Tackling the data conundrum: Normalising data across NHS Trusts with OMOP
Written by Paul Martin, UK Public Sector Healthcare Lead Cognizant
The lack of a standard data model is holding back potentially game-changing clinical research in the UK. The OMOP model offers a promising solution.
In theory, the data entrusted to the NHS could solve all kinds of pressing public health issues. One example is if we want to understand the impact over time of different Long Covid treatments, the insights are hidden across the data collected by individual Trusts and GPs.
But, as every clinical research team knows, extracting genuine insights from large volumes of healthcare data is rife with challenges. One persistent stumbling block is the lack of a common data model (CDM). With systems and Trusts employing different vendors with their own model, the difficulty of standardising data and vocabulary from different sources means many AI and ML projects struggle to deliver their true value.
One solution that is gathering ground, with organisations including Genomics England, HDR UK and NICE all using it to standardise large-scale anonymised data for research and analytics projects is OMOP, the open-source common data model (CDM). recently had the pleasure of showcasing the possibilities of OMOP for UK healthcare researchers at Digital Leaders Innovation Week along with our partner Sensyne Health — and these are some of the key points from that session.
OMOP takes the heavy lifting out of data curation
OMOP brings an elegant solution to the challenge of fragmented data across the UK healthcare ecosystem. While EMRs have gone a long way towards making large-scale population data available for clinical studies, much of the data that clinical researchers sometimes need reside outside of EMRs or across several EMRs and departmental systems.
The result is that even when data is fully anonymised and approved for research use, analytics teams still spend 80% of their time consolidating it, standardising it and preparing it for analysis, leaving only 20% for the research.
OMOP addresses that challenge by doing that curation work for researchers before they start. It brings together datasets from different sources and institutions and applies common vocabularies, allowing researchers to simply select the specific cohort they need for their study. As an example, a research team can select a UK-wide cohort of patients with diabetes, without having to worry about stitching together data from SNOMED, ICD-10 and other source systems.
OMOP accelerates time to insight for Sensyne Health and its NHS Trust partners
Cognizant partner Sensyne Health is using OMOP today to vastly accelerate research projects for the 13 NHS Trusts it works with. Sensyne uses AI and machine learning to generate insights from massive quantities of anonymised patient data — for example, to help NHS Trusts to manage ICU resources more efficiently.
Rather than creating a new data warehouse or data mart for every new project, it is moving to OMOP as a pre-existing standardisation layer for the data, enabling its research teams to quickly get to work on building machine learning models to extract insights from it.
The beauty of OMOP, according to Sensyne Health CIO Alan Payne , is that it’s open-source, meaning that its user community is continually building tools to allow researchers to build sophisticated cohorts based on hundreds of criteria, rather than just a handful. And because it observes strict data federation principles, the underlying data stays in each contributing Trusts’ own environment, rather than being pooled in a cloud data lake.
Secondary care stands to benefit most from OMOP
Alan Payne believes that secondary care is the area of healthcare that can benefit most from OMOP. It enables NHS Trusts to identify trends and patterns much faster, and thus find faster and better solutions to urgent issues like reducing wait times for surgery.
Trusts that are currently looking at building one or more data marts for specific studies should consider using OMOP as an alternative, he says, as its open-source means no licensing fees, and its general-purpose design and wide range of vocabularies means it can be used and re-used for any type of study. And because it can ingest data from emerging and non-traditional sources, like connected devices, genomics and imaging, its value will only increase over time.
See OMOP in action at HDR UK
One of the best ways to get an idea of the power of OMOP to create very sophisticated cohorts is to try it out. HDR UK, the national institute for health data science, offers an online Innovation Gateway portal , which allows researchers to discover and request access to anonymised UK health datasets that are fully permissioned and clinically approved for research. OMOP provides the standardisation layer and toolset behind this portal, so it’s a great way to assess its capabilities.
At Cognizant, we believe that OMOP represents an extremely valuable common data model that promises to unlock the full value of UK healthcare datasets — not just for clinical research, but also to identify potential efficiency gains and cost savings in healthcare delivery. If you would like to know more about how you could put OMOP to work in your own research, please do get in touch .
Originally published at https://digileaders.com on January 5, 2022.