2: Knowledge Structure

All knowledge is structured. Radiation Oncology knowledge is structured, and very highly structured to boot. All trainees are inculcated into a strict way of thinking which is assessed in the final exams (so the sooner you get it the better!).

The pattern of

Diagnosis > Stage > Biological Prognostic factors > Treatment Intent > Therapy Choice > Therapy Preparation > Therapy Delivery > Follow Up (cycle until recurrence at which time you return to stage)

is well known to you. And while it is a workflow, it also mirrors the knowledge flow. For example, there is no point staging some one without cancer! (Duh! It's so obvious you say!) If you are in Therapy Preparation, ther must be data that clearly defines the patient's particular Diagnosis, Stage, Biological Prognostic factors, Treatment Intent and Therapy Choice.

Not only is there a structure to our knowledge but there is also the recurring theme of defining the structure of this knowledge. This knowledge structure may take the form of vocabularies (words to use), nomenclatures & classifications (e.g., ICD10) and ontologies.

They all share common problems.
  1. Purpose of the knowledge structure will determine the way that the structure develops. The problem with many of these efforts is that the people undertaking the classification is not the same group who are required to apply the classification. In hospitals, the classification of medical and nursing work is not undertaken, by and large, using classifications built by nurses and doctors. Because we are happy to document in free text which is of little use in analysis without recoding, the classification schemes have been built down stream of the body of the profession by people who want to analyse the data, predominantly governmental and managerial efforts. Unfortunately, these people do not understand the knowledge structures of the medical and nursing domains well enough to produce a classification system that is clinically relevant and useful.
  2. All classifications need to be maintained. The process of maintenance also has ramifications. The normal paradigm is that a professional body undertakes the task by forming an administrative empire devoid of the end users (they are called in episodically and infrequently) and diluted by vested interests (these interests want the data, and want it in their own special ways not having learnt to understand the clinical data and then work out how to suck it back into their system). The paradigms of knowledge structure used are usually basic and rudimentary. Many items of importance are left out because the non-clinicians are not interested.
Which is best?

The best knowledge structure will be the one decided on by the domain experts. In the case of Radiation Oncology, who is the domain expert? We have a specification of a "standard" nomenclature for planning by the ATC (Santanam, et al) that describes the naming pattern of contours and volumes. That's fine, but there isn't an oncologist mentioned in th report! Now last time I looked, defining and naming structures was my responsibility, not that of physicists.

Does their schema make sense? Well, it does in parts, but not in others. And you know what doctors are like, if they didn't agree to it, and it doesn't make complete sense, they won't use it. Even the definitions of ICRU50 (1993!) and ICRU 62 (1999) are not absolutely entrenched in use (and who predominantly defined those things?)

I am writing in very black & white terms to reinforce a point. If we oncologists don't get involved in the proper classification of our knowledge, someone else who wants to know some radiation oncology things will do it badly for us.

The best definition of the expert domain knowledge structure is an ontology. This is not a mispelling of the word 'oncology', and unfortunately even though 7/8 of the letters are identical, expert status in one means nothing in the other (that statement works both ways!)! Adding to the problems are the structure of the ontology which is convoluted and not like anything you have seen in Medicine. And no one has really demonstrated a good way to have oncologists verify the knowledge structure. All bar one previous attempt has been achieved by processing oncology text. This has ended up with 'palliative radiotherapy', 'photon radiotherapy', 'large field radiotherapy', 'hypofractionated radiotherapy' and 'nodal radiotherapy' all being grouped together. If you asked these people whether 'yellow car', 'rusted car', 'dented car', 'rental car' and 'Toyota car' should be all lumped together, the answer would be an obvious "NO", but lack of domain knowledge in radiotherapy prevents this error from being appreciated. There is a large tome on this here if you are interested.

What should I do?

Doctors in general say and believe that they are interested in their data. However when you look at the quality of data that you meet in a retrospective review (and I encourage you to do one to work out just how bad it is), you have to doubt this belief. Our Information Systems are generally unhelpful in storing our knowledge in a way that means a retrospective review only requires a database report. We generally do not persist with electronic systems to see if they can be made to work (the salesman wants to take your money, so he will tell you any rubbish about what their system will do, AND he has no specific knowledge of oncology to be able to show you how it works and how it can save time). So most clinicians stay with paper until they forced to use the electronic equivalent of paper.

Through all this, the profession is subject to what others want to do, and then try to impose it. The challenge is for you to educate yourself on these issues and combine with like minded colleagues to develop a knowledge structure. There are people working in this area both in Australia and overseas. There is a Journal of Radiation Oncology Informatics.

You don't need to be able to code to be of use in Radiation Oncology Informatics.

Topics

Various knowledge structures related to radiation oncology will be discussed in detail below.

i) Coding

ii) Knowledge Structure of Treatments

Links