Questions about the GALEN Common Reference Model
4.1 What is the GALEN Common Reference Model?
4.2 What is it a model of?
4.3 What’s in it?
4.4 How big is it?
4.5 How hard is it to build? How is the modelling managed?
4.6 Is it finished?
4.7 Where can I get it from?
4.8 How much does it cost?
4.9 Can I change it to suit my application?
4.10 Why doesn’t the GALEN Common Reference Model cover all medical knowledge?
4.1 What is the GALEN Common Reference Model?
The GALEN Common Reference Model is the model of medical concepts (or ‘clinical terminology’) being built in GRAIL. This model forms the underlying structural foundation for the services provided by the GALEN Terminology Servers.
4.2 What is it a model of?
The short answer is ‘medical concepts’ — the ideas we use to talk about medicine. The model aims to represent ‘all and only sensible medical concepts’. It is not a repository of every kind of information used in the practice of medicine — rather it aims to represent the underlying conceptual model of medicine that we believe is shared across national boundaries — and hence, we believe, can be usefully represented in a language- and application-independent way. For example, whilst knowledge of human anatomy is represented in the model, details of the specific protocol of care within a single organisation will not be present. The conceptual model can be used as a framework for specific knowledge about protocols or decision support — indeed that’s one of the key uses of the GALEN Common Reference Model. However, representations of such knowledge use the model; they are not part of it.
Informally, if doctors agree on the medicine, you may well find it in the GALEN Common Reference Model. If, on the other hand, there’s no overall consensus, it won’t be there. For example, most doctors agree that rheumatoid arthritis is a kind of “inflammatory arthritis” so that relationship is in the model. Doctors disagree on the diagnostic criteria for rheumatoid arthritis so the diagnostic criteria are not part of the model. Doctors can talk about rheumatoid arthritis with each other without agreeing on the exact diagnostic criteria. The GALEN representation might even help them express their disagreements more clearly. Hence the GALEN slogan “managing diversity, without imposing uniformity”. We often speak of the model as being at the level of detail two specialists need to talk about medicine outside their speciality.
4.3 What’s in it?
The GALEN Common Reference Model is made up of formal expressions in GRAIL equivalent to sentences in English such as:
“Ulcer is a kind of inflammatory lesion”,
“The process whose outcome is an ulcer is called ulceration”,
“The stomach is a part of the GI tract”,
“It is sensible to talk about ulcers located in the stomach”,
“Ulcers located in the stomach are called ‘Gastric Ulcers’”.
“Ulcers located in the stomach are actually located on the mucosa of the wall of the stomach”.
Just to give a flavour, the first two statements in GRAIL are:
InflammatoryLesion newSub Ulcer.
(Process which hasSpecificOutcome Ulcer) name UlcerationProcess
A GALEN Terminology Server will contain the additional information needed to translate the GRAIL names into various languages, e.g. that the usual English expression for UlcerationProcess is ‘ulceration’. (Names in the GALEN Common Reference Model itself must be unique and unambiguous so tend to be awkward and pedantic. One job of the GALEN Terminology Server is to provide expressions users find natural in their own languages.)
4.4 How big is it?
GRAIL is compositional, so this is not a trivial question to answer – there are at least three questions:
How many elementary entities (concept representations) are there?
How many sanctions have been added, implying how many composite entities which could be made?
How many entities have been made and named for one purpose or another by the current applications, mappings, etc.?
We can give numbers for the number of asserted elementary entities and sanctions, but the more useful indication is the scope that is covered: at present, there is roughly half the coverage that we would eventually expect to find. However, there are important caveats: firstly, the depth and complexity in those areas that we have covered is greater than would be found in existing coding and classification schemes, and secondly, the system as a whole is well-developed and usable in specific areas. For example, the coverage of surgical procedures is extensive and virtually complete.
4.5 How hard is it to build? How is the modelling managed?
It is hard. GRAIL is a relatively new language, and, although it is similar to other concept representation languages, none had previously been used to build a model of the anticipated scope and detail that is present in the GALEN Common Reference Model. New techniques and tools for building the GALEN Common Reference Model have had to be developed. There are three parts to its development.
The first is a small group of GRAIL modelling experts working very closely together, developing the high-level ontology, and defining structures and styles for the model. They work directly in GRAIL, in which it takes on the order of 3 months to become proficient. GRAIL is now treated as the ‘assembly language’, and it is not expected that large numbers of clinicians will need to become proficient in its use.
The second part is a loosely-coupled collection of clinicians working and contributing separately in the model. They do not work directly in GRAIL, but in a higher level abstraction (called an ‘intermediate representation’) that hides a good deal of the complexity. Typically this group use tools that map onto the structure that the GRAIL-experts produce, but these tools only take a day or two to learn. This group typically produces large quantities of volume, but based upon, and coherent with, the overall structure that is previously defined.
The third part is the integration of these two, very different activities.
4.6 Is it finished?
Yes and No.
Yes, in that a version exists and can be licenced for use (and is being used) to support clinical applications.
No, in that it does not cover the breadth of scope that we eventually expect to find (because that work has not yet been resourced). We’re roughly half way there.
4.7 Where can I get it from?
The GALEN Common Reference Model is licensed under open source by Open GALEN. The model is delivered and used through a GALEN Terminology Server, a variety of which are available from different vendors; Open GALEN can put you in touch with them.
4.8 How much does it cost?
You can download it free from the Open GALEN web site, or you may choose to obtain it in a more packaged form – for example, with a service or training package – from 3rd parties, in which case you may have to pay.
4.9 Can I change it to suit my application?
It depends what you mean by ‘change’ (but the answer either way is some kind of ‘yes’).
If ‘change’ means ‘remove some details that I’m not interested in’, then the answer is an unqualified ‘yes’; we do expect that this will be required in virtually every case of use of the GALEN Common Reference Model. Because it is intended as a generic, application-independent resource, it contains a level of detail that is typically too great for any specific application. Tools and techniques have been developed to enable ‘perspectives’ to be created and applied: they act as a filter to change the external view of the model to suit a particular application’s needs.
If ‘change’ means ‘extend to cover areas, or levels of complexity which my application needs’, then the answer is also yes. Depending on the specific kinds of extensions required, they can be made locally using available tool-kits, or you may wish to have someone else make those extensions for you. The GALEN Common Reference Model is a much firmer base from which to build extensions than existing coding systems because it limits what can be done — if it is allowed, it will fit. Conversely, changing the basic constraints in the model would defeat the entire purpose. For example, concepts already implied by the model can be given new names or new high level organisations can be defined without any fundamental change to the model. The model can be extended to provide more detail in specialist areas. and the new developments checked for overall coherence with the model as a whole.
Another common kind of extension is to map the GALEN Common Reference Model to local nomenclatures, term-lists or other coding or classification schemes that may be in local use. This enables the application of GALEN Technology for re-organising or maintaining those local schemes, or to auto-encode into those schemes.
4.10 Why doesn’t the GALEN Common Reference Model cover all medical knowledge?
Because that would not be appropriate: the GALEN Technology offers a way to represent a clinical terminology, an underlying framework of medical knowledge describing what things are and how they are organised. It does not describe medical knowledge such as recommended treatment regimens for diseases because a) not everyone in a specific medical field will necessarily agree, and b) even if they did agree, the specific technology that is GALEN may not be the most appropriate way to represent such knowledge.
We do intend that the GALEN Common Reference Model be a useful framework on which to hang other kinds of medical information, and we can demonstrate ways in which that can be done, but the GALEN model itself cannot, and does not, contain all medical knowledge.
From the beginning we have attempted to be tightly focused on a single task. GALEN does not do everything — or as one project slogan puts it: “GALEN does not make the coffee – deliberately”.