Contact us Request demo
Link to Home

View navigation

Graphnet Health completes major population health analytics transformation, paving the way for faster NHS insights

18 November 2026

As NHS organisations increasingly rely on population health intelligence to plan services, identify patients at risk and support more proactive care, the technology underpinning that intelligence needs to keep pace.

Graphnet Health has completed a major transformation of its population health analytics platform, moving its NHS partners to a Delta Lakehouse architecture designed to support growing data volumes, faster analysis and more advanced analytical capabilities.

With SQL queries now running at least 66% faster, the transition marks an important milestone in Graphnet's population health journey. It also provides a foundation for new capabilities, from neighbourhood-level analytics to predictive risk modelling.

But delivering the transformation while preserving live services was no small undertaking. Here, we share what prompted the change, what we've learned and what it means for NHS data and analytical teams.

The case for change 

Population health analytics is central to NHS planning and delivery. Both the NHS 10 Year Health Plan and the shift towards neighbourhood care have increased the need for timely, integrated analytical insight and the data is big! From our own platform insights, we knew that we had to change because

  • Demand is growing: We saw that our user access has increased by 40% in the last 12 months.
  • High‑volume longitudinal data size is increasing: Event‑level histories across millions of patients and billions of records. Over the last four years, record volumes have increased by 134%, taking the platform data beyond 80 billion records.
  • Our health and care partners want advanced analytics: Risk stratification, segmentation, clustering and machine learning driven prediction are now standard techniques. (and you can’t have a technical blog these days without mentioning the rise of AI!)
  • The data is so diverse and we need to get innovative with the way we store and process heterogeneous data sets: Primary care, acute, mental health, community, social care, genomics, remote monitoring and increasing uses for unstructured clinical text force us to process data differently according to the contents and the use cases that our clinical and non-clinical users need.

Moving away from Azure SQL

Graphnet’s Population Health journey started in January 2018 when public cloud was just getting traction. We chose Microsoft Azure as our hosting platform as we had long been a Microsoft partner. We built our platform around Azure SQL as the primary storage, with a mixture of Azure Data Factory / SSIS as the ETL engine and Power BI as the visualisation layer.

Fast forward to today and as the world has changed, so has the tech Azure SQL which has gone from being a dependable workhorse, which was structured, reliable, and familiar, to having significant architectural limitations.

Population Health outgrew the relational database

Graphnet’s mission is to create the most insightful linked data set for the NHS. Our customers linked data sets grew by 21% (14 billion data row!) in the last 12 months.

Our use of Azure SQL was starting to really struggle with these volumes. This led to:

  • Data ingestion competing with data use
  • Analytics workloads competing with operational queries
  • Rigid and costly scaling options

Implementing Delta Lakehouse has resolved these concerns.

How we did it

There was no point in replacing one monolithic platform with another, so we created a modular architecture designed to preserve familiar tools while creating more headroom for scale and advanced analytics including

  • Delta parquet open standard data storage – this gave us portability and reduced lock-in.
  • Microsoft Fabric’s SQL endpoint – providing continuity for SQL-based analysts and existing assets.
  • Fabric visualisation layer - preserving our existing reporting investment.
  • Databricks – which supports more advanced engineering and data science workloads
  • Implementing our Population Health workbench - which gives end users governed access to trusted data while creating more flexibility for local analysis.

The gain for NHS analysts and data teams is real

Although the benefits of the transition to Delta Lakehouse are not always immediately visible to end users it creates an important foundation for the future because

  • Data can both be loaded and replayed much faster.
  • The whole platform is even more reliable than before as data ingestion is not competing with data use for the same compute resources.
  • SQL queries are at least 66% faster, with larger and more complex queries showing even greater performance gains.
  • The platform now supports both standardised reporting and more flexible analytical workflows using SQL, Spark, Python and R.

Most importantly, this architectural shift creates the foundation for the next generation of population health analytics capabilities which provide data and insight to enable a shift towards proactive, preventive care.

I'm incredibly proud that we’ve just finished converting all our NHS partners to using Delta Lakehouse architecture. It’s a fundamental shift in how population health data can be stored, processed, governed and exploited by Integrated Care Boards (ICB) and their data sharing partners. It was intensive work for our team, but we kept the transition periods to a minimum, preserving live use cases and avoided interruption to operational or clinical workflows

Next generation population health analytics – what’s next for us?

Neighbourhood population health management is one area where the new architecture is critical. It supports the combination of system-wide longitudinal data with more flexible local analysis, helping ICBs and place teams identify cohorts, target interventions and evaluate impact at a neighbourhood level.

 We’ve already been able to

  • Integrate population health calculated data with our Shared Care Record such as patient segmentation and risk stratification scores meaning population health insight is available at point of care
  • Created a population health data model which is now live in multiple ICBs enabling point and click creation of analytics using Power BI. This demystifies the complexities of NHS data and makes it easier for analysts to embrace population analytics
  • Introduced the customer workspace. This gives each ICB access to its own governed Data Lakehouse environment, including the ability to run Fabric notebooks and develop bespoke analysis. One early example is the support of winter pressures intervention analysis in Kent.

And we’re currently working on

  • Predictive risk scoring for Frailty category progression and other data science initiatives both in house and working with partners (always interested in more partners)
  • Benefits analysis services driven by propensity score matching.
  • Advanced cohort creation - increasing the complexity and scope of our cohort matching engine
  • Fabric copilot as a query mechanism and augmented development for analysts

Taking our team on the journey was really important

Our Population Health team is made up of numerous specialist roles and skillsets. The move to Data Lakehouse technology involved everyone upskilling in all aspects of the transition. This has been as much a people and skills project as a technical one. The journey required:

  • Architectural and data engineering redesign
  • Upskilling of everyone in Lakehouse concepts, Python, Databricks and the rapidly evolving Microsoft Fabric.
  • Extensive cross-functional collaboration
  • Preserving live services while changing the technology platform which underpins it

We are proud of our team and the positive way they have embraced this new technology, committing to professional development.

Conclusion: It’s a journey, but it’s one built for the future

Azure SQL gave Graphnet a strong foundation for the first phase of population health analytics. Microsoft Fabric and Delta Lakehouse architecture now provide the scale, flexibility and analytical headroom needed for what comes next.

For organisations looking to strengthen integrated care coordination, improve case finding, support benefits analysis and enable more advanced segmentation and predictive analytics, this creates a platform that is ready for both today’s needs and tomorrow’s opportunities.

If you would like to explore what this could mean for your system, we would love to talk. Please contact us.