Part of SNOMED CT PaLM Mapping Best Practice
Mapping creation, assurance, and governance
The mapping workflow involves creation, review, and maintenance. For a given clinical domain, it is recommended that the review and assurance of automated mappings is undertaken by subject matter experts (SMEs) and that all information regarding review and ownership of mapping is captured.
For the pathology and laboratory medicine terminology mapping requirement, as labs are ultimately responsible for their data, they must clinically assure their own maps to SNOMED CT PaLM. This means they would always be involved in the review, but other parts of the process could be performed or supported, in part or in full, by third parties.
Roles and responsibilities
The services or people required for a mapping project could include:
- mapping management
- clinical and lab IT SMEs
- mapping support service
The involvement and capacity of mapping management and the mapping support service can be minimised in proportion to the capabilities and functionality built into mapping software. This will empower the lab user to drive the process.
Responsibilities of mapping management
The aim of this is to create the maps and manage the process from start to finish. Ideally, it should minimise the workload placed upon clinical SMEs so they can focus on assurance and clinical sign off.
Pathology lab IT managers are well placed to provide mapping management as they understand local processes and procedures and how these are reflected in their labs’ reportable data. Alternatively, a central NHS England service or commercial vendor could provide it.
Responsibilities of clinical and lab IT SMEs
Pathologists and biomedical scientists currently working in labs are required to review and assure maps created and provide clinical sign off.
If existing maps to Read PBCL are used as a strategy to map to SNOMED CT PaLM, they are able to provide assurance that these maps are current and correct.
Responsibilities of mapping support service
To deal with the technical demands of more complex mapping, commercial mapping software may be required. The vendor may also offer a mapping support service, providing custom algorithm generation and custom workflows to facilitate mapping and subject matter expertise in terminology mapping.
Change management and version control
As described in Mapping maintenance and implementation, maps will change over time. 2 elements of functionality that support this are change management and version control. Change management supports clinical audit, as it enables users to see changes to maps, when these were made, and by whom. Version control underpins change management and supports the production process between authoring, review, publishing and maintenance, as users can see the status of a map at a particular point in time.
Managing the mapping process
This section explores some of the issues involved in the mapping process via examples of using 2 different techniques to map a lab’s local reportables to SNOMED CT PaLM.
Managing the process of using referential terminology maps
This process involves a chain of referential terminology maps. Firstly, local reportable to Read PBCL maps, then NHS England’s assured Read PBCL to SNOMED CT PBCL maps, and finally SNOMED CT PBCL to SNOMED CT PaLM maps to reach appropriate SNOMED CT PaLM map targets. This requires an appropriate level of assurance to ensure reliability and minimise errors.
As a chain of maps are key to the output, as seen in the diagram below, this increases the requirement for manual and technical assurance by SMEs at appropriate steps to address sources of error.
Local code to Read PBCL maps
The most likely sources of error are the maps between local codes and Read PBCL. During testing, the PaLM Mapping project team found both duplicate local codes and variation in local code to PBCL mappings between Pathology Networks. As noted in Historical mapping, errors could have been introduced at inception, particularly if the curation and assurance process was not captured. Further variation may have been introduced if either informatic or clinical processes have changed since the maps were created. Consequently, the maps might not reflect current practice and will require clinical review.
Read PBCL to SNOMED CT PBCL maps
As the NHS England assured mapping table from Read PBCL to SNOMED CT PBCL is used in this step, it should not require clinical review. However, it may require technical review to ensure mapping has occurred correctly.
SNOMED CT PBCL to SNOMED CT PaLM maps
One to one mapping between SNOMED CT PBCL reportables and SNOMED CT PALM reportables is not always possible. As an example, SNOMED CT PBCL content often uses the unspecific term 'level' to represent a lab test property, something interpreted and represented in more granularity in SNOMED CT PaLM. A typical consequence of this is SNOMED CT PBCL reportables could map to either 'substance concentration' or 'mass concentration' SNOMED PaLM reportables, as per the example below.
Consequently, some of these ambiguities require resolving via UoM data (or other supporting data), whilst others will require clinical subject matter expertise to resolve.
The mapping process would therefore involve:
- load local code to Read PBCL maps
- identify local code to Read PBCL mapping anomalies*
- validate local code to Read PBCL maps to resolve anomalies with SMEs
- apply Read PBCL to SNOMED CT PBCL maps
- apply SNOMED CT PBCL to SNOMED CT PaLM maps
- use UoM and other supporting data to resolve ambiguities*
- use clinical SME to resolve ambiguities
- review output with laboratory network SMEs
* These validation steps could be performed algorithmically.
Managing the process of using terminology component building blocks
This process involves using source data to create terminology component building blocks, to generate SNOMED CT PaLM-like strings for ingestion into a mapping tool, that will in turn map to SNOMED CT PaLM targets.
In this example, the source data is gleaned from:
- Property – lab source data
- Component – local code description
- Specimen – mapped Read PBCL/SNOMED CT PBCL description
Whilst this method is generally more reliable than relying on using referential terminology maps alone, there are some instances where it can prove problematic, requiring manual and technical assurance from SMEs.
Missing specimen type - Where a specimen type cannot be derived from the Read PBCL/SNOMED CT PBCL description, expert domain knowledge is required to predict the mostly likely option. For example, the Read PBCL code 'Epithelial cell count' (4KH0.) is generally measured in urine, whilst the Read PBCL code 'Lymphocyte count' (42M..) is generally measured in blood.
Where there is an incorrect map between the Local code and the Read PBCL code - This requires clinical validation and fixing.
Where the mapped Read PBCL code is broader in meaning than the local code (for example, certain microbiology reportables) - This requires clinical validation and fixing.
The mapping process would therefore involve:
- collate all required source data
- cleanse the data
- rectify any data issues and combine into ‘master codes’ with SNOMED CT PaLM-like strings '{property} of {component} in {specimen}'
- load ‘master codes’ as source data
- map to SNOMED CT PaLM
- use clinical SME to resolve ambiguities
- review output with SMEs
Last edited: 22 May 2025 4:59 pm