A data revolution is sweeping healthcare and life sciences, offering new opportunities to improve patient journeys and operational agility. For hospitals, this means moving beyond fragmented data to unified analytics, enabling data-driven decisions that enhance care quality, reduce costs, and foster groundbreaking research for a healthier future.
And yet, there is still an issue with data in Irish healthcare, said Philip Schenck, technology manager data and artificial intelligence (AI), at OpenSky Data Systems.
“Put simply: there are still silos,” he said.
IT systems built over decades for continuity often create legacy issues and data silos, leading to: delayed decision-making, clinician burnout from inefficient information retrieval, compromised patient safety via incomplete data, and missed research or operational improvements.
Change is coming, however, as Schenck notes hospitals are starting to embrace data democratisation.
“What they want is to use data analytics to get a holistic view of what is going on in terms of hospital operations, patients, etc, so they really do see the need to modernise. I would say it is a hot topic in the sector,” Schenck said.
OpenSky helps Irish hospitals address this by creating Unified Data Analytics Platforms using the Microsoft ecosystem for reporting from a single source of truth. Confidentiality is crucial, including compliance with EU regulations like GDPR, NIS2, and the EU AI Act.
With semantic models, I can see how the data is defined and calculated, which helps me confirm that the reports are accurate
Key to OpenSky’s work is Microsoft Fabric, an end-to-end analytics platform unifying data movement, processing, ingestion, transformation, real-time event routing, and reporting.
“When the data is pulled into Microsoft Fabric it is essential to ensure that compliance is taken very seriously.
“This means using the ‘least-privileged’ principle, so only people who need it have access to the data. However, with Fabric the data is all kept in the same data lake, known as a ‘OneLake’, but we create workspaces as containers for departmental access that allow different people and groups to have access and implement role-based access at the workspace and dataset/report level for different levels of access,” he said.
For instance, with Fabric, clinical and finance data can be accessed by relevant departments or by management needing both, crucially without data duplication. This supports Schenck’s point:
“In a sense, we are all trying to report on the same data, but we are trying to answer different questions. By giving access to shared workspaces, we can make sure everyone is working with the same trusted data and doing their calculations from a consistent source. As the manager, I can either see the results myself, or request a report showing who used the data and how,” he said.
Data lineage helps authorised users trace how data flows from source systems to final reports.
“With semantic models, I can see how the data is defined and calculated, which helps me confirm that the reports are accurate,” Schenck said.
Sometimes, however, a different data type is needed: synthetic data. Synthetic data, artificially created by algorithms rather than from real events, mimics the patterns found in real-world datasets without using actual patient records.
“Synthetic data can be an important tool in terms of predicting power in smarter hospitals. There is a risk that, without synthetic data, you could put patient privacy at risk during development,” said Schenck.
“Synthetic data is a game changer. You already have these trends, so you can create synthetic data created on the trends [but] using dummy patients,” he said.
OneLake acts as a secure, unified foundation for all data workloads in Microsoft Fabric—covering ingestion, transformation, analytics, and AI. It simplifies governance and enables diverse teams to work securely on shared data.
However, the practical implementation requires care.
“What we see, and we see it very often, is organisations say they want to work in an agile manner, but at the end of the day you typically see a hybrid mode: an agile methodology within a waterfall development. Yes, it delivers in an agile manner, but nothing is really being productionised until you hit the ‘go live’ button,” said Schenck.
OpenSky promotes a truly agile model for rapid delivery and immediate value-add, Schenck added.
“You can get value quickly and build on top of that, meaning users get what they need, when they need it, and their work is aligned to company strategy,” he said.
Roadmaps can then evolve based on what the organisation learns along the way. “When you fall into the trap of the hybrid model, you find that user or business needs have moved on by the time you hit the go-live date 12 or 18 months on; that's a problem,” Schenck said.
For a data and AI roadmap, Schenck says that while hospitals can define requirements for later analysis and roadmap delivery, productive collaboration is preferable.
“I would say the better method is the collaborative one, where we work together with the hospital, with subject matter experts, to define it. The roadmap need not lead you twelve or eighteen months down the road; it could just be three months, or even six, and you deliver iteratively to meet that goal,” he said.