Computed-aided Decision Support (CDS)
- John Svirbely
Description
A CDS pattern is a specialization of the multi-perspective pattern. It is used to model a (clinical) decision made by a human agent that is informed by a decision made by a computer. The computer decision aids the human by making the decision and proposing a recommendation.
Problem
Distributed cognition is common in knowledge work. Human decisions are often aided by computerized decision algorithms. The computer decisions may be overridden when the human agent takes into account inputs that are not available to or used by the computer.
In healthcare, clinical decision support (CDS) systems present alerts, reminders, and recommendations to clinicians at the point of care to enforce desired behaviors or prevent errors and omissions. For example, there are well-accepted guidelines on prescribing anticoagulation therapy for patients with atrial fibrillation that take into account a patient's likelihood of stroke (CHA2DS2-VASc) and likelihood of major bleeding (HAS-BLED). CDS systems can calculate these scores and make a recommendation on whether to prescribe anticoagulation treatments. The clinician takes that recommendation into account, along with other inputs (including data found and also not found in the patient's record as well as data contained in the patient's record but not used in the logic of the CDS rule), to make the decision whether to prescribe an anticoagulant.
Applicability
This pattern applies under the following conditions:
· There are two distinct decision-makers trying to make the same decision: The human and the computer system;
· Both decision-makers base their decisions on the same premises to the degree possible. (data, information, and knowledge); and
· The human is the decision-maker who is ultimately responsible and accountable for the decision.
Pattern
Figure 1 – DMN: Computer-aided decision support (CDS)
Create two decision elements with the same questions and answers: one with the human (clinician) decision- maker and one with the computer (CDS) as decision-maker. The decision made by the CDS system informs, and thus is a sub-decision of, the clinician's decision. The same knowledge source should inform both the CDS decision and the clinician's decision. The actual logic used by the CDS system is computable by definition and should be modeled explicitly using the business knowledge model (BKM) element.
Knowledge sources include evidence-based guidelines, policies, etc. If a local adaptation of a more general guideline exists, the local version should be used. Best practice is to use the actual authority that was used as the basis for the CDS rule.
The BKM can either embed the logic if the decision model is intended for computation or can reference some external location where the logic is available in a computable format. This mechanism could allow a direct connection between the decision model and the organization's CDS rules library and insulates the overall model from changes to the CDS system.
It is assumed that the clinician has access to all of the data elements (usually from the electronic health record or similar sources) that the CDS system uses. These data elements are modeled as input data shared by both decisions. The converse is usually not true; the clinician may have access to other information that is not available to or used by the CDS system.
Specific Example
Figure 2 – DMN: Computer-aided decision support (CDS) Specific
The example illustrates the long-term anticoagulation decision described above. In this case, the CDS system calculates the CHA2DS2-Vasc score and the Has-Bled score and on the basis of those scores, recommends long- term anticoagulation.
In some scenarios, the CDS system will recommend anticoagulation. The clinician will consider the recommendation and also the patient's occupation. Should their job put the patient at high risk for injury, the clinician may decide not to initiate anticoagulation.
Metadata
<blank>
Discussion
<blank>
Related patterns
<blank>
