A Simple Checklist Helps Doctors Spot the Sickest Newborns — And It Works in More Than One Hospital

How Researchers Tested an Ethiopian Newborn Risk Score in a Second Hospital to See If It Still Held Up

Researchers in Ethiopia built a simple, five-question checklist that helps hospital staff spot which newborns in intensive care are most at risk of dying — using only things a nurse can observe, no lab tests needed. When they tested it at a second, different hospital, it worked just as well, a reassuring sign that the tool could help far beyond the hospital where it was built.

Why This Question Matters

When a newborn is admitted to a neonatal intensive care unit (NICU) — the part of a hospital that cares for sick or very small babies — one of the first and most important jobs for the medical team is figuring out just how sick that baby really is. Getting this right shapes everything that follows: how closely the baby is watched, how the team plans care over the coming hours, and how honestly and clearly the family can be told what to expect.

A newly published study, carried out by a team of Ethiopian and American researchers, tested whether a simple, bedside-only checklist for judging newborn risk could work not just at the hospital where it was invented, but at a completely different one too [1]. The answer matters well beyond the two hospitals in the study, because it speaks to a much bigger problem in newborn care worldwide.

The Problem With Borrowed Tools

In big, well-funded hospitals, doctors often use blood tests and machines to help judge how sick a newborn is. But in many hospitals around the world — including many in Ethiopia, where this research took place — those blood tests and machines are not always available. Doctors and nurses need a way to judge risk using only what they can see and measure at the bedside: how a baby is breathing, how alert the baby seems, how big and how early the baby was born. This is not a small problem: an estimated 2.3 million newborns die worldwide each year, and the risk falls disproportionately on small and vulnerable babies born in lower-resource settings [2], with newborn and preterm-related causes making up a growing share of all deaths of children under five as other causes of child death have fallen faster [3].

For a long time, the risk-scoring tools available to neonatal teams were built and tested somewhere else — mostly in well-resourced hospitals in wealthier countries. One well-known tool from the early 1990s, called the Clinical Risk Index for Babies (CRIB), was genuinely useful, but it needed a blood gas test — a lab test that measures oxygen and acid levels in the blood — within a baby's first several hours of life [4]. A later tool, the Score for Neonatal Acute Physiology, known as SNAP-II, also leaned on laboratory results that are simply not available everywhere [5].

Back in 2020, the same team of researchers behind this new study tried to close that gap. Working with records from one hospital in the capital, Addis Ababa, they built a checklist — the Neonatal Mortality Score — using only clinical observations: a baby's gestational age (how many weeks along the pregnancy was at birth), birth weight, breathing rate, whether the baby showed signs of breathing difficulty, and how alert the baby was. No blood tests required. That first version worked well at the hospital where it was built [6]. But there was an obvious next question, one that comes up with almost any medical tool built at a single hospital: would it still work somewhere else? A checklist that only works where it was invented is much less useful than one that travels. This question carries real weight in Ethiopia specifically, where recent national data show that babies remain most vulnerable to dying in their very first days of life [7], and where a large survey of 208 Ethiopian hospitals found big differences from one hospital to the next in what equipment, staff, and newborn-care routines were actually in place [8].

What the Researchers Did

The team refined their checklist into an updated version, called the uNMS, and tested it in three groups of babies. Two of those groups came from the original Addis Ababa hospital: one set of 2,004 newborns used to build the refined checklist, and a second set of 669 newborns used to double-check it at the same hospital [1]. The third group was the real test: 812 newborns from a completely different hospital, in the city of Gondar, that had nothing to do with building the tool.

The uNMS adds up points across the same five simple observations to produce a single number from 0 to 24 — the higher the number, the higher the estimated risk. It takes only basic arithmetic to calculate, with no software or laboratory equipment needed [1].

What They Found

The checklist worked well in all three groups of babies — including, importantly, the Gondar hospital it had never seen before. Statisticians measure how well a tool like this tells apart babies who will survive from babies who will not using a score called the AUC, where 0.50 would mean the tool is no better than a coin flip and 1.00 would mean perfect prediction. The uNMS scored 0.85 in the group used to build it, 0.87 in the follow-up group at the same hospital, and 0.87 in the brand-new Gondar hospital — essentially just as accurate in a hospital it had never encountered [1]. That is a meaningful result, because the babies admitted in Gondar were, on the whole, considerably sicker: about 1 in 4 of the Gondar babies in the study died, compared with roughly 1 in 7 at the Addis Ababa hospital. Despite that real difference in how sick the two groups of babies were, the checklist kept telling higher-risk and lower-risk babies apart about equally well in both places.

What This Means for Families

For a family whose baby is admitted to a NICU that uses a tool like this, a score calculated from simple bedside observations — not an invasive test — can give the medical team, and eventually the family, a clearer, steadier sense of how serious the situation is. It does not replace the judgment of an experienced doctor or nurse, and the researchers themselves are careful to note that more testing is needed before it is used widely; but a validated tool like this, checked at more than one hospital, is more trustworthy than any single clinician's individual impression on its own, precisely because it produces the same measured answer no matter who is calculating it or which day of the week it happens to be. Families should also know that this study only looked at whether the checklist could predict risk correctly — not at whether using it changed the care babies actually received or the outcomes they had; that is exactly the kind of question researchers say needs to be studied next, alongside making sure hospital staff at all levels — not just those at large referral hospitals like the two studied here — have the training and time to use it consistently [9].

The Historical Backdrop

For decades, families of critically ill newborns in lower-resource hospitals had to rely almost entirely on their medical team's individual experience and instinct to understand how serious their baby's condition was, because the internationally recognized scoring tools of the era assumed a level of laboratory testing that simply was not available. That began to change only gradually, first with the 2020 Addis Ababa checklist built specifically for a setting like this one, and now with this study's demonstration that a version of that same idea can hold up in a second hospital entirely. Researchers who build tools like this are increasingly expected to follow international standards for testing and reporting them honestly — including checking, as this study did, whether a tool still works somewhere it was not built, rather than only reporting results from its home hospital [10]. It is a small, incremental step — but in newborn medicine, tools that travel reliably from one hospital to another are rare enough that each confirmed example matters.

What Researchers Are Working On Next

The next step the researchers themselves call for is testing the checklist in real time, as part of everyday hospital routine, rather than reconstructing it afterward from medical records — and doing so at a wider range of hospitals, including smaller, less-resourced ones, not only the two larger referral hospitals studied here. Separately, some research teams are exploring whether more advanced computer models, trained on much larger sets of newborn health data pooled from many countries, can eventually offer even sharper predictions than a simple five-item checklist [11]. But for hospitals without the technology or data infrastructure such approaches require today, a checklist like the uNMS — simple enough to calculate by hand, and now shown to work in more than one hospital — represents real, usable progress toward giving every sick newborn, wherever they are born, a fairer chance at being recognized as high-risk in time to help.

References

  1. Dickinson M, Workneh R, Phibbs CS, Abayneh M, Mediratta RP. Developing and validating an updated neonatal mortality score on admission to neonatal intensive care units in Ethiopia. BMC Pediatrics. 2026;26:7511. doi:10.1186/s12887-026-07511-0 ↩
  2. Lawn JE, Ohuma EO, Bradley E, et al. Small babies, big risks: global estimates of prevalence and mortality for vulnerable newborns to accelerate change and improve counting. Lancet. 2023;401(10389):1707–1719. doi:10.1016/S0140-6736(23)00522-600522-6) ↩
  3. Perin J, Mulick A, Yeung D, et al. Global, regional, and national causes of under-5 mortality in 2000–19: an updated systematic analysis with implications for the Sustainable Development Goals. Lancet Child Adolesc Health. 2022;6(2):106–115. doi:10.1016/S2352-4642(21)00311-400311-4) ↩
  4. International Neonatal Network. The CRIB (clinical risk index for babies) score: a tool for assessing initial neonatal risk and comparing performance of neonatal intensive care units. Lancet. 1993;342(8865):193–198. doi:10.1016/0140-6736(93)92296-692296-6) ↩
  5. Richardson DK, Corcoran JD, Escobar GJ, Lee SK. SNAP-II and SNAPPE-II: Simplified newborn illness severity and mortality risk scores. J Pediatr. 2001;138(1):92–100. doi:10.1067/mpd.2001.109608 ↩
  6. Mediratta RP, Amare AT, Behl R, et al. Derivation and validation of a prognostic score for neonatal mortality in Ethiopia: a case-control study. BMC Pediatrics. 2020;20:238. doi:10.1186/s12887-020-02107-8 ↩
  7. Tebeje TM, Seifu BL, Asgedom YS, et al. Trends and determinants of early neonatal mortality in Ethiopia: evidence from the Ethiopian demographic and health survey data. BMC Pediatrics. 2025;25:704. doi:10.1186/s12887-025-06022-8 ↩
  8. Endalamaw LA, Estifanos AS, Medhanyie AA, et al. Readiness to Provide Neonatal Care Services in 208 Ethiopian Hospitals Prior to Implementation of the Saving Little Lives Program. Children. 2026;13(4):481. doi:10.3390/children13040481 ↩
  9. Penzias RE, Ohuma EO, Odedere O, et al. Neonatal unit human resources: coverage for six cadres and trends for staff-to-baby ratios in 65 neonatal units implementing with NEST360 in Kenya, Malawi, Nigeria, and Tanzania. Hum Resour Health. 2025;23:64. doi:10.1186/s12960-025-01031-1 ↩
  10. Collins GS, Reitsma JB, Altman DG, Moons KGM. Transparent Reporting of a Multivariable Prediction Model for Individual Prognosis or Diagnosis (TRIPOD): The TRIPOD Statement. Ann Intern Med. 2015;162(1):55–63. doi:10.7326/M14-0697 ↩
  11. Silva GFS, Wichmann RM, Silva Junior FC, Chiavegatto Filho ADP. Development and evaluation of machine learning training strategies for neonatal mortality prediction using multicountry data. Sci Rep. 2025;15:24278. doi:10.1038/s41598-025-04066-5 ↩