Separating data from applications that use it

I’d like to share work we are doing to shape the information architecture of the NHS.  I joined an eclectic group of experts from the UK and US to discuss how we can achieve the policy commitment to “separate the data from applications”.  In effect, to unlock the data held in operational systems (care records, rostering, HR, finance etc.) and make those data available to be use in other applications or for analytics.

The key is to move from the range of syntactic interfaces designed for specific use cases towards a common semantic record format secured in containers called online data stores.  A person’s record (care record and duty roster) could be copied into one or data stores (perhaps one care record data store for each care provider).

How would all this work – a common data model, a common set of ontologies and a common design for the data containers.  We already have the candidate ontologies, such as SNOMED CT, dm+d, UCUM.  There are candidates for the care record format though the group concluded FHIR was not a natural structure to persist/store data.  FHIR is a candidate for the APIs that applications would use.

The three days were wonderful, exhausting and thought-provoking.  Its clear that we can make progress but that much still needs to be done.  In many cases, the meaning locked in a care record depends on the templates used to record a patient’s circumstances and the care planned and given.  The configuration of the record systems is going to be as important as the records. 

Micro-clinical models

For us to integrate care for each patient we are going to need to share forms, scales, assessments, templates, archetypes etc. across the health and care system.  Some of these will be large structures, with the PRSB Core Information Model possibly being the most inclusive; many more will be small and specific, such as a height measurement used as a component of a body surface area and body mass index calculation.  Each of our clinical systems holds these reusable components in diverse ways.  Understanding, cataloguing and mapping the diversity will depend on a federation of catalogues working as one.  We’ve been demonstrating that capability with our new tooling for the NHS Data Model and Dictionary, Mauro Data Mapper. 

The last couple of years have been transformational for the NHS and digital technology.  I’m continuously amazed by how NHS Digital’s Information Representation Services team delivers and how we find our products are at the heart of NHS operations.  A good example is finding that our Bronze FHIR API for organisational data is being used in the COVID passport application.  As with many of our products and services, its only when things go wrong that we are valued.  In that case, the COVID passport app failed when the API was overloaded earlier this week.  The service is being uplifted to Platinum and we’re looking at alternative ways of providing it.