Read codes have been in use by the NHS since the 80s. While it has played a useful part there have been issues with the system. Snomed CT will replace it and be implemented throughout the UK in 2017
Read codes, a coded thesaurus of clinical terms, have been with us since the earliest days of computer systems in primary care. They have been added to and adapted over the last three decades and incorporated new diseases such as swine flu and SARS (severe acute respiratory syndrome) as well as the rigors of the quality outcomes framework (QOF) or admissions avoidance care plans. Some have become etched into our brains and other, more bizarre codes involving spaceships or bathtubs have provided light relief.
With perhaps surprisingly little ceremony, the last ever set of the version 2 read codes – used by the majority of practices – were issued in April and will be officially retired from practice computers by the end of the year. The replacement system, called Snomed CT, will be the basis for QOF submissions and enhanced service claims in March 2017, so why are theses changes being made and what differences will it make to your practice?
When read codes were first devised computers were much less capable than they are now. At this time, Mark Zuckerberg, the creator of Facebook, was just coming up to his second birthday. Computer memory and hard discs were expensive and this was reflected in the design of read codes. There is a simple structure to the codes themselves, which are a maximum of five characters long. While this could, in theory, allow nearly a billion possible codes we are running out because of the second feature of the codes. They have a hierarchy based on the code itself; similar concepts have similar codes. Some sections of read codes are full while many others are completely empty.
There was an attempt to solve these problems with version three of read codes (CTV3), which is currently used by the System One computer system. However, despite its improvements CTV3 was still largely anchored in primary care and was quickly merged with the American Snomed RT system to become Snomed CT and provide a more comprehensive terminology for all of healthcare.
In the UK
Snomed CT is an international collaboration and is used in many countries although, UK general practice will be one of the biggest implementations. Despite its international base its roots in read codes and extensive customisation for the UK should allow a reasonably smooth transition. Some GP computer systems are already using Snomed in the background alongside read codes. Snomed will cover both primary and secondary care, so it should make it much easier to share data throughout the health system.
Eventually hospital discharge information should be as easy to enter into practice records as blood test results already can be.
Because of its more modern structure CTV3 will continue to be supported for another couple of years before it too is replaced by Snomed.
The conversion to Snomed by the computers should be entirely automatic. The Health and Social Care Information Centre (HSCIC) and its predecessors have spent a great deal of time making a map to translate between the two systems.
This is not a particularly simple thing to do, largely due to the complicated patchwork that read codes have become. In particular read codes used synonyms – different text terms that were both attached to the same code. These were not always strictly synonymous and may be mapped onto different Snomed codes. For instance, read uses the same code for top manager and top administrator while Snomed has separate concepts.
The need to transfer all read codes means that some anachronisms persist. In the same area the synonym business – top men becomes its own concept in Snomed. There is no similar concept for women.
More practically there can be cases where two read codes with similar meanings become a single Snomed concept. In the early days of QOF patients with type 1 diabetes were coded with C108. Later the appropriate code was C10E. They will be merged into a single concept in Snomed, which should make life a little easier for practices – if a patient’s diagnosis is coded C108 they will now reappear on the register.
The emphasis in this translation is to preserve the actual descriptions that appear in the clinical record. What appears on the screen should not change much. The text of the record will remain largely the same although the context of the concepts that the text describes could change quite a bit.
The power of Snomed
Snomed CT does make use of synonyms although the process for approval seems to be rather more rigorous than has been the case with read codes. The synonyms in Snomed are more likely to actually be synonymous. These synonyms are called terms and can cover different languages, or countries. Different sets of acceptable terms can be defined according to the country or context in which they will be used. In theory it is possible to use Snomed to translate a patient record to several other languages.
There is a huge amount of other customisation that can be done and this will be important to making the system usable. There are concepts that are core to general practice and many that we are likely to use only very rarely. An effective implementation will prioritise the former over the latter.
Read codes included much of the International Classification of Diseases (ICD) 10. With these came many terms that ended not otherwise specified (NOS) or not elsewhere classified (NEC), which were designed to be used respectively where there was no further information or no more specific concept existed. These codes will not appear in Snomed (to be precise they exist but have been inactivated). Similarly codes that state unspecified will no longer be used.
In theory NOS or NEC concepts gave no more information than was already available one level up the hierarchy, although the constraints of read codes meant that this frequently was not the case. The improved hierarchy and range of concepts in Snomed should mean that there are genuine alternatives available.
Translation to the new concepts and codes is the minimum that practice computer systems can do to implement Snomed. All of the benefits of communication between care providers and the improved range of concepts will be available at this level of implementation. However, Snomed has a lot more to offer as more of its features are implemented.
As well has having more concepts than read codes, Snomed has a more sophisticated structure that should make things easier for users. That may feel like a contradiction but it is analogous to a modern computer tablet, which is both vastly more complicated and simultaneously much easier to use than a 1980s PC with a disk operating system (DOS) prompt. In a similar way much of the mechanics of Snomed will be hidden but provide a better experience for users.
Many GP computer users will have read codes memorised or written in lists next to the computer. This is fairly easy when a code is at most five characters long. Snomed codes can be up to 18 digits long and thus are pretty useless to humans. They are simply too long to memorise or to type in reliably. The meaning of the concept should be clear from its description and its relationships to other concepts.
Snomed groups types of concepts together into main branches such as observations, procedures, organisms or substance. The full description for each concept will contain the name of the branch in brackets at the end so it is easy to tell the difference between depressed mood (finding) and depressive disorder (disorder). Additonally, to see that the latter would qualify as a diagnosis for QOF but the former is simply a clinical observation without defining a formal diagnosis.
The relationships in Snomed are more helpful as well. Each read codes concept was the ‘child’ of one other concept. For instance pneumococcal pneumonia was a type of pneumonia. This meant that it was fairly easy to search for all types of pneumonia. This relationship was built into the code – pneumococcal pneumonia had the code H21 with pneumonia being H2 and respiratory disease having the code H. The code became shorter as you worked up the tree. These are sometimes described as ‘is a’ relationships as pneumococcal pneumonia is a pneumonia, which in turn is a respiratory disease.
The relationships in Snomed are defined separately to the numeric codes. These codes don’t carry any meaning about the concept itself. They are like the bar codes on supermarket products – unique identifiers but they can’t tell you whether you have breakfast cereal or toilet cleaner without looking up the database.
Snomed takes relationships a stage further by allowing each concept to have more than one ‘parent’. Pneumococcal pneumonia is now not only pneumonia but also a pneumococcal disease and will be picked up by a search for either. Unlike people, concepts can have as many parents as they need.
MMR vaccination, for example, is a measles, mumps and a rubella vaccination. Under Snomed it is possible to easily see all patients who have had a rubella vaccination whether this has happened via MMR or single vaccinations. This is just part of a larger network of relationships that concepts can have although the details of those are beyond the scope of this article.
Medicine has always evolved, not just in the treatments that we give or the evidence that we use but also in how we view diseases. Infectious diseases arise and, particularly in mental health, other diagnoses are no longer helpful or reflect current thinking. Snomed has a way to mark old concepts as no longer active and can even give hints to concepts that would be more appropriate. This will enable us to have a relevant coding system at all times and avoid the build up of out-of-date concepts that clutter up read code. The same is true of the relations between concepts that can be updated without needing to change the practice record at all.
The transition from read codes to Snomed has received a large amount of planning and it is expected to be a smooth one. It won’t be completely invisible to practices as read code synonyms are mapped to different concepts or separate read codes are merged into a single Snomed concept.
This will be most apparent when these are codes used by rule-based searches. Miquest, an approach to common data access, will not work with Snomed and this will in turn affect programs such as Primis, which extracts, collects, analyses and interprets clinical data from patient records held by GPs, that rely on it. More significantly it will affect the automatic extractions for the QOF and enhanced service payments. The business rules, which govern these, rely heavily on the current read code hierarchy. A single higher-level code and all its descendents is often specified, particularly when compiling disease registers or other denominator populations.
Practices would be wise to keep tabs on their recorded achievement against these rules before and after the switch to Snomed towards the end of 2016.
Gavin Jamie, full-time GP in Swindon with an interest in health informatics. He runs the QOF database website.