KMC Cost Effectiveness

in Bahraich, India

A collaboration of r.i.c.e., PHI, and the Government of Uttar Pradesh to provide KMC to many very-low-birthweight babies, in a setting where the counterfactual is very limited care (not conventional intensive neonatal care).

Executive Summary

• We staff, manage, and implement the government’s KMC policy in a large public hospital and through home-visits. Almost all very-low-birthweight babies we help otherwise would not receive conventional intensive care.
• Cost-effectiveness is a product of two terms: 1 death averted per $1,910 to $4,391, depending on assumptions ≈ (mortality reduction per baby helped) × (babies helped per dollar).
• Our number of babies helped per $ is exceptionally high (more babies already than any Cochrane study) because of our context at a large public hospital in a high-mortality population: $434 per average baby, taking in >11 babies less than 1800 gm per week.
• Our average mortality reduction per baby served is much higher than in the Cochrane studies because background neonatal mortality is greater in Bahraich than in the places where Cochrane studies were done.
• We started without dedicated funding. For as long as we have about $5,000 per week we can continue to pay salaries and other marginal costs to keep operating, saving lives, and potentially informing and inspiring replications.

Preview of four estimates from four empirical strategies

Sources: KMC Bahraich estimates are computed ahove, in these slides. AMF from 2020 GiveWell cost effectiveness analysis — version 4

Two notes:

• Because this prevents death at the very start of life, many life-years are saved, relative to later-life mortality reduction.
• We have not included the human capital and health effects of weight-gain in early life because we think the case is strong enough based on mortality, but we could include these benefits if GiveWell is interested.

To understand our program, the first step is to refresh our understanding of what we use KMC for in our context.

• GiveWell summary: “Kangaroo mother care (KMC) is intended to be a low-cost alternative to conventional neonatal intensive care for low birth weight infants and has been hypothesized to reduce neonatal mortality compared to conventional neonatal intensive care.”

​

​

​

• Conventional care is not the alternative here. In our setting, care is very constrained, even for those very-low-birthweight babies who receive any.
• Many would otherwise go home shortly after birth (as in the picture to the right) likely to die.

The problem we address

• The Governments of India and Uttar Pradesh have adopted KMC as the newborn care policy for stable low-birthweight babies in public hospitals
• Prior to our project, very low birthweight babies were unlikely to receive adequate, conventional intensive neonatal care (the “control” in the Cochrane Review studies, which include studies in US, UK, Australia…)
• Most went home upon being able to breathe on their own, even if they were unable to suck, swallow, gain weight, or regulate their temperatures.
• The public hospital where we operate had 3 beds designated for KMC (due to our program, there are now 17, which are almost always full); staffing and oversight for the 3 beds was incomplete
• There was no home-visit program to identify babies in distress, promote KMC, and promote breast milk feeding once babies left the hospital. Phone follow-ups were infrequent.
• We documented about ⅓ mortality.
• Because neonatal mortality in Bahraich district is very high, maternal nutrition is poor, and prenatal care is incomplete, there is a steady arrival of many very low birthweight babies

We provide staffing to the hospital-based KMC program, record-keeping, management, and a program of home visits and phone calls to babies who go home.

Outline

1. Overview of cost-effectiveness
• We show that our competitive cost-effectiveness is plausible even using independent estimates of background mortality rates and using the Cochrane Review’s effect size.
2. Why our effect-per-baby on neonatal survival probably exceeds Cochrane
• Our counterfactual is not conventional neonatal intensive care.
3. How we can help so many very-low-birthweight babies
• Where we operate, there are many VLBW babies and they have no alternative medical care
4. Our best estimates of cost-effectiveness
• It depends on how you classify “referrals” in the government’s data system
• There is an uncertainty interval, but all possibilities are cost-effective
5. Our funding gap and upside potential
• We started this program without dedicated funding. Each week we continue the program running, rather than dissolve it, costs about $5,000 and saves more than on life, on average.
• Keeping the program running preserves the opportunity for demonstration effects on other public hospitals, for an impact evaluation, and for advocacy to policy-makers for scale-up.

1. Overview of cost-effectiveness

before

same baby, after

How cost effective is our program at saving lives?

Cost-effectiveness, in general, is the product of these two dimensions.

​

Impact evaluations, like the Cochrane Review, only estimate lives saved per baby helped, and may not have external validity to the most constrained settings.

Our program helps many babies inexpensively.

As we will show, it has already helped more babies than all 8 Indian studies in the Cochrane Review combined, and more babies than in the largest Cochrane study.

• Not assessed by the Cochrane review
• Highly dependent on context and variable between elsewhere in India, on the one hand, and Rhode Island, Connecticut, and Colorado, on the other (all of these are example Cochrane studies)
• This is where our program excels
• We have carefully chosen a hospital and population with many VLBW babies
• High throughput, building on our decade of experience managing human-intensive projects in rural north India

Our program helps many babies inexpensively.

​

1/$434

(11.5 babies per week)

Cochrane studies deeply understate our effect per baby.

In this setting, preventable neonatal mortality is exceptionally high, by any standard. That is in part because very few of our babies would otherwise receive conventional care.

• The Cochrane Review highlights that their numbers would underestimate the value of our program:

"Given that the control group in studies evaluating continuous KMC was kept in incubators or radiant warmers, the potential beneficial effects of KMC on morbidity and mortality of LBW infants would be expected to be greatest in settings in which conventional neonatal care is unavailable." (Authors conclusions p. 26)

• Neonatal mortality is very high where we operate

Before we get into the details of our setting and program, let’s do a simple computation, using the Cochrane Review effect size of 0.6, to show that our basic cost effectiveness claim is plausible—even though we believe that the effect per baby is larger.

Ballpark estimate not using our own mortality numbers, instead using Cochrane Review’s estimate of a 0.6 relative risk, which is a 40% reduction

Counterfactual mortality * 0.4

​

So what is the counterfactual neonatal mortality rate for VLBW infants in a setting that lacks medical care for such babies?

What would neonatal mortality be without medical care?

​

That is hard to answer. Here are three population neonatal mortality rates for comparably disadvantaged babies that did receive some less-intensive medical care (which will bias against our cost-effectiveness).

0.133

0.2

0.233

An alternative approach that we are prepared to discuss uses babies for which birthweight is observed in the NFHS-5. The concern here is that babies who die are unlikely to be measured.

Ballpark estimate not using our own mortality numbers, instead using Cochrane Review’s estimate of a 0.6 relative risk, which is a 0.4 reduction

0.2 × 0.4

(NNM) (Cochrane)

1/$434

(11.5 babies per week)

The result is cost-effective, even using the Cochrane effect size, because background neonatal mortality is common and because many babies are helped.

0.08

1/$434

1/$5,424 =

But, in fact, we believe that our effect per baby is larger than 8 percentage points. So we prevent neonatal deaths at less than an average cost of $5,424 per death averted. The remaining slides explain why.

×

An alternative (older) systematic review uses only RCTs from low/middle income countries with <2,000g babies.

(0.2 × 0.51)

1/$434

1/$4,254 =

Because its effect size is larger (0.49 relative risk, or 51% reduction), this more-applicable study yields a more favorable cost-effectiveness estimate for our program, $4,254 per life saved.

×

With this background, we will build towards our own best estimates of cost effectiveness, by first discussing why our effect size is large and our number helped is large.

​

Here, we ignore the important benefits of early-life weight gain for the human capital of survivors, because we believe a strong enough case can be made with mortality, but we would be happy to add in these benefits if GiveWell is interested.

2. Why our effect-per-baby on neonatal mortality likely exceeds Cochrane

Because we observe the neonatal mortality rate in our program, the open question is how high neonatal mortality would be without our program. Evidence suggests it would be very high.

before

same baby, after

Counterfactual mortality is very high, reason 1 of 2:

Neonatal health is exceptionally poor in this part of India, relative to the full global distribution

​

Sources: CDC Wonder for US States, NFHS-5 for Indian states and Bahraich, UNICEF Child Mortality Estimation for other countries. We use most-recent rates for comparability.

Among all countries, only Lesotho (44 deaths per 1,000 births, pop. 2.1 million) has a neonatal mortality rate higher than Bahraich, a UP district of 3.6 million people and a neonatal mortality rate of 43 per 1,000.

settings of studies in Cochrane review

Counterfactual mortality is very high, reason 2 of 2:

Unlike in the Cochrane studies, these babies would not otherwise receive conventional neonatal intensive care

Representative quotations from Cochrane Review control groups:

• “Infants were kept in a warm room in Addis Ababa, with open cribs and the possibility of rewarming in a bulb-heated cot, and in incubators in the other two hospitals.”
• “Infants were kept in a warmer or incubator. Mothers were allowed to visit their babies and were encouraged to perform infant care activities such as diaper change, oil massage, and paladai feeding.” (Hyderabad, India study)

In contrast, many of the babies in our study would otherwise go home without professional care or support, when able to breathe without support:

• Although KMC was the government policy, there were few beds and no active, dedicated staff support.
• This constrained situation is common in nominally-designated KMC spaces throughout north India.

In summer 2021, we found that about a third of eligible babies died, as we expanded the number of dedicated beds and staff size.

Counterfactual: hypothermic at home

Now: KMC with mom

3. We help very many babies

before

same baby, after

Many, many very low birthweight babies are born in Bahraich

• The hospital with which we partner reported 8,309 births in 2019, or 23 per day.
• Would make it the 11th busiest hospital for births in US.
• This hospital also serves as a referral facility for VLBW babies born at other hospitals in Bahraich
• Through home visits and phone calls, our team has helped the hospital increase referrals to the facility, a positive externality that brings more babies to be helped
• 4.2% of babies in Bahraich are born below 1,800 grams, according to the most recent demographic survey.
• 4.2% × 8,309/365.25 would equal 0.95 born per day.
• The true population fraction is higher than 4.2% because babies who die have their weight under-reported.
• Because this public hospital is a referral facility, it is intended to care for VLBW babies born elsewhere also.
• Taking these facts together, it makes sense that we intake > 11 VLBW babies per week.

Our program excels at cost-effective throughput

• Many very-low birthweight babies are born in or come to ASMC Bahraich.
• Our team of 15 nurses, 4 managers, and 11 support staff handle them inexpensively.
• Average nurse salary: $450/mo
• Essentially all of our costs are marginal costs (salaries, baby supplies, fuel) per time period that the program operates.

​

Our program is already much larger than the size of the studies in the Cochrane review.

• There are 8 studies from India in the Cochrane review. They have an average size of 54 infants.
• Among all 21 Cochrane studies, including several from developing countries, the average size is 73 infants.
• We have already helped 540 infants under 1,800g, plus more who qualified for our program in other ways.
• Our program is larger than the largest study in the Cochrane review, which treated 396 babies in Colombia.

4. Our funding gap and upside potential

before

same baby, after

Our funding situation: We have almost no dedicated funds, we started without a funder, counting on EA to step up if we succeeded!

• We started this project with money left over from a survey that we didn’t end up needing.
• We use hospital infrastructure and some existing hospital staff.
• Any expansion to other public hospitals would, too.
• Founders Pledge contributed $125,000.
• They seemed confident that larger funders in the EA community would be excited to support us.
• Because new VLBW babies are born at or referred to ASMC Bahraich every day, every day that the program continues is more lives saved.
• Salaries are our main cost. If we run out of money, we’ll have to lay off the nurses!
• We will use a lot of the leadership’s time and attention (which could otherwise be spent on improving and advocating for the program) to keep that from happening, but it very well might.

summer

2021

summer

2022

summer

2023

neonatal deaths prevented

Because we are asking for funds for variable, marginal costs of continuing to operate, lives saved is an approximately linear function of funding.

Inexpensive way of generating excellent upside potential

• In principle, many hospitals in Uttar Pradesh or Bihar could benefit from a similar program, especially because an important state-level official is our ally.
• We look forward to advocating for this when our program becomes mature and sustainable.
• If our program runs out of funds and stops, we will lose the opportunity to use its success to inspire and inform this possible spread.
• A formal impact evaluation may be a useful input to advocacy and political support. This, too, would require the program to continue running.
• The political and institutional determinants of whether the program spreads probably do not depend on the same sort of econometric details of an impact evaluation that EA decision-makers would be interested in, but it is nevertheless a necessary condition that the program be continuing and be perceived to be succeeding for anyone to be interested in expanding it.

We invite conversation about funding a formal, matching-based econometric impact evaluation, which would require additional staff to collect data from comparable public hospitals.

Executive Summary (again)

• We staff, manage, and implement the government’s KMC policy in a public hospital and through home-visit follow-ups. Almost all very-low-birthweight babies we serve otherwise would not receive conventional neonatal care.
• Cost-effectiveness is a product of two terms: 1 death averted per $1,910 to $4,391, depending on assumptions ≈ (mortality reduction per baby helped) × (babies helped per dollar).
• Our core contribution is management and logistics. So, our number helped per dollar is exceptionally high (more babies already than any Cochrane study) because of our context at a public hospital in a high-mortality population: $434 per average baby, taking in >11 per week.
• Our average mortality reduction per baby served is much higher than in the Cochrane studies because background neonatal mortality is greater than any country’s but Lesotho’s and the counterfactual care is not up to conventional neonatal care (the control in Cochrane studies).
• We started without dedicated funding. For as long as we have about $5,000 per week we can continue to pay salaries and other marginal costs to keep operating, saving lives and potentially inspiring and informing replications.