Triage & Decision trees

How to set up solid decision trees

How to Sort and Order an unsorted situation

Triage is French.  Trier means sorting, ordening. A triage orders (and or sorts) a disordered and unsorted situation. (like trauma patients on a battle field or the ER department of a hospital.)

The first two S’es in the 5S workplace organisation method (Sort and Set) in fact are a triage.

Triage systems –and other decision trees– are commonly used to direct actions in a process.

Mostly it is a series of ‘Binary questions’ leading to a certain point, sometimes it is a simple sieve; in other cases it is a multi-dimensional ‘funneling system’ that leads the user through a series of questions to bring him to the one and only possible answer.

Remember in biology lessons, the plant determination guide? A beautiful example of a highly structured decision tree, containing an immens amount of knowledge, enabling even kids to find out the correct name of any plant or tree.

Triage – Decision trees: a form of knowledge management

A solid decision tree is difficult to make and requires deep understanding of the matter it tries to cope with. Thus making such trees can be seen as knowledge management: Whenever a tree has proven to be correct, the knowledge to make such decisions is ‘stored’in the tree… Now complex and difficult decisions can be made by less knowledgeable staff, assuming they understand the questions but this can be trained relatively easy. 

The pitfalls & Requirements

1. Ease of use

Many triage procedures are ‘written text’ and may be absolutely correct. However since our brain is much more powerful in visual handling than text handling, it would help to present the ‘decision making process’ in a visual manner.

The following example is based on a Dutch tax form.

A triage in form of a One Point Lesson First the whole process was precisely and in detail designed, including all ‘what if scenarios’. Next, this was brought into a PSD that was very much correct and complete, but also became quite difficult to read. With the now gained insights, the relevant decision points where transformed into understandable and unambiguous questions. In this way, the questions could now be presented in a visual manner to the final user.

In the design-state, the use of PSD can dramatically ease up the insight of the process; compare those two examples:

A simple sieve in a flowchart…. ….the same sieve in PSD
A sieve displayed as flowchart A sieve displayed as PSD (Process Structure Diagram)

2. Triage Questions are unambiguous

“The patient can go to blood test” Yes – NoDoes this mean “The patient is able go there?” or does it mean “we are ready to send the patient there”.

This means you might receive different answers, depending the perspective the user is answering the question from! This should be prevented at all circumstances!

3. Triage Questions are easy to understand

“Do you want to prevent to get no sugar with cream?”

Now what are you going to prevent exactly?

Sometimes it is already difficult enough to answer the question… So please do not make a riddle from the question?

4. Clear Standards and Criteria

Defining unambiguous questions and unambiguous answers requires very clear standards and criteria. These can be very obvious things like “what is left and what is right?” (left-right confusions are still a notorious problem in health care!).

Discussions about when to choose A and when should it be B can consume a lot of time; however this is part of knowledge gathering in this stage which I consider to be highly valuable!

Given a certain situation, any body making a decision at that point should come up with the same decision

4. Structure of Triage design

A good triage is full- ánd fool proof… This means it should work at all circumstances. Things tend not to go wrong when everything is OK. It is precisely when things gó wrong, under pressure, when there is panic, that decision making should work fast and flawlessly!

  • What do I have to do if I can not answer one of the questions?
  • And what if I made a wrong decision at any point?
  • What should I do if the answer is neither A or B but an unexpected C?
  • Is it possible to answer a specific question at this specific moment in the triage? (“If you do not like the product you may return it in an unopened package”)

5. Linearity vs Complexity

The average decision tree leads the user in a linear way from A to K to Z.

However sometimes decision are based on a multitude of parameters, and is the decision making process more like a three dimensional model. Statistical method’s like DOE can be helpfull and descriptive models like BRMS’es (business rules management systems using ‘business rules engines’) and Decision Model and Notation (DMN) in the end all meet at the same point: “If x….”

Complex decisions can rapidly have the disadvantage of going beyond the ‘human scale’ and thus become tricky since it is difficult to understand ‘what will happen when’ AND (most problematic)  to test whether te outcome WILL be correct in all and any situation (there are multiple examples of plains crashing due to this phenomenon; now what if a fail is far less obvious and visible?) 

A golden rule is still: “If you can not explain it to your neighbor, it probably sucks…”. Or in more friendly words: “Detect and untangle any complexity: Keep it Strait-forward and Simple (KISS)

Leaving Linear systems bring us into the fascinating world of non-linear dynamics.

Don’t worry, in my 40 years practice in improving processes and systems, I never met a situation where I needed it. So lets assume that the vast majority of your problems can be solved with straight forward methods. Also when you are working in complex environments lik building airplanes or in a hospital.

To achieve a robust and proven correct structure to the triage process, a Process Structure Diagram (PSD) is a tremendously strong tool!

6. The Decision process needs a System

A complex triage/decision tree is a process: from unsorted to sortet, from unknowing to knowing. Like any process it needs a system to be perfomed by; Only robust systems can perform robust processes. So not only putting knowledge in to the decision model is needed; also the way the decision model will be excecuted in order to spit out the right outcome is important! These are TWO different assignments when developing such a solution!!

Conclusion: Good decision making needs profound knowledge (which can be embedded in a decision tree) AND a system where this decision making process (using the tree) can be executed in order to bring the right result.


© 1996-2021 Arno Koch. Please contact Arno Koch if you want to use content of this site!