Team-based Decisions (BPMN and CMMN)
- Tracie Berardi
- John Svirbely
Description
Clinicians often make "chains" of decisions: complex cognitive processes such that the output of each decision informs the next decision and also what decisions to make next. The chain is implicitly aimed to optimize the cognitive load, so that attention is only focused on what matters, when it matters, and if – and only if – it matters.
For example, a guideline for the treatment of atrial fibrillation may recommend putting a patient on a long-term anticoagulant regime. A physician will eventually have to place an order, which requires a level of detail regarding the actual drug and (among other things) its form, dose and schedule. Way before then, however, the physician will determine whether the patient is an actual candidate for anticoagulation and only then what kind of anticoagulant to administer (e.g., warfarin vs. oral anticoagulant).
Problem
Even if decision models can be refactored without altering the overall semantics, some formulations can be more efficient than others. Part of what is called "expertise" is being able to optimize cognitive processes. In a decision model, this can be reflected by virtue of identifying, and properly separating, the decisions that deal with the problem at hand from those decisions that guide the cognitive process.
The optimal cognitive process can then be represented explicitly in a variety of ways. We propose two alternative representations: one based on BPMN and one based on CMMN. In this sense, this pattern specializes the Workflow-Interleaved Decision one.
A BPMN model, prescriptive by nature, can be used to specify the optimal, "natural" flow of decisions. In this model, each decision should be made, and made at a specific point in time (at least relatively to the other decisions).
In a CMMN model, adaptive by nature, the "control" decision determines the applicability (at planning time when the case is opened) and/or the eligibility (reactively, during the execution of the case) of other decisions.
Applicability
· A complex decision problem can be decomposed into a tree of Decisions
· Intermediary decisions influence what decision to make next
Specific Example
In the anticoagulation scenario, the clinician tries to determine as soon as possible whether the patient is eligible for anticoagulation. If that is not the case, no more effort will be invested in further trying to determine the details of an anticoagulation therapy that will never be prescribed. In this sense, this first decision is an "applicability decision."
The second decision, a "choice decision," resolves some of the degrees of freedom of the problem (i.e., which anticoagulant to use), but not all (e.g., the dose and schedule). In fact, it does not make sense to even frame the problem of dosing the drug until the choice of which drug has been made.
The BPMN representation of this process used branching logic (gateways with guards) to orchestrate the decisions. Notice that the applicability decision is Boolean (true/false), while the choice decision is categorical. With this model, only the relevant decisions are made – i.e., only the meaningful information is processes – and only at right point in time.
Figure 1 – DMN: Team-based BPMN decision
The CMMN formulation is analogous but has some important differences. The initial applicability decision serves to determine whether the entire anticoagulation stage shall be part of an individual case or not. If so, the choice of which anticoagulant is deferred to the proper time, as triggered by an event. Because the patient is eligible for anticoagulation, the type of anticoagulant must be chosen at some point before closing the case.
If and when an anticoagulant is chosen, depending on the choice, it is possible to make an appropriate decision regarding the actual dose.
Figure 2 – DMN: Team-based CMMN decision
Metadata
Discussion
Related patterns
