K | 59 | 50 |
| **Angola** | 52K | 47 | 36 |
| **Côte d'Ivoire** | 60K | 63 | 54 |
**The variation across countries reflects three drivers:**
* **Baseline neonatal and maternal mortality**, which determine the size of the avertable burden per birth reached;
* **Country-specific personnel and operating costs after purchasing power parity (PPP) adjustment**;
* **The number of births per country that happen in facilities**.
Countries with high baseline mortality and a high share of facility deliveries—Nigeria, Pakistan, Ethiopia—produce the most favorable ratios.
### Third-party models
Direct comparators for LDHF BEmONC training at primary facilities are scarce.
The most relevant peer-reviewed estimate is the program cost analysis underlying our cost inputs ([Willcox et al., 2017](https://doi.org/10.1186/s12992-017-0313-x)), which reported a cost of $1,300 per facility-year in Ghana but did not produce a $/DALY estimate.
GiveWell ([2022](https://docs.google.com/spreadsheets/d/1uUsn9jXnrq2XP0u40mIHo41KCtlo06uo49_apW4L14k/edit?gid=471914976#gid=471914976)) conducted a preliminary cost-effectiveness analysis of facility-based health-worker training in maternal and neonatal care, estimating cost-effectiveness of 12–16x its benchmark (across a range of countries), roughly corresponding to USD 42–56/DALY.
## 6.2 Modeling choices
[Our model](https://docs.google.com/spreadsheets/d/1qzLTQbqSK3VSZki6Y_nDRbJ0ArZuSpKk8ZYE9B19gvQ/edit?gid=929255937#gid=929255937) estimated the cost-effectiveness of two versions of the program:
* **Estimate 1** covers training and mentorship costs only, which follows the Gomez et al. [(2018)](https://doi.org/10.1186/s12884-018-1705-5) study.
* **Estimate 2** adds the cost of procuring and distributing the four medical commodities required for BEmONC signal functions—oxytocin, misoprostol, tranexamic acid, and magnesium sulfate—to facilities that do not currently stock them.
**Each participating facility receives two on-site training sessions, each four days long, separated by one month:**
* Session 1 (“LDHF1”) covers respectful maternity care, infection prevention, normal labor and birth, partograph use, immediate newborn care, and newborn resuscitation.
* During LDHF1, two to three skilled birth attendants (SBAs) per facility are oriented as **in-facility champions** who lead weekly skills practice sessions for the following 12 months. Each facility receives MamaNatalie and NeoNatalie birthing simulators, a NeoNatalie Resuscitator, and a Penguin Newborn Suction device, which remain permanently at the facility.
* Session 2 (“LDHF2”) covers complications: antenatal corticosteroids for preterm birth, severe pre-eclampsia and eclampsia, postpartum hemorrhage (PPH), maternal and newborn sepsis, and kangaroo mother care.
**Master mentors** deliver the training, conduct periodic on-site coaching visits, and provide phone-based mentorship to in-facility champions throughout the year. Master mentors themselves are trained by **in-house trainers** and supported by in-house **supervisors**.
**We model a scenario in which the master mentors are employed by the charity as full-time staff.** This contrasts with Jhpiego’s implementation in Ghana, where master mentors were Ghana Health Service staff drawn from regional health teams (Gomez et al., [2018](https://doi.org/10.1186/s12884-018-1705-5)). The charity-employed model trades higher personnel costs against greater per-master-mentor capacity and consistency of delivery.
### Costs
**We modeled costs using a bottom-up approach anchored in the published costing of the Ghana LDHF trial** ([Willcox et al., 2017](https://doi.org/10.1186/s12992-017-0313-x); Gomez et al., [2018](https://doi.org/10.1186/s12884-018-1705-5)). All costs are expressed in 2023 USD.
Unit costs sourced from the Ghana studies were inflation-adjusted and then converted to each candidate country using price-level conversion ratios. Following AIM’s internal costing guidance, we applied a 10% payroll adjustment to direct labor and a 13% supply-chain adjustment to commodities delivered by the organization.
Note that, since we assumed the same fixed costs for all countries, these costs may be under-estimated for very large countries (like India) and over-estimated for small countries (like Guinea-Bissau). Indeed, fixed overhead costs make up the majority of total costs in our model for Guinea-Bissau, which is the primary driver behind making this country look not cost-effective.
**Figure 5: Driver tree of the modeled** [**costs**](#user-content-fn-3)[^3]**.**
### Difference between Estimate 1 and Estimate 2
**Estimate 1 captures only the cost of training, mentorship, and supporting infrastructure.** It treats medical commodity availability as a constraint that is either already met by existing supply chains or not addressed as in the Gomez et al. (2018) study. The implicit assumption is that trained providers can use existing facility stocks.
**Estimate 2 adds the cost of procuring and distributing four medical commodities required to perform BEmONC signal functions.** The components are:
* **Oxytocin** for PPH prevention.
* **Misoprostol** as a backup uterotonic in settings without a reliable cold chain.
* **Tranexamic acid** for PPH treatment.
* **Magnesium sulfate** for pre-eclampsia and eclampsia.
**Non-training commodity costs are modeled per birth reached and then summed to a per-country annual total.**
**We do not assume that adding commodities increases the effect size in our model—only that it adds cost.** If commodity gaps are currently a binding constraint on signal-function performance at the target facilities, Estimate 2 understates the true marginal benefit of the bundled approach. We did not have evidence sufficient to model this effect quantitatively.
### Reach
**The CEA assumes a maximum steady-state reach equivalent to 75% of all facility-based births in small countries, 50% in medium-sized countries, and 37.5% in large countries.** This translates to a large range, from 52 and 297 facilities in Guinea-Bissau and Sierra Leone, respectively, to 3,915 and 14,555 in Nigeria and India, respectively.
**The number of births per trained facility is the parameter that drives most cross-country variation in cost-effectiveness.** We estimate it country by country from country level facility surveys and Demographic and Health Survey (DHS) data. The estimates range from 180 births per facility per year in Chad to 882 in Ghana.
### Effects
We measure impact in DALYs averted and model mortality effects only, through three pathways: neonatal deaths, intrapartum (“fresh”) stillbirths, and maternal deaths. We exclude non-mortality benefits (e.g., reduced maternal and neonatal morbidity, long-term effects for surviving children), which makes the estimate conservative.
### Tractability
We assumed an 100% probability that the charity reaches at-scale delivery. This is consistent with AIM defaults for charities pursuing a direct service model with an established intervention playbook.
Figure 6: Driver tree for modeled scale of impact
### Counterfactual adjustments
**We assumed a 3% per year probability that the intervention happens counterfactually at the same facilities the charity would reach.** Over the five-year scale-up period, this compounds to 14.1% probability that the change would happen anyway, reducing the marginal impact of the charity by that share.
The 3% figure reflects the partial but uneven presence of Jhpiego, the Helping Mothers Survive Safer Births Bundle of Care, PRONTO, the United Nations Population Fund (UNFPA), and ministries of health in BEmONC training across the candidate countries. None of these actors is currently delivering focused, philanthropically funded, cost-optimized LDHF BEmONC training at primary health facilities with long-term country commitments—the closest model, the Safer Births Bundle of Care, is concentrated in East Africa and at hospital-level facilities—but the space is active enough that some share of facilities the charity would target are likely to be reached by other actors over a five-year horizon. The 3% figure represents the lower end of plausible counterfactual displacement and could be higher in crowded settings (Ghana, Tanzania) and lower in neglected ones (Chad, Guinea-Bissau).
### What we have not yet done
* **Non-mortality effects.** Spillover effects on household members of deceased mothers, child morbidity from birth complications, and benefits of skin-to-skin contact and kangaroo mother care education are currently unmodeled. GiveWell applies a +25% upward adjustment for comparable effects.
* **Sensitivity analysis.** We have not yet run formal sensitivity analyses on key parameters (effect size discounts, births per facility, counterfactual rate, fixed-cost staffing). These would help bound the uncertainty around the central estimates.
**Table 8: CEA considerations**
| Reasons this intervention could be MORE cost-effective than modeled, all else equal. | Reasons this intervention could be LESS cost-effective than modeled, all else equal. |
| --------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- |
|
• Non-mortality effects (long-term morbidity in surviving children and averted maternal morbidity) are excluded. • Household spillovers of averted maternal deaths are also excluded. As discussed in Kudymowa et al. (2026), such spillovers may be large. • Income benefits to the beneficiaries of the intervention are not included. These may include, for instance, a benefit to the mother who averts suffering from an obstetric fistula. • Benefits are assumed to last for one year, with no persistence without retraining. This is likely conservative. • The multiplication of all the applied discounts may be overly conservative. • Fixed overhead costs are implicitly assumed to replicate with each additional country. In reality, an organization that expands to multiple countries may be able to split these costs across several countries.
| • Long-term health disbenefits for survivors not modeled. For instance, those surviving neonatal encephalopathy in intrapartum-asphyxia may suffer long-term health impairments (hence, the total DALYs averted by saving their lives is lower than modeled). |
[^1]: Reported as of 24 August 2026—note the models are live and may be subject to tweaks or (in rare occasions) large changes that may not be reflected in the text if carried out after publication.
[^2]: The list of countries modeled is based on our geographic WFM but doesn’t fully match Table 6 due to discrepancies between earlier and later versions of the WFM.
[^3]: Exact costs in this figure may be out of date. Please refer to the [live model](https://docs.google.com/spreadsheets/d/1qzLTQbqSK3VSZki6Y_nDRbJ0ArZuSpKk8ZYE9B19gvQ/edit?gid=919555071#gid=919555071) for the most recent numbers.
---
# Agent Instructions
This documentation is published with GitBook. GitBook is the documentation platform designed so that both humans and AI agents can read, navigate, and reason over technical content effectively. Learn more at gitbook.com.
## Querying This Documentation
If you need additional information that is not directly available in this page, you can query the documentation by asking a question.
Perform an HTTP GET request on the following URL with the `ask` and `goal` query parameters:
```
GET https://research.ambitiousimpact.com/bemonc/simulation-based-training-for-emergency-obstetric-and-newborn-care/6-cost-effectiveness-analysis.md?ask=&goal=
```
`ask` is the immediate question: it should be specific, self-contained, and written in natural language.
`goal` is what the user is ultimately trying to achieve, the reason they need the answer. Sharing it helps GitBook give you a better, more relevant answer. A goal is most helpful when it describes the outcome the user wants rather than restating the question. For example, with `ask=how do I create an API token`, a goal like `automate deployments from our CI pipeline` lets GitBook tailor the answer to that use case.
The response will contain a direct answer to the question and relevant excerpts and sources from the documentation.
Use this mechanism when the answer is not explicitly present in the current page, you need clarification or additional context, or you want to retrieve related documentation sections.
**Figure 5: Driver tree of the modeled** [**costs**](#user-content-fn-3)[^3]**.**
### Difference between Estimate 1 and Estimate 2
**Estimate 1 captures only the cost of training, mentorship, and supporting infrastructure.** It treats medical commodity availability as a constraint that is either already met by existing supply chains or not addressed as in the Gomez et al. (2018) study. The implicit assumption is that trained providers can use existing facility stocks.
**Estimate 2 adds the cost of procuring and distributing four medical commodities required to perform BEmONC signal functions.** The components are:
* **Oxytocin** for PPH prevention.
* **Misoprostol** as a backup uterotonic in settings without a reliable cold chain.
* **Tranexamic acid** for PPH treatment.
* **Magnesium sulfate** for pre-eclampsia and eclampsia.
**Non-training commodity costs are modeled per birth reached and then summed to a per-country annual total.**
**We do not assume that adding commodities increases the effect size in our model—only that it adds cost.** If commodity gaps are currently a binding constraint on signal-function performance at the target facilities, Estimate 2 understates the true marginal benefit of the bundled approach. We did not have evidence sufficient to model this effect quantitatively.
### Reach
**The CEA assumes a maximum steady-state reach equivalent to 75% of all facility-based births in small countries, 50% in medium-sized countries, and 37.5% in large countries.** This translates to a large range, from 52 and 297 facilities in Guinea-Bissau and Sierra Leone, respectively, to 3,915 and 14,555 in Nigeria and India, respectively.
**The number of births per trained facility is the parameter that drives most cross-country variation in cost-effectiveness.** We estimate it country by country from country level facility surveys and Demographic and Health Survey (DHS) data. The estimates range from 180 births per facility per year in Chad to 882 in Ghana.
### Effects
We measure impact in DALYs averted and model mortality effects only, through three pathways: neonatal deaths, intrapartum (“fresh”) stillbirths, and maternal deaths. We exclude non-mortality benefits (e.g., reduced maternal and neonatal morbidity, long-term effects for surviving children), which makes the estimate conservative.
### Tractability
We assumed an 100% probability that the charity reaches at-scale delivery. This is consistent with AIM defaults for charities pursuing a direct service model with an established intervention playbook.