NEODATA+: Connecting NICUs Across Borders to Turn Fragmented Vital Signs Into Life-Saving Algorithms

How did NEODATA+ come about?

Neolook started with a fundamentally different model from conventional medical innovation. It was actually an idea born at Philips. Philips was the number one patent filer in Europe at the time - a strong company. They would invest a billion per year in pipeline innovation: fundamental research, then clinical research, then they fund the pipeline further, further, further, and after a number of years the product comes out into the market and may have success or fail. That's a linear form of innovation. We flipped that model entirely, because it doesn't really serve neonatology and paediatrics. It doesn't work well for them as pediatric innovation is neglected and underserved by industry in the first place.

What we do is called "bedside to breakthrough." We walk the floor on neonatal intensive care - in the hospital, in the department, with the nurses, with the doctors, with the suits on, with the masks on. We work with them to see their day-to-day reality. And then we ask them: what do you need today? What do you need tomorrow? We write down what they say verbatim - the quotes, literal quotes - and those quotes we use to innovate. So instead of pipeline thinking, you have the verbalisation of a real lived NICU, and that wording is what we put on the roadmap.

The next step, when we find these unmet needs, is to check if more hospitals have the same problem. If they do, we put it on our agenda. We did interviews with all the Dutch hospitals, and we ended up with an innovation agenda created by the hospitals themselves.

And that's where the idea for NEODATA+ was born. We looked at it and thought: we could apply the same process in other countries. So we started to look abroad, to find accelerators to get the project going.

What are the concrete benefits of being part of UNITE?

UNITE gave us the framework and the opportunity to make this real at European scale. The whole logic of Regional Innovation Valleys - connecting strong innovators with emerging regions - is exactly what this project does. The Dutch, Italian, and Romanian partners are building something together that none of them could build alone.

Our initiative actually has two European branches: one going through UNITE to Italy and Romania, and another parallel project called NeoVitAI with Germany. Both are expansions of the same core idea — a spine dataset that every NICU must be able to produce, and that can be shared to drive algorithms and AI.

How does it work, in practice?

We pull together a leading Dutch hospital (UMC Groningen), an Italian hospital (Policlinico San Matteo Pavia), a Romanian academic institution, Tito Maiorescu University and a Romanian hospital (Filantropia Clinical Hospital), and we start with what we call the "spine dataset" - a basic dataset that is so common that every NICU must be able to produce it. These are the vital signs: heartbeat, breathing, temperature, SpO2. Collected at high frequency — maybe once per minute or once per second, depending on what each site can deliver.

The dataset anchors on apnoea, bradycardia, and desaturation events - what clinicians call ABD. These are episodes of cardiorespiratory instability that happen frequently in premature and critically ill newborns. Every neonatologist knows them.

But the data can do more than just detect ABD. We can look upstream - to pain, stress, and discomfort, because those are often the predictive triggers. And we can look downstream - to neurodevelopmental outcome. So one dataset, instead of doing one thing, can do three things: it can detect and predict ABD events, it can support the objective assessment of pain and stress, and it can be connected to long-term outcomes.

Take heart rate variability as an example. If you look at a heart rate reading, you'd say: great, looks fine. But if you look at the tiny variability - the small differences between one beat and the next - that's the real predictor. The signals are more subtle, but they're incredibly informative.

What role does each partner play?

Neolook provides the technology and innovation platform. UMC Groningen - this silent powerhouse in the north of the Netherlands - is the scientific leader. They serve three provinces and they're very strong in data, in big cohorts, longitudinal and intergenerational cohorts.

Then there’s Policlinico San Matteo hospital from Pavia. The Italian connection came through Burke and Burke, a medical device company based in Milano. They told us about a tender at Policlinico di Milano, we participated, and they won - partly because of the Neolook solutions. So when this project came along, I went to them and asked: would there be any hospitals interested? They said: Pavia could be a good fit. We went to visit, we got a good click, and that's how Pavia came on board.

Then we needed a third region. We were looking at Spain, Greece, Macedonia, but for some reason Romania, which we knew from a previous collaboration in the EU Horizon project BornToGetThere, raised a hand. One person in the call, two people in the call, one hospital, one University - Filantropia Clinical Hospital Bucharest, and Titu Maiorescu University - they were really on the ball. So you end up with a top hospital from the Netherlands, a strong one from Italy, and two partners in Romania. That's a nice train, as we call it.

We provide the technology. They provide the brains — very educated doctors and scientists — the patients, and the data.

How many patients will be involved in the project?

We do that with a small group of patients from each country.

But here comes the interesting part. Why is there a gap in innovation for neonatology? Because the departments are too small. An average NICU has 20 to 25 beds. The data power from a single unit is too small to develop robust algorithms. But if you combine two departments, you have maybe 40 to 60 beds, and you break the first statistical barrier. If you combine three departments, you break through to a qualitatively strong study - 50 to 60 patients - and the data becomes really powerful.

The key point is that these departments all collect data in a similar way. The data is highly comparable. Combining nurses’ and doctors’ knowledge of patients with scientific innovation and expertise gives you access to deeper, more comprehensive insights that do not emerge when data is treated in isolation.

You've said the biggest challenge is not technical. What is it?

Before we can do any of that exciting analytics, we need to understand how to do that in way that is mindful of EU regulations. These rules are there for good reason, to safeguard privacy and patients’ rights but the other side of the coin is that they can slow down innovation.

If you stack up the problems - data privacy, sharing of data, pseudonymisation of data - and then in each country the medical ethical committee has veto power... it gets complex and sluggish.

And the challenges go deeper. You need consent from the parents - that's a big one. But also: how long can you save the data? Who has access? It's very difficult to anonymise 100 percent. I'll give you a pragmatic example. A friend of mine had a rare disease. I knew the disease, I knew the area where she lived. Even with an anonymised dataset, I read a scientific publication and it occurred to me this was about her without me even seeking to know this. Being able to figure out who the patient is without intent. Now imagine that with newborns in a small NICU, even years and years afterwards.

So the question is not just about technology - it's about making the rules work across jurisdictions, across ethics committees, across legal systems. That's the real breakthrough we're trying to achieve.

How does the European Health Data Space come into play, here?

We are in a very regulated area. We have multiple frameworks: GDPR, ISO 27001, NEN 7510, MDR, ISO13485, NIS2 and AI Act. And then there is the EHDS - that's a new one that we're going to adopt. For us, this is one more framework, and it's one that everybody wants, because it creates a common European standard for health data sharing.

What we're trying to do with NEODATA+ is show that all of these frameworks can work together in practice, for the most sensitive patient group in healthcare. If we can solve the governance for neonatal data - where patients can't consent, where parents are under extreme stress, where the ethical scrutiny is at its highest - then we create a model that can be reused across Europe for many other clinical domains. It's very step by step, but it's very scalable.

What is the current status of the project?

We're on. We're in full motion.

The thing that's lagging is, again, the regulatory part. The contracts are not in place yet. But we're doing it anyway. We're moving forward on the technical and clinical preparation while the legal framework catches up. If we waited for every contract to be signed before starting, we'd lose months we can't afford to lose.

What will success look like in 18 months?

In eighteen months, we will have three hospitals in three countries with standardised datasets on a small patient cohort. The data will sit in separate, governed servers, locally - each hospital keeps its own data. We will have proven that it's all regulatory-proof: privacy, medical ethics, everything in order. And that the method is replicable.

And then we will have done a test run with AI. An algorithm that might reach, say, 50 percent accuracy if trained with data from just one hospital. That would reach 60 to 70 percent with two hospitals. But it reaches 90 percent - the level you actually need - when you combine three hospitals.

As for the data governance framework, once solved, it doesn't benefit just one project. It can scale to any NICU and pediatric department that wants to join. And if we can solve it for neonatal data - the most sensitive domain in healthcare - it works for almost everything else.