Summary

• Neonatal mortality is one of the leading contributors to DALYs worldwide.
• Training health workers to prevent newborn deaths may be a highly cost-effective intervention, with an average cost of $59 per DALY averted among the studies I could find.
• Highly scalable models (such as training of trainers and online courses) have excellent unit economics and could lower costs per DALY averted even further.
• There is a need for neonatal care training programs that can inexpensively yet effectively train tens of thousands of health workers.
• The world faces a major health worker shortage and neonatal care training could be a “proving ground” for developing new tools to address this gap.
• Future work should focus on identifying and funding health worker training programs that are most likely to be effective and scalable.

Disclaimer: This is a rapid, non-exhaustive investigation of this cause area and I’m not a domain expert in maternal and child health. There may be relevant evidence that I have missed or assumptions I have overlooked.

Neonatal mortality is a leading killer

The day of birth is the riskiest day of a child’s life. Neonatal disorders are the leading global cause of disability-adjusted life years (DALYs), accounting for 186 million DALYs or just over 7% of the global total. Fortunately, neonatal mortality declines as countries improve their health systems and develop economically. The overall rate of under-5 child mortality has dropped by almost 50% since 1990, while the rate of newborn deaths declined by 37% in the same period. Nonetheless, neonatal mortality remains a big problem in low- and middle-income countries (LMICs). Just 12 countries account for two-thirds of global newborn deaths. In light of such an unequal global distribution, this analysis focuses on interventions that prevent newborn deaths in LMICs. 

In addition to its direct effects as measured in years of life lost, neonatal mortality has many downstream effects, including the emotional trauma it inflicts on the mother and the rest of the family. Rates of depression are much higher in women who lose a newborn than among those who have healthy babies. There is also evidence that women who lose a child tend to have additional pregnancies in order to replace the lost child, and each pregnancy poses a substantial health risk to the mother. Cost-effectiveness studies of neonatal mortality interventions tend to account only for direct effects, so the analysis that follows does not estimate the substantial secondary benefits that flow from preventing a newborn death.

Some neonatal mortality is preventable

Not all deaths attributed to neonatal disorders are readily preventable. Nonetheless, the burden of preventable newborn deaths (estimated at 2.9 million per year) is extremely high. When health workers (HWs) are skilled and up-to-date on the standard of care before, during, and after birth, they provide better care and intervene appropriately to save babies’ lives. The following excerpt from a report by the World Health Organization (WHO) and UNICEF summarizes this well:

“We have solutions to address the main causes of newborn death: More than 80% of all newborn deaths result from three preventable and treatable conditions – complications due to prematurity, intrapartum-related deaths (including birth asphyxia) and neonatal infections. Cost-effective, proven interventions exist to prevent and treat each main cause. Improving quality of care around the time of birth will save the most lives, but this requires educated and equipped health workers, including those with midwifery skills, and availability of essential commodities.” 

There are a variety of different approaches that can be taken to prevent neonatal deaths. These range from HW training to community-based interventions and providing performance-based incentives to HWs. GiveWell recently reviewed interventions intended to reduce neonatal and maternal mortality, ultimately concluding that some programs are in the range of the most cost-effective charities though there are substantial differences between programs. GiveWell’s work led me to ask whether certain programs or types of interventions are consistently more effective than others.

Health worker training is a cost-effective way to prevent neonatal deaths

Based on data summarized in systematic reviews, I hypothesized that HW education is more cost-effective than other types of neonatal care interventions. I searched reviews and databases for any study of HW training in neonatal care meeting the following criteria:

• is published in a peer-reviewed journal,
• evaluates a HW training intervention in a low- or middle-income country (LMIC),
• focuses on neonatal deaths as a primary outcome, and
• reports costs in US dollars per DALY averted.

The six studies meeting those criteria are shown below. Two of these studies focus on pre-service programs (training new HWs), while the others focus on in-service programs (training existing HWs). 

These studies suggest that training HWs may be quite a cost-effective intervention. With an average cost per DALY averted of $59, training delivers an almost “1,000X” return on investment, which has been cited as a target for the most effective charities. For reference, in 2013 GiveWell estimated that programs to prevent malaria with insecticide treated bednets cost between $44-$114 per DALY averted.^[1]

Although existing studies on the cost-effectiveness of training look promising, these results should be interpreted carefully due to a number of limitations.

• Study design: only two of these studies are randomized controlled trials.
• Publication bias: there are only a handful of studies published and it may be that just the most favorable studies advanced to publication.
• Declining neonatal mortality rates: with global neonatal mortality rates declining, the effect of these interventions may be lower now than at the time the studies were conducted. It is worth noting, however, that mortality rates due to many causes have been declining for decades, so this is likely true of other cause areas with highly effective interventions.

An additional complication is that these studies differ in training method and the group of HWs trained. Different HW cadres trained included midwives, traditional birth attendants, “hospital staff” (likely a mix of nurses, doctors, midwives, and other HWs), and community health workers. This speaks to the wide array of different roles held by HWs present at birth and involved in neonatal care.

What is already being done

It is easy to overlook the fact that, even in LMICs, the majority of public health funding comes from national governments rather than international aid. Maternal and child health care is part of a package of “essential services” provided in primary health care facilities and in the community. The level of government involvement in HW training varies. In general, governments oversee the review and accreditation of pre-service training programs (for example, medical and nursing schools) and in-service training programs (continuing professional development). In any given country, it’s likely that the national government is the largest funder of HW training and the largest funder of neonatal care services. Though governments regulate and fund training, the training packages used for in-service neonatal care training frequently come from outside entities. 

The WHO and a number of non-governmental organizations (NGOs) have created neonatal care curricula. Two of the most commonly-used training packages are the Essential Newborn Care and Helping Babies Breathe programs. Both of these programs focus on key clinical knowledge and skills that HWs need when attending births, including essential steps to prevent infections and resuscitate babies who aren’t breathing. A meta-analysis of training interventions found substantial reductions in newborn deaths after these programs were implemented. Overall mortality (the sum of stillbirths and deaths after birth) declined by 18%. While existing training packages are effective, it is not clear that they can be scaled to reach many thousands of additional HWs in LMICs.

Promising approaches to training at scale

Many global health interventions that are now recognized as highly effective (such as insecticide treated bed nets, deworming, and childhood vaccines) share the following properties.

• Excellent unit economics: the cost per additional unit of treatment or intervention is very low.
• Lasting effects: benefits persist for a period of time beyond the intervention. For example, many vaccines protect against infection for years.

Scalable education approaches, such as online learning and training of trainers, also have these two properties, making them promising for further investigation. I look into two such approaches in more detail below.

Training of trainers

Training of trainer (ToT) methods are fairly common in health care. These typically involve identifying HWs to serve as “lead trainers” and bringing them to a central location for several days or weeks of intensive, in-person training. The lead trainers are often given supplies, which might include consumables and equipment (such as PPE and a mannequin used to simulate newborn resuscitation). The lead trainers then fan out to health facilities to train other HWs. In Zambia, a ToT approach was used to train 18 midwives, who went on to train a total of 123 midwives working in primary health facilities. The estimated cost of this ToT program was just $5 per DALY averted. Another ToT program run in Tanzania had a cost of $23 per DALY averted.

Potential challenges to ToT approaches include:

• Quality control: lead trainers need to be well-trained and prepared to deliver training and some monitoring mechanisms should be in place to ensure that trainings are conducted properly.
• Complex logistics: ToT programs may involve paying honoraria, managing transportation logistics, and acquiring and disseminating teaching supplies. However, these costs are likely lower in ToT programs than comparable programs that involve flying in outside “experts.”
• Travel restrictions: this has been an issue during the COVID-19 pandemic but can have other causes (such as civil unrest or armed conflict).

These challenges are clearly surmountable in the sense that they are true of many other global health programs, including programs that are highly cost-effective. The low cost per DALY averted and strong unit economics (cost per HW trained) of ToT programs suggest that they are a particularly promising class of intervention.

Online training

There is emerging evidence that online learning, when properly and thoughtfully designed, can be quite effective for training HWs. My own research on an infection control course delivered in Nigeria demonstrates that self-paced, case-based, mobile-optimized online training that specifically addresses the learning needs of frontline HWs can be very engaging and effectively teach essential clinical concepts. This provides proof of concept that high-quality online HW training is possible with current technology.

In the absence of existing data on the cost-effectiveness of fully online neonatal care training, I modeled two different scenarios:

• Initial build: the cost of creating a fully online training in neonatal care and piloting it for three years to reach 50,000 HWs engaged in neonatal care.
• Sustained delivery: the cost of maintaining such a training after the three-year build and reaching 10,000 HWs engaged in neonatal care per year.

I focused on in-service HW training because all pre-service training (nursing school, for example) requires a large in-person component. By contrast, there are many examples of fully online in-service training packages and I have experience in creating such programs and estimating their costs. For audience sizing (a very conservative estimate of the number of HWs who can be reached), I focused on the WHO Africa region which has a relatively high burden of neonatal mortality. The potential audience size is much larger given that several countries in other regions have a high burden of newborn deaths. Two pre-existing cost-effectiveness studies reported DALYs averted and the total number of HWs trained. I used the ratio of these numbers as a baseline, but to account for a high degree of skepticism about online training, I assumed that online approaches would avert many fewer DALYs per HW trained than in-person approaches. The estimates derived from the model are shown below.

This simple modeling exercise demonstrates the excellent unit economics of online training. Even if it is substantially less effective at averting DALYs on a per-HW basis, the very low cost of training each additional HW could make online training very cost-effective. This modeling does not account for secondary benefits, which may ultimately outweigh the direct effects of neonatal care training. For example, if this work drives the development of online platforms that are very effective at training HWs at scale, this could become a new tool for averting death and disability in priority areas other than neonatal care.

Online training mitigates some of the challenges of ToT training approaches because no travel is required and quality control can potentially be more centralized. However, training HWs online creates new challenges:

• Lack of hands-on simulation exercises: neonatal care training packages include clinical simulations that usually involve using mannequins to practice procedures and resuscitation. It is reasonable to assume that even well-designed online training would be less effective for teaching some of these skills.
• Access to devices and internet: not all HWs have devices or a reliable internet connection. Given rapid growth in device ownership and cellular network coverage worldwide, this challenge is diminishing with time.

These challenges justify skepticism in line with the assumption that online training averts 1/5^th to 1/100^th the number of DALYs per HW trained in-person. One might assert that it’s best to assume online training has no effect at all (which would clearly mean it is not cost-effective). I would argue that it’s reasonable to estimate a small positive effect based on research showing that a phone app can improve HWs’ birth attendant skills and work from other fields showing that well-designed app-based training can effectively teach complex skills.

Nuances in neglectedness

At a first glance, it may not be clear that neonatal mortality is a neglected cause space: many high-profile philanthropies and intergovernmental organizations (such as the Bill and Melinda Gates Foundation and the WHO) already work on maternal and child health. However, as discussed above, most health care funding does not come from international aid, and most neonatal care is financed by public expenditures. Primary health care funding in LMICs is insufficient to ensure adequate coverage of the essential health services that include neonatal care. Though this indicates a need for more primary health care funding, it doesn’t answer the specific question of whether HW training programs could effectively deploy more funding.

Existing training options

I have searched extensively for existing online training solutions and it isn’t clear to me that any are optimized to meet the learning needs of HWs. Given the major role of the public sector in neonatal health care, it would be ideal if governments had access to high-quality, effective, and scalable courses that they could adopt or adapt with minimal modifications to train their health workforces. There isn’t a consensus solution that most governments and HWs currently use, but there are some existing initiatives that merit a closer look.

• The Safe Delivery App, created by the Maternity Foundation and the University of Copenhagen, is currently available and supported by some promising efficacy data.
• The Helping Mothers and Babies Survive course, offered by the American Academy of Pediatrics, the NGO Jhpiego, and Laerdal (a medical simulator company), provides hybrid online and in-person training.
• The WHO sets international health guidance in many fields, including maternal and child health. The OpenWHO massive open online course platform currently offers one course on infection control in neonatal care, but does not appear to have other neonatal health courses.

Substantial room for growth

As of 2013, there was a global shortage of 17 million HWs, even though there were 43 million HWs around the world at the time. Relative to the size of the health workforce, all existing online trainings for HWs operate on a small scale. Even OpenWHO, with more than 500,000 learners enrolled in some courses, only reaches a fraction of the global health workforce. At the same time, we have proof that online learning platforms can effectively reach huge audiences – Duolingo has over 40 million active users per month. Yet there is no “Duolingo for health” – a simple, effective, engaging, gamified, mobile-optimized, free learning app for in-service and pre-service HW training. Such a “digital public good” wouldn’t take the place of in-person training, but it could help to alleviate major global HW shortages and improve quality of care across the board.

Why doesn’t such an app already exist? I’ve written elsewhere about why it’s so hard (but entirely possible) to make online learning excellent. Creating engaging, effective, easy-to-use, highly scalable learning experiences requires a blend of expertise in technology, business and marketing, learning sciences, human-centered design, and the subject matter taught. It isn’t cheap or trivial to make it happen. Many different players are engaged in the HW training space, including governments, NGOs, intergovernmental organizations, universities, and private sector firms. While all of these stakeholders invest in training, much of this investment is not coordinated. In other words, it’s a public goods problem in which total investment is high but spread thin.

I say all of this to make the following point: high rates of neonatal mortality are one symptom of a much bigger gap in health workforce development. While neonatal care training may stand on its own in terms of cost-effectiveness, it also could be a “proving ground” for scalable training solutions. Workforce development is a cross-cutting tool that averts death and disability caused by a wide array of ailments. “Moonshots” in this area would require a lot of funding and are not likely to show substantial return on investment for many years. Better mechanisms to pool and direct funding (such as those that fund education projects in other cause areas) might be needed to accomplish this.^[2]

I was tempted to title this piece “New cause area: health workforce development,” but the reality is that there’s much better data to quantify cost-effectiveness on a dollars-per-DALY basis in the niche of neonatal care training.^[3] I believe that this much narrower cause area is just the tip of the iceberg – it’s easy to see and quantify but also indicative of a much bigger underlying and unaddressed problem. Fortunately, modest investments in neonatal care training might be cost-effective on their own while also generating transferable lessons applicable to the bigger problem. 

Thoughts on next steps

This shallow investigation is subject to my own biases and it certainly isn’t comprehensive. Here are some recommendations for next steps.

Conduct additional research by doing a deep search of the literature, speaking with experts, and dissecting the arguments made in this piece. This stage might include some additional modeling to estimate cost-effectiveness of HW training using the GiveWell or Open Philanthropy frameworks, in order to make direct comparisons with other interventions. It would also be useful to investigate the properties of different training interventions (HW cadres trained, methods used, skills taught, etc.) to determine if some types of training are consistently more effective than others. 

Identify candidate interventions by looking into specific projects that have the potential to effectively train HWs at scale. This would involve seeking out promising organizations, including those listed above, and speaking with their representatives to understand their work. ToT, online, in-person, and hybrid approaches should all be considered. The most promising initiatives should:

• build off an evidence base of what works for training HWs,
• have potential to scale up to train many thousands of HWs,
• commit to ongoing evaluations of efficacy, and
• have an approach that complements government initiatives to reduce neonatal mortality.

While I have opinions about approaches that are the most promising, I’d rather not make specific funding recommendations now as I think more due diligence is needed.

Invest in pilot and scale-up projects after identifying the most promising interventions. In spite of the magnitude of this problem, it isn’t clear to me that any project in this space has been built to maximize learning and engagement while also being poised to achieve global or regional scale. Funders may need to patiently spend a few years learning more and taking a “hits-based” funding approach. This could include supporting a mixture of projects at different stages of development (pilots of new programs and scale-ups of existing programs, for example). Those initial investments would build up a stronger evidence base while also helping organizations develop scalable models with the capacity to absorb more funding.

Conclusions

1. Training HWs at scale may be a highly cost-effective approach to reducing the burden of neonatal mortality in LMICs. More work is needed to test this tentative conclusion.
2. The most cost-effective global health and well-being interventions share the properties of (1) excellent unit economics and (2) effects that last beyond the intervention. It may be useful to look for these properties when evaluating new and untested interventions.

Transparency about my biases

I design, deliver, and study online HW training, but I do not currently work on any neonatal care courses. Like many people, I want to believe my work has meaning so I’m probably less skeptical than the average person about the potential of online learning. While I believe many of the criticisms of online learning as a category are fair, I’m also convinced it can be excellent if well designed. That said, I tried to control for my biases by presenting a range of skeptical estimates of the efficacy of online training approaches.

Acknowledgements

I’d like to thank Spencer Queen for excellent feedback on this piece and Akhil Bansal who encouraged me to investigate cost-effectiveness in this cause area.

As of June 28, 2022, this article has a certificate of impact fully owned by the author.

1. ^{^}

   This estimate of DALYs averted by malaria prevention programs may be a bit out-of-date. GiveWell now uses a methodology that does not rely upon DALYs to compare the effectiveness of different charities. 

2. ^{^}

   While I’m confident that better pooled funding mechanisms could coordinate investments and concentrate it on promising workforce development initiatives, this is well outside of my expertise so I don’t speculate in this piece on how pooled funding might be structured.

3. ^{^}

   Some may contest my categorization of HW training in neonatal care as a "cause area." I think the reality is that, within any given broad cause area - even those that are extremely important - different interventions have a wide range of efficacies. I chose to focus on a class of interventions that seem particularly promising from a cost-effectiveness standpoint.